Almut Schuez, Robert Miller - Cortical Areas - Unity and Diversity (Conceptual Advances in Brain Research) - CRC Press (2002)

Cortical Areas:
Unity and Diversity
© 2002 Taylor & Francis

Conceptual Advances in Brain Research
A series of books focusing on brain dynamics and information processing systems of the
brain.
Edited by Robert Miller, Otago Centre for Theoretical Studies in Psychiatry and
Neuroscience, New Zealand (Editor-in-chief), Günther Palm, University of Ulm, Germany
and Gordon Shaw, University of California at Irvine, USA.
Volume 1
Brain Dynamics and the Striatal Complex
edited by R. Miller and J.R. Wickens
Volume 2
Complex Brain Functions: Conceptual Advances in Russian Neuroscience
edited by R. Miller, A.M. Ivanitsky and P.M. Balaban
Volume 3
Time and the Brain
edited by R. Miller
Volume 4
Sex Differences in Lateralization in the Animal Brain
by V.L. Bianki and E.B. Filippova
Volume 5
Cortical Areas: Unity and Diversity
edited by A. Schüz and R. Miller
This book is part of a series. The publisher will accept continuation orders which may be cancelled
at any time and which provide for automatic billing and shipping of each title in the series upon
publication. Please write for details.

Cortical Areas:
Unity and Diversity
Edited by
Almut Schüz
Max-Planck-Institute for Biological Cybernetics,
Tübingen, Germany
and
Robert Miller
University of Otago, Dunedin, New Zealand
London and New York

First published 2002
by Taylor & Francis
11 New Fetter Lane, London EC4P 4EE
Simultaneously published in the USA and Canada
by Taylor & Francis Inc,
29 West 35th Street, New York, NY 10001
Taylor & Francis is an imprint of the Taylor & Francis Group
Typeset in Times by
Integra Software Services Pvt. Ltd, Pondicherry, India
Printed and bound in Great Britain by
TJ International Ltd, Padstow, Cornwall
All rights reserved. No part of this book may be reprinted or reproduced or utilised in any
form or by any electronic, mechanical, or other means, now known or hereafter invented,
including photocopying and recording, or in any information storage or retrieval system,
without permission in writing from the publishers.
Every effort has been made to ensure that the advice and information in this book is true
and accurate at the time of going to press. However, neither the publisher nor the authors
can accept any legal responsibility or liability for any errors or omissions that may be
made. In the case of drug administration, any medical procedure or the use of technical
equipment mentioned within this book, you are strongly advised to consult the manufac-
turer’s guidelines.
British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library
Library of Congress Cataloging in Publication Data
A catalog record for this book has been requested
ISBN 0–415–27723–X

CONTENTS
Preface vii
List of Contributors ix
1 Introduction: Homogeneity and Heterogeneity of Cortical Structure:

A Theme and its Variations 1
Almut Schüz
Part I
THE EMPIRICAL STATUS OF CORTICAL MAPS
2 Cyto- and Myeloarchitectonics: Their Relationship and Possible

Functional Significance 15
Bernhard Hellwig
3 Architectonic Mapping of the Human Cerebral Cortex 29
Katrin Amunts, Axel Schleicher and Karl Zilles
4 Topographical Variability of Cytoarchitectonic Areas 53
Jörg Rademacher
5 Mapping of Human Brain Function by Neuroimaging Methods 79
Rüdiger J. Seitz
Part II
CORTICAL AREAS: CORRELATION WITH CONNECTIVITY
6 Regional Dendritic Variation in Primate Cortical Pyramidal Cells 111

Bob Jacobs and Arnold B. Scheibel
7 Intrinsic Connections in Mammalian Cerebral Cortex 133
Jonathan B. Levitt and Jennifer S. Lund
8 Thalamic Systems and the Diversity of Cortical Areas 155
Catherine G. Cusick
9 Cortical Areas and Patterns of Cortico-Cortical Connections 179
Jon H. Kaas
v
vi Contents
Part III
CONSTANCY AND VARIATION ACROSS SPECIES
10 The Cerebral Cortex of Mammals: Diversity within Unity 195

Facundo Valverde, Juan A. De Carlos and Laura López-Mascaraque
11 Laminar Continuity between Neo- and Meso-Cortex:
The Hypothesis of the Added Laminae in the Neocortex 219
Robert Miller and Rupa Maitra
Part IV
FUNCTIONAL EQUIVALENCE BETWEEN AREAS
12 Cross-Modal Plasticity as a Tool for Understanding the Ontogeny and

Phylogeny of Cerebral Cortex 245
Sarah L. Pallas
13 Do Primary Sensory Areas Play Analogous Roles in Different
Sensory Modalities? 273
Hubert R. Dinse and Christoph E. Schreiner
14 Plastic-Adaptive Properties of Cortical Areas 311
Hubert R. Dinse and Gerd Boehmer
Part V
MORPHOLOGICAL SUBSTRATES OF SEGREGATION
AND INTEGRATION
15 Connectional Organisation and Function in the Macaque Cerebral Cortex 351

Malcolm P. Young
16 The Human Cortical White Matter: Quantitative Aspects of Cortico-Cortical
Long-Range Connectivity 377
Almut Schüz and Valentino Braitenberg
17 Fundamentals of Association Cortex 387
Stewart Shipp
18 Wheels within Wheels: Circuits for Integration of Neural Assemblies
on Small and Large Scales 423
Robert Miller
19 Discussion Section 459

Robert Miller and Almut Schüz

PREFACE
Since the time when Bell and Magendie first showed the different functions of dorsal and
ventral roots of the spinal cord, idea that different functions can be identified with differ-
ent locations in the central nervous system has been central to attempts to understand the
brain. The possibility that different psychological functions might in some way “reside” in
different locations of the cerebral cortex was also an attractive idea, even when scientific
study of the cerebral cortex was in its earliest infancy, as shown by the popularity of
phrenology in the first half of the nineteenth century. This possibility came to have a
firmer empirical basis in the latter half of that century, as a result of the studies of neuro-
logists such as Broca, Wernicke and others. Development of ideas of cortical localization
of function was given further impetus by results of cortical stimulation experiments, and,
in the early twentieth century, from the study of cortical cytoarchitectonics. Nowadays,
a localizationist view of the cortex is also favoured by the widespread use of functional
imaging techniques.
Throughout this long history, an alternative perspective has been advocated periodic-
ally, placing emphasis of the fact that many psychological functions appear not to be
localized in specific cortical regions, or if they are associated with particular cortical
areas, these areas are multiple, and distributed, rather than single and discrete. In the
lesion studies of memory conducted by Lashley it was even concluded that functional loss
depends more on the size of the lesions, rather than its exact location. A modern expres-
sion of this perspective comes from some of those using functional imaging methods, who
are also concerned with widely distributed functions, and document networks of several
cortical areas activated together when particular psychological functions are employed.
Modern morphological work on the cerebral cortex, to which one of us has contributed
also fits into this alternative tradition, cortical connectivity being described and analysed
in terms of broad statistical constraints which might generalize across the whole neocorti-
cal mantle.
These two perspectives might seem antithetical, but this is appearance rather than
reality. It is not a contradiction to believe that some functions have a strict association with
particular cortical areas, while others are based on more widely-distributed cortical
regions. Which of these two perspectives emerges as prominent in an experiment depends
on the way the experimenters frame their questions.
In the chapters below, many aspects of this complex topic are explored. These include
the actual evidence that the cortex can be subdivided into morphologically different areas,
the correlation between such parcellation and patterns of connectivity of various sorts, the
degree to which there is nevertheless an underlying uniformity to the cortex, generalizing
across areas and between species, the functional equivalence of different areas, as well as
the large-scale patterning of cortical functioning, and the overall integration of cortical
vii
viii Preface
functions by interplay with other forebrain structures. all these issues have been discussed
many times in the past. However, we believe it is timely to revisit them, and thus to put
some of these long-standing debates in the context of modern evidence about the structure
and function of the cerebral cortex.
We would like to express thanks to a few people. Claudia Holt was very helpful in the
handling of electronic material and Nicola Arndt in technical assistance with the manu-
scripts. In particular, we thank Valentino Braitenberg for valuable advice and discussions.
The planning of this book was in part done at the Institute for Advanced Studies in
Delmenhorst, Germany. R. Miller expresses his thanks to Professor Gareth Jones, of
Otago University, and to the Schizophrenia Fellowship of New Zealand for continuing
support, and to the Max Planck Institute for Biological Cybernetics, for support during visits
to Tübingen, during the planning and development of this book.
R. Miller, Dunedin
A. Schüz, Tübingen
April 2001

CONTRIBUTORS
Katrin Amunts Hubert R. Dinse

Institut für Medizin Institute for Neuroinformatics
Forschungszentrum Jülich GmbH Dept of Theoretical Biology
52425 Jülich Group Experimental Neurobiology
Germany Ruhr-University Bochum
ND 04
Gerd Boehmer D-44780 Bochum
Institute of Physiology and Germany
Pathophysiology
Gutenberg-University Jon H. Kaas
55099 Mainz Vanderbilt University
Germany Dept of Psychology
111 21st street Avenue South
Valentino Braitenberg 301 Wilson Hall
Max-Planck-Institut für biologische Nashville
Kybernetik TN 37240
Spemannstr. 38 USA
72076 Tübingen
Germany Bernhard Hellwig
Neurologische Universitätsklinik
Catherine G. Cusick Neurozentrum
Dept of Structural and Cellular Biology Breisacher Str. 64
and Neurosciences Programme 79106 Freiburg
Tulane University School of Medicine Germany
1430 Tulane Avenue
New Orleans Bob Jacobs
Louisiana Laboratory of Quantitative
USA 70112 Neuromorphology
Dept of Psychology
Juan A. De Carlos The Colorado College
Instituto Cajal (CSIC) 14 East Cache La Poudre
Avenida del Doctor Arce 37 Colorado Springs
28002 Madrid CO 80903
Spain USA
ix
x Contributors
Laura López-Mascaraque Jörg Rademacher

Instituto Cajal (CSIC) Neurologische Klinik
Avenida del Doctor Arce 37 Heinrich-Heine-Universität
28002 Madrid Düsseldorf
Spain Moorenstrasse 5
40225 Düsseldorf
Jennifer S. Lund Germany
Department of Ophthalmology
Moran Eye Center Arnold B. Scheibel
University of Utah Department of Neurobiology
50 North Medical Drive Brain Research Institute
Salt Lake City UT 84132 University of California
USA Los Angeles
CA 90024-1769
Jonathan B. Levitt
Dept of Biology Axel Schleicher
City College of the City University Institute of Neuroanatomy
of New York Heinrich Heine University
138th Street & Convent Avenue 40225 Düsseldorf
New York NY10031 Germany
USA
Christoph E. Schreiner
Rupa Maitra Coleman Laboratory
Department of Anatomic Pathology W.M. Keck Center for Integrative
Wellington Hospital Neuroscience
Wellington Sloan Center for Theoretical
New Zealand Neurobiology
University of California San Francisco
Robert Miller San Francisco
Otago Centre for Theoretical Studies USA
in Psychiatry & Neuroscience
Dept of Anatomy and Structural Almut Schüz
Biology Max-Planck-Institut für biologische
School of Medical science Kybernetik
University of Otago Spemannstr. 38
PO Box 913 72076 Tübingen
Dunedin Germany
New Zealand
Rüdiger J. Seitz
Sarah L. Pallas Dept of Neurology
Dept of Biology University Hospital
Georgia State University Düsseldorf
PO Box 4010 Moorenstrasse 5
Atlanta GA 30302 40225 Düsseldorf
USA Germany

Contributors xi
Stewart Shipp Malcolm P. Young

Wellcome Dept of Cognitive Neurology Neural Systems Group
University College Dept of Psychology
Gower Street Claremont Place
WC1E 6BT London NE1 7RU Newcastle upon Tyne
England England
Facundo Valverde Karl Zilles

Instituto Cajal (CSIC) C. and O. Vogt Institute of Brain Research
Avenida del Doctor Arce 37 Heinrich Heine University
28002 Madrid 40225 Düsseldorf
Spain Germany

Part I
THE EMPIRICAL STATUS

OF CORTICAL MAPS

2 Cyto- and Myeloarchitectonics: Their Relationship
and Possible Functional Significance
Bernhard Hellwig
Neurologische Universitätsklinik, Neurozentrum, Breisacher Str. 64, 79106 Freiburg, Germany
Tel: ++49 761 270 5001; FAX: ++49 761 270 5390
e-mail: hellwigb@nz11.ukl.uni-freiburg.de
In the human cerebral cortex, a number of cortical areas can be distinguished by anatomical methods.
Two classical types of cortical parcellation have been described, based on cyto- and myeloarchitecton-
ics. In cytoarchitectonics, the definition of areas relies on variations in the sizes and packing densities
of cell bodies. Myeloarchitectonic parcellation is based on the layering, the distribution and the amount
of intracortical myelinated fibres. It is shown here that cyto- and myeloarchitectonics are closely related.
Two simple assumptions are sufficient to transform quantitative cytoarchitectural data into the
corresponding myelin picture. The rules linking cyto- and myeloarchitectonics seem to be essentially
uniform throughout the neocortex. It is also well known that characteristic functional specializations can
be attributed to cortical areas. However, beyond the localization of function, the functional significance
of areal variability in the cortex is largely unclear. For instance, it remains to be clarified why certain
areal adaptations of the basic cortical network seem to be particularly appropriate for the execution
of specific tasks. It is argued that this issue will only be understood when the wiring schemes of each
area are known. Since it is difficult to infer connectivity patterns from cyto- and myeloarchitectonics,
their significance for a functional interpretation of cortical anatomy seems to be limited. The
paper suggests, however, possible strategies that may allow one to describe cortical architectonics in
terms of connectivity.
KEYWORDS: areas, connectivity, cytoarchitectonics, human cerebral cortex, myelin, myelo-

architectonics
1. INTRODUCTION
It has been known for a long time, i.e. since the discovery of the stripe of Gennari in the
primary visual cortex (Gennari, 1782), that the cerebral cortex is not uniform. A number
of histological methods allow one to distinguish cortical areas which are defined by char-
acteristic variations of the basic cortical architecture. Interestingly, this anatomical parcel-
lation of the cortex is not merely descriptive, but is somehow related to cortical function.
Different types of information (visual, auditory, motor, etc.) are processed in different
cortical areas. As yet, this relation between structure and function has been elucidated
mainly in just one respect: Certain functions can be localized in certain areas. Reaching
this conclusion is an important achievement, useful, for example, for a clinical neurologist
who can associate symptoms in a patient with lesions visible in a CT scan. However,
15
16 Bernhard Hellwig
localization of function is not the whole story. Knowing where a certain type of information
is processed does not explain how this is done. The functional significance of areal varia-
tions in the cortex will not be understood until the mechanisms of information processing
as well as its localization can be related to cortical anatomy. For instance, it would be
interesting to know why a piece of cortex that is involved in motor control looks like the
motor area, and not like the primary visual cortex.
The present paper will not be able to solve this problem. However, it will consider two
classical approaches in cortical parcellation, cyto- and myeloarchitectonics, and discuss
whether a functional interpretation is possible beyond the mere localization of function.
1.1. Cytoarchitectonics
In cytoarchitectonics, cortical areas are defined on the basis of cell body stains such as the
Nissl stain. Cortical parcellation relies on variations in the sizes and packing densities of
neurones leading to characteristic patterns of layering (Figure 2.1). The most prominent
maps of the human cerebral cortex worked out on the basis of cytoarchitectural observa-
1o
I 1a
1b
1c
II 2
1 1
III 3
2
III 3
2
3
III 3
3
IV 4
5a
Va
Vb
5b
6aα
VIa
6aβ
VIbα 6bα
VIbβ 6bβ
Figure 2.1. Schematic drawing of a piece of association cortex: cytoarchitectonics (left) and myeloarchitec-
tonics (right). The stripes of Baillarger correspond to the horizontal bands of myelinated fibres in layers 4 and
5b. From Vogt and Vogt (1919).

Cyto- and Myeloarchitectonics 17
Figure 2.2. Brodmann’s map (1909) of the human cerebral cortex (lateral view).
tions are those by Brodmann (1909) (Figure 2.2) and von Economo and Koskinas (1925)
(see Figure 3.1(B) in chapter by Amunts et al., this volume). It is not the aim of this paper
to present all the anatomical details of each area. These can be found in the monographs
by Brodmann and von Economo mentioned above as well as in a more recent treatise by
Braak (1980) which combines cyto- and myeloarchitectural observations with studies on
the pigmentoarchitectonics of the human cerebral cortex. However, in order to give a basic
idea of the cytoarchitectural organization of the neocortex, some general principles should
be mentioned. Following a suggestion by von Economo and Koskinas (1925) the different
cortical areas can be collected into larger groups. Most areas, in particular the association
areas, show the typical six-layered cortex schematically illustrated in Figure 2.1. They are
referred to as homotypical. Areas in which six layers cannot be clearly discerned are called
heterotypical. They come in two forms. First, there is the agranular cortex in which layers
2 and 4 with small, densely packed neurones are not well developed. Examples for the
agranular cortex are the Brodmann areas 4 and 6, i.e. the motor and premotor areas (Figure
2.2). The second type of heterotypical cortex is the granular cortex, which is characterized
by strongly developed layers 2 and 4 with many densely packed, small neurones. This
type of cortex is mainly found in the primary sensory cortices, e.g. in the Brodmann areas
17, 41 and 3 (primary visual, auditory and somatosensory cortex).
1.2. Myeloarchitectonics
The myeloarchitectonics of the human cerebral cortex, based on the layering, the distribu-
tion and the amount of intracortical myelinated fibres, has been described in detail by
Vogt and his co-workers (e.g. Vogt, 1910, 1911; Vogt and Vogt, 1919; Strasburger, 1937;

18 Bernhard Hellwig
Hopf, 1954; Batsch, 1956). Myelin preparations show three types of intracortical fibres
(Figure 2.1): (1) radial fibres (vertical to the cortical surface), (2) oblique fibres, and
(3) horizontal fibres (parallel to the cortical surface). The horizontal fibres are particularly
useful for cortical parcellation. In most areas they form two conspicuous horizontal bands,
the so-called stripes of Baillarger (Baillarger, 1840), which are usually located in layers 4
and 5b respectively (Figure 2.1). The stripes of Baillarger vary from area to area. Again, it
is not the aim of this paper to describe the areal variability of myeloarchitectonic patterns
in detail. The reader is referred to the papers by Vogt and his co-workers mentioned above
as well as to the treatise by Braak (1980). However, some general remarks can be made.
The homotypical cortex, as it was defined for cytoarchitectonics, usually exhibits both
stripes of Baillarger. In some regions, such as the frontal and temporal pole, or areas
located medially in the interhemispheric cleft, only the outer stripe of Baillarger may be
discernible. In the heterotypical agranular cortex the stripes of Baillarger are concealed in
a dense feltwork of fibres, either completely as in the primary motor cortex or partially as
in the premotor cortex where only the outer stripe of Baillarger is visible. In the hetero-
typical granular cortex two myelin patterns can be distinguished. On the one hand, there
is the primary visual cortex with its conspicuous band of horizontal myelinated fibres in
layer 4b, the so-called stripe of Gennari. On the other hand, in the primary somatosensory
cortex or the primary auditory cortex both stripes of Baillarger are present, the inner one
being distinctly more prominent than the outer one.
It is also interesting to consider the areal variability of the total amount of myelin in the
cortex. The degree of myelination diminishes with increasing distance from the primary
areas. It is particularly low in the region of the frontal and temporal pole as well as in areas
located medially in the interhemispheric cleft.
1.3. Cyto- and Myeloarchitectonics: Different Aspects of the Same Underlying

Cortical Network?
Before discussing possible functional implications of the areal variability described by
cyto- and myeloarchitectonics, it seems worthwhile to consider whether there is a relation
between patterns of cell bodies and patterns of myelinated fibres. Finding such a relation
may elucidate aspects of the underlying cortical network. According to Brodmann (1909),
maps of the human cerebral cortex based on either cyto- or myeloarchitectonics are essen-
tially identical. This is corroborated by Sanides’ monograph (1962) on the frontal lobe of
the human brain. Here, the investigation of a series of sections alternatively stained by
a cell body and a myelin stain yielded only one map of areal diversity in the frontal cortex.
Thus, there seems to be a close relationship between cyto- and myeloarchitectonics in
the sense that they obviously reflect different aspects of the same underlying cortical
network. However, what is the nature of this relation? Inspection of Figure 2.1 reveals
that an answer to this question is by no means obvious. While the outer stripe of Bail-
larger, situated in layer 4, corresponds to densely packed, small cell bodies, the inner
stripe of Baillarger, located in layer 5b, coincides with less densely packed, large cell
bodies.
Braitenberg (1962, 1974) put forward a hypothesis as to how cyto- and myeloarchitec-
tonics might be related. He suggested that horizontal intracortical myelinated fibres, i.e.
those fibres forming the stripes of Baillarger, correspond mainly to local axonal ramifica-
tions of pyramidal neurons, the most frequent cell type in the cerebral cortex (Braitenberg,

Figure 2.3. Camera lucida drawing of a Golgi-stained pyramidal cell in the cerebral cortex of the mouse.
A number of horizontally directed axon collaterals originate slightly below the cell body. Bar, 50 µm.
1978; Braitenberg and Schüz, 1998). The bulk of horizontal axon collaterals of pyramidal
cells leave the descending main axon 200 to 300 µm below the cell body (Figure 2.3)
(cf. Cajal, 1911; Gilbert and Wiesel, 1979, 1983; Landry et al., 1980; Martin and Whitteridge,
1984; DeFelipe et al., 1986; Schwark and Jones, 1989). The pyramidal cells which in the
majority of areas are most conspicuous in layers 3 and 5 would thus produce two maxima
of horizontal fibres. These maxima, shifted downwards relative to layers 3 and 5 by 200 to
300 µm, could account for the two stripes of Baillarger. The assumption that horizontal
myelinated fibres in the cortex consist mainly of local axonal ramifications of pyramidal
cells (and not of thalamic or cortico-cortical afferent fibres) is supported by degeneration and
tracer studies (Le Gros Clark and Sunderland, 1939; Fisken et al., 1975; Creutzfeldt et al.,
1977; Gatter and Powell, 1978; Colonnier and Sas, 1978; Levitt et al., 1993). Starting out
from Braitenberg’s hypothesis, Hellwig (1993) showed in a computational study that,
provided quantitative data on the cell body picture of a certain area are given, two simple
assumptions are sufficient to predict correctly the corresponding myelin picture. Part of
this work is reviewed in the following section.

20 Bernhard Hellwig
2. SIMPLE RULES RELATE THE CYTO- AND

MYELOARCHITECTONICS OF THE HUMAN CEREBRAL CORTEX:
UNIFORMITY IN AREAL DIVERSITY
2.1. Cytoarchitectural Data

It was the aim of Hellwig’s study (1993) to compute myelin pictures from quantitative
data on the cytoarchitectonics of different areas. Cytoarchitectural data were taken from
the treatise on the human cortex by von Economo and Koskinas (1925) which contains
detailed descriptions of all areas. Three types of data were considered: (1) layer thick-
nesses; (2) neurone sizes in each layer (specified as the width of a cell body); (3) the volume
density of neurones in each layer (specified as numbers of neurones per 0.001 mm3).
2.2. Two Basic Assumptions

Two basic assumptions were used to transform von Economo’s cytoarchitectural data into
myelin pictures:
2.2.1. First assumption

Large neurones contribute more to the intracortical myelin content than small ones. This
relation can be represented by the sigmoid curve in Figure 2.4. The assumption is hypo-
thetical, but was inspired by observations on Nissl, myelin and Golgi stained sections
through human and non-human cortices.
2.2.2. Second assumption

The average distribution of horizontal axon collaterals of pyramidal neurones can be
quantified by the histogram of Figure 2.5. This histogram is derived from a Golgi study on
pyramidal cells in the rat visual cortex by Paldino and Harth (1977). It was used as a model
of the distribution of horizontal axon collaterals with respect to the cell body. Only one
modification was introduced for the computations: the histogram was scaled to the
0.8
myelin value
0.6
0.4
0.2
0
0 10 20 30 40
diameter of the cell body [µm]
Figure 2.4. A hypothetical curve that transforms the diameter of a neurone’s cell body into a “myelin value”.

100
number of collaterals
80
60
40
20
0
–200 0 200 400 600 800
distance from the cell body [µm]
Figure 2.5. Modified diagram from a study by Paldino and Harth (1977) on pyramidal neurones in the rat
visual cortex. Distances (vertical to the cortical surface) between the endpoints of axon collaterals and the cell
body were measured (positive distances: below the cell body; negative distances: above the cell body). Note that
the bulk of collaterals is located below the cell body.
thickness of each area. Note that the second assumption concerns only pyramidal neu-
rones. For the computation this means that the few neurones in layer 1 which are all of the
non-pyramidal type (e.g. Peters and Kara, 1985) were discarded. In addition, below layer 1,
where the non-pyramidal neurones account for only about 15% of the whole neurone
population (Peters and Kara, 1985; Braak and Braak, 1986; Braitenberg and Schüz, 1998),
all neurones were considered as pyramidal cells.
2.3. Procedure and Results

In all, 14 neocortical areas were chosen for this study. They comprise areas focussed on by
many investigators, and include the motor cortex, the primary sensory areas or the speech
centres. Moreover, they give a fair impression of the variability of myeloarchitectonic
patterns across the human neocortex. Here, the results of just three areas are presented, the
Brodmann areas 4, 7 and 17. The architecture of area 7, a field in the parietal association
cortex, is paradigmatic for the homotypical cortex. Area 4 (primary motor cortex) and area
17 (primary visual cortex), on the other hand, represent the two extremes of the heterotypical
cortex (Nissl sections shown in Figure 2.6a).
Myelin pictures were computed in two steps. First, using data from von Economo and
Koskinas (1925), the average size of neurones in each layer was transformed into a myelin
value by means of the curve in Figure 2.4. The myelin value was then multiplied by the
corresponding number of neurones per unit volume. The procedure yields, for each layer,
a single value which can be considered as the layer-specific contribution to the population
of horizontal myelinated fibres (Figure 2.6b).
In the second step of the computation, it was taken into account that myelin is distrib-
uted along axonal arborizations. This was done by convolving the diagrams in Figure 2.6b
with the histogram of Figure 2.5. This simply provides for shifting the myelin into the
appropriate position (Figure 2.6c).
The densities of myelin thus obtained were represented as shades of grey (Figure 2.6d)
in order to facilitate comparison with real myelin preparations (Figure 2.6e). The simulated

22 Bernhard Hellwig
a b c d e
0
1
Area 7
2
3
5
distance from cortical surface [mm]
1
Area 17
2
3
2
Area 4
5
0 50 100 0 50 100
amount of myelin [%]
Figure 2.6. Computation of myelin pictures for areas 4, 7 and 17 and comparison with real myelin prepara-
tions. (a) Nissl pictures. (b) First step of the computation: the layer-specific amounts of myelin are shown as
a function of the cortical depth. (c) The second step of the computation: the diagrams of Figure 2.6b are convolved
with the histogram of Figure 2.5. (d) Figure 2.6c, transformed into shades of grey. (e) Real myelin preparations.
Bar, 1 mm.
myelin pictures are remarkably close to the real ones. In area 7, both stripes of Baillarger
are visible, in agreement with Vogt’s (1911) original description. In area 17 only one band
of myelinated fibres is conspicuous, the so-called stripe of Gennari (cf. Vogt and Vogt,
1919). In area 4 the comparison between simulation and reality is complicated by the fact
that the myeloarchitectonic patterns differ in two subfields. The simulation is close to the
anterior part of area 4 where, according to Vogt (1910), only the outer stripe of Baillarger
is visible, while the inner one grades into the white matter.
2.4. Conclusion
The findings presented above support the assumption that the stripes of Baillarger consist
mainly of horizontal axon collaterals of pyramidal cells. The two assumptions relating
cyto- to myeloarchitectonics apply also to the other areas investigated in Hellwig’s study
(1993). This suggests that the distribution of horizontal axon collaterals of pyramidal
neurones and the principles of their myelination are remarkably similar in different areas.
Thus, there is obviously both diversity and unity in the cortex. Despite areal variability,

the rules linking cyto- and myeloarchitectonics seem to be essentially uniform throughout
the neocortex.
3. CAN CYTO- AND MYELOARCHITECTONICS BE INTERPRETED

IN FUNCTIONAL TERMS?
3.1. Connectivity and Function

A functionally or computationally relevant description of cortical anatomy will focus on
the connectivity between neurones. It is obvious that the wiring scheme in the cortex
strongly influences how information is processed. Considering the enormous number of
synapses in the cortex, connectivity certainly has to be described in a statistical way. The
cortical wiring scheme can probably be adequately grasped by parameters such as connec-
tion probabilities, number of synapses involved in a connection, the amount of divergence
and convergence or the relative importance of short- and long-range connections. Once
these parameters are known, one should be able to specify the connections of an arbitrarily
selected neurone to other neurones in the cortex in a probabilistic way.
Braitenberg and Schüz (1998) have described the basic machinery of the mouse cortex
on the basis of a statistical analysis of its components. To some extent, these data can be
extrapolated to the human cerebral cortex. However, it is still largely unclear how the
basic cortical wiring scheme varies from area to area. This leads to the question discussed
in the next section: Can the variability of the connectivity scheme in different areas of the
human cerebral cortex be inferred from cyto- and myeloarchitectonics and does this lead
to a better understanding of the mechanisms of information processing in these areas?
3.2. Discussion
In cytoarchitectonics, the local variations of size and packing density of cell bodies are
used for cortical parcellation. The number of synapses on cell bodies is small, rarely
exceeding 200 (Peters and Kaiserman-Abramof, 1970; White and Rock, 1980; Müller et al.,
1984). This is not much compared to the overall number of synapses carried by a cortical
neurone: about 8000 in the mouse and about 40 000 in the human cortex (Braitenberg and
Schüz, 1998). In other words, a method that stains the cell bodies of a neurone cannot be
very helpful for elucidating cortical connectivity. Connections are predominantly located
in the neurophil, i.e. in those parts of the cortical tissue that remain unstained in cell body
preparations.
Nevertheless, a few general statements about connectivity can be made, since the size
of a cell body is positively correlated to the length of its dendritic arborizations (Bok,
1959). For instance, small perikarya which are densely packed indicate that the dendritic
processes are relatively short, thus occupying only a small volume. This applies to layer 4
of the primary sensory areas where the thalamic afferents arrive. The dense packing of cell
bodies points to a local preprocessing of the incoming thalamic information.
Some layers contain large cell bodies which are not so densely packed, e.g. layers 3 and 5
in Figure 2.1. This indicates large and richly ramified dendritic trees, i.e. information is
sampled from a relatively extended piece of cortex. Pyramidal neurones in layers 3 and 5
are the origin of important long-range projections to other cortical or subcortical structures.

24 Bernhard Hellwig
Their large dendrites seem to ensure that the information which is projected contains
a relatively general overview of cortical activity, and not some specialized data about
the processes in small cortical patches.
Beyond this, cytoarchitectonics does not tell us much about patterns of dendritic or axonal
arborizatons which carry the bulk of synapses and are thus most important for cortical
connectivity. In this respect, myeloarchitectonics might be more interesting because it shows
patterns of intracortical fibres.
The main function of myelin is probably to increase conduction velocities. However, it
is doubtful that this is an important property for all intracortical axonal fibres. Many of
them are so short that the actual conduction times are within the range of a few milli-
seconds, even if the whole variability of conduction velocities encountered in the nervous
system is taken into consideration. This is unlikely to be significant (Hellwig, 1993). An
important function of intracortical myelin might be its ability to insulate axonal fibres, in
the sense that myelinated segments of axonal ramifications are unable to form synapses.
This is an interesting property of myelin, since it means that myelination imposes a spatial
structure on axonal trees: in some places they are capable of interacting with other neu-
rones, in others they are not.
The distribution of myelin over the axonal tree is probably not random. The time course
of maturation in the primary visual cortex of the cat suggests that myelination is related to
early learning processes. In the first postnatal weeks, there is a period of extraordinary
plasticity in the visual cortex, the so-called critical period. Plasticity, for example the
susceptibility to the effects of monocular deprivation, is high until some time between the
sixth or eighth week after birth (Hubel and Wiesel, 1970; Olson and Freeman, 1980). On
the level of pyramidal neurones this period is characterized by the emergence and refine-
ment of axonal arborizations (Callaway and Katz, 1990). By pruning of inappropriate
axon collaterals, axonal ramifications are formed in which long horizontal axonal fibres
give off clusters of axon collaterals that preferably contact certain target regions, namely
columns of similar orientation preference (Gilbert and Wiesel, 1989). The process of
shaping axonal trees is experience-dependent (Löwel and Singer, 1992). Axonal arboriza-
tions attain an adult appearance by the end of the critical period, i.e. about 7 weeks after
birth (Callaway and Katz, 1990). Interestingly, this is the time when myelination starts.
The first myelinated fibres in the primary visual cortex of the cat appear by the end of the
sixth postnatal week, myelination is moderate until the end of the eighth week, and then
undergoes an enormous, almost explosive increase (Haug et al., 1976).
In conclusion, two processes seem to coincide at the end of the critical period in the
visual cortex of the cat: the termination of the experience-dependent shaping of axonal
branching patterns and the onset of myelination. Observations on individual pyramidal
cells suggest that the myelinated parts of axonal trees are mainly those horizontal fibre
segments that interconnect clusters of collaterals (DeFelipe et al., 1986) (Figure 2.7).
Thus, myelin would insulate predominantly axonal segments which failed to establish
functional relations with other cortical neurones during the critical period. In other words,
myelin would be a sort of memory trace, a tool to store information about early learning
processes.
The interpretation of myelin as a memory trace by which early experiences are fixed
may explain why the overall amount of myelin is higher in the heterotypical areas, i.e. in
the primary motor and sensory cortices, than in the homotypical association cortex. The
primary areas are in a close relation to the outside world, and a repertoire of information

dendrite
myelin
axon
Figure 2.7. Schematic drawing of a cortical pyramidal cell. Long myelinated horizontal axon collaterals
emanate below the cell body and give off clusters of non-myelinated collaterals.
processing steps fixed by early experiences may be an efficient way to deal with
ever-recurring standard tasks. In the association cortex the tasks to be expected are less
predictable. Thus, a less rigid wiring scheme may be useful in which associations of all
kinds can be learned. In this context, it is also interesting to note that the onset of myelina-
tion is much earlier in the primary areas than in the association cortex (Flechsig, 1920,
1927). All in all, it is, however, most difficult to interpret areal variability as revealed by
myeloarchitectonics in functional terms. This is mainly due to the fact that myelin prepa-
rations, although showing axonal fibres, do not reveal the cortical wiring scheme, since
those axonal fibre segments are stained that, insulated by myelin, are unable to contact
other neurones. In a way, myelin preparations display the “negative” of intracortical
connectivity.
3.3. Outlook
It is an important task for neuroanatomists (and for neuroscientists in general) to relate
structure to function. As far as the parcellation of the cortex into areas is concerned, this
goal has been achieved mainly in one respect: Certain functional specializations can be
attributed to certain areas. However, beyond the localization of function, the functional
interpretation of areal variability in the cortex is largely unclear. In particular, it remains to
be clarified why areal adaptations of the basic cortical network seem to be particularly
appropriate for the execution of specific tasks. For instance, one wonders why the struc-
ture of the motor cortex is obviously useful for the control of movements, but not for other
tasks such as the processing of visual information. In other words, the relation between the
mechanisms of information processing and the areal variability of cortical anatomy is

26 Bernhard Hellwig
unclear. This is due to the fact that the connectivity patterns in each area are largely
unknown. As pointed out above, cyto- and myeloarchitectonics are not very helpful in this
respect.
How can variations of wiring schemes in different areas be elucidated? Knowing the
variability of neuronal arborizations in different areas would in itself be helpful. Unfortu-
nately, this type of information is scarce. For the human cortex, one of the main sources is
still Cajal’s (1911) treatise on the nervous system, which yields some qualitative, but no
quantitative data on the variability of Golgi-stained neurones in different areas. More
recent material is reviewed in this book in chapters by Jacobs and Scheibel (2002) and
Valverde et al. (2002). In all, studying the architectonics of the cortex as revealed by
Golgi or similar methods is still a worthwhile research program.
An approach by which the local connectivity between pyramidal neurones in a given
area can be quantitatively estimated has been suggested by Hellwig (2000). Pyramidal
neurones in layers 2 and 3 of the rat visual cortex were intracellularly stained and three-
dimensionally reconstructed using a computer-based camera lucida system. In a computer
experiment, pairs of pre- and postsynaptic neurones were formed and potential synaptic
contacts, i.e. spatial contacts between axons and dendrites, were calculated. For each pair,
the calculations were carried out for a whole range of distances (0 to 500 µm) between the
pre- and the postsynaptic neurone, in order to describe cortical connectivity as a function
of the spatial separation of neurones. It was also possible to differentiate whether neurones
were situated in the same or in different cortical layers. The data thus obtained were used
to compute connection probabilities, the average number of contacts between neurones
or the frequency of specific numbers of contacts. It could be shown by comparison with
independent data that the local cortical connectivity between pyramidal neurones esti-
mated in this way was a good approximation to reality. In principle, this approach can be
extended to other layers as well as to other areas. This makes it possible to investigate
cortical architectonics in terms of connectivity.
The interpretation of functional processes in cortical areas will certainly be promoted
by knowledge about the underlying wiring scheme. However, data on connectivity could
also be important in another context: They could actually be used to build artificial
neuronal networks with a biologically realistic structure. In such networks, the areal vari-
ability of information processing could be studied. Thus, describing cortical architectonics
in terms of connectivity would not just be an analytic undertaking, but could also serve as
a basis for a synthetic approach.
REFERENCES
Baillarger, J.G.F. (1840) Recherches sur la structure de la couche corticale des circonvolutions du cerveau.
Mémoires de l’academie royale de Médecine, 8, 149–183.
Batsch, G. (1956) Die myeloarchitektonische Untergliederung des Isocortex parietalis beim Menschen. Journal
für Hirnforschung, 2, 225–270.
Bok, S.T. (1959) Histonomy of the Cerebral Cortex. Amsterdam, London: Elsevier.
Braak, H. (1980) Architectonics of the Human Telencephalic Cortex. Berlin, Heidelberg, New York: Springer
Verlag.
Braak, H. and Braak, E. (1986) Ratio of pyramidal cells versus non-pyramidal cells in the human frontal iso-
cortex and changes in ratio with ageing and Alzheimer’s disease. In: D.F. Swaab, E. Fliers. M. Mirmiran,
W.A. van Gool and F. van Haaren (eds), Aging of the brain and Alzheimer’s disease (Progress in Brain
Research, Vol. 70), Amsterdam: Elsevier, pp. 185–212.
Braitenberg, V. (1962) A note on myeloarchitectonics. Journal of Comparative Neurology, 118, 141–146.

Braitenberg, V. (1974) Thoughts on the cerebral cortex. Journal of Theoretical Biology, 46, 421–447.
Braitenberg, V. (1978) Cortical architectonics: general and areal. In: M.A.B. Brazier and H. Petsche (eds),
Architectonics of the Cerebral Cortex, New York: Raven Press, pp. 443–465.
Braitenberg, V. and Schüz, A. (1998) Cortex: Statistics and Geometry of Neuronal Connectivity. Berlin,
Heidelberg, New York: Springer Verlag.
Brodmann, K. (1909) Vergleichende Lokalisationslehre der Großhirnrinde. Leipzig: Johann Ambrosius Barth.
Cajal, S Ramón y (1911) Histologie du système nerveux de l’homme et des vertébrés. Madrid: Consejo superior
de investigaciones cientificas, Instituto Ramón y Cajal.
Callaway, E.M. and Katz, L.C. (1990) Emergence and refinement of clustered horizontal connections in cat
striate cortex. Journal of Neuroscience, 10, 1134–1153.
Colonnier, M. and Sas, E. (1978) An anterograde degeneration study of the tangential spread of axons in
cortical areas 17 and 18 of the squirrel monkey (Saimiri sciureus). Journal of Comparative Neurology, 179,
245–262.
Creutzfeldt, O.D., Garey, L.J., Kuroda, R. and Wolff, J.-R. (1977) The distribution of degenerating axons after
small lesions in the intact and isolated visual cortex of the cat. Experimental Brain Research, 27, 419–440.
DeFelipe, J., Conley, M. and Jones, E.G. (1986) Long-range focal collateralization of axons arising from
corticocortical cells in monkey sensory-motor cortex. Journal of Neuroscience, 6, 3749–3766.
Fisken, R.A., Garey, L.J. and Powell, T.P.S. (1975) The intrinsic, association and commissural connections of
area 17 of the visual cortex. Philosophical Transactions of the Royal Society London, Series B, 272, 487–536.
Flechsig, P. (1920) Anatomie des menschlichen Gehirns und Rückenmarks auf myelogenetischer Grundlage.
Leipzig: Thieme.
Flechsig, P. (1927) Meine myelogenetische Hirnlehre. Berlin: Springer.
Gatter, K.C. and Powell, T.P.S. (1978) The intrinsic connections of the cortex of area 4 of the monkey. Brain,
101, 513–541.
Gennari, F. (1782) De peculiari structura cerebri nonnullisque eius morbus. Parma.
Gilbert, C.D. and Wiesel, T.N. (1979) Morphology and intracortical projections of functionally characterised
neurons in the cat visual cortex. Nature, London, 280, 120–125.
Gilbert, C.D. and Wiesel, T.N. (1983) Clustered intrinsic connections in cat visual cortex. Journal of Neuro-
science, 3, 116–1133.
Gilbert, C.D. and Wiesel, T.N. (1989) Columnar specificity of intrinsic horizontal and corticocortical connec-
tions in cat visual cortex. Journal of Neuroscience, 9, 2432–2442.
Haug, H., Kölln, M. and Rast, A. (1976) The postnatal development of myelinated nerve fibres in the visual
cortex of the cat. Cell and Tissue Research, 167, 265–288.
Hellwig, B. (1993) How the myelin picture of the human cerebral cortex can be computed from cytoarchitectural
data. A bridge between von Economo and Vogt. Journal für Hirnforschung, 34, 387–402.
Hellwig, B. (2000) A quantitative analysis of the local connectivity between pyramidal neurons in layers 2/3 of
the rat visual cortex. Biological Cybernetics, 82, 111–121.
Hopf, A. (1954) Die Myeloarchitektonik des Isocortex temporalis beim Menschen. Journal für Hirnforschung,
1, 208–279.
Hubel, D.H. and Wiesel, T.N. (1970) The period of susceptibility to the physiological effects of unilateral eye
closure in kittens. Journal of Physiology (London), 206, 419–436.
Jacobs, B. and Scheibel, A.B. (2002) Regional dendritic variation in primate cortical pyramidal cells. In:
A. Schüz and R. Miller (eds), Cortical areas: unity and diversity (Conceptual Advances in Brain Research
series), London: Taylor and Francis Publishers.
Landry, P., Labelle, A. and Deschênes, M. (1980) Intracortical distribution of axonal collaterals of pyramidal
tract cells in the cat motor cortex. Brain Research, 191, 327–336.
Le Gros Clark, W.E. and Sunderland, S. (1939) Structural changes in the isolated visual cortex. Journal of
Anatomy, 73, 563–574.
Levitt, J.B., Lewis, D.A., Yoshioka, T. and Lund, J.S. (1993) Topography of pyramidal neuron intrinsic con-
nections in macaque monkey prefrontal cortex (areas 9 and 46). Journal of Comparative Neurology, 338,
360–376.
Löwel, S. and Singer, W. (1992) Selection of intrinsic horizontal connections in the visual cortex by correlated
neuronal activity. Science, Washington, 255, 209–212.
Martin, K.A.C. and Whitteridge, D. (1984) Form, function and intracortical projections of spiny neurones in the
striate visual cortex of the cat. Journal of Physiology (London), 353, 463–504.
Müller, L.J., Verwer, R.W.H., Nunes Cardoso, B. and Vrensen, G. (1984) Synaptic characteristics of identified
pyramidal and multipolar non-pyramidal neurons in the visual cortex of young and adult rabbits. A quanti-
tative Golgi-electron microscope study. Journal of Neuroscience, 12, 1071–1087.
Olson, C.R. and Freeman, R.D. (1980) Profile of the sensitive period for monocular deprivation in kittens.
Experimental Brain Research, 39, 17–21.
Paldino, A. and Harth, E. (1977) A computerized study of Golgi-impregnated axons in rat visual cortex.
In: R.D. Lindsay (ed.), Computer Analysis of Neuronal Structures, New York: Plenum Press, pp. 189–207.

28 Bernhard Hellwig
Peters, A. and Kaiserman-Abramof, I.R. (1970) The small pyramidal neuron of the rat cerebral cortex. The
perikaryon, dendrites and spines. American Journal of Anatomy, 127, 321–355.
Peters, A. and Kara, D.A. (1985) The neuronal composition of area 17 of rat visual cortex. II. The nonpyramidal
cells.
Sanides, F. (1962) Die Architektonik des menschlichen Stirnhirns. In: M. Müller, H. Spatz and P. Vogel (eds),
Monographien aus dem Gesamtgebiete der Neurologie und Psychiatrie, Vol. 98, Berlin, Göttingen, Heidel-
berg: Springer Verlag.
Schwark, H.D. and Jones, E.G. (1989) The distribution of intrinsic cortical axons in area 3b of cat primary
somatosensory cortex. Experimental Brain Research, 78, 501–513.
Strasburger, E.H. (1937) Die myeloarchitektonische Gliederung des Stirnhirns beim Menschen und Schimpansen.
I. Teil. Myeloarchitektonische Gliederung des menschlichen Stirnhirns. Journal für Psychologie und Neurol-
ogie, 47, 461–491.
Valverde, F., De Carlos, J.A. and López-Mascaraque, L. (2002) The cerebral cortex of mammals: diversity
within unity. In: A. Schüz and R. Miller (eds), Cortical areas: unity and diversity (Conceptual Advances in
Brain Research series), Taylor and Francis Publishers, London, New York.
Vogt, C. and Vogt, O. (1919) Allgemeinere Ergebnisse unserer Hirnforschung. Journal für Psychologie und
Neurologie, 25, 279–461.
Vogt, O. (1910) Die myeloarchitektonische Felderung des menschlichen Stirnhirns. Journal für Psychologie und
Vogt, O. (1911) Die Myeloarchitektonik des Isocortex parietalis. Journal für Psychologie und Neurologie, 18,
379–390.
von Economo, C. and Koskinas, G.N. (1925) Die Cytoarchitektonik der Hirnrinde des erwachsenen Menschen.
Wien, Berlin: Springer Verlag.
White, E.L. and Rock, M.P. (1980) Three-dimensional aspects and synaptic relationships of a Golgi-impregnated
spiny stellate cell reconstructed from serial thin sections. Journal of Neurocytology, 9, 615–636.

3 Architectonic Mapping of the Human Cerebral Cortex
Katrin Amunts1, Axel Schleicher3 and Karl Zilles1,2,3
1
Institut für Medizin, Forschungszentrum Jülich, Germany and 2C. and O. Vogt Institute
of Brain Research and 3Institute of Neuroanatomy, Heinrich Heine University,
Düsseldorf, Germany
Correspondance: Dr. Katrin Amunts, Institut für Medizin, Forschungszentrum Jülich,
GmbH, D-52425 Jülich, Germany
Tel: +49 2461 614300; FAX: +49 2461 618307
e-mail: k.amunts@fz-juelich.de
The classical cyto- and myeloarchitectonic maps of the human cerebral cortex considerably influ-
enced the concept of localization of function. Presently, these maps serve as anatomical references
in functional imaging studies. However, the classical maps suffer from drawbacks such as the highly
observer-dependent definition of areal borders; the fact that they present only a single aspect of
architectonic organization of the cortex (e.g. only cytoarchitecture), and the lack of information on
intersubject variability of location and size of a cortical area in a spatial reference system. Recent
methodological progress in computerized image analysis of histological specimens, the introduction
of markers which reflect various architectonic aspects of cortical organization (e.g. receptor auto-
radiography), and the development of warping techniques to compensate for intersubject variability
of brain structure in 3D made it possible to overcome these drawbacks. We propose a new concept
of architectonic mapping which is based on: (i) a definition of areal borders by using multivariate
statistical analysis, and not by highly subjective judgements; (ii) a quantitative analysis of similarity
and dissimilarity in architecture between cortical areas; and (iii) a multimodal characterization of
cortical organization based on cyto-, myelo- and receptor-architectonic mapping. The comparison of
architectonic maps with functional imaging data in a common standard reference space allows, for
the first time, a direct analysis of correlations between structure and function in the living human
brain, and provides new insights into the architecture of the cerebral cortex.
KEYWORDS: architecture, brain mapping, human cerebral cortex, intersubject variability, transmitter
receptors
1. INTRODUCTION
The classical cytoarchitectonic maps of the human cerebral cortex published by Brodmann
(1909), Campbell (1905), Elliot Smith (1907), von Economo and Koskinas (1925) and the
Vogts (Vogt and Vogt, 1919) have recently gained considerable attention, since they
present mandatory structural data for the microanatomical interpretation of functional
imaging data. These maps, however, do not fulfil the requirements of an anatomical refer-
ence system for functional human brain mapping. For instance, they present only schematic,
simplified drawings of a single, individual brain or hemisphere in a two-dimensional view
without any descriptions of the intersubject variability of cortical architecture. The same is
true for more recent architectonic maps, e.g. by the Russian school (Sarkisov et al., 1949),
29
30 Katrin Amunts, Axel Schleicher and Karl Zilles
Sanides (1962, 1964), Bailey and von Bonin (1951) and Braak (1979). Moreover, these
maps differ between each other with respect to the number, location and extent of cortical
areas (Zilles, 1990).
Campbell subdivided the human cerebral cortex on the basis of cell body- and myelin-
staining into 14 regions, amongst them precentral, frontal, visuo-sensory, and the audito-
psychic areas (Campbell, 1905). Elliot Smith studied the regional and laminar distribution
of myelinated fibers in unstained sections, and proposed a different map containing about
50 areas (Elliot Smith, 1907). Nowadays, the most widely used map is that of Brodmann,
which relies on extensive studies of cell body-stained (Nissl-stain) histological sections
(Brodmann, 1903, 1905, 1908, 1909). He subdivided the cortex on the basis of cytoarchitec-
tonic criteria into approximately 40 cortical areas. Unfortunately, he never described his
criteria for parcellation of most of the areas in sufficient detail. This is true in particular for
so-called higher associative areas like areas of the prefrontal cortex, posterior parietal
lobe, and of the inferior temporal cortex.
Brodmann’s schematic surface drawing of an architectonic map was used by Talairach
and Tournoux as basis of the architectonic parcellation in their stereotaxic atlas (Talairach
and Tournoux, 1988). They simply transferred Brodmann’s areas to their own brain atlas
by trying to identify corresponding sulcal patterns in both brains, assuming a strong
association between the sulcal pattern and borders of cortical areas. Such an association,
however, was already doubted by Brodmann. He mentioned that “ . . . a schematic drawing
can reflect only the major spatial relationships, and therefore, precise topographical
associations1 cannot be considered in general or only in a distorted manner; this is true in
particular for all those cortical regions which have borders in the neighborhood of sulci and
those regions which are located in the depth of such a cortical region” (Brodmann, 1908).
The basis of Brodmann’s research was the working hypothesis that the cerebral cortex
is composed of numerous cortical areas, each of them characterized by a distinct cyto-
architecture and function. Following this concept, the cytoarchitecture of a cortical area
should be more or less constant within a cortical area, but changes considerably at its
border. For example, Brodmann’s area 4 was conceptualized as the anatomical equivalent
of the primary motor cortex which guides voluntary movements (Fritsch and Hitzig, 1870)
and Broca’s region was regarded as the anatomical correlate of the functionally defined
center of speech (Broca, 1861). Although for the vast majority of cortical areas such
as microstructural-function relationship could not be rigorously tested at that time, both
Brodmann and Campbell took architectonic localization of function for granted. The strict
localizationist approach culminated in a map of the human cortex of Kleist (1934) in
which complex functions were assigned to a distinct cytoarchitectonic area. Brodmann’s
area 18 (for instance) was associated with visual attention, perception of spatial position,
and eye movements toward the upper and lower visual field. Brodmann himself did not
represent such an extreme localizational concept (Brodmann, 1909). In order to avoid
a confusion of histological data and unproven evolutionary and functional speculations, he
created his system of a “neutral” nomenclature by numbering different cytoarchitectonic
areas mainly according to their dorso-ventral sequence. Older studies (Vogt and Vogt, 1919)
and more recent electrophysiological studies in nonhuman primates have demonstrated
that the basic idea of Brodmann was true: Neurones with similar receptive fields and
1
i.e. between sulci and areal borders [Au].

Architectonic Mapping 31
Figure 3.1. Cytoarchitectonic maps of the lateral surface of human brain adapted from [A] Brodmann (1909),
[B] von Economo and Koskinas (1925) and [C] the Russian school (Sarkisov et al., 1949). Cytoarchitectonic
areas are marked by different hatches and classified according to Brodmann’s nomenclature by Arabic numerals
[A, C] or according to that of von Economo and Koskinas by letters and numerals [B]. Note differences in sulcal
pattern as well as in shape and extent of the areas (e.g. in the frontal lobe with respect to areas 45, 9 and 46;
compare A with C).

response properties lie within the same cytoarchitectonic area, as found when the same
brain is sectioned and cell-stained following recording experiments, and correlations are
sought between penetration sites and the cytoarchitectonic pattern. Conversely, response
properties of neurones change across cytoarchitectonic borders (Luppino et al., 1991;
Matelli et al., 1991; Tanji and Kurata, 1989).
Although later studies extended and supplemented the maps of Brodmann and Campbell,
they followed the concept of a cortical area implied by these maps. von Economo and
Koskinas (1925) introduced an even more complex subdivision of the human cortex into
cortical areas with regional peculiarities (= subareas). They defined cortical areas on the
basis of their topography (frontal, parietal, occipital, etc.), and in terms of their cytoarchi-
tecture and local peculiarities. As an example, area FAg is characterized by its location in
the frontal lobe (“F”), an agranular cytoarchitecture (“A”) with giant pyramidal cells (“γ ”).
Area FCB, e.g. has common features of area FB and FC, etc. Discussion of this concept,
however, raised controversy in the scientific community. von Economo and Koskinas
applied quantitative criteria (e.g. size of cells, thickness of layers) in order to provide
a more precise characterization of a cortical area, to formalize the cytoarchitectonic
description of cortical areas, and to make it more independent of the experience of the
observer. Finally, the Russian school published another map of the human cerebral cortex
which was, however, based mainly on Brodmann’s approach. Additionally, they tried to
overcome one of the unsolved problems of Brodmann’s map, i.e. the neglect of inter-
subject variability. Their atlas considered intersubject variability in the extent and position
of cortical areas by analyzing a sample of dozens of hemispheres (Filimonoff, 1932;
Kononova, 1935, 1938; Sarkisov et al., 1949).
The increasing number of available architectonic atlases revealed a further problem of
architectonic mapping. Although all the cytoarchitectonic maps were based on the same
concept of a cortical area as an architectonically distinct and homogeneous region, and all
were the result of the same methodical approach, their areal patterns do not match, for
example, with respect to the number of cortical areas, their relationship to sulci and gyri,
as well as to the neighbouring cortical areas. Even if we compensate for interindividual
differences in the macroscopical anatomy of the brains, numerous differences between the
maps can hardly be explained. Thus, in the frontal lobe, area 46 has a common border with
areas 44 and 45 in Brodmann’s map (Brodmann, 1909), but this border is absent in the
map of the Russian school, since here area 9 separates completely area 46 from 44 and 45
(Sarkisov et al., 1949). In a more recent study, transitional areas were defined which
exhibited mixed architectonic features of areas 46 and 45 (Rajkowska and Goldman-Rakic,
1995b). Considerable differences between the maps can also be found with respect to the
anterior border of area 4, the extrastriate visual cortex, and the parietal cortex, where
Brodmann found only a few areas, but recent observations revealed a much higher number
of areas.
What might be the reasons for differences between the maps? One reason concerns
differences in parcellation criteria of the different observers. The most important criteria
used in all studies are the density and size of nerve cells, their distribution within cortical
layers, the absolute and relative thicknesses of cortical layers, the radial and horizontal
arrangement of neurones, the presence of special cells (e.g. giant Betz cells of Brodmann’s
area 4), and locally specific subdivisions of layers into sublayers (e.g. the subdivision of
layer IV of Brodmann’s area 17 into sublayers IVA–C). For the vast majority of cortical
areas, not only one, but a whole complex of criteria is used for its definition. Very often,

these criteria are weighted relative to each other, in a different way by each observer. In
addition, the criteria are sometimes difficult to formalize objectively. This can be illustrated
by the example of such a “simple” area as Brodmann’s area 4. Typical for this area are
giant pyramidal cells in layer V (Betz cells), which were discovered by Betz as character-
istic cells of the motor area of man, chimpanzee, other primates and dog (Betz, 1874).
However, how big is a Betz cell? The height of these cells may vary between different
individuals from 60–120 µm, their width from 30–60 µm. Moreover, comparably large-
sized cells can be found outside area 4 in the area postcentralis gigantopyramidalis
(von Economo and Koskinas, 1925). Furthermore, the distance between single Betz cells
increases towards subarea 4a (Geyer et al., 1996; Zilles et al., 1995). Thus, the border
between area 4 and the rostrally adjoining area 6 is difficult to define on the basis of the Betz
cells-criterion. If giant pyramidal cells are defined not by their absolute size, but by their
relative size (i.e. comparison with cells in neighbouring areas), such cells can be also found
in layers III and V of areas 44 and 45, in the extrastriate visual cortex and in the temporal
cortex (Bailey and von Bonin, 1951). Consequently, a reliable definition of area 4 requires
not only this, but also additional criteria, e.g. the absence of an inner granular layer.
Bailey and von Bonin further followed this line of discussion and asked if there is any
objective basis for a detailed cytoarchitectonic map at all. They came to the final con-
clusion that “ . . . vast areas are so closely similar in structure as to make any attempt at sub-
divisions unprofitable, if not impossible”. As a consequence, their cytoarchitectonic map is
based only on a parcellation into a few main types of cortical regions: regions with numer-
ous granular cells (koniocortex), without granular cells (agranular cortex), with large
pyramids in layer III, and the allocortex, as well as 4 combinations between these main
types. In contrast to the previously mentioned maps of Brodmann and others, their map does
not show sharp borders but gradual transitions between areas (Bailey and von Bonin, 1951).
The question arises about which cortical map is the most appropriate. Is it that of Bailey
and von Bonin with 8 subdivisions, that of Campbell, Brodmann and the Russian school
with about 20 to 40 subdivisions, or that of von Economo and Koskinas with about 100
areas and subareas? One way to answer this questions may be the combination of cyto-
architectonic mapping with other architectonic mapping techniques (multimodal mapping).
Flechsig was the first to gave a detailed subdivision of the neocortex into 40 cortical areas
by his myelogenetic method, i.e. by studying the heterochronous development of myeli-
nation in the white matter immediately below the cortex during foetal and early postnatal
periods (Flechsig, 1898). The Vogts and their co-workers subdivided the human cortex on
the basis of myeloarchitectonic criteria (distribution and density of myelinated axons
within the cortex) into more than 150 fields (Lungwitz, 1937; Riegele, 1931; Strasburger,
1938; Vogt and Vogt, 1919; Vogt, 1919). Their map and the underlying nomenclature
were quite complex and difficult to verify for other observers. This might be one reason
why it did not reach general acceptance in subsequent years. More recent methods of
cortical mapping, e.g. by immunohistochemistry (Bidmon et al., 1997; Campbell and
Morrison, 1989; Hendry et al., 1994; Tootell and Taylor, 1995; Zilles et al., 1991c),
histochemistry (Burkhalter and Bernardo, 1989; Clarke, 1994; Wong-Riley et al., 1993),
pigmentoarchitecture (Braak, 1977, 1979) and regional and laminar distribution of dif-
ferent transmitter receptor binding sites (Dietl et al., 1987; Jansen et al., 1989; Zilles and
Clarke, 1997; Zilles and Schleicher, 1995; Zilles et al., 1988, 1991d) proved to be valu-
able alternatives in architectonic research. Most importantly, the maps based on different
histological and histochemical techniques frequently show a perfect spatial coincidence of

many areal borders, thus corroborating the position of an areal border by multimodal
imaging. Moreover, since a single receptor may not reveal all borders demonstrated by
other markers, this finding can be used to define a family of neurochemically related areas
by studying the regional pattern of one transmitter receptor, and comparing its distribution
with the maps revealed by other receptors or by cytoarchitecture. We think that such a
multimodal concept of cortical mapping improves and supplements classical cytoarchitec-
tonic analysis.
We will show below, that the architectonic analysis of any histological or histochemical
specimen can also be improved considerably by using quantitative measurements and
statistically testable image analysis procedures:
(i) Borders between cortical areas can be identified by observer-independent statistical

analysis of local changes in cytoarchitecture (Schleicher et al., 1999). We will illustrate
this approach in cytoarchitectonic specimens, although it can also be applied to receptor
architectonic and myeloarchitectonic specimens (Zilles and Schleicher, 1993).
(ii) We will present a method for quantifying cytoarchitectonic differences between
cortical areas. This method defines the similarity or dissimilarity between cortical
areas in terms of numerical distance measures. Using this approach, it can be tested
statistically whether differences in cytoarchitecture (or any other architecture) are
significant. Furthermore, it allows one to test the long-standing hypothesis of the
gradual, rather than distinct character of the majority of cytoarchitectonic borders
(Bailey and von Bonin, 1951).
(iii) In contrast to previous cortical maps which were based on only one technique, multi-
modal architectonic analysis will be performed. We will discuss the correspondence
and differences of architectonic borders which are revealed by receptor autoradio-
graphy of numerous different receptor binding sites, as well as by cytoarchitecture.
Human striate and extrastriate areas, as well as Brodmann’s areas 44 and 45 (Broca’s
region) will serve as examples for multimodal mapping of the cerebral cortex.
(iv) We will conclude with some perspectives on the application of these maps in a three-
dimensional probabilistic atlas system.
2. OBSERVER-INDEPENDENT DEFINITION
OF CYTOARCHITECTONIC BORDERS
One of the key features of the neocortex is its organization in layers running parallel to the
pial surface. Cortical layers differ by their absolute and relative widths and cell densities.
The laminar pattern of a cortical area is represented by its sequence of layers, varying in
cell density. Our observer-independent approach to the definition of cortical borders
considers these architectonic features. It is based on the assumption that each area has
a unique, homogeneous laminar pattern, which distinguishes it from those of neighbouring
cortical areas. Several methods have been applied in the past for quantifying the laminar
pattern. An early approach was described by Hudspeth and colleagues (1976). They analyzed
optical density profiles to describe the distribution of staining intensity across cortical
layers in the human primary visual cortex. Although the optical density is an easy and fast
measurable parameter, it has the major disadvantage of being sensitive to differences in
staining intensity of nerve cells (and of the background) in different brains and sections.

For technical reasons, such differences are almost inevitable in histological specimen.
Variations in intensity are influenced by factors like age, clinical history, cause of death,
post mortem delay, autopsy conditions, and histological techniques (Blinkov and Glezer,
1968; Haug, 1980; Skullerud, 1985; Vierordt, 1893).
Based on this experience, we used the volume density of nerve cells in order to quantify
the laminar pattern, a parameter with a long tradition in quantitative neurobiology (Haug,
1956; von Economo and Koskinas, 1925). It has the advantage that, within reasonable
limits, it is not affected by either staining, or anisotropy (Weibel, 1979). The volume
density of nerve cells was estimated as the areal fraction of all stained cellular profiles in
square measuring fields of 20–30 µm and defined as gray level index (GLI) ranging from
0% to 100% (Schleicher and Zilles, 1990). Other stereological parameters (e.g. the numer-
ical density) are also available, but we focussed on the volume density since this robust
stereological parameter can be automatically estimated from existing histological series in
large samples. This parameter is highly correlated with the volume density of neurones,
since the density of endothelial and glial cells does not vary systematically throughout the
cortical layers (Wree et al., 1982). Using a computerized image analyzer, the GLI was
measured in cortical regions of interest (Figure 3.2). GLI images were achieved from
which GLI profiles (= density profiles) reaching from the border between layers I and II to
the border between cortex and white matter were extracted. The shape of these density
profiles describes quantitatively the laminar pattern, i.e. the cytoarchitecture of a cortical
area. Dissimilarities between cortical areas and their laminar patterns were reflected by
differences in shape of the density profiles. The shape of a profile was numerically
described by a set of ten features: the mean of the amplitude (i.e. the mean GLI; meany.o),
the center of gravity in the x-direction (meanx.o), the standard deviation (sd.o), the skew-
ness (skew.o), the kurtosis (kurt.o), and the analogous parameters for the first derivative of
each profile (meany.d, meanx.d, sd.d, skew.d, kurt.d). Features are based on central
moments (Dixon et al., 1988) of the original density profile, and on its first derivative, by
treating the profile as a frequency distribution, whereby the cortical depth is the x-value
and the GLI is the frequency value at that x-value. Features were normalized in order to
weight them equally. Some features can be interpreted directly in terms of cytoarchitecture:
the mean GLI increases with increasing density of cell bodies. The feature meanx.o will be
smaller than 50% if the supragranular layers have a higher GLI than the infragranular
layers. Vice versa, if the infragranular layers show more densely packed cell bodies than
the supragranular layers, the meanx.o will be shifted to a value greater than 50%.
Multivariate statistical analysis was then used in order to quantify differences in shape
between profiles. The Mahalanobis distance D was used as a multivariate measure of
differences in shape between neighbouring profiles for detecting cytoarchitectonic borders
(Schleicher et al., 1998, 1999). The basic idea was that profiles are more or less similar in
shape within a cortical area (homogeneity criterion), and the shape changes abruptly at the
border of two neighbouring areas (Schleicher et al., 1995). In order to detect the position of
the border, cortical regions of interest were covered by a sequence of equidistant density
profiles (Figure 3.2A). The Mahalanobis distance was then calculated between two
neighboring sets (= blocks) of profiles (Figure 3.2C). If these two blocks belong to one
and the same area, the Mahalanobis distance was small, since differences in the laminar
pattern between these two groups of profiles were small. Vice versa, if these two blocks
were located exactly at opposite sides of a cortical border, the Mahalanobis distance was
maximal since differences in the laminar pattern of these two groups of profiles were

Figure 3.2. Observer-independent definition of cytoarchitectonic borders of the visual cortex. The GLI as
a measure of neuronal packing density (Wree et al., 1982) was obtained in a histological section stained for cell
bodies (Zilles and Schleicher, 1980; Zilles et al., 1986; Amunts et al., 2000; Schleicher and Zilles, 1990). As a
result, a GLI image [A] was produced, in which each pixel corresponds to a GLI value measured with a spatial
resolution of 25 µm. Light pixels correspond to a low packing density, dark pixels to a high density. The cortical
region of interest was covered by a sequence of profiles, indexed consecutively from 1 to 242 [A]. Each profile
quantifies the course of the GLI from the border between layers I and II to the border between the cortex and the
white matter (along a line perpendicular to the cortical surface). A multivariate distance measure, the Mahalanobis
distance D, was calculated (Schleicher et al., 1998) [C]. D is a measure of difference in profile shape between
neighbouring blocks of profiles; e.g. D at the position of profile 20 was calculated as the difference in shape
between profiles 1–20 and profiles 21–40 [C]. Since 20 profiles of one block were compared with 20 profiles
of the neighbouring block, the block size in this case was 20. D was calculated for different block sizes ranging
from 8 to 24 [D]. The dots mark the positions of significant Mahalanobis distances for each block size and each
position of the profile. For block size 20, significant distances were obtained from the graph of [C]. Significant
values of the Mahalanobis distance are marked by red circles and lines. In this histological section, borders
were quantitatively defined between areas V1 and V2d (large arrowhead at position 61), within area V2d
(small arrowheads at positions 103 and 153), and between areas V2d and V3 (large arrowhead at position 173),
and transferred to the original histological section [B]. The border between areas V1 and V2d corresponds to
the border between Brodmann’s areas 17 and 18, that between V2d and V3 to the border between Brodmann’s
areas 18 and 19 (Amunts et al., 2000; Gattass et al., 1981; Newsome and Allman, 1980; Newsome et al., 1986;
Zilles and Clarke, 1997). Scal—Sulcus calcarinus. (see Color Plate 1)

large. After the calculation of the Mahalanobis distance for the two adjacent blocks of
profiles, both blocks were shifted simultaneously by ≈128 µm (i.e. by the width of one
profile) to the next position. In this manner, the Mahalanobis distance was calculated
continuously for all sequential positions of all possible blocks of profiles in the region
studied (Figure 3.2C). Distances were calculated for different block sizes ranging from
8 to 24 profiles per block (Figure 3.2D). They were calculated between blocks of profiles,
and not between single profiles, in order to improve the signal to noise ratio. A subsequent
Hotelling’s T2 test (with a Bonferroni correction of the p-values) was applied for testing
the significance of each value of the Mahalanobis distance. Borders were defined at those
positions of profiles where the following criteria were fulfilled. The Mahalanobis distance
D is significant (Hotelling’s T2-test; α = 5%), positions with significant D were stable across
different block sizes and could be followed up through neighboring histological
sections.
In the histological section shown in Figure 3.2, the Mahalanobis distance reaches sig-
nificant values at the areal borders between areas V2d and V1 at position 61, and between
areas V2d and V3 at position 173. Within area V2, the distance shows local maxima at
positions 103 and 153. In our sample, internal subparcellations of V2 were associated with
the presence or absence of large pyramidal cells in deep layer III. Borders within area V2
have been described in the past by several authors using cytoarchitectonic criteria
(Amunts et al., 2000; von Economo and Koskinas, 1925), myeloarchitecture (Lungwitz,
1937; Sanides and Vitzthum, 1965a,b) as well as on the basis of cytochrome oxidase stain-
ing (Burkhalter and Bernardo, 1989; Clarke, 1993; Clarke and Miklossy, 1990; Gattass
et al., 1997; Lewis and Olavarria, 1995; Merigan et al., 1993; Tootell and Taylor, 1995).
Thus, this approach not only confirms borders between well known cytoarchitectonic
areas according to Brodmann’s map, but it also detects new subdivisions. A further example
is the subdivision of Brodmann’s area 4 into an anterior and a posterior part (Geyer et al.,
1996). Recently, areas 3a and 3b were confirmed in cytoarchitectonic specimens (Geyer
et al., 1999) using this method. These areas were first mentioned by Brodmann (1909)
and later explicitly described by the Vogts in their myeloarchitectonic map (Vogt and
Vogt, 1919).
3. HOW DIFFERENT ARE TWO CORTICAL AREAS IN THEIR

CYTOARCHITECTURE?
Whether a cortical region is homogeneous in architecture and thus constitutes a single

cortical area or, alternatively, consists of two or more cortical areas, has been a matter of
controversy between different observers. Consequently, the different cortical maps display
a different number of cortical fields. The number reaches from 8 (Bailey and von Bonin,
1951) to more than 100 (von Economo and Koskinas, 1925; Vogt and Vogt, 1919). The
analysis of interareal differences in cytoarchitectonics becomes even more complicated
due to intersubject variability in architecture of the same cortical area in different brains.
Cytoarchitectonic variability has been described since the early days of architectonic
research (Kononova, 1938; von Economo and Koskinas, 1925). Other authors mentioned
it as “considerable”, but the degree of variability was not quantified. Intersubject vari-
ability in microstructure makes it often difficult or even impossible to detect reliably subtle
differences in cytoarchitecture between areas. Finally, the statement of whether several

cortical areas are more similar (or different) in cytoarchitecture cannot be verified by
using pure visual inspection.
Thus, analysis of interareal differences has to be based on measurements. Most studies
in the past relied on the measurement of single morphometric parameters within a certain
sample, e.g. the dendritic length (Hayes and Lewis, 1996; Huttenlocher, 1979), cell sizes
(Blinkov and Glezer, 1968; Hayes and Lewis, 1996; von Economo and Koskinas, 1925),
the layer thickness (Amunts et al., 1995; Harasty et al., 1996; Zilles et al., 1986), the sizes of
cortical areas, subcortical structures and fibre bundles (Andrews et al., 1997; Filimonoff,
1932; Geyer et al., 1999; Haug, 1987a; Kononova, 1935, 1938; Rajkowska and Goldman-
Rakic, 1995b; Stensaas et al., 1974). Stereological parameters have also been applied
successfully (Brody, 1955; Gundersen et al., 1988; Haug, 1984, 1987a,b; Henderson
et al., 1980; Pakkenberg and Gundersen, 1997; Schmitz et al., 1999; Terry et al., 1987;
West, 1993; Zilles et al., 1986). Altogether, these parameters represent important, quanti-
tative data of cortical microstructure. However, they often reflect only a single aspect of
cortical microstructure (e.g. cell density) and do not consider that the cortex is a layered
structure with local changes in cell density, size and number within cortical layers and
sublayers. In a more recent study on the human frontal lobe, several cytoarchitectonic
parameters of areas 9 and 46 were analyzed (Rajkowska and Goldman-Rakic, 1995a,b).
Hereby, a three-dimensional counting method (Williams and Rakic, 1988) was applied to
measure total cortical and relative laminar thicknesses, neuronal packing density per
0.001 mm3 in individual cortical layers, and sizes of neuronal somata in selected cortical
layers. The analysis of these morphometric parameters revealed differences between both
areas in the thickness of layer IV, in the packing density of neurones as well as in the size
distribution of neurones. The authors concluded that objective cytometric methods can
clearly distinguish two adjacent areas within the human prefrontal lobe. In this study,
different morphometric parameters were treated and interpreted separately.
It is also possible to combine cytoarchitectonic parameters and to analyze them by
multivariate statistical analysis. Such an approach has been used by us in defining areal
borders (see above). The multivariate approach offers the advantage that different
morphometric parameters can be normalized by compensating for different scales and can
be combined into one feature vector. The feature vectors are then used in a comprehensive
statistical test. In addition, multivariate analyses (e.g. discriminant analysis) take into
account the correlations – often high – between parameters and offer procedures to detect
those parameters which contribute most to the dissociation between areas.
We illustrate this multivariate approach and discuss its implication for a group of five
cortical areas: Brodmann’s areas 6, 44, 45, V1 and V2. These areas were selected by the
following considerations:
(i) The terms Broca’s region and Broca’s area are based on functional concepts. They
are used inconsistently with respect to cytoarchitecture. It is widely accepted that
areas 44 and 45 constitute Broca’s region (Aboitiz and Garcia, 1997; Amunts et al.,
1999; Kononova, 1949; Petrides and Pandya, 1994; Roland, 1993; Uylings et al.,
1999), but the terms Broca’s region and Broca’s area are also applied for areas 44, 45
and 47 (Riegele, 1931; Vogt, 1910), as well as for area 44 only (Galaburda, 1980; von
Economo and Koskinas, 1925). In some recent functional imaging studies, Broca’s
region (or area) refers to a cortical region which includes area 44 and, sometimes,
adjacent area 6 (Paulesu et al., 1993; Petrides et al., 1993). Recent fMRI studies

have provided evidence that the posterior part of Broca’s region, area 44, and the
homologue region of the right hemisphere might be involved in imagery of movement
(Binkofski et al., 1999, 2000). This would associate area 44 functionally with area 6.
Originally, the definition was based on gross macroscopical markers, i.e. by gyri and
sulci (Broca, 1861; Herve, 1888). In this context, we wanted to check whether areas
44 and 45 can be grouped together on the basis of cytoarchitectonic similarity.
(ii) If areas 44 and 45 constitute Broca’s region, they should be more similar in cytoarchi-
tecture to each other than to neighboring cortical areas, e.g. to the adjacent ventral
part of area 6.
(iii) It was expected that areas 44 and 45 would differ even more from areas V1 and V2
than from area 6, because V1 and V2 belong to the visual cortex and are characterized
by a completely different organization of their input and output, reflected by different
laminar patterns.
(Dis-)similarities between these areas were quantified by calculating a multivariate

distance measure between the density profiles in these areas. Profiles were obtained from
cytoarchitectonic areas 6, 44, 45, V1 and V2 of ten human post mortem brains (Amunts
et al., 1999, 2000). Fifteen to thirty profiles were obtained from three randomly selected
sections of each area, hemisphere and brain. Thus, a total of about 3000 profiles were
processed. Ten features were extracted from each of the profiles, as described above, for
the definition of borders. In contrast to the latter approach, the Euclidean distance (and not
the Mahalanobis distance) was used as multivariate distance measure. The advantage of
the Euclidean distance for this type of analysis is that it is more sensitive in detecting the
dissimilarity in architecture between cortical areas. i.e. the Euclidean distance measures
the absolute distance between two centroids of the ten-dimensional space (= dissimilarity
between areas), whereas the Mahalanobis distance depends not only on this distance, but
also on the variability within an area. Thus, the Mahalanobis distance becomes smaller
with increasing variance (Schleicher et al., 2000).
The Euclidean distance was calculated in each individual brain for all ten possible
combinations of two areas from the areas 44, 45, 6, V1 and V2 (= interareal differences).
It was also calculated between profiles from corresponding areas of the left and the
right hemisphere (= interhemispheric differences). Corresponding distances were aver-
aged across the whole sample size. Multidimensional scaling (Systat® for Windows, Ver-
sion 9, SPSS, USA) was applied for data reduction and visualization of distances between
the cortical areas. Interhemispheric differences were tested statistically against differences
between randomly selected profiles from one and the same area. The results are shown in
Figure 3.3.
The analysis showed a high degree of similarity in cytoarchitecture of areas 44 and 45.
Both areas differed considerably from areas 6 as well as V1 and V2 (large distances
between the centroids). Area 6 showed shorter distances to areas 44 and 45 than to V1
and V2, which may correspond to the close topographical and functional relationship
between areas 44/45 and 6. On the basis of the cytoarchitectonic similarity of areas 44
and 45, these data provide an anatomical argument to combine areas 44 and 45, but not
44 and 6 into a region. Based on classical cytoarchitectonic descriptions, area 44 (which
is dysgranular) takes a transitional position between area 45 (which is granular) and area
6 (which is agranular). This relationship is kept in the arrangement of the centroids in
the graphs.

Figure 3.3. Dissimilarities of Brodmann’s areas 44, 45, 6, as well as areas V1 and V2 based on quantitative
cytoarchitecture (K. Amunts, unpublished observations). Euclidean distances were calculated as multivariate
measures of dissimilarity between profiles of different cortical areas (= interareal differences) and of the two
hemispheres of one and the same area (= interhemispheric differences). Euclidean distance is based on features,
which characterize the shape of the profiles of an area (see text). Multidimensional scaling was applied for
visualization of interareal differences and for data reduction to a two-dimensional plane defined by dimension-1
and dimension-2 (Schleicher et al., 2000). The larger the dissimilarity in cytoarchitecture between two areas, the
larger the distance between them in the graph. Error bars indicate standard errors. Results in this upper graph
quantify interareal differences of the left hemisphere. Whereas areas 44 and 45 were found to be very similar in
cytoarchitecture, both areas differed considerably from area 6 as well as from visual areas V1 and V2.
Interhemispheric differences in cytoarchitecture (lower graph) were significant (marked by an asterisk) for
areas 44 and 45, but not for areas 6, V1 and V2. The line marks the level of intersubject variability in cytoarchi-
tecture. It was calculated as the average Euclidean distance between corresponding areas across different
subjects (i.e. the distances in shape within the sample of 10 brains between all areas 44, all areas 45, etc. were
calculated and then averaged across the brains and areas). The analysis supplemented previous findings on
asymmetry in volume of area 44 (Amunts et al., 1999) and demonstrates that cytoarchitectonic asymmetry in
areas 44 and 45 might be a microstructural correlate of brain lateralization and dominance for language, in
particular.

Finally, interhemispheric distances between corresponding areas of both hemispheres

were significant for areas 44 and 45, but not for areas 6, V1 and V2. That is, areas 44 and
45 revealed significant left/right differences in cytoarchitecture. Thus, in addition to
asymmetry in volume of area 44 which was reported previously (Amunts et al., 1999;
Galaburda, 1980), interhemispheric asymmetry was demonstrated at a microstructural
level. This cytoarchitectonic asymmetry may contribute to the functional phenomenon of
cerebral lateralization and dominance for language, in particular.
Information obtained from analysis of interareal differences might be applied for
creating hierarchies and families of cortical areas, as described for the cortex of nonhuman
primates (Fellemann et al., 1997; Fellemann and Van Essen, 1991; Gattass et al., 1997;
Hubel and Wiesel, 1972; Kaas, 1989; Kötter and Sommer, 2000; Nakamura et al., 1993;
Peterhans and von der Heydt, 1993; Stephan et al., 2000; Xiao et al., 1999; Zeki, 1978;
Zilles and Clarke, 1997).
Multivariate distance analysis has been applied to detect interhemispheric cytoarchitec-
tonic differences of the motor cortex (Amunts et al., 1996), its developmental changes
(Amunts et al., 1997), interareal, interhemispheric and intersubject differences of Broca’s
region (Amunts et al., 1999), as well interareal differences in receptor architecture of the
mesial motor and premotor cortex in the macaque (Geyer et al., 1998) and of the human
the somatosensory cortex (Geyer et al., 1997, 1999). A quantitative analysis of interareal
differences in cytoarchitecture is also relevant for detecting architectonic differences
between normal and pathologically altered cortical tissue. Finally, the criterion of similar-
ity in architecture might be valuable with respect to a comparative analysis of homologies
between humans and non-human primates (Petrides and Pandya, 1994).
4. MULTIMODAL MAPPING – CORRESPONDENCES AND DIFFERENCES

BETWEEN CYTOARCHITECTONIC AND RECEPTOR ARCHITECTONIC
BORDERS
Architectonic analysis using multivariate statistics can be even more decisive when they
incorporate other modalities of architecture, e.g. receptor architecture. The comparisons of
receptor- and cytoarchitectonic maps provided evidence that in several cortical regions
receptor architecture reveals similar architectonic parcellations as compared to cytoarchi-
tecture. We will present here recent data from our analysis of the human visual cortex
for discussing the correspondence between cytoarchitectonic and receptor-architectonic
parcellation schemes.
Numerous observations on the regional and laminar distribution of transmitter receptors
in the human primary (V1) and secondary visual cortex (V2) have been published. Reports
on receptors in other extrastriate areas are rare. (For an overview see Zilles and Clarke,
1997). This is due to the facts, that (i) human extrastriate areas are difficult to identify in
cytoarchitectonic sections, (ii) the classical architectonic maps of the human occipital lobe
show a much less detailed parcellation than corresponding maps of nonhuman primates,
and (iii) receptor architectonics of the human occipital lobe require extraordinary large
cryostat sections of unfixed, frozen brains, which are difficult to handle. Our own observa-
tions in nonhuman primates and human post-mortem brains, as well as recent data from
the literature, provide evidence that receptor architectonic mapping is a promising approach
to human brain mapping (Bonaventure et al., 2000; d’Argy et al., 1988; Geyer et al., 1997,

1998; King et al., 1995; Pazos et al., 1987a,b; Thoss et al., 1996; Zezula et al., 1988;
Zilles et al., 1991a,b,d, 1995, 1988).
As an example, results of receptor-architectonic mapping of receptor binding sites of
the glutamatergic kainate receptor (ligand: [3H]kainate), the muscarinic M1 (ligand:
[3H]pirenzipine) and M2 (ligand: [3H]oxotremorine-M) receptors as well as the 5-HT2
receptor (ligand: [3H]ketanserin) of human areas V1, V2v, V2d, V3d and V3a are shown
(Figure 3.4, K. Zilles, unpublished observations). We used standardized protocols for the
visualization of a variety of different receptors and of cytoarchitecture in a series of adja-
cent sections (Zilles et al., 1995), thus providing reliable data for cortical parcellation and
for evaluation of areal borders. After contrast enhancement and colour coding, differences
in regional and laminar receptor distributions reveal borders between visual areas which
match to those in cyto- and myeloarchitecture, and provide insight into the neurochemical
aspects of cortical organization in the occipital lobe.
In detail, the border between areas V1 and V2 was characterized by a prominent
change in density and laminar pattern in almost all ligands as well as in cyto- and
myeloarchitecture (Figure 3.4A–F). In correspondence to cytoarchitectonic mapping, V1
was located on the upper and the lower bank of the calcarine sulcus and extended to the
mesial surface of the cuneus (Amunts et al., 2000; Filimonoff, 1932; Stensaas et al.,
1974). The cuneal extension was highly variable between different brains. For kainate
receptors (Figure 3.4A), the infragranular layers displayed a higher density in V1 than in
V2, whereas the supragranular layers had a lower density in V1 than in V2. The dorsal
and ventral borders of V1 coincided precisely with the borders found in the autoradio-
graphs of 5-HT2 (Figure 3.4B), M1 (Figure 3.4C) and M2 (Figure 3.4D) receptors as well
as in cyto- (Figure 3.4E) and myeloarchitecture (Figure 3.4F). The border between areas
V2d and V3d was clearly established by kainate, 5-HT2 and M2 receptor autoradio-
graphy, but less obviously by the M1 receptor. M2 and 5-HT2 receptor densities and
laminar patterns showed a clear border between V2v and the adjacent ventral V3,
whereas more subtle changes were seen in kainate and M1 receptor autoradiographs. M2
receptors revealed a distinct subdivision of area V3 into V3d and V3A, which was absent
in the other receptor autoradiographs and detected only with difficulty by simple visual
inspection in cytoarchitectonic sections. In the myeloarchitectonic section, this border
could be identified by a higher myelin density in supragranular layers of area V3d than of
V3A (Figure 3.4F).
The different receptors revealed corresponding areal borders, which coincided with
the cyto- or myeloarchitectonic borders. Coincidence in position of borders defined by
receptor autoradiography has been proven in an observer-independent fashion using the
same method as described above for cytoarchitecture (Schleicher et al., 1998). Some
receptors, however, did not reveal all borders seen with other receptors. The 5-HT2
receptor and the 5-HT1A receptors (images of the regional distribution of the latter are
not presented here) are examples, which display a more homogenous distribution in the
occipital lobe than that of other receptors. For instance, only minor differences in
receptor density were found between areas V2 and VP/V3. These receptors, however,
showed a pronounced heterogeneity in the motor cortex, where they make visible the
border between motor and premotor areas (Zilles et al., 1995). In contrast, the GABAa
receptor is heterogeneously distributed even within area V1 (Zilles and Schleicher,
1993). It shows periodically distributed patches in layers II/IVc, which might be related
to cytochrome oxidase blobs and ocular dominance columns (Hendrickson et al.,
1981).

3
[ H] kainate binding sites 3
5-HT2 receptor [ H] ketanserin binding sites
3
M1 receptor [ H] pirenzepin binding sites 3
M2 receptor [ H] oxotremorine binding sites
Figure 3.4. Regional distributions of kainate [A], 5-HT2 [B] and muscarinic M1 [C] and M2 [D] receptors
(demonstrated with [3H]kainate, [3H]ketanserin, [3H]pirenzipine and [3H]oxotremorine-M binding; K. Zilles,
unpublished observations) in coronal sections through the human occipital lobe. Grey value images were scaled
to receptor densities, enhanced in contrast, and colour coded. The colour scales indicate the receptor densities in
fmol/mg protein. The receptors are heterogeneously distributed, and therefore allow mapping of striate and
extrastriate areas including V1, V2v, V2d, V3d, V3A. Borders obtained by receptor achitectonic mapping were
compared with borders revealed by cytoarchitectonic [E] and myeloarchitectonic [F] mapping. Note the pro-
nounced differences in receptor density [A–D] as well as in the laminar pattern of cell packing density [E] and
myeloarchitecture [F] at the borders of area V1 to V2. The border between area V2d and V3d is most pro-
nounced in A, B and D mapping. In E and F, it is recognizable only at higher magnification; it can be detected
by the observer-independent approach for border definition. The border between V3d and V3A is clearly
marked in D and F and can be verified quantitatively also in E. Thus, different architectonic techniques supple-
ment each other and are the basis of multimodal architectonic mapping. SCal –Sulcus calcarinus, SLin—Sulcus
lingualis, SCol—Sulcus collateralis, SOS—Sulcus occipitalis superior. (see Color Plate 2)

Finally, the parcellations revealed by quantitative cyto-, myelo- and receptor architectonic
techniques revealed a more subtle subdivision of the human occipital lobe than shown in
the classical architectonic maps from the beginning of the 20th century. They seem to cor-
respond to parcellations observed in functional imaging (DeYoe et al., 1996; Engel et al.,
1997; Gulyas and Roland, 1994; Hadjikhani and Tootell, 2000; Haxby et al., 1994;
Larsson et al., 1999; Sereno et al., 1995; Shipp et al., 1995; Tootell et al., 1995, 1997;
Watson et al., 1993) and nonhuman primate studies (Fellemann et al., 1997; Fellemann
and Van Essen, 1991; Gattass et al., 1997; Hubel and Wiesel, 1972; Kaas, 1989, 1993;
Nakamura et al., 1993; Peterhans and von der Heydt, 1993; Xiao et al., 1999; Zeki, 1978;
Zilles and Clarke, 1997).
5. CONCLUSION AND PERSPECTIVES
Although the parcellation of the human cerebral cortex is still far from complete, receptor
and quantitative cytoarchitectonic findings provide new criteria for a detailed mapping,
which cannot be achieved by cytoarchitectonic analysis alone. If the borders between
receptor architectonic areas are established, receptor densities for distinct areas and layers
can be calculated. This information can be used to create receptor “fingerprints” of cor-
tical areas (Geyer et al., 1998), which represent the densities of a set of several receptors
in a defined cortical area as a multivariate polar plot of receptor densities. First results
have shown that these fingerprints have a similar shape in functionally-related areas, but
a different shape if functionally differing areas are compared.
What is a cortical area? This question provides an important argument for multimodal
architectonic mapping, since not all subparcellations of the cerebral cortex constitute
a cortical area. For instance, the subdivision of areas V1 and V2 into blob and interblob
regions (Livingstone and Hubel, 1987; Roe and Tso, 1995; Tootell et al., 1983; Wong-Riley
et al., 1993) reflects differences in colour and orientation selectivity (V1) and receptive
field properties (V2), but these regions do not constitute cortical areas. Additional examples
are the somatotopy of the motor and somatosensory cortex, the tonotopical organisation of
the auditory cortex, each of which represents a functional segregation without represent-
ing an architectonic entity. The isolated analysis of only one aspect of cortical organiza-
tion without consideration of other mapping techniques, would lead to an over-parcellation
of the cerebral cortex. The multimodal approach proposed here avoids this problem by
providing an overview of the different hierarchical levels (e.g. cytoarchitectonic or receptor
architectonic families of cortical areas) of the cortical organization.
An important perspective for a functionally-relevant architectonic parcellation of the
cortex arises from the integration of architectonic maps with recent PET, fMRI and MEG
studies in a common spatial reference system and database, e.g., the European Computer-
ized Human Brain Database ECHBD (Roland et al., 1999; Roland and Zilles, 1994,
1996a,b, 1998). Borders verified in an observer-independent manner for the sensorimotor
(Geyer et al., 1996, 1999; Grefkes et al., 2001) and auditory cortex (Morosan et al., 2001),
Broca’s region (Amunts et al., 1999), and the visual cortex (Amunts et al., 2000) have
been defined in ten human post mortem brains. The areal borders were labeled in
corresponding digitized histological sections, and subjected to 3-D reconstruction. These
reconstructions of the histological sections and of the cytoarchitectonic areas were warped
to the format of the standard reference brain of the ECHBD (Roland et al., 1999; Roland

and Zilles, 1996a,b; Schormann et al., 1999b). A modified “principal axes” theory and a
movement model for large deformations were applied for the warping of the 3-D datasets
of post mortem brains to the standard brain, in order to compensate for intersubject variab-
ility in extent, shape and sulcal pattern of the brains, and to achieve a maximal anatomical
overlap between the different post mortem brains and the reference brain (Schormann et al.,
1997, 1999a; Schormann and Zilles, 1997, 1998). Individual cortical maps were super-
imposed in the standard reference brain (Schormann et al., 1999b). The overlapping
cortical areas of individual brains and their architectonic areas in the standard reference
brain result in probability or population maps, which display the intersubject variability in
the extent, shape and topography of cortical areas. Variability of macroscopical features
(e.g. hemispheric shape or sulcal pattern) has been analyzed in numerous previous studies
(Dumoulin et al., 1998; Kennedy et al., 1998; Rademacher et al., 1993; Thompson et al.,
1996, 1998; Westbury et al., 1999; Zilles et al., 1997). (For a more comprehensive review
concerning this aspect see chapter by Rademacher in this book.)
In contrast to traditional brain atlases and to the atlas of Talairach and Tournoux (1988),
the ECHBD is (i) a real 3-D representation of cortical areas, (ii) it is based on an observer-
independent architectonic definition of cortical areas, and (iii) it provides quantitative
information on intersubject variability of the topography of each cortical area. Using its
common spatial reference system for combined analysis of PET and cytoarchitectonic
data, it has been shown that the processing of both real and illusory contours activates area
V2 (Larsson et al., 1999). Another example is the cytoarchitectonic subdivision of area 4
into an anterior and a posterior part, which were first observed in receptor architectonic
sections (Zilles et al., 1995), then defined in cytoarchitectonic sections and superimposed
with PET data, which allowed one to correlate these areas with functional differences
(Geyer et al., 1996). Further combined cytoarchitectonic/functional imaging studies have
been performed in the sensorimotor and visual cortices (Bodegard et al., 2000a,b; Naito
et al., 1999, 2000). Since the ECHBD has an open structure, it allows the integration of
additional modalities of architectonic data (e.g. receptor architectonic, myeloarchitectonic)
and of further functional imaging data.
In conclusion, architectonic brain mapping has become more objective and less observer-
dependent. Recent marker techniques, e.g. receptor-autoradiography of different receptors,
immuno- and enzyme-histochemical methods have added functional meaningful informa-
tion. Multimodal mapping and quantitative analysis of interareal differences promote a new
and more complex concept of a cortical area. Finally, by applying recent 3-D probabilistic
atlas systems, the combined analysis of architectonic maps and functional imaging studies
allows the testing of the functional significance of architectonic parcellations and a system-
atic search for new, functionally relevant cortical areas. Thus, architectonic brain mapping
has changed considerably during the last 100 years and opens exciting perspectives for the
future.
ACKNOWLEDGEMENTS
Published and unpublished work by the authors reviewed in this chapter was supported by
the Deutsche Forschungsgemeinschaft (SFB 194/A6), the BioTech program of the EC and
the Human Brain Project (P20-MHDA52176) funded by the National Institute of Mental
Health, National Institute for Drug Abuse, and the National Cancer Institute. The authors

thank Aleksandar Malikovic, Nicola Palomero-Gallagher, Ursula Blohm, Brigitte Machus

and Renate Dohm for assistance in preparing the histological sections and the autoradio-
graphs.
REFERENCES
Aboitiz, F. and Garcia, G.L. (1997) The evolutionary origin of language areas in the human brain. A neuro-
anatomical perspective. Brain Research Reviews, 25, 381–396.
Amunts, K., Istomin, V.V., Schleicher, A. and Zilles, K. (1995) The postnatal development of the human primary
motor cortex: a quantitative cytoarchitectonic analysis. Anatomy and Embryology, 192, 557–571.
Amunts, K., Schlaug, G., Schleicher, A., Steinmetz, H., Dabringhaus, A., Roland, P.E. and Zilles, K. (1996)
Asymmetry in the human motor cortex and handedness. Neuroimage, 4, 216–222.
Amunts, K., Schmidt-Passos, F., Schleicher, A. and Zilles, K. (1997) Postnatal development of interhemispheric
asymmetry in the cytoarchitecture of human area 4. Anatomy and Embryology, 196, 393–402.
Amunts, K., Schleicher, A., Bürgel, U., Mohlberg, H., Uylings, H.B.M. and Zilles, K. (1999) Broca’s region
revisited: Cytoarchitecture and intersubject variability. Journal of Comparative Neurology, 412, 319–341.
Amunts, K., Malikovic, A., Mohlberg, H., Schormann, T. and Zilles, K. (2000) Brodmann’s areas 17 and 18
brought into stereotaxic space – where and how variable? Neuroimage, 11, 66–84.
Andrews, T.J., Halpern, S.C. and Purves, D. (1997) Correlated size variations in human visual cortex, lateral
geniculate nucleus, and optic tract. Journal of Neuroscience, 17, 1859–2868.
Bailey, P. and von Bonin, G. (1951) The isocortex of man. Urbana: University of Illinois Press.
Betz, W. (1874) Anatomischer Nachweis zweier Gehirncentra. Centralblatt für die medizinischen Wissen-
schaften, 12, 578–580, 594–599.
Bidmon, H.J., Wu, J.Y., Godecke, A., Schleicher, A., Mayer, B. and Zilles, K. (1997) Nitric oxide synthase-
expressing neurons are area-specifically distributed within the cerebral cortex of the rat. Neuroscience,
81, 321–330.
Binkofski, F., Buccino, G., Posse, S., Seitz, R., Rizzolatti, G. and Freund, H.-J. (1999) A fronto-parietal circuit
for object manipulation in man: evidence from an fMRI-study. European Journal of Neuroscience, 11,
3276–3286.
Binkofski, F., Amunts, K., Stephan, K.M., Posse, S., Schormann, T., Freund, H.-J., Zilles, K. and Seitz, R.
(2000) Broca’s region subserves recognition of action. Human Brain Mapping, 11, 273–285.
Blinkov, S.M. and Glezer, I.I. (1968) Das Zentralnervensystem in Zahlen und Tabellen. Jena: Fischer.
Bodegard, A., Geyer, S., Naito, E., Zilles, K. and Roland, P.E. (2000a) Somatosensory areas in men activated
by moving stimuli: cytoarchitectonic mapping and PET. NeuroReport, 11, 187–191.
Bodegard, A., Ledberg, A., Geyer, S., Naito, E., Zilles, K. and Roland, P.E. (2000b) Object shape differences
reflected by somatosensory cortical activation in human. Journal of Neuroscience, 20 RC 51, 1–5.
Bonaventure, P., Hall, H., Gommeren, W., Cras, P., Langlois, X., Jurzak, M. and Leysen, J.E. (2000) Mapping of
serotonin 5-HT(4) receptor mRNA and ligand binding sites in the post-mortem human brain. Synapse, 36,
35–46.
Braak, H. (1977) The pigment architecture of the human occipital lobe. Anatomy and Embryology, 150,
229–250.
Braak, H. (1979) The pigment architecture of the human frontal lobe. Anatomy and Embryology, 157, 35–68.
Broca, M.P. (1861) Remarques sur le siége de la faculté du langage articulé, suivies d’une observation d’aphemie
(Perte de la Parole). Bulletins et memoires de la Societe Anatomique de Paris, 36, 330–357.
Brodmann, K. (1903) Beiträge zur histologischen Lokalisation der Grosshirnrinde. II: Der Calcarinustyp.
Journal für Psychologie und Neurologie, II, 133–159.
Brodmann, K. (1905) Beiträge zur histologischen Lokalisation der Groβhirnrinde. V: Über den allgemeinen
Bauplan des Cortex pallii bei den Mammaliern und zwei homologe Rindenfelder im besonderen. Zugleich
ein Beitrag zur Furchenlehre. Journal für Psychologie und Neurologie, VI, 275–303.
Brodmann, K. (1908) Beiträge zur histologischen Lokalisation der Groβhirnrinde. VI: Die Cortexgliederung des
Menschen. Journal für Psychologie und Neurologie, X, 231–246.
Brodmann, K. (1909) Vergleichende Lokalisationslehre der Groβhirnrinde in ihren Prinzipien dargestellt auf
Grund des Zellenbaues. Leipzig: Barth JA.
Brody, H. (1955) Organisation of the cerebral cortex. Part 3. A study of aging in the human cerebral cortex.
Journal of Comparative Neurology, 102, 551–556.
Burkhalter, A. and Bernardo, K.L. (1989) Organization of cortico-cortical connections in human visual cortex.
Proceedings of the National Academy of Scence, U.S.A., 13, 1916–1931.
Campbell, A.W. (1905) Histological studies on the localisation of cerebral function, Cambridge: Cambridge
University Press.

Campbell, M. and Morrison, J.H. (1989) Monoclonal antibody to neurofilament protein (SMI-32) labels a sub-
population of pyramidal neurons in the human and monkey neocortex. Journal of Comparative Neurology,
282, 191–205.
Clarke, S. (1993) Callosal connections and functional subdivision of the human occipital lobe. In: B. Gulyas,
D. Ottoson and P. Roland (eds), Functional organization of the human visual cortex. Oxford: Pergamon
Press, pp. 137–149.
Clarke, S. (1994) Modular organization of human extrastriate visual cortex: Evidence from cytochrome
oxidase pattern in normal and macular degeneration cases. European Journal of Neuroscience, 6,
725–736.
Clarke, S. and Miklossy, J. (1990) Occipital cortex in man: Organization of callosal connections, related cyto-
and myeloarchitecture, and putative boundaries of functional visual areas. Journal of Comparative Neuro-
logy, 298, 188–214.
d’Argy, R., Gillberg, P.G., Stalnacke, C.G., Persson, A., Bergstrom, M., Langstrom, B., Schoeps, K.O. and
Aquilonius, S.M. (1988) In vivo and in vitro receptor autoradiography of the human brain using an 11C-
labelled benzodiazepine analogue. Neuroscience Letters, 85, 304–310.
DeYoe, E.A., Carman, G.J., Bandettini, P., Glickman, S., Wieser, J., Cox, R., Miller, D. and Neitz, J. (1996)
Mapping striate and extrastriate visual areas in human cerebral cortex. Proceedings of the National
Academy of Science, U.S.A., 93, 2382–2386.
Dietl, M.M., Probst, A. and Palacios, J.M. (1987) On the distribution of cholecytokinin receptor binding in the
human brain: an autoradioraphic study. Synapse, 1, 169–183.
Dixon, W.J., Brown, M.B., Engelman, L., Hill, M.A. and Jennrich, R.I. (1988) BMDP. Statistical software
manual, Berkley: University of California Press.
Dumoulin, S.O., Bittar, R.G., Kabani, N.J., Baker, C.L., LeGoualher, G., Pike, G.B. and Evans, A.C. (1998)
Quantification of the variability of human area v5/MT in relation to the sulcal pattern in the parieto-
temporo-occipital cortex: a new anatomical landmark. Neuroimage, 7, S319-S319.
Elliot Smith, G. (1907) A new topographical survey of the human cerebral cortex, being an account of the distri-
bution of the anatomically distinct cortical areas and their relationship to the cerebral sulci. Journal of
Anatomy, 41, 237–254.
Engel, S.A., Glover, G.H. and Wandell, B.A. (1997) Retinotopic organization in human visual cortex and the
spatial precision of functional MRI. Cerebral Cortex, 7, 181–192.
Fellemann, D.J. and Van Essen, D.C. (1991) Distributed hierarchical processing in the primate cerebral cortex.
Cerebral Cortex, 1, 1–47.
Fellemann, D.J., Burkhalter, A. and Van Essen, D.C. (1997) Cortical connections of areas V3 and VP of
Macacque monkey. Extrastriate visual cortex. Journal of Comparative Neurology, 379, 21–47.
Filimonoff, I.N. (1932) Über die Variabilität der Groβhirnrindenstruktur. Mitteilung II – Regio occipitalis beim
erwachsenen Menschen. Journal für Psychologie und Neurologie, 44, 2–96.
Flechsig, P. (1898) Neue Untersuchungen über die Marbildung in den menschlichen Grosshirnlappen. Neuro-
logisches Zentralblatt (Leipzig), 21, 977–996.
Fritsch, G. and Hitzig, E. (1870) Über die elektrische Erregbarkeit des Grosshirns. Archiv für Anatomie, Physio-
logie und wissenschaftliche Medizin, Jg. 1870, 300–332.
Galaburda, A.M. (1980) La region de Broca: Observations anatomiques faites un siècle après la mort de son
découvreur. Revue Neurologique (Paris), 136, 609–616.
Gattass, R., Gross, C.G. and Sandell, J.H. (1981) Visual topography of V2 in the macaque. Journal of Comparative
Neurology, 201, 519–539.
Gattass, R., Sousa, A.P., Mishkin, M. and Ungerleider, L.G. (1997) Cortical projections of area v2 in the
macaque. Cerebral Cortex, 7, 110–129.
Geyer, S., Ledberg, A., Schleicher, A., Kinomura, S., Schormann, T., Bürgel, U., Klingberg, T., Larsson, J.,
Zilles, K. and Roland, P.E. (1996) Two different areas within the primary motor cortex of man. Nature
(London), 382, 805–807.
Geyer, S., Schleicher, A. and Zilles, K. (1997) The somatosensory cortex of human: Cytoarchitecture and
regional distributions of receptor-binding sites. Neuroimage, 6, 27–45.
Geyer, S., Matelli, M., Luppino, G., Schleicher, A., Jansen, Y., Palomero-Gallagher, N. and Zilles, K. (1998)
Receptor autoradiographic mapping of the mesial and premotor cortex of the macaque monkey. Journal of
Comparative Neurology, 397, 231–250.
Geyer, S., Schleicher, A. and Zilles, K. (1999) Areas 3a, 3b, and 1 of human primary somatosensory cortex:
I. Microstructural organisation and interindividual variability. Neuroimage, 10, 63–83.
Grefkes, C., Geyer, S., Schormann, T., Roland, P. and Zilles, K. (2001) Human somatosensory area 2: Observer-
independent cytoarchitectonic mapping, interindividual variability, and population map. Neuroimage, 14,
617–632.
Gulyas, B. and Roland, P.E. (1994) Processing and analysis of form, color and binocular disparity in the
human brain: functional anatomy by positron emission tomography. European Journal of Neuroscience,
6, 725–736.

Gundersen, H.J.G., Bendtsen, T.F., Korbo, L., Marcussen, N., Moeller, A., Nielsen, K., Nyengaard, Pakkenberg,
B., Sörensen, F.B., Vesterby, A. and West, M.J. (1988) Some new, simple and efficient stereological
methods and their use in pathological research and diagnosis. Acta Pathologica Microbiologica et Immuno-
logica Scandinavica, 96, 379–394.
Hadjikhani, N.K. and Tootell, R.B.H. (2000) Projection of rods and cones within human visual cortex. Human
Brain Mapping, 9, 55–63.
Harasty, J., Halliday, G.M. and Kril, J.J. (1996) Reproducible sampling regimen for specific cortical regions:
application to speech-related areas. Journal of Neuroscience Methods, 67, 43–51.
Haug, H. (1956) Remarks on the determination and significance of the gray cell coefficient. Journal of Compar-
ative Neurology, 104, 473–492.
Haug, H. (1980) Die Abhängigkeit der Einbettungsschrumpfung des Gehirngewebes vom Lebensalter. Verhand-
lungen der Anatomisches Gesellschaft, 74, 699–700.
Haug, H. (1984) Macroscopic and microscopic morphometry of the human brain and cortex. A survey in the
light of new results. Brain Pathology, 1, 123–149.
Haug, H. (1987a) Brain sizes, surfaces, and neuronal sizes of the cortex cerebri: a stereological investigation of
man and his variability and a comparison with some mammals (primates, whales, marsupials, insectivores,
and one elephant). American Journal of Anatomy, 180, 126–142.
Haug, H. (1987b) Die menschliche Hirnrinde—Architektur und Daten. In: E.H. Graul, S. Pütter and D. Loew
(eds), Natürliche und künstliche Intelligenz, Kunstfehler—Regresse—Sterbehilfe, (series: Das Gehirn und
seine Erkrankungen, Vol. 17). Iserlohn: Medice, pp. 137–164.
Haxby, J.V., Horrwitz, B., Ungerleider, L.G., Maisog, J.M., Pietrini, P. and Grady, C.L. (1994) The functional
organization of human extrastriate cortex: a PET-rCBF study of selective attention to faces and locations.
Journal of Neuroscience, 14, 6336–6353.
Hayes, T.L. and Lewis, D.A. (1996) Magnopyramidal neurons in the anterior motor speech region. Archives of
Neurology, 53, 1277–1283.
Henderson, G., Tomlinson, B.E. and Gibson, P.H. (1980) Cell count in human cerebral cortex in normal adults
throughout life using a image analysing computer. Journal of Neurological Science, 46, 113–136.
Hendrickson, A.E., Hunt, S.P. and Wu, J.Y. (1981) Immunocytochemical localization of glutamic acid decarb-
oxylase in monkey striate cortex. Nature (London), 292, 605–607.
Hendry, S.H.C., Huntsman, M., Vinuela, A., Möhler, H., DeBlas, A. and Jones, E. (1994) GABA a receptor
subunit immunoreactivity in primate visual cortex: Distribution in macaques and humans and regulation
by visual input in adulthood. Journal of Neuroscience, 10, 2438–2450.
Herve, G. (1888) La circonvolution de Broca. Etude de morphologie cerebrale. Paris: Davy, A.
Hubel, D.H. and Wiesel, T. (1972) Laminar and columnar distribution of geniculo-cortical fibers in the macaque
monkey. Journal of Comparative Neurology, 146, 421–450.
Hudspeth, A.J., Ruark, J.E. and Kelly, J.P. (1976) Cytoarchitectonic mapping by microdensitometry. Proceed-
ings of the National Academy of Science, U.S.A., 73, 2928–2931.
Huttenlocher, P.R. (1979) Synaptic density in human frontal cortex—developmental changes and effects of
aging. Brain Research, 163, 195–205.
Jansen, K.L.R., Faull, R.L.M. and Dragunow, M. (1989) Excitatory amino acid receptors in the human
cerebral cortex: A quantitative auroradiographic study comparing the distribution of [3H]TCP,
[3H]glycine, L-[3H]glutamate, [3H]AMPA and [3H]kainic acid binding sites. Journal of Neuroscience, 32,
587–607.
Kaas, J.H. (1989) Why does the brain have so many visual areas? Journal of Cognitive Neuroscience, 1,
121–135.
Kaas, J.H. (1993) The organization of visual cortex in primates: problems, conclusions, and the use of comparat-
ive studies understanding the human brain. In: B. Gulyas, D. Ottoson and P.E. Roland (eds), Functional
organization of the human visual cortex. Oxford: Pergamon Press, pp. 1–12.
Kennedy, D.N., Lange, N., Makris, N., Bates, J., Meyer, J. and Caviness, V.J. (1998) Gyri of the human neo-
cortex: an MRI-based analysis of volume and variance. Cerebral Cortex, 8, 372–384.
King, P.R., Gundlach, A.L. and Louis, W.J. (1995) Quantitative autoradiographic localization in rat brain of
alpha 2-adrenergic and non-adrenergic I-receptor binding sites labelled by [3H]rilmenidine. Brain Research,
675, 264–278.
Kleist, K. (1934) Gehirnpathologie. Leipzig: Barth.
Kononova, E.P. (1935) Structural variability of the cortex cerebri. Inferior frontal gyrus in adults (Russian). In:
S.A. Sarkisov and I.N. Filimonoff (eds), Annals of the Brain Research Institute. Vol. I. Moscow-Leningrad:
State Press for Biological and Medical Literature, pp. 49–118.
Kononova, E.P. (1938) Variability in the structure of the human cerebral cortex. Frontal lobe of adult man
(Russian). In: S.A. Sarkisov and I.N. Filimonoff (eds), Annuals of the Brain Research Institute. Vol. III–IV.,
Moscow-Leningrad: State Press for Biological and Medical Literature, pp. 213–274.
Kononova, E.P. (1949) The frontal lobe (Russian). In: S.A. Sarkisov, I.N. Filimonoff and N.S. Preobrashen-
skaya, The Cytoarchitecture of the Human Cortex Cerebri. Moscow: Medgiz, pp. 309–343.

Kötter, R. and Sommer, F.T. (2000) Global relationship between anatomical connectivity and activity propa-
gation in the cerebral cortex. Philosophical Transactions of the Royal Society, London, series B, 355,
127–134.
Larsson, J., Amunts, K., Gulyas, B., Malikovic, A., Zilles, K. and Roland, P.E. (1999) Neuronal correlates of real
and illusory contour perception: functional anatomy with PET. European Journal of Neuroscience, 11,
4024–4036.
Lewis, J.W. and Olavarria, J.F. (1995) Two rules for callosal connectivity in striate cortex of the rat. Journal of
Livingstone, M.S. and Hubel, D.H. (1987) Connections between layer 4B of area 17 and the thick cytochrome
oxidase stripes of area 18 in the squirrel monkey. Journal of Neuroscience, 7, 3371–3377.
Lungwitz, W. (1937) Zur myeloarchitektonischen Untergliederung der menschlichen Area praeoccipitalis (Area
19 Brodmann). Journal für Psychologie und Neurologie, 47, 607–638.
Luppino, G., Matelli, M., Camarda, R.M., Gallese, V. and Rizzolatti, G. (1991) Multiple representations of body
movements in mesial area 6 and the adjacent cingulate cortex: An intracortical microstimulation study in
the macaque monkey. Journal of Comparative Neurology, 311, 463–482.
Matelli, M., Luppino, G. and Rizzolatti, G. (1991) Architecture of superior and mesial area 6 and the adjacent
cingulate cortex in the macaque monkey. Journal of Comparative Neurology, 311, 445–462.
Merigan, W.H., Nealey, T.A. and Maunsell, J.R. (1993) Visual effects of lesions of cortical area v2 in mac-
acques. Journal of Neuroscience, 13, 3180–3191.
Morosan, P., Rademacher, J., Schleicher, A., Amunts, K., Schormann, T. and Zilles, K. (2001) Human primary
auditory cortex: Cytoarchitectonic subdivisions and mapping into a special reference system. Neuroimage,
13, 684–701.
Naito, E., Ehrsson, H.H., Geyer, S., Zilles, K. and Roland, P.E. (1999) Illusory arm movements activate cortical
motor areas: a positron emission tomography study. Journal of Neuroscience, 19, 6134–6144.
Naito, E., Kinomura, S., Geyer, S., Kawashima, R., Roland, P.E. and Zilles, K. (2000) Fast reaction to different
sensory modalities activate common fields in the motor areas, but the anterior cingulate cortex is involved
in the speed of reaction. Journal of Neurophysiology, 83, 1701–1709.
Nakamura, H., Gattass, R., Desimone, R. and Ungerleider, L.G. (1993) The modular organization of projections
from areas V1 and V2 to areas V4 and TEO in macaques. Journal of Neuroscience, 13, 3681–3691.
Newsome, W.T. and Allman, J.M. (1980) Interhemispheric connections of the visual cortex in the owl monkey,
Aotus trivirgatus, and the bushbaby. Journal of Comparative Neurology, 194, 209–233.
Newsome, W.T., Maunsell, J.H.R. and Van Essen, D.C. (1986) Ventral posterior visual area of the macaque:
visual topography and areal boundaries. Journal of Comparative Neurology, 252, 139–153.
Pakkenberg, B. and Gundersen, H.J.G. (1997) Neocortical neuron number in humans: Effect of sex and age.
Paulesu, E., Frith, C.D. and Frackowiak, R.S.J. (1993) The neural correlates of the verbal component of working
memory. Nature (London), 362, 342–345.
Pazos, A., Probst, A. and Palacios, J.M. (1987a) Serotonin receptors in the human brain—III. Autoradiographic
mapping of serotonin-1 receptors. Neuroscience, 21, 97–122.
Pazos, A., Probst, A. and Palacios, J.M. (1987b) Serotonin receptors in the human brain—IV. Autoradiographic
mapping of serotonin-2 receptors. Neuroscience 21, 123–139.
Peterhans, E. and von der Heydt, R. (1993) Functional organization of area V2 in the alert monkey. European
Petrides, M., Alivisatos, B., Meyer, E. and Evans, A.C. (1993) Functional activation of the human frontal cortex
during the performance of verbal working memory tasks. Proceedings of the National Academy of Science,
U.S.A., 90, 878–882.
Petrides, M. and Pandya, D.N. (1994) Comparative architectonic analysis of the human and the macaque frontal
cortex. In: F. Boller and J. Grafman (eds), Handbook of Neuropsychology. New York: Elsevier, pp. 17–58.
Rademacher, J., Caviness, J., Steinmetz, H. and Galaburda, A.M. (1993) Topographical variation of the human
primary cortices: implications for neuroimaging, brain mapping, and neurobiology. Cerebral Cortex, 3,
313–329.
Rajkowska, G. and Goldman-Rakic, P.S. (1995a) Cytoarchitectonic definition of prefrontal areas in the normal
human cortex: I. Remapping of areas 9 and 46 using quantitative criteria. Cerebral Cortex, 5, 307–322.
Rajkowska, G. and Goldman-Rakic, P.S. (1995b) Cytoarchitectonic definition of prefrontal areas in the normal
human cortex: II. Variability in locations of areas 9 and 46 and relationship to the Talairach coordinate
system. Cerebral Cortex, 5, 323–337.
Riegele, L. (1931) Die Cytoarchitektonik der Felder der Broca’schen Region. Journal für Psychologie und
Roe, A.W. and Tso, D.Y. (1995) Visual topography in primate v2: multiple representation across functional
stripes. Journal of Neuroscience, 15, 3689–3715.
Roland, P.E. (1993) Brain activation. New York: Wiley-Liss.
Roland, P.E. and Zilles, K. (1994) Brain atlases – a new research tool. Trends in Neuroscience, 17, 458–467.

Roland, P.E. and Zilles, K. (1996a) Functions and structures of the motor cortices in human. Current Opinion
in Neurobiology, 6, 773–781.
Roland, P.E. and Zilles, K. (1996b) The developing European computerized human brain database for all
imaging modalities. Neuroimage, 4, S39–S47.
Roland, P.E. and Zilles, K. (1998) Structural divisions and functional fields in the human cerebral cortex. Brain
Research Reviews, 26, 87–105.
Roland, P.E., Fredriksson, J., Svensson, P., Amunts, K., Cavada, C., Hari, R., Cowey, A., Crivello, F., Geyer, S.,
Kostopoulos, J., Mazoyer, B., Poppelwell, D., Schleicher, A., Schormann, T., Seppa, M., Uylings, H.,
deVos, K. and Zilles, K. (1999) ECHBD, A Database for functional-structural and functional-functional
relations. Neuroimage, 8, 128.
Sanides, F. (1962) The architecture of the human frontal lobe and the relation to its functional differentiation.
International Journal of Neurology, 5, 247–261.
Sanides, F. (1964) The cyto-myeloarchitecture of the human frontal lobe, and its relation to phylogenetic
differentiation of the cerebral cortex. Journal für Hirnforschung, 6, 269–282.
Sanides, F. and Vitzthum, H.G. (1965a) Die Grenzerscheinungen am Rande der menschlichen Sehrinde.
Deutsche Zeitschrift für Nervenheilkunde, 187, 708–719.
Sanides, F. and Vitzthum, H.G. (1965b) Zur Architektonik der menschlichen Sehrinde und den Prinzipien ihrer
Entwicklung. Deutsche Zeitschrift für Nervenheilkunde, 187, 680–707.
Sarkisov, S.A., Filimonoff, I.N. and Preobrashenskaya, N.S. (1949) Cytoarchitecture of the human cortex cerebri
(Russ.). Moscow: Medgiz.
Schleicher, A. and Zilles, K. (1990) A quantitative approach to cytoarchitectonics: analysis of structural
inhomogeneities in nervous tissue using an image analyser. Journal of Microscopy, 157, 367–381.
Schleicher, A., Amunts, K., Geyer, S., Simon, U., Zilles, K. and Roland, P.E. (1995) A method of observer-
independent cytoarchitectonic mapping of the human cortex. Human Brain Mapping, Suppl. 1, 77.
Schleicher, A., Amunts, K., Geyer, S., Kowalski, T. and Zilles, K. (1998) An observer-independent
cytoarchitectonic mapping of the human cortex using a stereological approach. Acta Stereologia, 17,
75–82.
Schleicher, A., Amunts, K., Geyer, S., Morosan, P. and Zilles, K. (1999) Observer-independent method for
microstructural parcellation of cerebral cortex: a quantitative approach to cytoarchitectonics. Neuroimage,
9, 165–177.
Schleicher, A., Amunts, K., Geyer, S., Kowalski, T., Schormann, T., Palomero-Gallagher, N. and Zilles, K.
(2000) A stereological approach to human cortical architecture: Identification and delineation of cortical
areas. Journal of Chemical Neuroanatomy, 20, 31–47.
Schmitz, C., Schuster, D., Niessen, P. and Korr, H. (1999) No difference between estimated mean nuclear
volumes of various types of neurons in the mouse brain obtained on either isotrophic uniform random
sections or conventional frontal or sagittal sections. Journal of Neuroscience Methods, 88, 71–82.
Schormann, T. and Zilles, K. (1997) Limitations of the principle axes theory. IEEE Transactions in Medical
Imaging, 16, 942–947.
Schormann, T. and Zilles, K. (1998) Three-dimensional linear and nonlinear transformations: an integration of
light microscopical and MRI data. Human Brain Mapping, 6, 339–347.
Schormann, T., Dabringhaus, A. and Zilles, K. (1997) Extension of the Principal axis theory for the determin-
ation of affine transformations. In Informatik aktuell, edited by Anonymous, pp. 384–391. Springer.
Schormann, T., Henn, S. and Zilles, K. (1999a) Ein sequentielles System zur Anpassung medizinischer
Bilddaten. In Proceedings of the DAGM, Berlin: Springer.
Schormann, T., Posse, S. and Zilles, K. (1999b) The new reference brain of the ECHBD database. Neuroimage,
9, 40.
Sereno, M.I., Dale, A.M., Reppas, J.B., Kwong, K.K., Belliveau, J.W., Brady, T.I., Rosen, B.R. and Tootell,
R.B.H. (1995) Borders of multiple visual areas in humans revealed by functional magnetic resonance
imaging. Science (Washington), 268, 889–893.
Shipp, S., Watson, J.D.G., Frackowiak, R.S.J. and Zeki, S. (1995) Retinotopic maps in human prestriate visual
cortex: the demarcation of areas V2 and V3. Neuroimage, 2, 125–132.
Skullerud, K. (1985) Variations in the size of the human brain. Acta Neurologica Scandinavica, 71, 1–94.
Stensaas, S.S., Eddington, D.K. and Dobelle, W.H. (1974) The topography and variability of the primary visual
cortex in man. Journal of Neurosurgery, 40, 755.
Stephan, K.E., Hilgetag, C.C., Burns, G.A., O’Neill, M.A., Young, M.P. and Kotter, R. (2000) Computational
analysis of functional connectivity between areas of primate cerebral cortex. Philosophical Transactions of
the Royal Society, London series B, 355, 111–126.
Strasburger, E.H. (1938) Vergleichende myeloarchitektonische Studien an der erweiterten Brocaschen Region
des Menschen. Journal für Psychologie und Neurologie, 48, 477–511.
Talairach, J. and Tournoux, P. (1988) Coplanar stereotaxic atlas of the human brain, Stuttgart: Thieme.
Tanji, J. and Kurata, K. (1989) Changing concepts of motor areas of the cerebral cortex. Brain Development, 11,
374–377.

Terry, R.D., De Teresa, R. and Hansen, L.A. (1987) Neocortical cell counts in normal human adult aging. Annals
of Neurology, 21, 530–539.
Thompson, P., Schwartz, C. and Toga, A.W. (1996) High-resolution random mesh algorithms for creating
a probabilistic 3D surface atlas of the human brain. Neuroimage, 3, 19–34.
Thompson, P.M., Moussai, J., Zohoori, S., Goldkorn, A., Khan, A.A., Mega, M.S., Small, G.W., Cummings, J.L.
and Toga, A.W. (1998) Cortical variability and asymmetry in normal aging and Alzheimer’s disease.
Thoss, V.S., Piwko, O. and Hoyer, D. (1996) Somatostatin receptors in the rhesus monkey brain: localization and
pharmacological characterization. Naunyn Schmiedebergs Archives of Pharmacology, 353, 648–440.
Tootell, R.B.H., Silverman, M.S., De Valois, R.J. and Jacobs, G.H. (1983) Functional organization of the second
cortical visual area im primates. Science, Washington, 220, 737–739.
Tootell, R.B.H., Reppas, J.B., Kwong, K.K., Malach, R., Born, R.T., Brady, T.J., Rosen, B.R. and Belliveau,
J.W. (1995) Functional analysis of human MT and related visual cortical areas using magnetic resonance
imaging. Journal of Neuroscience, 15, 3215–3230.
Tootell, R.B.H. and Taylor, J.B. (1995) Anatomical evidence for MT and additional cortical visual areas in
humans. Cerebral Cortex, 5, 39–55.
Tootell, R.B.H., Mendola, J.D., Hadjikhani, N.K., Ledden, P.J., Liu, A.K., Reppas, J.B., Sereno, M.I. and Dale,
A.M. (1997) Functional analysis of V3A and related areas in human visual cortex. Journal of Neuroscience,
17, 7060–7078.
Uylings, H.B.M., Malofeeva, L.I., Bogolepova, I.N., Amunts, K. and Zilles, K. (1999) Broca’s language area
from a neuroanatomical and developmental perspective. In: P. Hagoort and C. Brown (eds), Neurocognition
of language processing, Oxford: Oxford University Press, pp. 319–336.
Vierordt, H. (1893) Anatomische, physiologische und physikalische Daten und Tabellen zum Gebrauch für
Mediziner, Jena: Fischer Verlag.
Vogt, C. (1910) Die myeloarchitektonische Felderung des menschlichen Stirnhirns. Journal für Psychologie und
Vogt, O. (1919) Die myeloarchitektonische Felderung des menschlichen Stirnhirns. Journal für Psychologie und
Neurologie, XV, 221–232.
von Economo, C. and Koskinas, G.N. (1925) Die Cytoarchitektonik der Hirnrinde des erwachsenen Menschen,
Berlin: Springer.
Watson, J.D.G., Frackowiak, R.S.J. and Zeki, S. (1993) Functional separation of colour and motion centres in
human visual cortex. In: B. Gulyas, D. Ottoson and P.E. Roland (eds), Functional organization of the
human visual cortex, Oxford: Pergamon Press, pp. 317–328.
Weibel, E. (1979) Stereological methods. Vol 1: Practical methods for biological morphometry, London:
Academic Press.
West, M.J. (1993) New stereological methods for counting neurons. Neurobiology of Aging, 14, 275–285.
Westbury, C.F., Zatorre, R.J. and Evans, A.C. (1999) Quantifying variability in the planum temporale: A probab-
ility map. Cerebral Cortex, 9, 392–405.
Williams, R.W. and Rakic, P. (1988) Three-dimensional counting: an accurate and direct method to estimate
numbers of cells in sectioned material. Journal of Comparative Neurology, 278, 344–352.
Wong-Riley, M.T., Hevner, R.F., Cutlan, R., Earnest, M., Egan, R., Frost, J. and Nguyen, T. (1993) Cytochrome
oxidase in the human visual cortex: distribution in the developing and the adults brain. Visual Neuroscience,
10, 41–58.
Wree, A., Schleicher, A. and Zilles, K. (1982) Estimation of volume fractions in nervous tissue with an image
analyzer. Journal of Neuroscience Methods, 6, 29–43.
Xiao, Y., Zych, A. and Fellemann, D.J. (1999) Segregation and convergence of functionally defined V2 thin
stripe and interstripe compartment projections to area V4 of Macaques. Cerebral Cortex, 9, 792–804.
Zeki, S. (1978) Functional specialization of the visual cortex of the rhesus monkey. Nature (London), 274,
423–428.
Zezula, J., Cortés, R., Probst, A. and Palacios, J.M. (1988) Benzodiazepine receptor sites in the human brain:
autoradiographic mapping. Neuroscience, 25, 771–795.
Zilles, K. (1990) Cortex. In: G. Paxinos (ed.), The human nervous system. San Diego: Academic Press, pp. 757–802.
Zilles, K. and Schleicher, A. (1980) Quantitative Analyse der laminären Struktur menschlicher Cortexareale.
Verhandlungen der Anatomisches Gesellschaft, 74, 725–726.
Zilles, K. and Schleicher, A. (1993) Cyto- and myeloarchitecture of human visual cortex and the periodical
GABA-A receptor distribution. In: B. Gulyas, D. Ottoson and P. Roland (eds), Functional organization of
the human visual cortex. Oxford: Pergamon Press, pp. 111–120.
Zilles, K. and Schleicher, A. (1995) Correlative imaging of transmitter receptor distributions in human cortex.
In: W. Stumpf and H. Solomon (eds), Autoradiography and correlative imaging. San Diego: Academic
Press, pp. 277–307.

Zilles, K. and Clarke, S. (1997) Architecture, connectivity and transmitter receptors of human extrastriate visual
cortex. Comparison with non-human primates. In: K.S. Rockland, J.H. Kaas and A. Peters (eds), Cerebral
Cortex. Vol. 12. New York: Plenum Press, pp. 673–742.
Zilles, K., Schleicher, A., Rath, M. and Bauer, A. (1988) Quantitative receptor autoradiography in the human
brain. Histochemistry, 90, 129–137.
Zilles, K., Werner, L., Qü, M., Schleicher, A. and Gross, G. (1991a) Quantitative autoradiography of 11 different
transmitter binding sites in the basal forebrain region of the rat—evidence of heterogenity in distribution
patterns. Neuroscience, 42, 473–481.
Zilles, K., Gross, G., Schleicher, A., Schildgen, S., Bauer, A., Bahro, M.S., Zech, K. and Kolassa, N. (1991b)
Regional and laminar distribution of alpha-adrenoreceptors and their subtypes in human and rat hippocam-
pus. Neuroscience, 40, 307–320.
Zilles, K., Hajos, F., Kalman, M. and Schleicher, A. (1991c) Mapping of glial fibrillary acidic protein-immuno-
reactivity in the rat forebrain and mesencephalon by computerized image analysis. Journal of Comparative
Neurology, 308, 340–355.
Zilles, K., Qü, M., Schröder, H. and Schleicher, A. (1991d) Neurotransmitter receptors and cortical architecture.
Journal für Hirnforschung, 32, 343–356.
Zilles, K., Schlaug, G., Matelli, M., Luppino, G., Schleicher, A., Qü, M., Dabringhaus, A., Seitz, R. and
Roland, P.E. (1995) Mapping of human and macaque sensorimotor areas by integrating architectonic,
transmitter receptor, MRI and PET data. Journal of Anatomy, 187, 515–537.
Zilles, K., Schleicher, A., Langemann, C., Amunts, K., Morosan, P., Palomero-Gallagher, N., Schormann, T.,
Mohlberg, H., Bürgel, U., Steinmetz, H., Schlaug, G. and Roland, P.E. (1997) A quantitative analysis of
sulci in the human cerebral cortex: development, regional heterogeneity, gender difference, asymmetry,
intersubject variability and cortical architecture. Human Brain Mapping, 5, 218–221.
Zilles, K., Werners, R., Büsching, U. and Schleicher, A. (1986) Ontogenesis of the laminar structures in area 17
and 18 of the human visual cortex. A quantitative study. Anatomy and Embryology, 174, 129–144.

4 Topographical Variability of Cytoarchitectonic Areas
Jörg Rademacher
Neurologische Klinik, Heinrich-Heine-Universität, Düsseldorf, Moorenstrasse 5, 40225,
Düsseldorf, Germany
Tel: (0049)2461-612107; FAX: (0049)211-81-18485; e-mail: j.rademacher@fz-juelich.de
There is growing interest in the pattern of convolutions in the human brain, in the context of modern
neuroimaging studies of relationships between structure and function, because the sulcal landmarks
can be reliably visualized in vivo by magnetic resonance imaging techniques. It is generally assumed
that the pattern of gyri and sulci is a morphological feature of the brain which is strictly related to the
structural constraints of functional organization. However, the assumption that macroanatomic topo-
graphic landmarks may serve as a guide to functional imaging is problematic, because such a system
depends upon constant relationships of cytoarchitectonic field boundaries to the gyral pattern and
sulci of the brain. Although consistent correlations between the positions of certain architectonic
field boundaries and the primary brain sulci have been reported occasionally, the classical templates,
including the Brodmann map, do not provide such information. At the very best, these templates
include general guidelines for using macroanatomic landmarks that provide the framework for
specific cytoarchitectonic areas. They lack information about the course and size of gyri and sulci,
and do not permit prediction of how these landmarks relate to the architectonic areas. Thus, the early
students of cortical cytoarchitecture do not explain the extent to which cortical areas correlate with
the individual gyral and sulcal pattern. Individual anatomic variations have not been systematically
charted, but there is evidence suggesting that, at least for some cytoarchitectonic areas, there is con-
siderable variability. For practical purpose, one can distinguish “class 1 variability”, which is closely
predictable from the gyral and sulcal landmarks, and “class 2 variability” which is not predictable
from the visible landmarks. The latter introduces a significant error to the localization of function,
if the mapping technique is based on gross anatomical landmarks.
KEYWORDS: brain development, brain mapping, cerebral cortex, cytoarchitecture, gyrification,

morphometry
1. INTRODUCTION
The gyrencephalic human cerebral cortex is distributed over a folded cerebral surface,
thereby allowing for a larger area of cortical surface in the same volume than in case of
a lissencephalic brain with a smooth cerebral surface (Zilles, 1990). Early neuroanatomical
and electrophysiological studies supported the hypothesis that cytoarchitectonically-defined
cortical areas may represent distinct functional units, and that the cerebral sulci may
coincide with the areal borders (Brodmann, 1909; Vogt and Vogt, 1919). Consequently,
the Brodmann map has become the most influential anatomical reference system for
the analysis of structure-function correlations in the human brain. It is used as a two-
dimensional template for a cytoarchitectonically-based parcellation of the cerebral cortex,
which is defined by regional heterogeneities of the cellular and laminar organization.
53
54 Jörg Rademacher
Modern neuroimaging techniques, including positron emission tomography and func-

tional magnetic resonance imaging, subdivide the human cerebral cortex with increasing
spatial resolution. For analysis, the resulting foci of activation are related to visible macro-
anatomic landmarks, because even the most advanced imaging protocols do not permit
direct visualization of the laminar heterogeneities which define the cytoarchitectonic
pattern. It is the general assumption of many brain mapping studies that these gyral and
sulcal landmarks coincide with the borders of cytoarchitectonic areas, as shown in the
Brodmann map. It has also been hypothesized that the intrasulcal cortices may play
a distinctive role in higher cognitive processing, because the most rapid changes in
neuronal activity are frequently observed in the sulcal fundi (Markowitsch and Tulving,
1995). However, the available knowledge about the precise relationship between the topo-
graphy of specific cytoarchitectonic fields and the sulcal and gyral patterns in the general
population is not adequate compared to the demands of structural and functional brain
mapping techniques (Rademacher et al., 1992; Zilles et al., 1995; Geyer, 1996; van Essen
et al., 1998; Roland and Zilles, 1998; Schormann and Zilles, 1998; Amunts et al., 1999).
Anatomic variability may obscure and distort structure-function relationships in several
ways, if the range of individual macroanatomic variations (Eberstaller, 1890; Cunningham,
1892; Galaburda et al., 1987; Steinmetz et al., 1989; Ono et al., 1990; Paus et al., 1996;
Penhune et al., 1996; Thompson et al., 1996; Westbury et al., 1999) and microanatomic
variations (Tables 4.1 and 4.2) is underestimated. First, interindividual gross anatomic
variation regarding the course and the extent of the major (primary) gyri and sulci may
lead to mismatching of anatomy with function (Steinmetz et al., 1990). Caution has to be
urged where map locations from a coordinate system such as the Talairach atlas (Talairach
and Tournoux, 1988) or the Damasio atlas (Damasio and Damasio, 1989) are referenced
to a “standard” brain. This caution is needed especially because the transfer of Brodmann
areas to a three-dimensional reference system represents a “best guess” topography,
derived from Brodmann’s two-dimensional template of one hemisphere and is not based
on systematic anatomical data. Second, while there is considerable overlap in the cyto-
architectonic parcellation of the human brain between the published brain maps (Brodmann,
1909; von Economo and Koskinas, 1925; Sarkisov et al., 1949; Braak, 1980), there is also
considerable variation (Zilles, 1990). These discrepancies between schemes of parcellation
in the identification of areal borders may result from the use of descriptive non-quantitative
criteria by different authors, but they can also be expected to reflect microanatomic vari-
ability. Such variation will probably increase the range of mismatching between anatomy
and function. Third, interhemispheric asymmetries have been shown to exist for many
macroanatomically or cytoarchitectonically defined cortical regions, both in the general
population and in individuals (Geschwind and Levitsky, 1968; Galaburda et al., 1978;
Falzi et al., 1982; Eidelberg and Galaburda, 1984; Steinmetz et al., 1990; Witelson and
Kigar, 1992; Ide et al., 1996; Hutsler et al., 1998; Amunts et al., 1999; Rademacher et al.,
1999). In contrast, the classical cytoarchitectonic maps and the Talairach atlas are based
on the assumption that one hemisphere always represents the topographical mirror
image of its contralateral “homologue”. As a consequence, asymmetries of functional
activation are difficult to interpret. They may reflect different cognitive strategies
between the hemispheres, if bilateral but topographically discrepant foci map onto dif-
ferent cytoarchitectonic areas. However, they may also represent identical functional
units, if they simply follow asymmetries in cytoarchitectonic topography and map onto
the same areas.

Topographical Variability of Cytoarchitectonic Areas 55
In the following sections evidence will be presented on the topographical relationship

between the visible gyri and sulci and the borders of cytoarchitectonic areas, including
developmental aspects and experimental data on gyrus formation. Essentials of gyrus
formation will be described first, because they relate to the more general principles that
underlie this aspect of cortical organization. Results from the making of maps are dis-
cussed thereafter. This dual strategy is suggested to represent a useful concept for imaging
studies, because one may want not only to construct detailed anatomic maps, but also to
characterize the principles that govern them (Friston, 1998).
2. GYRUS FORMATION
Knowledge of the phylogenetic and ontogenetic mechanisms of gyrus formation may help
to improve understanding of the relationship between sulcal landmarks and architectonic
areal borders, as well as the biological range of variations. Several questions arise: Do
these anatomic parameters vary independently or do they show a strict covariation? Are
there systematic differences in the degree of correlation between distinct cortical regions?
If so, does this difference reflect organizational principles, for example, a difference
between the primary somatosensory areas and the association areas? Is there a fundamental
difference between the primary (high incidence rate) and tertiary (low incidence rate)
brain sulci with respect to cytoarchitectonic areal borders? Can cytoarchitectonic (and
functional) homologies or asymmetries be deduced from the sulcal topography? To what
extent are these patterns genetically determined? The ultimate answers to these questions
cannot be given yet, but the increasing research interest in these issue, which is important
to neuroimaging, have already provided interesting new insights.
Cortical folding has been proposed to represent the brain’s solution to the problem of
packing a phylogenetically increasing cortical surface into the restricted volume of the
cerebral vault (Zilles et al., 1988; Welker, 1990). Similar to the general phylogenetic
pattern from lissencephalic to gyrencephalic brains (Zilles et al., 1989), the human cortex
changes from a smooth surface to a highly convoluted structure during ontogenesis (Chi
et al., 1977; Armstrong et al., 1995). The continuous increase in gyrification begins in
ontogenetic week 22 after the majority of neurones have finished their migration into the
cortex. In general, the degree of cortical folding appears to be closely associated with the
size of the brain, if different species are compared (Armstrong et al., 1995). Nevertheless,
anatomical data from primates have shown that the size of individual cytoarchitectonic
areas cannot be predicted simply by regression analysis on the basis of brain weight
(Holloway, 1979) suggesting that additional intrinsic mechanisms may be active. In the
human brain, interindividual differences in the size of a given cytoarchitectonic area have
a wide range, up to a factor ten (Filimonoff, 1932; Stensaas et al., 1974; Galaburda et al.,
1978; Rademacher et al., 1993; Rajkowska and Goldman-Rakic, 1995; Amunts et al.,
1999) and obviously cannot be explained by the size of the individual brain.
It has also been speculated that gyrus formation correlates with the topography of
cytoarchitectonic areas and their respective borders, which tend to coincide with the
course of the cerebral sulci (Welker, 1990). In this model, gyri and sulci (i.e. macroana-
tomic parameters) are interpreted as the expression of specific structural principles and
constraints reflecting intracortical organization (Sisodiya et al., 1996). In fact, the degree
of cortical folding found in adult human brains appears to be a rather constant phenomenon,

56 Jörg Rademacher
with distinct and stable local changes in the anterior-to-posterior direction, showing the
highest values over the prefrontal and parieto-temporo-occipital association cortices and
the lowest values over the primary sensory cortices (Zilles et al., 1988). This finding could
imply that the cytoarchitectonic patterns show a similar constancy in their distribution and
that their relationship to the cerebral gyri is fixed genetically. However, the formation of
cytoarchitectonically defined cortical areas in the brains of mammals does not necessarily
lead to sulcus formation. Lissencephalic monkeys show the same main architectonic sub-
divisions as do other primate brains, including the human cortex (Brodmann, 1909), and
the highly convoluted brains of dolphins or whales do not reflect an increase in cytoarchi-
tectonic areas. Phylogenetically, cytoarchitectonic areal formation and the development of
a convoluted brain are not interdependent. They probably represent distinct processes with
partially overlapping genetic and epigenetic mechanisms, thereby allowing for variations
not only of their respective patterns, but also of their relationship to each other.
Is this phylogenetic evidence paralleled by similar insights from studies of ontogenesis?
Shortly after birth, the convolutedness of the human brain reaches its adult configuration
while the brain itself continues to grow (Chi et al., 1977). Consequently, the postnatal
changes in the degree of cortical folding match those of brain growth thereby maintaining
a constant degree of convolutedness. On the basis of this observation, Armstrong et al.
(1995) have made a qualitative distinction between the primary and secondary brain sulci,
which together establish the degree of convolutedness that characterizes the human brain,
and the tertiary sulci which appear postnatally and only maintain the degree of convolu-
tedness that was previously established. Interestingly, only the topography of the deep and
ontogenetically early primary sulci can be reliably observed in all brains (Ono et al., 1990)
and this appears to be strongly determined genetically (Lohmann et al., 1999). However,
the heritability of the precise overall gyral pattern may be less than 20% (Bartley et al.,
1997), and surprising discordances for sulcal shape have been found among monozygotic
twins (Steinmetz et al., 1995). Thus, one may postulate that an ontogenetic process exists
that can produce profound morphological shifts as determined by random environmental
factors without much genetic change. If this principle is also valid for the relationship
between sulci and cytoarchitectonic borders, one would expect a topographical coinci-
dence in only a few examples, i.e. by chance. In contrast, the available evidence shows that
the rate of a topographical overlap between these two anatomical parameters is much
higher than would be expected, if such relationship was governed by chance (Sanides,
1962; Rademacher et al., 1993; White et al., 1997a,b; Amunts et al., 1999; Geyer et al.,
1999). However, dissociations between the sulcal topography and the borders of the
cytoarchitectonic areas would not be surprising.
In this context, it is a major challenge to analyze how genetically determined intrinsic
control and non-genetically determined extrinsic control have a complementary or con-
current influence on the final amount of overlap between the cytoarchitectonic topography
and the individual gyral pattern. Three standard theories have been proposed. First, the
“mechanical model” of brain development of brain convolutions (Richman et al., 1975)
which can explain the general degree of cortical folding, but not the distinct placement and
orientation of gyri and sulci (Armstrong et al., 1995). In brief, the mechanical model
assumes that the relative amount of cortical folding is the result of the mechanical forces
internal to the cortex and generated by having two differently sized cortical strata (i.e.
supragranular layers I–III vs granular and infragranular layers IV–VI). Second, a “tension-
based” theory of mammalian brain morphogenesis has been proposed that stresses the

importance of mechanical tension along axons and dendrites in the white matter (van
Essen, 1997). According to this model, tension along axons can explain how and why the
cortex folds in a distinct pattern. The underlying phylogenetic (and ontogenetic) principle
would be to keep total axonal length low, thus contributing to the compactness of neural
connectivity in the adult brain. Third, an active process termed “gyrogenesis” is postu-
lated, whereby internal growth processes including cytoarchitectonic differentiation,
ingrowth of thalamic and cortical afferents, selective neuronal death and progressive
myelination move the gyral crowns outward (Rakic, 1988; Kostovic and Rakic, 1990;
Welker, 1990). The concept of gyrogenesis suggests that the formation of gyri and sulci
must bear a close relationship to the cytoarchitectonic parcellation of the cerebral cortex.
It has also been hypothesized to explain better the individual placement, orientation, and
depth of the cerebral convolutions (Armstrong et al., 1995) and this has been supported by
experimental data on gyrus formation (Smart and McSherry, 1986; Ferrer et al., 1988).
The comparative experimental evidence for genetic control of these mechanisms respons-
ible for mammalian cortical development has been presented recently by Rubenstein and
Rakic (1999). In agreement with the concept of gyrogenesis, it takes the complex and ordered
array in functionally distinct cytoarchitectonic areas as depending on species-specific
interactions between intrinsic properties of cortical cells and connectivity between cortical
or subcortical structures (Rubenstein et al., 1999). Defective genes are recognized as
a cause for cortical abnormalities at the microanatomic (cytoarchitecture) and macroana-
tomic (gyri and sulci) level (Raymond et al., 1995). For example, bilateral enucleation in
the foetal monkey leads to a decrease of Brodmann area 17 (primary visual cortex) and to
an increase of Brodmann area 18 (visual association cortex), paralleled by the induction of
supplementary sulci and gyri of the occipital lobe (Dehay et al., 1996). The modified
borders of the striate cortex often coincided with a new and deep sulcus. While these and
the aforementioned results support the concept that the architectonic subdivisions and the
sulcal pattern may show a high degree of anatomical (and functional) plasticity, it also
supports the hypothesis that there is some degree of covariation between the cytoarchitec-
tonic borders and the framework of sulcal landmarks. To date, it is impossible to clarify
the precise impact of the genotype on these biological parameters. Quantitative analyses
of the anatomical phenotype are mandatory to improve our understanding of the relation-
ship between cytoarchitecture and macroanatomy.
3. CORTICAL MAPS
A comprehensive analysis of the relationship of cytoarchitectonic areal borders to the

individual gyral pattern and surrounding sulci needs to be based on larger series of brain
specimens in order to consider both interhemispheric and interindividual variations. The
most relevant anatomic studies which fullfill this criterion are summarized in Tables 4.1
and 4.2. Based on this criterion the present selection of cytoarchitectonic areas is focused
on the primary areas, their direct neighbours, and the language regions. There is a lack of
comparable data relating to other brain regions such as the superior parietal lobe, the infer-
omedial temporal lobe, and others. To evaluate the findings, two classes of variability
were distinguished. Class 1 variability is closely predictable from visible landmarks.
It reflects topographical variations which can be predicted by the framing gyri and sulci
even if these landmarks are variable in a stereotactic system. It does not degrade the

58 Jörg Rademacher
confidence of mapping of a system keyed to landmarks visible in MR images. Class 2

variability, is not predictable from visible landmarks, thereby degrading the confidence of
mapping. The incidence rates in % (IR) of the relevant sulci (for their topography see
Figures 4.1 and 4.2) were taken from the atlas of Ono et al. (1990) which is based on 25
autopsy specimen brains. Given the individual variability described above, attention has
also been payed to differences in the location of specific cytoarchitectonic areas between
the classical maps. The systematic presentation follows well-established criteria of sub-
dividing the cerebral cortex. On the basis of neuroanatomical studies in primates including
the human brain, the cortex is subdivided into the frontal, parietal, temporal and occipital
lobes which contain the primary sensorimotor areas and the association areas (Zilles,
1990). Cytoarchitectonic areas are labeled according to the Brodmann (1909) nomenclature
(i.e. numbers), unless otherwise stated. In addition, the nomenclature of von Economo and
Koskinas (1925) is given in brackets (i.e. letters). Anatomical descriptions are kept short,
and extensive definitions can be found in the original literature (see Tables 4.1 and 4.2;
Ono et al., 1990; Rademacher et al., 1992).
3.1. Frontal Lobe: Primary Motor Cortex

3.1.1. Cytoarchitectonic criteria and macroanatomic landmarks
Area 4 (area FAg) is distinctive for the presence of giant pyramidal cells in layer V, rela-
tively large cells in sublayer IIIc, and a low cell packing density through all layers without
sharp laminar definition. Related macroanatomic landmarks (Figures 4.1 and 4.2) include
the precentral gyrus which lies between the precentral (IR 100%) and the central sulci
(IR 100%). On the medial hemispheric surface, the paracentral lobule extends from the
ascending and descending paracentral sulci (IR 92%) anteriorly to the terminal upswing
of the cingulate sulcus (IR 100%) posteriorly.
3.1.2. Class 1 variability

The identification of area 4 with an observer-independent method (Schleicher et al., 1999)
has shown that it always occupies the posterior bank of the precentral gyrus (Geyer et al.,
1996) thereby supporting earlier observations (Rademacher et al., 1993). Its presence on
the exposed surface of the lateral convexity is restricted to the dorsal-most portion of the
precentral gyrus. The superior frontal sulcus represents the inferior level at which area 4
may be found on the convexity surface. There is no extension of area 4 onto the postcentral
gyrus and little or no extension onto the superior frontal gyrus, so that the central and
precentral sulci represent the posterior and anterior borders of dorsolateral area 4, respect-
ively. More ventrally, area 4 is exclusively localized in the depth of the central sulcus
where it shows a constant extent on its anterior bank (White et al., 1997a). Medially, area
4 is located in the central portion of the paracentral lobule, anterior to the termination of
the central sulcus. The bihemispheric surface topography of area 4 shows roughly sym-
metric patterns (shapes) in individual brains (Rademacher et al., 1993; White et al.,
1997a,b). It has also been shown that certain features visible in MR images provide reli-
able evidence for structural-functional organization. The functionally-defined motor hand
area can be localized to a knob on the middle third of the precentral gyrus (Yousry et al.,
1997) and the asymmetry in the depth of the central sulcus may represent a morphological

Figure 4.1. Sulci of the lateral brain surface. aar, anterior ascending ramus of the Sylvian fissure; ag, angular
sulcus; ahr, anterior horizontal ramus of the Sylvian fissure; ao, anterior occipital sulcus; ce, central sulcus; if,
inferior frontal sulcus; im, intermediate sulcus; ip, intraparietal sulcus; it, inferior temporal sulcus; lo, lateral occi-
pital sulcus; par, posterior ascending ramus of the Sylvian fissure; phr, posterior horizontal ramus of the Sylvian
fissure; poc, postcentral sulcus; prc, precentral sulcus; sf, superior frontal sulcus; st, superior temporal sulcus.
Figure 4.2. Sulci of the medial brain surface. ca, callosal sulcus; calc, calcarine sulcus; cc, corpus callosum;
ce, central sulcus; ci, cingulate sulcus; ma, marginal sulcus; pa, paracingulate sulcus; po, parietooccipital sulcus;
sp, subparietal sulcus; sr, superior rostral sulcus.

60 Jörg Rademacher
correlate of asymmetry in the size of the motor cortex and hand preference (Amunts et al.,
1996).

While the general topography of area 4 appears to be predictable from the gyral land-
marks, the precise borders for example between primary motor area 4 and premotor area 6
differ between hemispheres and do not always match macrostructural landmarks
(Rademacher et al., 1993; White et al., 1997b). Medially, on the paracentral lobule, the
ventral border of area 4 varies considerably with respect to the cingulate sulcus and there
is no constant limiting sulcal landmark for the anterior border of area 4. The ascending and
descending paracentral sulci are mostly located within area 6. Laterally, the intrasulcal size
of area 4 on the anterior lip of the central sulcus shows individual variations by a factor of
up to two. Consequently, the total cortical volume of area 4 cannot be inferred reliably
from the gyral pattern. Similarily, the subdivision of area 4 into areas 4a and 4p, which has
been recognized by Geyer et al. (1996) is not indicated by sulcal borders.
3.1.4. Differences between classical maps

On the dorsal portion of the precentral gyrus area 4 correlates well with Sanides’ motor
area 42 (Sanides, 1962). In contrast, the ventral portions of their homologues in the maps
of Brodmann and von Economo and Koskinas have been depicted with a much larger
surface extent towards the Sylvian fissure. This discrepancy exceeds the range of the
observed anatomic variations as described above and may result from the authors’ inten-
tion to visualize the intrasulcal extent of area 4 on the two-dimensional brain templates.
Medially, area 4 may be relatively large and extends down to the cingulate sulcus
(Brodmann, 1909) or it may be relatively small (i.e. <50%) with only a triangular tongue
reaching the sulcus (von Economo and Koskinas, 1925). In this case, similar discrepancies
can be expected from the maximal range of normal variability (Rademacher et al., 1993).
A location of area 4 beyond the cingulate sulcus onto the cingulate gyrus has not been
described. The results of pigmentoarchitectonic studies are within the range of these
observations (Braak, 1979, 1980). There are conflicting reports on a putative asymmetry
in the human primary motor cortex (Rademacher et al., 1993; Amunts et al., 1996; White
et al., 1997b).
3.2. Frontal Lobe: Broca’s Region

Areas 44 and 45 (areas FCBm and FDG) are characterized by large pyramidal cells in the
depth of layer III. Area 44 is relatively thick, shows distinct cellular columns, a thin layer
II, and a blurring of the border between gray matter and white matter. Area 45 is thinner
than area 44, its layer IV is distinct and layer V is bilaminar and quite narrow. Macro-
anatomically, Broca’s region is localized at the posterior end of the inferior frontal gyrus.
It is subdivided by the ascending branch of the Sylvian fissure (IR 88%) into its triangular
portion anteriorly and its opercular portion posteriorly. Reliable local sulcal landmarks
(IR 100%) include the inferior frontal sulcus, the precentral sulcus and the anterior hori-

zontal branch of the Sylvian fissure (IR 88%) (Figure 4.1). The diagonal sulcus is a more
variable landmark (68%).

Constant relationships between the variable sulcal topography of Broca’s region and the
underlying cytoarchitectonic parcellation have been described in the classical literature
(Stengel, 1930; Riegele, 1931; Sanides, 1962). The borders of areas 45 and 44 have also
been reproduced with an observer-independent method (Amunts et al., 1999). Area 45 is
always localized on the triangular part of the inferior frontal gyrus, anterior to the ascending
part of the Sylvian fissure. The anterior horizontal branch of the Sylvian fissure represents
the anterior and ventral borders of area 45 in most of the brains. Area 44 is constantly
found posterior to area 45 on the opercular part of the inferior frontal gyrus and anterior to
the precentral sulcus. The inferior frontal sulcus always marks the dorsal border of both
areas, and they do not reach the convexity surface of the middle frontal gyrus. Macro-
anatomically defined asymmetries of the inferior frontal gyrus towards the left side (Falzi
et al., 1982) are paralleled by a leftward asymmetry of area 44 (Galaburda, 1980; Amunts
et al., 1999) in a given population.

The precise localization of cytoarchitectonically defined Broca’s region is not always
indicated by the surrounding sulci (Stengel, 1930; Riegele, 1931; Amunts et al., 1999).
For example, its dorsal border may be found in the ventral or dorsal lip of the inferior
frontal sulcus and its posterior border may be detected in the anterior or posterior wall of
the precentral sulcus. With varying frequencies, area 45 reaches onto the orbital surface of
the inferior frontal gyrus. The border between areas 44 and 45 varies up to 2 cm around the
fundus of the ascending branch of the Sylvian fissure. With respect to the interareal
border, the diagonal sulcus does not represent a useful anatomical landmark. The volumes
of area 44 differ across subjects by a factor up to 10, while such variability cannot be
observed for gyral size (Amunts et al., 1999). In addition, it cannot be expected that
the aforementioned class 1 asymmetries towards the left side in a given population show
a strict correlation between micro- and macroanatomy in every individual. For example,
no systematic interhemispheric differences have been detected for area 45 because half of
the brains had larger volumes on the left side and half of the brains showed an inverse
asymmetry (Amunts et al., 1999). However, consistent volumetric asymmetries (left larger
than right) have been found for area 44 in the same brains. This shows that one gyral
region of interest (here: posterior inferior frontal gyrus) may contain divergent asymmetry
patterns of two cytoarchitectonic areas (i.e. area 45 vs area 44).

The available cytoarchitectonic maps differ considerably in the topography, sizes and
sulcal borders of areas 44 and 45. Brodmann (1909) localized area 46 dorsal to area 45,
and area 9 dorsal to area 44, while Sarkisov et al. (1949) localized area 9 dorsal to area 45,
and area 8 dorsal to area 44. While these and other authors (von Economo and Koskinas,
1925) agree on a subdivision of Broca’s area into two cytoarchitectonic areas, those of the

62 Jörg Rademacher
Vogt school (Vogt and Vogt, 1919; Riegele, 1931) described up to five areas in the same
region, defined by myelo- or cyto-architectonics. These differences may result from an
observer-dependent bias in the interpretation of smaller cellular and laminar changes as
reliable borders of distinct architectonic areas. Furthermore, the relative proportions of
areas 44 and 45 vary from apparently equal sizes (Brodmann, 1909; von Economo and
Koskinas, 1925) to a significant preponderance of area 45 (Sarkisov et al., 1949). This
inconsistent pattern can be expected to reflect the normal range of variability in the popu-
lation (Amunts et al., 1999).
3.3. Frontal Lobe: Prefrontal Association Cortex

Area 9 (area FD) is thick and characterized by a low cell-packing density, a blurring of the
laminar borders of layer IV, and a pale sublayer Vb. The border of layer VI to the under-
lying white matter is indistinct. Compared to area 9, area 46 (area FDd) shows a higher
neuronal packing density and its horizontal lamination has sharper borders between all
layers. Characteristically, there is a pronounced radial arrangement of cells and the cortex/
white matter border is sharper than in area 9. Macroanatomically, the superior frontal
gyrus and the middle frontal gyrus serve as landmarks. They are separated by the superior
frontal sulcus (IR 100%). The middle frontal sulcus (IR 86%) subdivides the middle
frontal gyrus into two portions. Laterally, the inferior frontal sulcus (IR 100%) is the ventral
border of the middle frontal gyrus and medially, the cingulate sulcus (IR 100%) is the
ventral border of the superior frontal gyrus (Figures 4.1 and 4.2). The frontomarginal
sulcus (IR 100%) separates the prefrontal and fronto-orbital cortices.

Typically, area 9 is located on the middle third of the superior frontal gyrus, covering its
dorsolateral and dorsomedial portions, while area 46 occupies the central portions of the
middle frontal gyrus (Rajkowska and Goldman-Rakic, 1995). Area 46 is always surrounded
by area 9, and never extends onto the medial brain surface, where the cingulate sulcus is a
reliable ventral border for area 9. Laterally, the inferior frontal sulcus represents a reliable
border for the maximal ventral extent of both areas, and the frontomarginal sulcus may
serve as the border for their maximal anterior extent.

The exact position and size of areas 9 and 46 are characterized by considerable differences
between individuals (Rajkowska and Goldman-Rakic, 1995). Area 9 may also occupy
varying portions of the middle frontal gyrus and its length may vary by a factor of up to
two. The size of area 46 on the exposed surface of the middle frontal gyrus may vary by
a factor of up to five. It also has a variable extent in the depth of the middle frontal sulcus,
including cases with an almost exclusively intrasulcal distribution. Rarely, area 46 occupies
small portions of the superior frontal or inferior frontal gyri. These variations in size of
area 9 cannot be predicted from those of area 46 and vice versa. Needless to say, the size
of the gyri, and the course of the sulci do not permit precise localization or measurement

Table 4.1. Variability of cytoarchitectonic areas: frontal and parietal lobes
Brain region Study Hemispheres Brodmann areas

(new parcellation)
Frontal lobe
Primary motor cortex Rademacher et al., 1993 20 4
Geyer et al., 1996 9 4(4a, 4p)
Zilles et al., 1996 12 4
Amunts et al., 1997 40 4
White et al., 1997a,b 40 4
Broca’s area Riegele, 1931 16 44, 45(57–66)
Kononova, 1949 20 44, 45
Galaburda, 1980 20 44
Amunts et al., 1999 20 44, 45
Prefrontal cortex Kononova, 1949 20 9, 46
Rajkowska and Goldman-Rakic, 1995 10 9, 46
Parietal lobe
Primary somatosensory Rademacher et al., 1993 20 3b
cortex White et al., 1997a,b 40 3
Geyer, Schleicher and Zilles, 1999 20 3a, 3b, 1
Inferior parietal cortex Schulze, 1962 6 40(88, 89), 39(90)
Eidelberg and Galaburda, 1984 16 40(PF, PFG),
39(PG, PEG)
of the underlying cytoarchitecture. Furthermore, inconsistent left-right asymmetries have

been observed in individual brains for areas 9 and 46 (Kononova, 1949).

Variations in location and size of areas 9 and 46 have also been documented in the clas-
sical brain maps. In the Brodmann map, area 46 is located ventrally on the middle frontal
gyrus and extends onto the inferior frontal gyrus, bordering area 45. In contrast, the von
Economo and Koskinas template maps area 46 dorsally on the middle frontal gyrus, sur-
rounded by area 9. Nevertheless, the interindividual differences observed here between
cases exceeds the variability noted in the aforementioned classical maps (Kononova,
1949; Amunts et al., 1999). Consequently, it can be assumed that most of the variability
results from biological diversity. Inter-observer variability may result from the fact that
the cytoarchitectonic borders between areas 9 and 46 exhibit transitional features with the
surrounding areas, but these phenomena appear to introduce only a minor source of
variability.
3.4. Parietal Lobe: Primary Somatosensory Cortex

Area 3 (area PB) is recognized by the high packing density of the small neurones of layers
II–IV, extremely low density of neurones in layer V, and a sharp border between gray and
white matter. Area 3 can be subdivided into areas 3a and 3b (areas PA and PB) with the

64 Jörg Rademacher
latter having a higher cell density and lack of columnar arrangement. Large pyramidal
cells of lower layer III are characteristic of area 1 (area PC). This band of layer III pyra-
mids is reduced in area 2 (area PD). Precentral gyrus, postcentral gyrus and the central
sulcus represent the relevant topographic landmarks (Figures 4.1 and 4.2).

The borders of cytoarchitectonic areas 3a, 3b, 1 and 2 can be recognized by statistically
significant changes in cytoarchitecture (Geyer et al., 1999). Area 3a is located in the
fundus of the central sulcus and area 3b is located in the anterior bank of the postcentral
gyrus (IR 100%). Neither area extends onto the convexity surface of the postcentral gyrus
or onto major parts of the precentral gyrus. The locations of the borders of area 3 (com-
bined areas 3a and 3b) along the central sulcus are remarkably consistent among indi-
vidual brains (White et al., 1997a). Posteriorly, area 1 is consistently found on the crown
of the postcentral gyrus and area 2 is always located in the depth of the postcentral sulcus,
usually on the posterior bank of the postcentral gyrus. The total amount of primary somato-
sensory cortex can be expected to correlate with the size of the postcentral gyrus, because
there is only a small degree of topographical variation and the surface areas usually show
symmetry in size and shape (Rademacher et al., 1993; White et al., 1997a,b). Function-
ally, the interdigitation of the inner walls of the central sulcus halfway between the mid-
line and the Sylvian fissure seems to be reliably related to digit- and hand-related activity
(White et al., 1997a). Interestingly, this “presumptive hand region” can be characterized
macroanatomically, but distinct cytoarchitectonic features have not been observed.

The cytoarchitectonic borders between areas 3, 1 and 2 do not always match the macro-
structural landmarks. For example, the extent of area 3 onto the medial hemispheric sur-
face shows the highest degree of interindividual variability in the primary somatosensory
cortex (Rademacher et al., 1993; White et al., 1997a). Area 1 may be restricted to the
crown of the postcentral gyrus or it may also extend up to 1cm into the depth of the post-
central sulcus in some cases (Geyer et al., 1999).

On the lateral brain surface, the primary somatosensory areas are depicted in a relatively
uniform pattern on the postcentral gyrus (Brodmann, 1909; von Economo and Koskinas,
1925; Sarkisov et al., 1949). Their topography does not appear to vary in the anterior-
to-posterior or dorsal-to-ventral direction. On the templates, the intrasulcal areas 3 and 2
(areas PA, PB, PD) seem to be localized on the crown of the postcentral gyrus, but again
this is an artefact of the two-dimensional drawings. The authors describe an exclusively
intrasulcal extent for these areas. On the medial brain surface, Brodmann (1909) and
Sarkisov et al. (1949) depict areas 3, 1, and 2 in the anterior-to-posterior direction. In
contrast, von Economo and Koskinas (1925) do not map the putative homologue of area 2,
i.e. area PD, onto the medial surface. Their map characterizes area PA, i.e. area 3a,
as horseshoe-shaped and surrounding areas PB and PC, i.e. areas 3b and 1, while the
Brodmann map shows area 5 in the same location.

3.5. Parietal Lobe: Inferior Parietal Cortex

In comparison with the superior parietal lobe, areas 40 and 39 belong to the more granular
cortices with a relative accentuation of layer IV. Area 40 (area PF including areas PFop,
PFT, PFcm, PFm; areas 88 and 89 of Schulze, 1962) is characterized by a prominent
horizontal layering, a thick layer III with small pyramidal cells, decreasing cell volumes
in layers V and VI, and the appearance of radial columns perpendicular to the cortical
surface extending from layers III to VI. Area 39 (area PG; combined areas PEG and PG of
Eidelberg and Galaburda, 1984) is thinner with a granular but narrow layer IV, has larger
cells in layer III, a lower cell density in layer V, and a more prominent horizontal layering.
Macroanatomically, the opercularizing intraparietal sulcus (IR 100%) takes a transaxial
course which represents the dorsal border of the inferior parietal lobe (Figure 4.1). The
posterior segment of the superior temporal sulcus (IR 100%) marks its ventral border.
Anteriorly, the postcentral sulcus (IR 100%) indicates the border. The intermediate sulcus of
Jensen (IR 52%) takes its origin from the intraparietal sulcus and separates supramarginal
and angular gyri.

The depth of the intraparietal sulcus always represents the dorsal border of areas 40 and
39, and the superior temporal sulcus is never crossed in the ventral direction (Brodmann,
1909; Schulze, 1962; Eidelberg and Galaburda, 1984). Consequently, the cytoarchitec-
tonic areas of the inferior parietal lobe are not found on the convolutions of the superior
parietal lobe or the middle temporal gyrus. The major portion of area 40 is localized on the
supramarginal gyrus including its opercular portion in the depth of the Sylvian fissure, and
most of area 39 is localized on the angular gyrus. The postcentral sulcus represents the
anterior border of area 40, and posteriorly the border is marked by the descending portion
of the intraparietal sulcus. The overall size of combined areas 40 and 39 can be estimated
from the size of the inferior parietal lobe with some confidence.

There is no reliable sulcal landmark indicating the precise cytoarchitectonic border
between areas 40 and 39, and their relative sizes compared with each other vary consider-
ably. When present, the intermediate sulcus is roughly correlated with the interareal border
(Eidelberg and Galaburda, 1984), but up to three downward-projecting sulci of the intra-
parietal sulcus may make the interpretation difficult (Ono et al., 1990). Anteriorly, area
40 has been found to cross or to fall short of the postcentral sulcus in single hemispheres.
Posteriorly, the cytoarchitectonic differentiation between area 39 and the occipital cortices
appears to be difficult because of rather small and gradual changes (Schulze, 1962).
Ventrally, an inconstant portion of area 40 occupies the superior temporal gyrus. Minor
variations have also been noted dorsally where both areas are localized at varying positions
of the ventral lip of the intraparietal sulcus (Eidelberg and Galaburda, 1984). Furthermore,
macroanatomic and cytoarchitectonic asymmetries of varying extent and directions exist
in individual brains (Schulze, 1962; Eidelberg and Galaburda, 1984). Although there

66 Jörg Rademacher
are no systematic side differences, mirror images in size and topography of both areas between
the hemispheres of one brain are exceptional (Schulze, 1962).

Brodmann has described two areas on the inferior parietal lobe, area 40 anteriorly and area
39 posteriorly. In contrast, von Economo and Koskinas (1925) and Sarkisov et al. (1949)
have subdivided the homologue of area 40 into four subareas and Schulze (1962)
described two areas in the same region. Area 39 has been subdivided into two or three
areas by Eidelberg and Galaburda (1984) and Sarkisov et al. (1949), respectively. All
classical maps lack information on the intrasulcal extent of area 40 into the Sylvian fissure,
and of areas 40 and 39 into the intraparietal sulcus. The extent of area 39 onto the superior
temporal gyrus and onto the occipital lobe varies considerably between the maps, probably
due to the normal range of variations and difficulties in the identification of sharp borders
in these locations (Eidelberg and Galaburda, 1984). Nevertheless, the relative sizes of both
areas appear to be quite similar, occupying approximately equal portions of the inferior
parietal lobe (Brodmann, 1909; von Economo and Koskinas, 1925; Sarkisov et al., 1949).
3.6. Temporal Lobe: Primary Auditory Cortex

Area 41 (areas TC and TD) is characterized by the extremely high packing density of
small cells in layers II–IV and the paler-staining, low-density neurones in layer V. It lacks
prominent pyramidal cells in layer III. Heschl’s gyrus represents the relevant macro-
anatomic landmark (Figure 4.3). It is the rostral-most transverse gyrus on the superior
Figure 4.3. Posterior superior temporal plane. H, Heschl’s gyrus; left and right cytoarchitectonic areas 41
(line marks on Heschl’s gyrus), 42 (light hatch marks on the central planum temporale), 22 (dark hatch marks on
the lateral planum temporale and lateral Heschl gyrus) and medial prokoniocortex (black); view from above, the
left hemisphere is on the left side; anterior is at the top.

temporal plane and extends from the retroinsular region to the lateral rim of the superior
temporal gyrus. Heschl’s gyrus is bordered anteriorly by the first temporal transverse
sulcus (IR 100%) and posteriorly by Heschl’s sulcus (IR 100%). An intermediate sulcus
(IR 41%) may indent Heschl’s gyrus along its long axis.

In all brains, area 41 is localized on the central portion of Heschl’s gyrus in the depth of
the Sylvian fissure (Rademacher et al., 1993; Morosan et al., 1996), confirming earlier
observations on the relationship between primary auditory cortex and the anterior trans-
verse gyrus (Pfeifer, 1920). Its major portion is limited by the first transverse sulcus
rostrally and Heschl’s sulcus posteriorly. Laterally, area 41 does not extend upon the con-
vexity surface of the temporal lobe and medially, it does not extend beyond the insular
margin (Figure 4.3). The most medial part of Heschl’s gyrus is consistently occupied by
prokoniocortex (Galaburda and Sanides, 1980). Area 41 never extends beyond the second
transverse gyrus (Rademacher et al., 1993).
Table 4.2. Variability of cytoarchitectonic areas: temporal and occipital lobes
Brain region Study Hemispheres Brodmann areas

(new parcellation)
Temporal lobe
Primary auditory cortex von Economo and Horn, 1930 14 41(TC, TD)
Galaburda and Sanides, 1980 6 41(KAm, KAlt)
Rademacher et al., 1993 20 41
Morosan et al., 1996 20 41(Te1.1, 1.0, 1.2)
Clarke and Rivier, 1998 10 41
Wernicke’s region von Economo and Horn, 1930 14 42(TB), 22(TA1)
Braak, 1978 12 42(magn.py.m.),
22(magn.py.c.)
Galaburda, Sanides and 8 22(Tpt)
Geschwind, 1978
Galaburda and Sanides, 1980 6 42(PaAi, PaAe),
22(Tpt)
Witelson, Glezer and Kigar, 1995 18 22(TA1)
Rivier and Clarke, 1997 10 42, 22(AA, PA, LA,
MA, AIA, STA)
Rademacher et al., 1999 22 42, 22
Occipital lobe
Primary visual cortex Brodmann, 1903 4 17
Filimonoff, 1932 13 17
Stensaas et al., 1974 52 17
Rademacher et al., 1993 20 17
Visual association cortex Filimonoff, 1932 13 18, 19
Heinze, 1954 12 18, 19
Clarke and Miklossy, 1990 6 18, 19(V2, V3, VP,
V4, V5)

68 Jörg Rademacher

The hypothesis that there is a typical “standard” brain with two transverse gyri on the right
side and one on the left (Pfeifer, 1920) does not prove to be a useful guideline for topo-
graphic mapping, because variations in the position and configuration of Heschl’s gyrus
complicate the framing of area 41 (Rademacher et al., 1993; Morosan et al., 1996). The
position of Heschl’s sulcus with respect to the temporal pole may vary up to 35 mm (von
Economo and Horn, 1930). Most importantly, there can be multiple transverse gyri which
are frequently indented by an intermediate transverse sulcus. In some hemispheres, the
intermediate sulcus represents the posterior border of area 41, but area 41 may extend
much further in the posterior direction, beyond Heschl’s sulcus.
Bilateral symmetry of areal extent is present only in a minority of cases, while in the
majority (60%), there is an interhemispheric size difference of approximately one third
towards the larger left side (Rademacher et al., 1993). There appears to be no relationship
between asymmetry in the number of transverse gyri and asymmetry of architectonic area
41, i.e. the side differences in the number of transverse gyri are not correlated with similar
asymmetries in the size or topography of area 41.

The location of the primary auditory cortex may vary according to the architectonic
method used to define it. Thus, compared to the pigmentoarchitectonically-defined granular
core area that is reported exclusively on the medial half of Heschl’s gyrus (Braak, 1980),
the cytoarchitectonically defined primary auditory cortex extends more laterally and
covers about two thirds of Heschl’s gyrus (von Economo and Koskinas, 1925). Obviously,
considerable individual variations in the mediolateral extent of area 41 have also been
reported (von Economo and Horn, 1930). In the Brodmann map, the primary auditory
cortex is represented by a single cytoarchitectonic area, while in the map of von Economo
and Koskinas (1925) it contains two cytoarchitectonic fields (TC and TD). Two koniocor-
tical fields within the primary auditory cortex have also been differentiated by Galaburda
and Sanides (1980). With an observer-independent cytoarchitectonic method, three distinct
areas have been recognized in the same region (Morosan et al., 1996) and Clarke and
Rivier (1998) described up to eight additional compartments of unknown functional sig-
nificance based on cytochrome oxidase staining.
3.7. Temporal Lobe: Wernicke’s Region

Wernicke’s region includes areas 42 and 22 and is classically regarded as the morpho-
logical equivalent of the auditory association cortex. Area 42 (area TB) is characterized by
bulky pyramidal cells in layer IIIc, a well granularized layer IV, and a cell poor layer V.
Area 22 (area TA1) is characterized by cell dense layers IV and V with large pyramidal
cells in layer IIIc and smaller pyramids in layer V. There are continuous radial stripes
from layer II to V. Macroanatomically, Wernicke’s region includes the planum temporale
between Heschl’s sulcus and the terminal horizontal (left sided IR 56%; right sided IR
8%) or ascending (left sided IR 44%; right sided IR 92%) segment of the Sylvian fissure

(Figure 4.3). It also contains the posterior portion of the superior temporal gyrus above the
superior temporal sulcus (IR 100%).

Typically, area 42 occupies the major portion (>90%) of the planum temporale (von
Economo and Horn, 1930). Heschl’s sulcus represents in most cases its anterior border
and posteriorly, the end of the horizontal ramus of the Sylvian fissure may serve as the
border for its maximal variation zone (Rademacher et al., 1999). Area 42 does not extend
onto the insular cortex medially and it does not occupy the convexity of the superior tem-
poral gyrus laterally (Figure 4.3). Area 22 covers a cortical stripe of varying extent along
the lateral border of the planum temporale. Its major portion is always localized on the
convexity of the superior temporal gyrus where the superior temporal sulcus represents its
ventral border. The posterior end of the horizontal portion of the Sylvian fissure marks the
posterior border of area 22. Anteriorly, area 22 always reaches the coronal level which
passes through the anterolateral end of Heschl’s sulcus.

The precise cytoarchitectonic borders of areas 42 and 22 do not coincide with gyral or sul-
cal landmarks. Bordering on the medial boundary of area 42 parainsular prokoniocortex
occupies the planum temporale without a limiting sulcus (Galaburda and Sanides, 1980).
Similarily, lateral and immediately adjacent to area 42, area 22 is found without a limiting
sulcus. The anterior and posterior borders of area 22 are also not directly delineated by
brain sulci. In individual brains, area 22 may cross the coronal levels which indicate the
anterolateral end of Heschl’s sulcus or the posterior end of the horizontal segment of the
Sylvian fissure. In addition, the shape of areas 42 and 22 varies considerably and does not
always follow the expected rostrocaudal orientation as described by Galaburda and Sanides
(1980). The cortical volumes of area 42 vary by up to a factor of five and these inter-
subject differences cannot generally be inferred from the size of the planum temporale
(Rademacher et al., 1999).
In the modern literature, gross left-right asymmetry of the planum temporale has first been
described by Geschwind and Levitsky (1968). In a meta-analysis of 520 adult brains from
various studies, a left area larger than the right was found in 74% of all brains, with left/right
ratios ranging from 1.20 to 2.0 (values from Steinmetz et al., 1990). From their cytoarchitec-
tonic study in four brains Galaburda et al. (1978) concluded that there is a strong correlation
between asymmetry of the planum temporale and asymmetry of area 22 which explains the
full structural asymmetry. However, a leftward asymmetry has also been observed for area
42 in more than two thirds of cases in another brain series while area 22 showed leftward
asymmetry in only half of the brains (Rademacher et al., 1999). On the planum temporale,
area 42 was considerably larger than area 22 thus indicating that macroanatomic asymmet-
ries may not simply be reflections of the underlying asymmetry of one architectonic area.

In general, there is good overlap between the classical maps in the topography of area 42
on the planum temporale and area 22 on the middle and posterior third of the superior

70 Jörg Rademacher
temporal gyrus (Brodmann, 1909; von Economo and Koskinas, 1925). Anteriorly, at the
level of Heschl’s gyrus and posteriorly, at the caudal end of the planum temporale, area 22
extends into the depth of the Sylvian fissure. In contrast, Galaburda and Sanides (1980)
also depicted a portion of area 22 (area Tpt) which covers the parietal operculum and pos-
tulated a parietal extension of the auditory region. Topographically, this region is identical
with the posterior portion of area 40 (area PFcm). Witelson et al. (1995) reported that area
22 may in fact occupy the terminal ascending segment of the Sylvian fissure. These dis-
crepancies are further obscured by the finding of macroanatomic Sylvian fissure asym-
metries and planum temporale variability which have usually not been referred to in the
cytoarchitectonic maps (Witelson and Kigar, 1992; Ide et al., 1996; Westbury et al., 1999).
In addition the parcellation of cytoarchitectonic areas on the planum temporale varies
between different brain maps. For example, area 42 has been subdivided into three para-
koniocortical areas (i.e. caudo-dorsal, internal and external) by Galaburda and Sanides (1980).
The external parakoniocortex (so-defined) is supposed to share a major portion of the super-
ior temporal gyrus with area 22. Therefore, the new parcellation scheme includes changes in
the number of areas per macroanatomic unit and significant differences in the local topo-
graphy which cannot be explained by the range of biological variability. Rivier and Clarke
(1997) have described six auditory areas which are located in the posterior temporal region.
Significant controversies over the structural organization of the auditory cortex can be
expected to emerge if results from myeloarchitectonics (Hopf, 1954), pigmentoarchitecton-
ics (Braak, 1978), and cytochrome oxidase staining (Rivier and Clarke, 1997) are included
in a comparison between different areal maps of the superior temporal lobe.
3.8. Occipital Lobe: Primary Visual Cortex

Area 17 (area OC) is defined by its characteristic laminar composition with the granular
layer IV having three distinct sublaminae IVa, IVb, and IVc (stripe of Gennari). This
pattern ceases abruptly at its border with area 18. Macroanatomically, the calcarine sulcus
(IR 100%) is the ventral border of the cuneus’ and the parieto-occipital sulcus (IR 100%)
is its anterior border (Figure 4.2). The cuneal point marks the intersection of both sulci.
The sagittal sulci of the cuneus (IR 46%) and the intralingual sulci (IR 34%) course hori-
zontally across the cuneus and the lingual gyrus, respectively. The collateral sulcus (IR
100%) represents the boundary between lingual gyrus and fusiform gyrus. On the cerebral
convexity, the lateral occipital sulcus (IR 96%) takes a sagittal course (Figure 4.1).

Typically, area 17 is almost exclusively (>90%) localized posteriorly to the cuneal point
on the medial occipital brain surface, where it lies within the region outlined by the parieto-
occipital sulcus anteriorly and the occipital pole posteriorly (Rademacher et al., 1993)
(Figure 4.4). Bilaterally, approximately two thirds of striate cortex are buried in the cal-
carine sulcus, confirming earlier results (Stensaas et al., 1974). With the relative size of
the intracalcarine portion of area 17 compared to the total amount of striate cortex being
relatively constant, the length and depth of the calcarine sulcus can serve as an indicator
for the overall size of area 17 (Filimonoff, 1932). Direct visualization and identification of

Figure 4.4. Surface rendering of the medial occipital lobe. Left and right (L and R) cytoarchitectonic areas
17 (red) and 18 (green); dorsal is at the top; arrowheads, parietooccipital sulcus; (modified from Amunts et al.,
2000). (see Color Plate 3)
the calcarine sulcus are also mandatory for studies of structural-functional mapping,
because its variability zone in a stereotactic reference system measures 2 cm in the vertical
axis (Steinmetz et al., 1990).

Given the diversity of patterning of the cuneal, lingual, collateral and lateral occipital sulci,
it requires some surmise to distinguish the homologies between different hemispheres.
The exact topography of area 17 with regard to the occipital gross anatomical landmarks
and the absolute amount of area 17 show a considerable amount of individual variation
(Filimonoff, 1932; Stensaas et al., 1974; Rademacher et al., 1993; Amunts et al., 2000).
For example, functional activation which maps onto the cuneus may relate either to area
17 or to area 18, depending on the individual anatomy (Figure 4.4). Relative to the size of
the cuneus, the dorsal portion of area 17 may constitute between 12% and 74% of its area
(Rademacher et al., 1993). Also, the varying relationship between the cuneal and lingual
portions of area 17, showing up to fivefold individual differences, cannot be inferred from
the surface relief (Amunts et al., 2000). The total size of area 17 appears to be symmetrical
between the hemispheres for individual brains (Rademacher et al., 1993; Amunts et al.,
2000), but in a minority of cases there may be a relevant asymmetry, with differences
between sides of up to 400% (Stensaas et al., 1974). Interhemispheric differences have
been described for the spatial position of area 17, which is located more medially and
caudally on the left than on the right. Whether area 17 extends onto the lateral convexity
or not cannot be deduced from the sulcal pattern. When present, this portion usually com-
prises less than 10% of the total striate cortex (Rademacher et al., 1993).

The Brodmann map shows an area 17 which takes a sagittal course parallel to the cal-
carine sulcus from the cuneal point anteriorly to the occipital pole posteriorly. The von

72 Jörg Rademacher
Economo and Koskinas map shows the same anterior-to-posterior extent of area 17, but its
shape is triangular with the base at the occipital pole. In contrast with the Brodmann map,
which describes a close relationship between the cuneal and lingual extent of area 17 and
the course of the cuneal and lingual sulci respectively, the von Economo and Koskinas
map excludes such an overlap between limiting sulci and cytoarchitectonic topography.
In general, the modern maps show more irregular patterns in the shape of area 17 than the
classical maps (Rademacher et al., 1993; Amunts et al., 2000) (Figure 4.4).
3.9. Occipital Lobe: Visual Association Cortex

Areas 18 and 19 (areas OB and OA) represent larger parts of the extrastriate visual cortex.
Area 18 has a very prominent layer III and a lower cell packing density than area 17.
There is no stripe of Gennari. Delineation of the border between areas 18 and 19 by visual
inspection alone is problematic in Nissl-stained sections (Zilles, 1990). Filimonoff (1932)
reported that the most typical difference is a more compact layer III with less radial stripes
in area 18. At the parietal border, the remaining radial stripes of area 19 disappear and
there are larger cells in layer III. At the border with area 37 there is a characteristic
increase in the cell packing density of layer V. Recently, area 18 has been analyzed on the
basis of quantitative cytoarchitecture and multivariate statistics (Amunts et al., 2000).
Macroanatomically, the occipitotemporal sulcus (IR 100%) marks the border between the
fusiform gyrus and the inferior temporal gyrus on the ventral brain surface. On the dorso-
lateral brain convexity the external parieto-occipital sulcus (IR 98%) separates the superior
parietal lobe and the occipital lobe. The terminal downward projection of the intraparietal
sulcus (IR 100%) and the anterior occipital sulcus (IR 98%) mark the approximate border
between the inferior parietal lobe and the occipital lobe (Figure 4.1). The lateral occipital
(IR 96%) sulci take a horizontal or oblique course across the lateral occipital lobe and,
when present, the transverse occipital sulcus (IR 62%) is characterized by its vertical
extent near the occipital pole.

Typically, areas 18 and 19 are localized dorsally and ventrally to area 17 on the medial
hemispheric surface. Anteriorly, the parieto-occipital sulcus is a reliable border of area 18
(Amunts et al., 2000) (Figure 4.4) and it may also be so for area 19 (Filimonoff, 1932;
Braak, 1980). Ventrally, the collateral sulcus frequently represents the border between
area 19 and area 37. Considerable amounts of areas 18 and 19 are localized on the lateral
brain surface near the occipital pole with the external parieto-occipital sulcus and the
terminal descending segment of the intraparietal sulcus as the anterior borders of area 19
(Filimonoff, 1932).

There are no sulcal landmarks on the ventral and inferolateral brain surface which repre-
sent the precise macroanatomic or cytoarchitectonic border between the occipital and the
temporal cortices. On the lateral brain surface, the lateral occipital sulci do not represent

useful landmarks for the borders of areas 18 and 19 (Filimonoff, 1932; Amunts et al.,
2000). Similarily, the anterior extent of area 19 onto the parieto-occipital sulcus and the
intraparietal sulcus varies considerably (Filimonoff, 1932). For area 18, topographical
variations appear to be most prominent on the medial brain surfaces (Amunts et al., 2000).
Compared to area 17, the topographical variability of area 18 is considerably larger
(Amunts et al., 2000) and the extent of area 18 is underestimated in the stereotaxic atlas of
Talairach and Tournoux (1988). On the ventral brain surface, the occipitotemporal sulcus
does not appear to coincide systematically with cytoarchitectonic areal borders (Braak,
1980). The mean volumes of area 18 do not differ significantly between the hemispheres
(Amunts et al., 2000), but individual size differences of up to 100% have been reported
(Heinze, 1954). It is not known whether the consistent differences in gyrification between
the occipital association cortex (higher degree of cortical folding) and the occipital pole
(lower degree of cortical folding) may serve as a marker for the indirect localization of
visual association areas 18 and 19 (Gebhard et al., 1993).

Typically, the parieto-occipital sulcus is depicted as the anterior border of the visual associ-
ation cortices on the medial brain surface and the external parieto-occipital sulcus and the
posterior intraparietal sulcus are presented as the anterior border of the visual association
cortices on the lateral brain surface. However, there are also differences between the clas-
sical studies (Brodmann, 1909; von Economo and Koskinas, 1925; Sarkisov et al., 1949).
While the Brodmann map shows that the superior half of the parieto-occipital sulcus is
bordered by area 19 and its inferior half is bordered by area 18, the other maps indicate
that more than 80% of the length of the parieto-occipital sulcus borders area 19. The rela-
tive size of area 18 with respect to area 19 on the cuneus shows all possible variations
including symmetry (Sarkisov et al., 1949), asymmetry in favour of area 18 (Brodmann,
1909) and asymmetry in favour of area 19 (von Economo and Koskinas, 1925). These dif-
ferences and asymmetries can be expected to result from the normal range of variations
(Filimonoff, 1932; Amunts et al., 2000). While the Brodmann map shows the sagittal
lingual sulcus as the dorsal border of area 18 and the collateral sulcus as the ventral border
of area 18, such a clear relationship between the sulcal landmarks and the cytoarchitec-
tonic borders cannot be deduced from the other classical or modern maps (Amunts et al.,
2000). Even more important, the concept of a single area 19 is not supported by the
modern literature on human brain mapping (Clarke and Miklossy, 1990; Zilles, 1990;
Zeki, 1993). Braak (1980) subdivided the same region into 10 pigmentoarchitectonically
defined areas. With few exceptions, the borders of the occipital association cortices do not
show a systematic relationship to the local sulcal pattern (Amunts et al., 2000).
4. ANATOMIC LANDMARKS AND FUNCTIONAL ZONES
Since cortical areas reflect the organization of the cerebral cortex, the issue of parcellating
the cortex is fundamental for human brain mapping (Roland and Zilles, 1994). As described
above, the gyral and sulcal landmarks frequently indicate the approximate location of
a cortical area in the individual brain (class 1 variability), but the precise cytoarchitec-
tonic borders do not match reliably with gyral or sulcal landmarks (class 2 variability).

74 Jörg Rademacher
The aforementioned data from various brain maps indicate that neither do the macro- and
microanatomic parameters vary independently nor do they show a strict covariation. In
this context, there is a fundamental difference between the primary (high incidence rate)
and tertiary (low incidence rate) brain sulci with respect to cytoarchitectonic areal borders.
With varying frequencies, only the former showed relevant relationships with areal
borders. The data also appeared to reflect other systematic differences, in that there was
a higher degree of class 1 variability for the primary somatosensory areas, and a higher
degree of class 2 variability for the association areas. A higher biological variability of the
association cortices compared to relatively strict patterns for the primary cortices may
reflect a useful ontogenetic principle which allows for competition in the final develop-
ment of individual cortical organization. Interindividual discrepancies have been observed
for various anatomic parameters such as areal topography, size, laterality and parcellation.
Between individual brains these differences could vary up to a factor 10, the functional
implications of such enormous variations being unknown. In this context, it seems to be at
least problematic, if cytoarchitectonic and functional homologies or asymmetries are to be
deduced from the sulcal topography and the size of regions of interest as defined macro-
anatomically.
Functional imaging studies have traditionally relied on a “standard” topography derived
from a single brain (Damasio and Damasio, 1989) or on stereotactic reference systems
(Talairach and Tournoux, 1988) for estimating the location of regions of interest. Sub-
stantial anatomical variation must be ignored by these approaches, with inevitable incon-
sistencies and mismatching between anatomy and function (Steinmetz et al., 1990). The
significance of intersubject variability in the topography of functional signal changes
remains unclear, because it may reflect either individual differences in cognitive strategy,
or in anatomic topography. Obviously, a probabilistic atlas of the human brain including
macroanatomic and microanatomic population maps could account for the variance in
position of structures between individuals and hemispheres (Roland and Zilles, 1994;
Maziotta et al., 1995; Zilles et al., 1995). Population maps are defined as three-dimensional
representations of cytoarchitectonic areas in standard anatomical format in a population of
subjects.
Probabilistic cytoarchitectonic maps provide overlay maps for each area showing the
degree of interindividual microstructural variability in any point of the brain’s three-
dimensional space (Morosan et al., 1996; Amunts et al., 1999, 2000; Geyer et al., 1999;
Rademacher et al., 1999). The hypothesis that the cerebral sulcal patterns provide practic-
ally no information about the underlying anatomical organization of the neocortex
(Killackey, 1995) is not supported by these studies. However, also cytoarchitectonic and
macroanatomic parcellations are only interesting if they can be attributed to cerebral func-
tion. For example, many brains show asymmetries in the size of a multitude of gyri and of
cytoarchitectonic areas. Despite these findings, the mean asymmetry scores are not so
frequently asymmetrical, and the large variability of individual side differences indicates
that cortical morphological asymmetry may be present even in the absence of clear func-
tional asymmetry (Hutsler et al., 1998).
Other problems relate to the homologies between different parcellation systems. The
cytoarchitectonically based Brodmann map (1909) depicts about 50 cortical areas while
the myeloarchitectonically based Vogt map shows up to 200 cortical areas (Vogt and
Vogt, 1919). Similarily, it is not known how these subdivisions relate to those of Braak’s
pigmentoarchitectonic method (Braak, 1980). In the case of the human visual association

areas, the classical cortical maps do not describe as many subdivisions as in the well-
established map of the macaque (Zeki, 1993). In contrast, functional and anatomical
(Clarke and Miklossy, 1990) evidence suggests that the human primary visual cortex (V1)
is also surrounded by multiple functionally distinct association areas which can be mapped
onto the gyral pattern and the primary sulci. The visual colour area (V4) maps onto the
posterior portion of the fusiform gyrus (Allison et al., 1993) and the visual motion area
(V5) bears a constant relationship to the posterior segment of the inferior temporal sulcus
and the lateral occipital sulcus (Watson et al., 1993). This shows that discrepancies exist
between the results of any pair of mapping techniques. Therefore, it seems reasonable to
combine cytoarchitectonic and macroanatomic parcellation systems in order to attribute
functional properties to distinct cortical areas.
REFERENCES
Allison, T., Begleiter, A., McCarthy, G., Roessler, E., Nobre, A.C. and Spencer, D.D. (1993) Electrophysiological
studies of color processing in human visual cortex. Electroencephalography and Clinical Neurophysiology,
88, 343–355.
Amunts, K., Schlaug, G., Schleicher, A., Steinmetz, H., Dabringhaus, A., Roland, P.E. and Zilles, K. (1996)
Asymmetry in the human motor cortex and handedness. NeuroImage, 4, 216–222.
revisited: cytoarchitecture and intersubject variability. Journal of Comparative Neurology, 412, 319–341.
brought into stereotaxic space—Where and how variable? NeuroImage, 11, 66–84.
Armstrong, E., Schleicher, A., Omran, H., Curtis, M. and Zilles, K. (1995) The ontogeny of human gyrification.
Bartley, A.J., Jones, D.W. and Weinberger, D.R. (1997) Genetic variability of human brain size and cortical
gyral patterns. Brain, 120, 257–269.
Braak, H. (1978) On magnopyramidal temporal fields in the human brain—probable morphological counterparts
of Wernicke’s sensory speech region. Anatomy and Embryology, 152, 141–169.
Braak, H. (1979) The pigment architecture of the human frontal lobe. I. Precentral, subcentral and frontal region.
Anatomy and Embryology, 157, 35–68.
Braak, H. (1980) Architectonics of the Human Telencephalic Cortex. Berlin: Springer-Verlag.
Brodmann, K. (1909) Vergleichende Lokalisationslehre der Grosshirnrinde. Leipzig: Barth.
Chi, J.G., Dooling, E.C. and Gilles, F.H. (1977) Gyral development of the human brain. Annals of Neurology, 1,
86–93.
Clarke, S. and Miklossy, J. (1990) Occipital cortex in man: organization of callosal connections, related myelo-
and cytoarchitecture, and putative boundaries of functional visual areas. Journal of Comparative
Neurology, 298, 188–214.
Clarke, S. and Rivier, F. (1998) Compartments within human primary auditory cortex: evidence from cyto-
chrome oxidase and acetylcholinesterase staining. European Journal of Neuroscience, 10, 741–745.
Cunningham, D.J. (1892) Contribution to the Surface Anatomy of the Cerebral Hemispheres. Dublin: Royal Irish
Academy.
Damasio, H. and Damasio, A.R. (1989) Lesion Analysis in Neuropsychology. New York: Oxford University Press.
Dehay, C., Giroud, P., Berland, M., Killackey, H. and Kennedy, H. (1996) Contribution of thalamic input to the
specification of cytoarchitectonic cortical fields in the primate: effects of bilateral enucleation in the fetal
monkey on the boundaries, dimensions, and gyrification of striate and extrastriate cortex. Journal of Com-
parative Neurology, 367, 70–89.
Eberstaller, O. (1890) Das Stirnhirn. Ein Beitrag zur Anatomie der Oberfläche des Gehirns. Wien: Urban &
Schwarzenberg.
Eidelberg, D. and Galaburda, A.M. (1984) Inferior parietal lobule. Divergent architectonic asymmetries in the
human brain. Archives of Neurology, 41, 843–852.
Falzi, G., Perrone, P. and Vignolo, L.A. (1982) Right-left asymmetry in anterior speech region. Archives of
Neurology, 39, 239–240.
Ferrer, I., Hernandez-Marti, M., Bernet, E. and Galofre, E. (1988) Formation and growth of the cerebral convolu-
tions. I. Postnatal development of the median-suprasylvian gyrus and adjoining sulci in the cat. Journal of
Anatomy, 160, 89–100.

76 Jörg Rademacher
Filimonoff, I.N. (1932) Über die Variabilität der Grosshirnrindenstruktur. Mitteilung II: Regio occipitalis beim
erwachsenen Menschen. Journal of Psychology and Neurology, 44, 1–96.
Friston, K.J. (1998) Imaging neuroscience: principles or maps? Proceedings of the National Academy of
Sciences U.S.A., 95, 796–802.
Galaburda, A.M. (1980) La région de Broca: observations anatomiques faites un siècle après la mort de son
découvreur. Revue Neurologique, 136, 609–616.
Galaburda, A. and Sanides, F. (1980) Cytoarchitectonic organization of the human auditory cortex. Journal of
Galaburda, A.M., Sanides, F. and Geschwind, N. (1978) Human brain: cytoarchitectonic left-right asymmetries
in the temporal speech region. Archives of Neurology, 35, 812–817.
Galaburda, A.M., Corsiglia, J., Rosen, G.D. and Sherman, G.F. (1987) Planum temporale asymmetry,
reappraisal since Geschwind and Levitsky. Neuropsychologia, 25, 853–868.
Gebhard, R., Zilles, K. and Armstrong, E. (1993) Gross anatomy and gyrification of the occipital cortex in
human and non-human primate. In: B. Gulyas, D. Ottoson and P.E. Roland (eds), Functional Organization
of the Human Visual Cortex. Oxford: Pergamon Press, pp. 101–109.
Geschwind, N. and Levitsky, W. (1968) Human brain: Left-right asymmetries in temporal speech region.
Science, Washington, 161, 186–187.
Geyer, S., Ledberg, A., Schleicher, A., Kinomura, S., Schormann, T., Bürgel, U., Klingberg, T., Larsson, J., Zilles,
K. and Roland, P.E. (1996) Two different areas within the primary motor cortex of man. Nature, London,
382, 805–807.
Geyer, S., Schleicher, A. and Zilles, K. (1999) Areas 3a, 3b, and 1 of human primary somatosensory cortex.
NeuroImage, 10, 63–83.
Heinze, G. (1954) Zytoarchitektonische Untergliederung der Area occipitalis. Journal für Hirnforschung, 1,
173–189.
Holloway, R.L. (1979) Brain size, allometry, and reorganization: toward a synthesis. In: M.E. Hahn, C. Jensen
and B.C. Dudek (eds), Development and Evolution of Brain Size. New York: Academic Press, pp. 59–88.
Hopf, A. (1954) Die Myeloarchitektonik des Isocortex temporalis beim Menschen. Journal für Hirnforschung, 1,
208–279.
Hutsler, J.J., Loftus, W.C. and Gazzaniga, M.S. (1998) Individual variation of cortical surface area asymmetries.
Ide, A., Rodríguez, E., Zaidel, E. and Aboitiz, F. (1996) Bifurcation patterns in the human Sylvian fissure:
hemispheric and sex differences. Cerebral Cortex, 6, 717–725.
Killackey, H.P. (1995) Evolution of the human brain: a neuroanatomical perspective. In: M.S. Gazzaniga (ed.),
The Cognitive Neurosciences. Cambridge: MIT Press, pp. 1243–1253.
Kononova, E.P. (1949) The frontal lobe [in Russian]. In: S.A. Sarkisov, I.N. Filimonoff and N.S. Preobrashen-
skaya (eds), The Cytoarchitecture of the Human Cortex Cerebri, Moscow: Medgiz, pp. 309–343.
Kostovic, I. and Rakic, P. (1990) Developmental history of the transient subplate zone in the visual and somato-
sensory cortex of the macaque monkey and human brain. Journal of Comparative Neurology, 297, 441–470.
Lohmann, G., von Cramon, Y. and Steinmetz, H. (1999) Sulcal variability of twins. Cerebral Cortex, 9, 754–763.
Markowitsch, H.J. and Tulving, E. (1995) Cognitive processes and cerebral cortical fundi. NeuroReport, 15,
413–418.
Mazziotta, J.C., Toga, A.W., Evans, A., Fox, P. and Lancaster, J. (1995) A probabilistic atlas of the human brain:
theory and rationale for its development. NeuroImage, 2, 89–101.
Morosan, P., Schleicher, A., Schormann, T. and Zilles, K. (1996) Cytoarchitectonic mapping of cortical areas on
the first transverse temporal gyrus and intersubject variability. NeuroImage, 3, S141.
Ono, M., Kubik, S. and Abernathey, C.D. (1990) Atlas of the Cerebral Sulci. Stuttgart-New York: Georg Thieme
Verlag.
Paus, T., Tomaiuolo, F., Otaky, N., MacDonald, D., Petrides, M., Atlas, J., Morris, R. and Evans, A.C. (1996)
Human cingulate and paracingulate sulci: pattern, variability, asymmetry, and probabilistic map. Cerebral
Cortex, 6, 207–214.
Penhune, V.B., Zatorre, R.J., MacDonald, J.D. and Evans, A.C. (1996) Interhemispheric anatomical differences
in human primary auditory cortex: probabilistic mapping and volume measurement from magnetic resonance
scans. Cerebral Cortex, 6, 661–672.
Pfeifer, R.A. (1920) Myelogenetisch-anatomische Untersuchungen über das kortikale Ende der Hörleitung.
Abhandlungen der mathematisch physikalischen Klasse der sächsischen Akademie der Wissenschaften, 37,
1–54.
Rademacher, J., Galaburda, A.M., Kennedy, D.N. and Caviness, Jr., V.S. (1992) Human cerebral cortex: local-
ization, parcellation, and morphometry with magnetic resonance imaging. Journal of Cognitive Neuro-
science, 4, 352–374.
Rademacher, J., Caviness, Jr., V.S., Steinmetz, H. and Galaburda, A.M. (1993) Topographical variation of the
human primary cortices and its relevance to brain mapping and neuroimaging studies. Cerebral Cortex, 3,
313–329.

Rademacher, J., Werner, C., Schleicher, A. and Zilles, K. (1999) Planum temporale and language: cytoarchitec-
tonic variability. NeuroImage, 9, S1045.
Rajkowska, G. and Goldman-Rakic, P.S. (1995) Cytoarchitectonic definition of prefrontal areas in the normal
human cortex. II. variability in locations of areas 9 and 46 and relationship to the Talairach coordinate
system. Cerebral Cortex, 5, 323–337.
Rakic, P. (1988) Specification of cerebral cortical areas. Science, Washington, 241, 170–176.
Raymond, A.A., Fish, D.R., Sisodiya, S.M., Alsanjayri, N., Stevens, J.M. and Shorvon, S.D. (1995) Abnormalit-
ies of gyration, heterotopias, tuberous sclerosis, focal cortical dysplasia, microdysgenesis, dysembryoplastic
neuroepithelial tumour and dysgenesis of the archicortex in epilepsy. Clinical, EEG and neuroimaging
features in 100 adult patients. Brain, 118, 629–660.
Richman, D.P., Stewart, R.M., Hutchison, J.W. and Caviness, V.S. (1975) Mechanical model of brain convolu-
tional development. Science, Washington, 189, 18–21.
Riegele, L. (1931) Die Cytoarchitektonik der Felder der Brocaschen Region. Journal of Psychology and Neuro-
logy, 42, 496–514.
Rivier, F. and Clarke, S. (1997) Cytochrome oxidase, acetylcholinesterase, and NADPH-diaphorase staining in
human supratemporal and insular cortex: evidence for multiple auditory areas. NeuroImage, 6, 288–304.
Roland, P.E., and Zilles, K. (1994) Brain atlases—A new research tool. Trends in Neurosciences, 17, 458–467.
Roland, P.E. and Zilles, K. (1998) Structural divisions and functional fields in the human cerebral cortex. Brain
Research Brain Research Reviews, 26, 87–105.
Rubenstein, J.L.R., Anderson, S., Shi, L., Miyashita-Lin, E., Bulfone, A. and Hevner, R. (1999) Genetic control
of cortical regionalization and connectivity. Cerebral Cortex, 9, 524–532.
Rubenstein, J.L.R. and Rakic, P. (1999) Genetic control of cortical development. Cerebral Cortex, 9, 521–523.
Sanides, F. (1962) Die Architektonik des menschlichen Stirnhirns. Berlin: Springer-Verlag.
Sarkisov, S.A., Filimonoff, I.N. and Preobrashenskaya, N.S. (1949) Cytoarchitecture of the Human Cortex
Cerebri [in Russian]. Moscow: Medgiz.
Schleicher, A., Amunts, K., Geyer, S., Simon, U., Zilles, K. and Roland, P.E. (1999) Observer-independent
method for microstructural parcellation of cerebral cortex: a quantitative approach to cytoarchitectonics.
NeuroImage, 9, 165–177.
Schulze, H.A.F. (1962) Quantitative Untersuchungen zur Frage der individuellen Variation und der
Hemisphärendifferenzen der corticalen Areale des unteren Parietalläppchens. Journal für Hirnforschung, 5,
345–376.
Sisodiya, S., Free, S., Fish, D. and Shorvon, S. (1996) MRI-based surface area estimates in the normal adult
human brain: evidence for structural organisation. Journal of Anatomy, 188, 425–438.
Smart, I.H. and McSherry, G.M. (1986) Gyrus formation in the cerebral cortex of the ferret. II. Description of the
internal histological changes. Journal of Anatomy, 147, 27–43.
Steinmetz, H., Rademacher, J., Huang, Y., Hefter, H., Zilles, K., Thron, A. and Freund, H.-J. (1989) Cerebral
asymmetry: MR planimetry of the human planum temporale. Journal of Computer Assisted Tomography,
13, 996–1005.
Steinmetz, H., Fürst, G. and Freund, H.-J. (1990) Variation of perisylvian and calcarine anatomic landmarks
within stereotaxic proportional coordinates. American Journal of Neuroradiology, 11, 1123–1130.
Steinmetz, H., Rademacher, J., Jäncke, L., Huang, Y., Thron, A. and Zilles, K. (1990) Total surface of temporo-
parietal intrasylvian cortex: diverging left-right asymmetries. Brain and Language, 39, 357–372.
Steinmetz, H., Herzog, A., Schlaug, G., Huang, Y. and Jäncke, L. (1995) Brain (a)symmetry in monozygotic
twins. Cerebral Cortex, 5, 296–300.
Stengel, E. (1930) Morphologische und cytoarchitektonische Studien über den Bau der unteren Frontalwindung
bei Normalen und Taubstummen. Ihre individuellen und Seitenunterschiede. Zeitschrift für Experimentelle
und Angewandte Psychologie, 130, 630–677.
cortex in man. Journal of Neurosurgery, 40, 747–755.
Talairach, J. and Tournoux, P. (1988) Coplanar Stereotaxic Atlas of the Human Brain. Stuttgart: Thieme.
Thompson, P.M., Schwartz, C., Lin, R.T., Khan, A.A. and Toga, A.W. (1996) Three-dimensional statistical
analysis of sulcal variability in the human brain. Journal of Neuroscience, 16, 4261–4274.
van Essen, D.C. (1997) A tension-based theory of morphogenesis and compact wiring in the central nervous
system. Nature, London, 385, 313–318.
van Essen, D.C., Drury, H.A., Joshi, S. and Miller, M.I. (1998) Functional and structural mapping of human
cerebral cortex: solutions are in the surfaces. Proceedings of the National Academy of Sciences, U.S.A., 95,
788–795.
Vogt, O. and Vogt, C. (1919) Ergebnisse unserer Hirnforschung. Journal of Psychology and Neurology, 25, 277–462.
Berlin: Springer Verlag.

78 Jörg Rademacher
von Economo, C. and Horn, L. (1930) Über Windungsrelief, Maße und Rindenarchitektonik der Supratempo-
ralfläche, ihre individuellen und ihre Seitenunterschiede. Zeitschrift für Neurologie und Psychiatrie, 130,
678–757.
Watson, J.D.G., Myers, R., Frackowiack, R.S.J., Hajnal, J.V., Woods, R.P., Mazziotta, J.C., Shipp, S. and Zeki,
S. (1993) Area V5 of the human brain: evidence from a combined study using positron emission tomo-
graphy and magnetic resonance imaging. Cerebral Cortex, 3, 79–94.
Welker, W. (1990) Why does cerebral cortex fissure and fold? A review of determinants of gyri and sulci. In:
E.G. Jones and A. Peters (eds), Cerebral Cortex, Vol 8B, Comparative Structure and Evolution of Cerebral
Cortex, Part II. New York: Plenum Press, pp. 3–136.
Westbury, C.F., Zatorre, R.J. and Evans, A.C. (1999) Quantifying variability in the planum temporale: a probab-
ility map. Cerebral Cortex, 9, 392–405.
White, L.E., Andrews, T.J., Hulette, C., Richards, A., Groelle, M., Paydarfar, J. and Purves, D. (1997a) Structure
of the human sensorimotor system. I: morphology and cytoarchitecture of the central sulcus. Cerebral
Cortex, 7, 18–30.
White, L.E., Andrews, T.J., Hulette, C., Richards, A., Groelle, M., Paydarfar, J. and Purves, D. (1997b) Structure
of the human sensorimotor system. II: Lateral symmetry. Cerebral Cortex, 7, 31–47.
Witelson, S.F. and Kigar, D.L. (1992) Sylvian fissure morphology and asymmetry in men and women: bilateral
differences in relation to handedness in men. Journal of Comparative Neurology, 323, 326–340.
Witelson, S.F., Glezer, I.I. and Kigar, D.L. (1995) Women have greater density of neurons in posterior temporal
cortex. Journal of Neuroscience, 15, 3418–3428.
Yousry, T.A., Schmid, U.D., Alkadhi, H., Schmidt, D., Peraud, A., Buettner, A. and Winkler, P. (1997) Localiza-
tion of the motor hand area to a knob on the precentral gyrus. A new landmark. Brain, 120, 141–157.
Zeki, S. (1993) The multiple visual areas of the cerebral cortex. In: S. Zeki (ed.), A Vision of the Brain. Oxford:
Blackwell Scientific Publications, pp. 87–93.
Zilles, K. (1990) Cortex. In: G. Paxinos (ed.), The Human Nervous System. San Diego: Academic Press,
pp. 757–802.
Zilles, K., Armstrong, E., Schleicher, A. and Kretschmann, H.-J. (1988) The human pattern of gyrification in the
cerebral cortex. Anatomy and Embryology, 179, 173–179.
Zilles, K., Armstrong, E., Moser, K.H., Schleicher, A. and Stephan, H. (1989) Gyrification in the cerebral cortex
of primates. Brain, Behavior and Evolution, 34, 143–150.
Zilles, K., Schlaug, G., Matelli, M., Luppino, G., Schleicher, A., Qü., Seitz, R. and Roland, P.E. (1995) Mapping
of human and macaque sensorimotor areas by integrating architectonic, transmitter receptor, MRI and PET
data. Journal of Anatomy, 187, 515–537.

5 Mapping of Human Brain Function
by Neuroimaging Methods
Rüdiger J. Seitz
Department of Neurology, Heinrich-Heine-University Düsseldorf
Correspondence: Department of Neurology, University Hospital Düsseldorf
Moorenstraße 5, D-40225 Düsseldorf
Tel: 0049-211 81-18974; FAX: 0049-211 81-18485
e-mail: seitz@neurologie.uni-duesseldorf.de
This chapter gives an account of functional neuroanatomy of the human brain as revealed by neuro-
imaging methods. Metabolic autoradiography and optical imaging are the foundations of functional
brain mapping, providing robust insights into the topography and temporal dynamics of cerebral
metabolism in laboratory animals. The tomographic imaging methods provide excellent localising
information about activation-related haemodynamic changes, but cannot resolve the temporal
evolution of the underlying electromagnetic brain activity. Further, imaging of human brain function
is hampered by inter-subject variability of brain structure and function as well as by the “partial
volume effect” inherent in tomographic imaging. Nevertheless, the neuroimaging techniques allow
one to study brain function non-invasively in healthy people, making feasible the construction of a
physiological atlas of the different aspects of brain function. New technical developments have
opened avenues for exploiting the temporal aspect of brain activity and analysing functional–
structural relationships with reference to microanatomy. Examples will be described showing a good
correspondence of specific activation studies in healthy subjects to lesion studies in patients with
circumscribed neurological deficits. In contrast, activations which are abnormal in terms of quanti-
tation and localisation have been observed in pathological conditions, suggesting that cerebral
reorganisation can affect functional brain maps. Finally, evidence will be presented showing that the
cerebral activations are influenced by learning and follow remarkable temporal modulations upon
task performance. It is concluded that, the cortical representations of function appear as distributed
computation nodes with widespread cortical and subcortical connections, processing information in
task-related networks.
KEYWORDS: autoradiography, brain activation, brain lesions, cerebral blood flow, cerebral
metabolizm, cerebral reorganisation, deoxyhaemoglobin, electromagnetic activity, optical imaging,
partial volume effect, structural variability, tomographic imaging
1. INTRODUCTION
Functional neuroimaging is a powerful technology for mapping human brain function, and
has undergone enormous development during the past decade (Fox, 1997; Seitz et al., 2000).
Most widely used are measurements of stimulation-related haemodynamic changes. These
changes can be assessed with measurements of the regional cerebral blood flow (rCBF)
using positron emission tomography (PET), and of the blood oxygenation level-dependent
changes (BOLD) using functional magnetic resonance imaging (fMRI). These tomographic
79
80 Rüdiger J. Seitz
imaging tools can localize brain activity changes with relatively good spatial resolution of
approximately 5 to 9 mm (Frackowiak et al., 1994; Calamante et al., 1999). The temporal
resolution of PET and fMRI, however, is relatively poor, being in the range of approxi-
mately 6 s to 1 min, due, respectively, to the tracer kinetics and the haemodynamic charac-
teristics of the measurements. Nevertheless, the reconstructed tomographic imaging data
allow one to detect activity changes occurring simultaneously in different parts of the
brain, including the different parts of the cerebral cortex, subcortical structures as the
basal ganglia and thalamus, and the cerebellum. It should be born in mind, however, that
the observed haemodynamic changes represent only indirect measures of brain activity.
Although, under physiological conditions, there is a tight coupling of activation-related
metabolic and haemodynamic changes to increases in neural activity (Fox et al., 1984;
Blomqvist et al., 1994; Bandettini et al., 1997; Hoge et al., 1999), bioelectric neural activ-
ity has a time-course in the range of several milliseconds, faster than the haemodynamic
measures by three orders of magnitude. Therefore, one of the assumptions underlying
functional imaging with PET and fMRI is that a state of activation has to be kept constant
over a sufficiently long period of time in order to capture functional changes in the dif-
ferent parts of the brain, during a condition approaching a steady state.
Bioelectric activity of the human brain can be recorded directly from the surface of the
head using electroencephalography (EEG) and magnetoencephalography (MEG), the lat-
ter measuring the magnetic fields induced by electrical current flow. Temporal resolution
of these techniques lies in the range of milliseconds, optimally reflecting the dynamics of
brain activity (Hari and Lounasmaa, 1989; Näätänen et al., 1994). In comparison, spatial
resolution is relatively poor, being determined by the number and distribution of the
recording electrodes or sensors covering the head. The electrical potentials or magnetic
fields recorded on the surface of the head, respectively with EEG and MEG, do not how-
ever reflect the localization of the real electrical activity in the brain, because the region-
ally-varying degrees of volume conductance in the cerebrospinal fluid compartment, the
meninges, and in particular the skull severely distort the recorded data. Therefore, spatial
analysis of the recorded data is based on biomathematical models that explain the data,
recorded from the surface of the brain statistically, in terms of intracerebral sources
(Wood et al., 1985; Romani and Rossini, 1988; Scherg, 1990; Kristeva et al., 1991;
Snyder, 1991; de Peralta et al., 1997). In other words, the uncertainty of localization of the
calculated source, the so-called inverse problem, critically depends on the model assump-
tions of head shape and volume conduction. Recently, information obtained from structural
magnetic resonance imaging (MRI) has been used to create realistic head models for the
analysis of the cortical generators of bioelectric activity as recorded with EEG (Dale and
Sereno, 1993; Gevins et al., 1994; Marin et al., 1998). Magnetic fields are virtually unaf-
fected by volume conduction. Therefore, even sources as deep in the brain as the thalamus
or the hippocampal formation can be picked up by MEG (Ribary et al., 1991; Ebersole,
1997). However, dependent on the type of measuring devices, radial or tangential mag-
netic fields remain undetected in the MEG measurements, and this can profoundly effect
the interpretation of the recorded data (Hämäläinen et al., 1993). Furthermore, due to the
limited statistical power inherent in the biomathematical models, only a limited number of
equivalent dipoles can be identified to explain the recorded data. In consequence, either
well-determined experimental stimuli, such as electrical stimuli which evoke potentials, or
simple, phase-locked sensory stimuli or movements have been studied extensively in brain
research. More recently, localization of functional activity changes, as demonstrated by

Functional Brain Mapping 81
rCBF or fMRI measurements, as well as anatomical constraints, have been used for inter-
preting as well as restricting the “search space” in MEG recordings (Heinze et al., 1994;
George et al., 1995; Gerloff et al., 1996; Kinsces et al., 1999; Korvenoja et al., 1999).
Nevertheless, the basic assumption underlying these measures is that the task-specific,
regional neuronal activity has the same temporal and spatial pattern across successive
trials during the event-related recordings of EEG and MEG. Thereby, these methods will
allow one to capture the temporal sequence of events in the brain.
In spite of their methodological limitations, the neuroimaging tools can provide new
insights into the topographic and temporal organization of human brain functions.
Changes of brain activity were originally conceptualized as increased activity in a task-
specific stimulation condition, as compared to a specific control condition (Raichle,
1987). Later, task-specific decreases of brain activity also came into focus (Seitz and
Roland, 1992a; Drevets et al., 1995; Shulman et al., 1997). Such categorical comparisons
were based on the simplified model of a hierarchical implementation of activity in the
human brain, that could be disentangled by psychological subtraction (Petersen et al.,
1988; Kosslyn et al., 1995). Recently, more complex experimental designs have been
addressed analytically to assess the effect of a number of different effective factors. In
these factorial designs, main effects of each variable, as well as the interaction between
these variables in the different psychophysical tasks, can be measured explicitly (Price et al.,
1997). Also, biomathematical approaches have been developed to account for coherent
brain activity in different brain regions, in relation to defined task conditions (Alexander
and Moeller, 1994; Friston, 1994; McIntosh and Gonzalez-Lima, 1994; Büchel et al.,
1997). These statistical approaches can be subdivided into those that are hypothesis-driven
and region-based, thus restricting the search space by imposing an external model
(McIntosh et al., 1994; Azari et al., 1999) and those using “omnibus” statistics, evaluating
the functional connectivity in the entire set of image data (Friston et al., 1993; Seitz et al.,
2001). Most recently, fMRI has been developed further to accommodate also activity
recordings in an event-related fashion (D’Esposito et al., 1997; Buckner et al., 1998;
Friston et al., 1998; Beauchamp et al., 1999). These measurements combine high spatial
resolution with high temporal resolution (Menon and Kim, 1999). Common to these dif-
ferent types of image data analysis, however, is the fundamental idea that human brain
function can be localized in topographically defined maps. The physiological foundations,
technical limitations of functional neuroimaging, and the perspectives of the functional
parcellation of the human brain will be discussed in this chapter.
2. PHYSIOLOGICAL FOUNDATIONS
Microelectrode recordings in non-human primates have provided evidence for the modu-
lar organization of the cerebral cortex. Hubel and Wiesel (1963) as well as Mountcastle
and collaborators (1969) were the first to report that neuronal populations assemble in
vertical units extending from the upper to the lower laminae of the cerebral cortex. These
so-called cortical columns are spatially distinct, representing receptive fields of the different
sensory modalities or motor output modules (Hubel et al., 1978; Phillips et al., 1988; Fetz,
1993). A cortical area represents a local network of neuronal populations which are ana-
tomically organized in cortical units, as evident physiologically by their correlated neuronal
activity (Peters and Kara, 1987; Gray et al., 1990). In primary cortical areas these units

follow a topographical map of their receptive fields (van Essen, 1985; Kaas, 1993). In
motor cortex, however, representations of neighbouring units are intermingled such that
sub-regions represent parts of local networks with highly organized patterns of cortico-
cortical and cortico-subcortical afferents, and efferent connections allowing for specificity
of movement rather than of muscles (Schieber and Hibbard, 1993; Lemon, 1988). More
recently, it has been shown that correlated firing in neighbouring neurones, as recorded with
multielectrodes, provides higher-order features of neural activity and thus far more informa-
tion than the mere neuronal firing rate (Maynard et al., 1999). Furthermore, evidence was
obtained showing that adjacent receptive fields are not invariant, but can be reshaped in
relation to experimental manipulations. For example, repetitive stimulation of a receptive
unit induces a local expansion of the corresponding cortical area, while conversely dener-
vation results in shrinkage and even loss of the corresponding cortical representation (Jenkins
and Merzenich, 1987; Eysel, 1992; Nudo et al., 1996; Kaas and Florence, 1997). In the
motor cortex, the locally interconnected sub-fields provide avenues for reorganization
subsequent to local cortical damage (Stepniewska et al., 1993; Weiss and Keller, 1994).
Autoradiography provides quantitative means of studying the regional glucose con-
sumption (rCMRGLc), the rCBF, and neurotransmitter and neuroreceptor distributions in
laboratory animals. Such studies, which were pioneered by Sokoloff (Sokoloff et al.,
1977) revealed that the cerebral cortex metabolizes the glucose analogue deoxyglucose
in a highly organized, stimulus-related pattern (Kennedy et al., 1976; Juliano et al., 1981).
Figure 5.1 shows a column-like cortical labelling in the striate cortex, with the highest
intensity in cortical layer IV, during visual stimulation of one eye. In comparison to
binocular whole-field visual stimulation, which induced a homogenous labelling of the
occipital cortex, the regular metabolic pattern appeared to correspond to the ocular domin-
ance columns. Exceptions were the areas with exclusively monocular input probably
representing the blind spot (arrows). Likewise, cortical columns in somatosensory cortex
were mapped autoradiographically in relation to simple and complex somatosensory stimuli
(Juliano et al., 1983). Thus, it became obvious that metabolic mapping provided a means
of studying the functional organization of the brain in vivo, with high spatial resolution.
Also, autoradiography was used to demonstrate cortical plasticity in response to experi-
mental brain lesions and peripheral deafferentation (Dietrich et al., 1985; Gilman et al.,
1987; Kossut et al., 1988; Welker et al., 1992). It should be emphasized that these auto-
radiographic recordings demonstrate the total amount of glucose consumption of brain
tissue, which is brought about by the neurones, the surrounding glial cells and to a small
part by the endothelium of the intracerebral vessels.
In addition to the topographic information, autoradiography also provided information
about the driving forces of enhanced metabolic activity. Evidence was obtained showing
that labelling intensity of glucose metabolism correlated with stimulus intensity (Kadekaro
et al., 1985; Yarowsky et al., 1983), as did the intensity of glucose metabolism with the
rCBF (Kuschinski et al., 1981; Cremer et al., 1983; Ueki et al., 1988). More specifically,
regional metabolic activity was shown to reflect potassium ion fluxes at synaptic clefts
(Mata et al., 1980; Kadekaro et al., 1985). Apart from neurones, astrocytes are predominantly
responsible for metabolizing glucose (Magistretti and Pellerin, 1999). Recently, it was shown
that excitation and inhibition are spatially closely related in the cerebral cortex, both
resulting in a graded increase of glucose metabolism (Brühl and Witte, 1995). By applying
optical imaging techniques, it was found that electrically active cortical areas induce a locally
enhanced blood flow and enhanced oxygen consumption (Frostig et al., 1990; Malonek

5.0 mm
Figure 5.1. Autoradiographic mapping of the ocular dominance columns using 2D-deoxy-glucose in the
rhesus monkey. The coronal sections at the level of the striate cortex show intensive metabolic activity during
binocular vision (A), low metabolic activity during bilateral visual deprivation (B), and a patterned metabolic
activity after right eye occlusion (C). Note the maximal metabolic activity in cortical Layer IV and the alternate
dark and light striations during one-eye visual stimulation. The arrows indicate the location of the blind spot
with only monocular input. Taken from Kennedy et al. (1976) with permission.
and Grinvald, 1996). Thus, glucose, the only energy substrate of the primate brain, is metab-
olized in the presence of oxygen, which is supplied rapidly to the area of enhanced brain
activity by a local rise in rCBF, that exceeds quantitatively and spatially the area of enhanced
metabolism (Grinvald et al., 1991). In addition, due to the high temporal resolution of the

optical imaging techniques it became possible to uncover the sequential events of oxygen
supply, oxygen extraction, and haemodynamic response after stimulation onset (Frostig et al.,
1990). Accordingly, oxygen transport from the cerebral blood vessels into brain tissue is the
initial event followed by a graded haemodynamic response. In accord with theoretical pre-
dictions these results support the view that a local rise in blood flow is such that it allows
sufficient regional oxygen supply (Kislyakov and Ivanov, 1986; Buxton and Frank, 1997).
These experimental recordings with nearly microscopical resolution opened detailed
insight into the spatially and temporally organized metabolic-hemodynamic changes
related to brain activity. The mediators for these events are manifold, including potassium
ions, nitrous oxide, and possibly lactate but their individual contributions are still unclear
(Erecinska and Silver, 1989; Iadecola, 1993).
The functional imaging techniques, such as positron emission tomography (PET) and
functional magnetic resonance imaging (fMRI), provide means to study the active human
brain. However, they present only indirect indicators of neural activity, as they measure
the stimulation-related haemodynamic response. PET activation studies rely essentially on
the tracer technology measuring the local task-specific cerebral accumulation of the rCBF
tracer throughout the entire brain (Raichle, 1987). Depending on the biomathematical
quantification model chosen, sampling intervals of 40 to 100 seconds are used in rCBF
studies. Tracers labelled with positron-emitting isotopes such as [15O]-water or [15O]-
butanol allow one to estimate the rCBF in quantitative units by computation with the
time-activity curve of the arterial tracer concentration (Herzog et al., 1996) or semiquanti-
tatively after normalizing the cerebral tracer uptake to the global normal reference value
of 50 ml/100g/min (Friston et al., 1994; Votaw et al., 1999). With respect to the neuronal
processes that evolve in the time frame of milliseconds, an inherent limitation of the rCBF
technique is therefore the long duration of the measuring time needed. In fMRI, the
endogenous blood oxygenation level-dependent (BOLD) contrast is used as an indirect
marker of cerebral blood flow changes (Calamante et al., 1999). Due to the enhanced
blood flow in activated brain areas the amount of diamagnetic oxygenated blood becomes
locally enhanced (see above). Thus, the paramagnetic deoxy-haemoglobin decreases which
results in a signal increase in the fMRI images. In proportion to the fast evolving haemo-
dynamics (Sitzer et al., 1994; Deppe et al., 2000), this signal builds up in some 8 seconds
which can be followed with fast MR sequences including echo-planar imaging (Stehling
et al., 1991; Frahm et al., 1992; Bandettini et al., 1997). While image evaluation of rCBF-
PET capitalizes on identifying significant changes that persist in the steady-state during
the scanning interval compared to the control steady-state, fMRI exploits the consistency
of activation-related changes over a couple of subsequent activation-control cycles. The
areas that survive the different steps of image analysis (see also above) are converted to
pseudocoloured hot spots, which thereafter can be superimposed on co-registered structural
MR images or anatomical templates (Steinmetz et al., 1992a; Frackowiak, 1994). These
hot spots indicate those areas in the brain that are specifically activated by a given task.
However, such an area is not uniquely specialized for this task but rather may subserve
also related operations. Nevertheless, the more routinely a task is performed, the more
focal are the areas of cerebral activation (Roland, 1993).
Interestingly, there seems to be a good correlation between brain electrical activity
and the haemodynamic response as measured with rCBF-PET and fMRI, both during
physiological stimulation (Heinze et al., 1994; Gerloff et al., 1996; Korvenoja et al., 1999)
and under pathological conditions (Volkmann et al., 1998; Krakow et al., 1999). Nevertheless,

the mode of stimulation has been shown to affect the hemodynamic response profoundly. In
the visual cortex, a relation between the haemodynamic response and the stimulation fre-
quency was observed (Fox et al., 1984; Hoge et al., 1999). In contrast, evoked electrical
brain activity, as measured with EEG and MEG, becomes more clearly discernible in rela-
tion to an increasing interstimulus interval (Ibánez et al., 1995; Schnitzler et al., 1999).
Recently, it was shown by simultaneous electrophysiological recordings and fMRI scanning
during visual stimulation in the primate that the BOLD signal closely correlates with the
local field potentials but not with the activity of single neurones (Logothetis et al., 2001).
These data support the idea that neuroimaging can record information processing in the
human brain due to concerted activity of neuronal populations. While these findings hold
not only for the different lobes of the cerebral cortex and most likely also for the cerebellum,
task-related activations may be elusive in the basal ganglia, thalamus, and brainstem struc-
tures. This failure may be due to the diversity of neurotransmitter systems, the higher ana-
tomical packing density of diversified functional units in these structures (Chesselet et al.,
2000; Schmahmann, 2000), as well as to the need to apply specialized approaches to
image analysis (Bohm et al., 1991; Kleinschmidt et al., 1994). In addition, the reader should
be alerted to the observation that there are gender differences of the rCBF and oxygenation
response to physiological stimuli (Kastrup et al., 1999). Finally, it should be pointed out
that the regional activation-induced changes of metabolism and blood flow, though
strongly linked to each other (Roland et al., 1987; Fox et al., 1988; Seitz and Roland,
1992a; Blomqvist et al., 1994; Hoge et al., 1999), have different time constants of return
to baseline, that can be picked-up by neuroimaging techniques (Madsen et al., 1999).
3. INTER-INDIVIDUAL VARIABILITY OF THE HUMAN BRAIN
The most critical issue concerning the transposition of the results of animal experiments to
tomographic imaging of the human brain is the inter-individual variability of human brain
anatomy. This factor influences group image data which have been conceptualized as
probabilistic representations of brain functions as well as the anatomic correspondence of
activation foci related to identical tasks among different subjects.
Inter-individual variability of the human brain has recently come into focus again in
relation to in vivo morphometry from MRI studies of healthy subjects. Qualitative analysis
of MRI images had already revealed profound inter-individual differences in gyrus forma-
tion, even in the brains of monozygotic twins (Steinmetz et al., 1995). By in vivo MRI, it
was shown by quantitative means that gyral configuration of the human brain is highly
variable among different individuals, and largely determined by epigenetic factors
(Steinmetz et al., 1992b; Amunts et al., 1997; Kennedy et al., 1998). Also, there are inter-
hemispheric differences within and between the sexes, probably related to differences in
handedness, skills, and capacities (Schlaug et al., 1995; Gur et al., 1999). Still, after
proportional scaling of the brains of different individuals into stereotactic reference space,
inter-individual variability of the major cerebral sulci at the cerebral surface has a range of
1–2 cm (Steinmetz et al., 1989). These in vivo studies on high-resolution MR images
validated previous findings by Talairach and colleagues obtained from post-mortem inves-
tigations (Talairach et al., 1967). Likewise, other approaches of spatial standardization
using non-linear transformation procedures in addition to linear (affine) ones have also
revealed residual inter-subject variability of a similar magnitude, both in terms of structural

and functional inter-subject variability (Evans et al.., 1988; Seitz et al., 1990; Friston
et al., 1994; Thurfjell et al., 1995; Roland et al., 1997). This is illustrated in Figure 5.2. After
spatial standardization using affine and non-linear transformations, residual variability of
the central sulcus was approximately 4 mm. The structural variability was similar to the
variability of the rCBF increases related to sequential finger movements after applying the
same spatial standardization parameters to the rCBF images as to the MR images. This
residual functional variability was shown to be of such a magnitude that the areas of
activation related to finger movement in eight of nine different subjects overlapped in only
one pixel (Schlaug et al., 1994). In addition, there is considerable inter-subject variability
of the cerebral activation pattern, that appeared to be related to the individual task per-
formance (Schlaug et al., 1994). Similar data were also reported by other groups using
other methods for normalization of spatial image data (Hunton et al., 1996; Hasnain et al.,
1998). Thus, large rCBF changes in tomographic images appeared to be consistent across
subjects, this being different from small rCBF changes that had a relatively large inter-sub-
ject variability (Figure 5.2). This finding is a reflection of the observation that the variance
of rCBF changes is not stationary within the pixel matrix (Grabowski et al., 1996). That is,
in the case of high mean rCBF due to consistent performance across the subjects, the inter-
individual location of the activation focus is more similar than in case of low mean rCBF
related to low and less consistent performance.
The problem of inter-subject variability cannot be discussed without mentioning the
limited spatial resolution, the so-called partial volume effect in tomographic images. The
partial volume effect obscures anatomic resolution and attenuates quantitation of meas-
ured activity (Mazziotta et al., 1981; Bohm et al., 1991). Figure 5.3 shows that the partial
volume effect severely attenuates the signals recorded with functional imaging: The
smaller the object the smaller the signal. The activity of objects with a dimension greater
than three-times the optimal image resolution is recovered at 95% of the true activity only.
The problem of the partial volume effect is of an even greater implication for imaging the
human brain, because the cerebral cortex, and subcortical structures, in particular, are far
smaller than the image resolution of current rCBF and fMRI image data. In addition, the
cerebral gyri are lined by the cerebral sulci that contain inactive cerebrospinal fluid which
further accentuates the partial-volume effect in functional imaging data. The same holds
because of the close neighbourhood of the basal ganglia, thalamus and brain stem nuclei to
the ventricular system. Figure 5.3 also shows that the smaller the signal compared to the
image noise, the greater is the underestimation of the signal in magnitude and spatial
dimension due to the partial volume effect (Knorr et al., 1993). Both, image noise and
the partial volume effect may contribute to false negatives in functional imaging. Thus,
the true activity changes in cerebral cortex and subcortical nuclei are underestimated by
functional imaging, even if the partial volume effect is in the range of a few pixels, as is
the case in fMRI (Calamante et al., 1999).
Consequently, the partial volume effect adds to inter-subject variability of brain config-
uration, by diminishing stimulation-induced signal changes in functional imaging data.
Thus, group image data underestimate the real changes, calculated by taking the peak
changes in the individual subjects, by about of 30% (Seitz et al., 1990). Further, Figure 5.2
shows that inter-subject averaging may result in a slightly different position of the mean
activation centre compared to the arithmetic mean of the peak changes in the individual
subjects. Thus, inter-subject averaging can affect the resulting data not only quantitatively
but also qualitatively with respect to anatomic validity. The danger of mislocalization

Figure 5.2. Influence of the inter-individual variability on the delineation of an activation area in spatially stand-
ardized images of a group of subjects. (A) A great mean rCBF change has a small inter-individual variance
(expressed as standard error of mean rCBF change) in identical pixel locations suggesting consistency of the
activation-induced mean rCBF changes across subjects. Note that pixels with low activation-induced rCBF
changes are quite variable (including also the largest SEM), reflecting a high degree of functional variability
across subjects. (B) Comparison of structural and functional variability expressed as standard deviations. The tip
of the central sulcus (left) had a range of variability of 4 mm compared to the variability of the mean rCBF
increase related to finger movements of 5 to 10 mm (right). The arithmetic mean (♦) of the peak activations in the
spatially standardized images of the individual subjects (●) deviated from the location of the peak of the mean
rCBF increase in the statistical group image (䊊) by approximately 3 mm. For further details see Seitz et al. (1990).

Figure 5.3. Influence of the functional image data on the delineation of an activation area. (A) Profile of the
mean rCBF increase along a continuous row of pixels in the left motor cortex in a group of healthy subjects
during right-hand finger movements. The mean rCBF increase at a threshold of p < 0.01 (28.7%) is lower than
the peak change (41.2%) and smaller in extent than the entire area of mean rCBF increase. The entire area of the
mean rCBF change had an increase of 23.4% compared to the resting control state. (B) Attenuated recovery of
real activity in individual data in relation to object size and image noise. Real activity is recovered at approxi-
mately 95% for large objects with a mean background activity of 10% (open columns). Increasing image noise
progressively impairs object detection and object recovery progressively: 60% background activity (hatched
columns), 78% background activity (black column). Object size is given in relation to image resolution
(FWHM, full width at half maximum). A complete description of the simulation study is given by Knorr et al.
(1993).

A B
Figure 5.4. Superimposition of the mean rCBF increases related to finger movement imagery as compared to
preparation in 10 subjects (A) as yellow area onto the cytoarchitectonic subarea 4a of left motor cortex (B).
Common areas of overlap of the anatomical site of cytoarchitectonic areas 4a (red) and 4p (violet) of two brains
(n = 2) after standardization into reference space of the Human Brain Atlas (Roland et al., 1997). The centres of
gravity (COG) are indicated. Note, the additional activation in the supplementary motor area and the lack of
activation in subarea 4p of motor cortex (A). The location of the central sulcus is indicated as undulated yellow
(A) and black (B) line, respectively. Preliminary data have been communicated by Stephan et al. (1995b) and
were kindly provided by the authors. Left in the image is the lateral surface of the brain, right in the image the
interhemispheric cleft. (see Color Plate 4)
increases with the partial volume effect in the image data. It is well recognized that heavy
image smoothing may enhance the detectability of functional changes, but at the expense
of creating virtual activation centres by merging of small closely adjacent real activity
changes (Worsley et al., 1992). Increasing the partial volume effect artificially by image
smoothing may therefore result in activation areas in possibly unexpected locations.
To get a more realistic view of human brain function, in particular in cognitive tasks, it
is mandatory to use improved methods of image standardization and optimal image res-
olution, as is feasible in the latest generation of PET scanners, and in high-Tesla fMRI
scanners. These technical refinements will enhance the tomographic reflection of true activity
changes, reducing the liability to false negatives and false positives. What will remain is
the inter-individual variability due to anatomic differences in the microscopical dimension.
As is highlighted in the chapter by Rademacher in this volume, there is not even a definite
correspondence of cytoarchitectonic and chemoarchitectonic areas with macroanatomic
landmarks in the human brain. This has been shown for the central sulcus, the lateral and
the calcarine fissure (Rademacher et al., 1993). Accordingly, inter-subject variability also
has a microscopical dimension which can be revealed only when macroscopical variability
can be fully accounted for by spatial standardization procedures. To this end, new methods
have been developed which allow one to transform different brains into the same reference
with a point-to-point correspondence (Christensen et al., 1994; Schormann and Zilles,
1998; Fischl et al., 1999). These processes allow one to create maps of true microanatomic
variability between different individuals, thus increasing the areas of rCBF overlap in
different subjects in group image data, and improving the signal-to-background relation in
activated areas dramatically (Schormann, personal communication).

By these means it will also become possible to study functional subdivisions of cortical
areas. For example, electrophysiological evidence in the somatosensory domain suggests
that the cortical areas 3a and 2 are concerned with proprioceptive information, while areas
3b and 1 are processors of cutaneous input (Kaas, 1993). Likewise, area 4 of the motor
cortex was shown to be made up by two cytoarchitectonically different sub-fields (Geyer
et al., 1996). Co-registration of these cytoarchitectonic data with functional imaging maps
suggest that area 4p in the depth of the central sulcus is concerned with motor execution
while area 4a at the lateral outer aspect of the precentral gyrus is more concerned with
preparatory aspects of movement (Figure 5.4). A question closely related to the issue of
subdivisions within areas is whether identical neuronal populations have the capacity
to assemble in a task-related manner to different subsets affording different functions.
In conclusion, at the microscopical and microelectrode level, respectively, the issue of
structural-functional correspondence is unsettled. Further discussion of this topic is given
in the chapter by Amunts et al., in this volume.
4. LOCALIZATION OF HUMAN BRAIN FUNCTION
Study of localization of human brain function has a long tradition. Until the advent of
computed tomography (CT) this could only be done retrospectively by correlating neuro-
logical deficits with brain lesions at autopsy. CT provided for the first time a means of
making in vivo correlations of brain lesions with clinical deficits, allowing for prospective
analyses. However, since only a few Hounsfield units differentiate tissue compartments
within the brain, and because of the partial volume effect produced by the skull and the
geometrically complicated base of the skull, anatomical resolution is compromised in CT.
Thus, small brain lesions, particularly in white matter structures and the brain stem, are
usually elusive in CT scans. The lesion-based approach for studying the brain was dramat-
ically improved by MRI which is far more sensitive and less prone to the confounds of the
partial-volume effect than CT (Young et al., 1982; Prichard and Brass, 1992). Accordingly,
MRI has become the method of choice for study of correlations between lesions and
deficits.
Brain lesions, however, do not follow functional divisions, but develop within the
pathogenetic framework of the underlying disease. For example, brain infarctions have an
individual configuration following the territories of the cerebral arteries or their branches
(Bogousslavski et al., 1986; Ringelstein et al., 1992). Thus, superimposition of brain
lesions of different patients introduces noise into the group data. Nevertheless, a common
area of lesion overlap is expected to demonstrate a brain area critical for a certain function.
It was therefore surprising that lesion mapping of syndromes like hemineglect was shown
not to be functionally revealing but rather resulted in a reflection of the territory of the middle
cerebral artery (Kertezs and Ferro, 1984). Similarly, motor aphasia was not a sufficient
description of deficit to pinpoint the representative speech area in the brain (Poeck et al.,
1984; Alexander et al., 1990). One explanation for this failure could be the distributed
localization of brain function involving networks of different brain structures—a concept
pioneered by Mesulam (1981, 1990). Accordingly, only damage to a critical node within
such a network interferes with the function of the network. However, there is a large
number of different nodes within such a network, subserving different sub-functions and
allowing for partial rewiring after damage to one or a few of these nodes. Combined

behavioural, electrophysiological and anatomical studies have shown the multiplicity and
diversity of such networks, for example for voluntary control of action (Rizzolatti et al.,
1998). This concept would explain variability of lesions across different subjects present-
ing with a closely similar deficit. However, as was evidenced recently, these functional
deficits have to be fairly well defined in neurophysiological or neuropsychological terms
to reflect a cortical module or a critical node. Examples are the clinically similar but clearly
differentiable syndromes of mirror agnosia, mirror ataxia, and visuomotor ataxia (Rondot
et al., 1977; Binkofski et al., 1999a). Thus, slightly different lesion locations are character-
ized by slightly different patterns of neurophysiological or neuropsychological deficits,
which can be mapped to non-overlapping brain lesions even within one lobe.
A different approach to localizing human brain function is to stimulate the brain or to
record brain activity directly during open brain surgery. Careful documentation of
observed brain lesions and the results of intra-operative cortical stimulation studies have
been the basis for the elaboration of the topographical organization of motor and sensory
representations by Foerster (1936) and subsequently by Penfield and Jasper (1954). More
recently such a systematic approach was used to map brain structures relevant for produc-
tion of speech and music (Ojemann et al., 1989; Creutzfeld and Ojemann, 1989; Haglund
et al., 1992). This type of investigation has, however, a severe inherent confounding
factor, which is the fact that studies are performed on abnormal brains. As lesions interfere
with function, compensatory mechanisms are evoked, resulting in functional reorganiza-
tion which is greater the earlier the lesion was acquired and the more slowly it expands
(Seitz and Azari, 1999; Gadian et al., 1999). Recently, hypothesis-driven stimulation of
the normal cortex using transcranial magnetic stimulation (TMS) has been shown to be
appropriate to map the cortical representations of individual finger muscles in human
motor cortex (Wassermann et al., 1996; Classen et al., 1998a). Quite differently, TMS has
also been used as a probe to interfere with brain function. Thereby, the role of posterior
brain areas for mediating visuospatial functions and visual imagery can be evaluated
(Beckers and Hömberg, 1991; Paus et al., 1996; Kosslyn et al., 1999). The localizing
capacity of TMS is limited however, due to the uncertainty of the exact location and
spatial extent of the stimulated part of the brain.
The most direct approach to the study of human brain function is to measure brain
activity during brain work. This has become possible since the advent of the functional
imaging techniques such as PET, fMRI, and MEG. In most experimental designs and data
analyses of functional neuroimaging, a hierarchical concept of functional representation
has been assumed (Petersen et al., 1988; Kosslyn et al., 1995). This approach follows the
idea that neuropsychological processes can be identified by subtracting from a more com-
plex task, a similar task in which the cerebral computations differ by one aspect. Thereby,
functional units subserving this additional demand should become identifiable. In an ideal
situation, task A would require areas 1 and 2 and task B areas 1 and 3, while the control
task used in both comparisons would engage area 1 to the same degree. It is clear, how-
ever, that area 1 represents a specific activation in task A as in task B thus being elusive by
direct comparisons of tasks A and B. Conversely, since area 1 is of critical importance for
both tasks, it could be demonstrated as an area of a main effect by a conjunction analysis
(Friston, 1997; Price et al., 1997). When area 1 is activated in a different manner in task A
and B, there is a region-specific interaction. When delineating activation areas in functional
imaging data, it is usually tacitly assumed that the activation area reflects a functional
brain unit activated by a given task. However, stimulation-induced activity changes

obtained by matrix statistics of groups of subjects are critically influenced by the image
noise (Figure 5.3). The peak mean activity change exhibits the point with the greatest
significance level, while the area of change at a predefined significance threshold is
smaller than the entire area of change (Lueck et al., 1989). As evident from Figure 5.3,
definition of what is considered to represent the stimulation-related change is a critical
determinant of the spatial dimension and magnitude of this change. The currently available
image processing methods yield astonishingly similar results with an advantage for larger
compared to small samples (Grabowski et al., 1996; Missimer et al., 1999).
In addition to the categorical type of data analysis, multivariate statistics have been
applied for accounting for the concept of distributed cerebral functions, in the sense of
task-related networks involving sub-units in different cortical and subcortical areas.
Approaches of this sort without a priori assumptions are correlation and principal compon-
ents analyses, requiring additional statistical testing to make inferences about the identi-
fied patterns (Friston et al., 1993; Alexander and Moeller, 1994; Friston, 1994). These
results can be displayed as image data, but do not provide any clue about the direction of
information flow among the constituting areas, or about the strength of connectivity
among them. For this goal, the approaches of structural equation modelling can be applied.
These types of analysis incorporate external models to describe inter-regional interactions
(McIntosh and Gonzalez-Lima, 1994; Büchel and Friston, 1997). These methods are,
however, biased by the model-related selection of areas included in the calculation. Also,
in these approaches the number of regions should be less than the number of study subjects,
since otherwise the calculated solutions are not stable.
5. CORRESPONDENCE OF LESION AND ACTIVATION STUDIES
It is widely accepted that brain lesions, by definition, interfere with brain function but do
not necessarily show the site where a certain brain function is represented. On the contrary,
brain imaging usually shows a number of brain areas activated in relation to a certain task.
This becomes evident from neuroimaging studies on visual information processing (see
Gulyas, 1997, for review). Accordingly, network approaches to analysis of image data
have provided evidence that a specific function involves an entire network of brain struc-
tures. For example, object vision involves a number of striate and extrastriate cortical areas
(Horwitz, 1994; Büchel and Friston, 1997). The question, though, is whether a certain
brain structure which is essential for a given function can be identified by the correspond-
ence of lesion and activation studies. A lesion to such a structure would then lead to a per-
manent deficit and coincide with the activation site related to the corresponding activation
paradigm.
In the motor system, it is well established that the mid-dorsal portion of the precentral
gyrus contains the cortical representation of finger and hand movements of the opposite
side of the body. The first evidence for this was obtained from intra-operative stimulation
studies showing that low threshold electrical stimulation of this cortical area induces
muscle twitches of the contralateral forearm (Foerster, 1936; Penfield and Boldrey, 1953).
Likewise, in neuroimaging studies, circumscribed rCBF increases were observed in this
location during hand and finger movements of the contralateral arm (Seitz and Roland,
1992b; Rao et al., 1993; Stephan et al., 1995; Seitz et al., 1996, 1997; Rijntjes et al.,
1999). More importantly, even muscle relaxation activates the same cortical areas (Toma

et al., 1999). There are anatomical landmarks such as the omega-shaped configuration of
the central sulcus on axial MR images, as well as the endpoint of the superior frontal
fissure touching the precentral sulcus at this point, to indicate the location of the motor
hand representation (Ebeling et al., 1992; Yousry et al., 1995). Conversely, combined
morphological and electrophysiological evidence was obtained recently showing that the
degree of destruction of part of the precentral gyrus or its corticospinal projection is a
quantitative indicator of both the loss of individual finger movements of the contralateral
hand, and the capacity for recovery (Binkofski et al., 1996; Pendlebury et al., 1999). It
should also be emphasized that lesion effects are topographically specific. For instance,
leg movements were affected to a lesser degree, or not at all, when precentral lesions
affected arm function, and vice versa (Schneider and Gautier, 1994; Azari et al., 1996).
Furthermore, motor impairment does not correlate with somatosensory impairment of the
same body part (Azari et al., 1996). These results have their counterpart in the rCBF
increases in the frontomesial cortex related to leg movements and in the postcentral gyrus
related to somatosensory stimulation in healthy volunteers (Fox et al., 1987a; Fukuyama
et al., 1997). Similarly, bimanual co-ordination was shown to be significantly impaired in
patients with circumscribed lesions of the mid-part of the cingulate gyrus, while activation
studies in healthy volunteers demonstrated large areas of activation both in the supple-
mentary motor area and in the mid-cingulate cortex, probably involving the cingulate
motor area as well as the lateral premotor cortex (Stephan et al., 1999).
Quantitative investigations revealed persistent motor deficits in patients or primates
with lesions within the precentral gyrus in spite of apparently complete clinical recovery
(Friel and Nudo, 1998; Seitz et al., 1999). Apparently, clinical restoration goes along with
measurable minute deficits of the arm contralateral and ipsilateral to the lesion, and is
often accompanied by subtle, but clear-cut changes of task performance related to alternat-
ive strategies of task performance (Winstein and Pohl, 1995). It should be emphasized that
cortical lesions subsequent to focal ischemia affect not only the local cortical machinery
but also their cortical and cortico-subcortical afferents and efferents. This widespread
effect of cortical lesions can be visualized by metabolic mapping (Seitz et al., 1994,
1999). Consequently, it can be argued that persistent neurological impairments result not
only from the lesion itself but also from suppression or deafferentation of the affected
cortical node from other parts of the network (von Giesen et al., 1994; Classen et al.,
1995). By contrast, circumscribed experimental lesions of cortex sparing subcortical fibre
tracts remain free from associated remote effects (Buchkremer-Ratzmann et al., 1997).
Similar observations have also been made in the visual system, showing a mirror-like
correspondence of the retinotopic activation of the visual cortex upon visual stimulation
and the central or peripheral homonymous visual field defects after circumscribed lesions
of the calcarine cortex or the optic tract (Fox et al., 1987b; Zihl and von Cramon, 1985).
The correspondence of lesion and activation studies also appears to hold for higher order
visual areas. Specifically, there appears to be a close correspondence between clinical
deficits in motion and colour perception related to circumscribed occipito-temporal brain
lesions and the cerebral activations upon motion perception and colour vision in healthy
volunteers (Zihl et al., 1991; Watson et al., 1993; Ungerleider and Haxby, 1994; Tootell
et al., 1995).
Recently, it was demonstrated that such a close correspondence of lesions and sites of
activation is also pertinent for the parietal lobe. Evidence was obtained that unilateral
disturbances of forming a precision grip formation for grasping are associated with lesions

Figure 5.5. Topographic correspondence of an area of lesion overlap among different patients and the site of
activation in healthy volunteers. The patients had a severe impairment of object prehension, while the healthy
subjects specifically activated the same region during prehension movements as contrasted with reaching
(Binkofski et al., 1998). The lesion and the activation foci are localized in the lateral bank of the cortex lining
the anterior part of the intraparietal sulcus (IPS) thus probably corresponding to the anterior intraparietal area.
Note that the lesions of the patients differed in extent (shaded area). CS = central sulcus.
of the parietal cortex lining the anterior portion of the intraparietal sulcus (Binkofski et al.,
1998). Conversely, healthy volunteers required to perform grasping movements showed
a specific activation of the cortex lining the interior part of the intraparietal sulcus in
surprisingly close correspondence to the cortical lesions inducing the grasping deficit
(Figure 5.5). Similarly, lesions of this superior parietal lobule induce severe deficits of
object manipulation, resulting in tactile agnosia (Binkofski et al., 2001). Conversely, neuro-
imaging studies in healthy volunteers showed circumscribed activations of the superior
parietal cortex in a closely corresponding location, related to tactile information process-
ing during exploration of large geometric objects, as well as of textures (Seitz et al., 1991;
O’Sullivan et al., 1994; Binkofski et al., 1999b). Lesions of the inferior parietal cortex
including the parietal operculum have been shown to abolish the ability to identify objects
upon tactile exploration, a syndrome which was termed object agnosia (Caselli et al.,
1993). Recently, fMRI data in healthy volunteers provided evidence that the inferior
parietal cortex, including the secondary somatosensory area, becomes activated during
tactile exploration and identification of large complex geometric objects (Binkofski et al.,
1999b). These data support the notion that the secondary somatosensory area is involved
in processing intrinsic object information necessary for object recognition.
A specific location for speech production was demonstrated recently in a study on lan-
guage recovery, after post-stroke aphasia (Heiss et al., 1999). Patients with frontal and
subcortical infarctions recovered well within 8 weeks after stroke, while patients with an
infarction involving the superior temporal cortex did not. The cortical activation patterns
related to a verb generation task showed activation of the superior temporal cortex bilater-
ally, and of the right inferior frontal cortex in the patients who recovered, whereas the
patients who did not recover showed activation of the inferior frontal cortex in both hemi-
spheres but only of the right superior temporal cortex. These patterns demonstrated the
impact of the left superior temporal cortex for language recovery in post-stroke aphasia.
They correspond to the activation patterns related to verb generation in healthy volunteers,
as sampled across different imaging centres (Poline et al., 1996). Of particular interest in
this connection are also the observations by Friston et al. (1991) who showed that the
superior temporal activations were negatively correlated to the dorsolateral prefrontal

activations, suggesting a modulatory action of the dorsolateral prefrontal cortex on the

temporal language areas. Conversely, a remote functional disturbance in the superior
temporal cortex impaired language abilities in patients with left medial temporal lobe epi-
lepsy (Arnold et al., 1996).
These examples indicate that circumscribed structural lesions (as well as functional
irritations remote from the brain lesions) can interfere with functions that are represented
in exactly these locations as evident from functional imaging. In other words, lesion stud-
ies can provide specific information about critical cortical nodes, whereas corresponding
activation studies demonstrate a number of activated cortical areas inclusive of the critical
area as implicated from lesion studies. Accordingly, the notion that cerebral lesions inter-
fere with the network required to produce a neuronal function, while the network becomes
visible as a whole in neuropsychological activation studies in healthy volunteers, is
supported by the evidence available from combined lesion and activation studies.
6. DISCREPANCY OF LESION AND ACTIVATION STUDIES
In contrast to complete destruction of a cortical module, a partial lesion of this module can
be compensated by reorganization of the remaining parts (Nudo et al, 1996; Xerri et al.,
1998). Similarly, a lesion to one node within a cortical network can be compensated by
reshaping of the remaining parts of the network. Such mechanisms of plastic reorganisa-
tion are supposed to underlie the clinical recovery of function. If so, cerebral lesions and
activation studies are expected to be discrepant. Examples include slowly progressive
brain lesions. Indeed, evidence was obtained showing that low grade gliomas distort the
representation of a given brain function to such an extent that it is mapped to an atypical
location (Wunderlich et al., 1998). This kind of cortical reorganization took place pre-
dominantly within a functional unit, such as the motor cortex. Morphometric measures
performed in parallel to the activation studies revealed that the functional displacement
exceeded the mass effect of the growing tumor (Figure 5.6). In cortical lesions that are
acquired in utero, neonatally or during early infancy, reorganization of function has been
shown even to exceed the limits of functional borders, allowing for abnormal functional
representation in grossly abnormal locations including even the contralateral cerebral
hemisphere (Müller et al., 1998; Seitz and Azari, 1999).
In temporal lobe epilepsy in which a structural abnormality evolved during adolescence,
there may be also a discrepancy between a circumscribed cerebral lesion and the clinical
deficit. It was shown that patients with left temporal lobe epilepsy exhibiting atrophy and
sclerosis of the hippocampal formation are impaired in verbal recall, verbal fluency and
verbal intelligence (Frisk and Milner, 1990; Tranel, 1991). The metabolic changes related
to this disease condition occurred not only in the mesiotemporal region but also in remote
locations in the superior temporal and inferior frontal language areas and in prefrontal
cortex (Arnold et al., 1996; Jokeit et al., 1997). Network approaches to analysis of the data
revealed large-scale abnormalities in these patients involving a weakened relationship of
the bilateral prefrontal cortex but strengthened connectivity of the language areas and the
right thalamus (Figure 5.6). These observations are in accord with the finding that patients
with left temporal lobe epilepsy enjoy recovery of their language functions following
selective anterior temporal lobectomy and thus excision of the irritative lesion (Regard et al.,
1994).

Figure 5.6. Local and large-scale reorganization of functional representations in chronic brain lesions.
(A) Three-dimensional vector displacements of the rCBF increases in the affected cerebral hemisphere of five
patients with precentral gliomas. Origin corresponds to the location of motor hand area as determined for the
normal hand in the contralesional hemisphere. The displacement of the vectors from origin indicates the struc-
tural displacement of the central sulcus. d = dorsal, f = frontal, l = lateral, o = occipital, m = medial, v = ventral.
(B) Abnormal metabolic interactions in patients suffering from left mesiotemporal epilepsy and impairment of
verbal abilities. Compared to healthy controls, weak interhemispheric coupling of prefrontal cortex and
enhanced coupling between the metabolically depressed inferior frontal and superior temporal cortex and the
unaffected right thalamus. The individual image data have been spatially standardized to the stereotactice space
(Talairach and Tournoux, 1988) to allow for inter-individual comparisons.
Another condition in which brain lesions compromise brain function in a remote fashion
is neglect. In motor neglect the somatosensory afferent and motor efferent projections as
well as the corresponding primary cortical processing areas were shown to be functional.
However, circumscribed lesions in quite variable locations interfered with the network
mediating conscious behaviour, resulting in a widespread pattern of metabolic deficits
including premotor, parietal, cingulate and thalamic locations (von Giesen et al., 1994).
Interestingly, this behavioural deficit regressed as the exaggerated cortical inhibition of
the motor cortical apparatus after stroke normalized, as evidenced by transcranial magnetic
stimulation (Classen et al., 1997). These findings have been supplemented by network
analysis approaches showing that post-stroke motor recovery involves large-scale networks
inclusive of brain structures remote from the infarct lesion. Within the motor system there
was a resetting of the pathway from cerebellum to thalamus and supplementary motor area
in the sub-acute post-infarct period (Azari et al., 1996). Furthermore, a network including
thalamus and visual cortex, active with a more prolonged time course after stroke, sug-
gested supramodal compensation of the deficit (Seitz et al., 1999).
In acute stroke magnetic resonance imaging is the method of choice for demonstrating
perfusion abnormalities (Prichard, 1992; Calamante et al., 1999). It should be pointed out
that in such situations conventional and even modern neuroimaging techniques (as for
instance diffusion-weighted imaging) may reveal only small lesions or even be normal
(Tong et al., 1998; Neumann-Haefelin et al., 1999). In these situations, the affected
cerebral cortex is electrically inexcitable as demonstrated in humans and in animal
experiments, remaining profoundly abnormal for a long period of time (Dominkus et al.,
1990; Heald et al., 1993; Bolay and Dalkara, 1998). Specifically, after incomplete

ischemia, the motor cortex may be excitable with transcranial magnetic stimulation in
patients who recovered from stroke, while the haemodynamic response related to functional
activation remains negative (Seitz et al., 1998). Initial data suggest that this electrical-
haemodynamic decoupling develops in the subacute stage after infarction (Binkofski et al.,
1999c). At present it is unclear whether, and if so, when the electrical-haemodynamic
coupling will become restored again.
As outlined above, the assumption underlying rCBF and fMRI measurements is that
brain function (as characterized by the changing demands of electric activity upon region-
ally enhanced metabolism) can be picked up indirectly by the haemodynamic response.
Changes of the threshold of neuronal excitability have however been shown to remain
elusive in such measurements, while they can be recognized by transcranial magnetic
stimulation and MEG recordings (Schnitzler et al., 1995; Stephan and Frackowiak, 1996).
Specifically, it was shown that imagery of motor activity fails to induce rCBF changes in
the motor cortex of a magnitude, spatial extent or significance comparable to that in
premotor, cingulate, and parietal cortical areas (Stephan et al., 1995a; Porro et al., 1996;
Roth et al., 1996; Seitz et al., 1997). Rather, only a portion in the lateral subarea of motor
cortex appears to be activated (Figure 5.4). Nevertheless, excitability of the motor cortex
after motor imagery was enhanced, threshold for excitation being reduced at the same
time (Schnitzler et al., 1995; Stephan and Frackowiak, 1996). Likewise, visual imagery
induces only minute haemodynamic changes in primary visual cortex (Roland and Gulyas,
1994). Furthermore, it was shown that transcranial magnetic stimulation over the occipital
cortex eliminated visual imagery while it was ineffective in posterior temporal application
(Kosslyn et al., 1999). These data show that haemodynamic measurements may only
partially reflect activity changes in the brain, adding to possible dissociations of lesion and
activation studies.
7. CORTICAL REPRESENTATIONS OF FUNCTION
The cortical representations of function have been shown to be affected remarkably by

physiological learning. Using fMRI it was shown that learning of sequential finger move-
ments induces a spatial increase of the cortical finger representation as learning pro-
gressed (Karni et al., 1995). Group data of motor learning provided evidence of a common
area of representation of hand-finger movement, with a clear rate effect on the rCBF
increase, and irrespective of the degree of learning of finger movement (Seitz and Roland,
1992; Grafton et al., 1992; Jenkins et al., 1994). Using transcranial magnetic stimulation,
it was however demonstrated that the threshold for cortical excitability varies in relation to
task acquisition. The excitable area of motor cortex was most extensive when the task was
explicitly and routinely performed (Pascual-Leone et al., 1994, 1995). Moreover, it was
shown that these changes coded movement direction (Classen et al., 1998b). Also, during
skill learning the motor cortical activation becomes progressively lateralized (Seitz and
Roland, 1992b). This change was accompanied by a temporal reorganization of EMG
activity from the typical three-phasic burst discharges related to individual finger move-
ments to a two-burst discharge pattern after learning, probably reflecting the establishment
of a tremor-like activity in the cortico-subcortical loop.
Event-related fMRI has provided evidence that the cortical representations of function
may also be modulated in a highly organized fashion in the temporal dimension. For

example, it was shown that in working memory conditions the haemodynamic response,
in those cortical areas concerned with stimulus perception, preceded those in areas related
to response generation, while prefrontal areas showed enhanced haemodynamic responses
during the entire delay period between stimulus presentation and the response execution
(D’Esposito et al., 1997; Menon and Kim, 1999). Likewise, the perception of ambiguous
stimuli resulted in antagonistic response curves in different visual sub-fields, which were
related to the actual perception of the stimuli (Kleinschmidt et al., 1998).
As outlined in this chapter, activation studies in healthy subjects and lesion studies in
patients with well-defined clinical deficits provide means for elucidating the modular
organization of the human brain. Thus, construction of a physiological atlas of the differ-
ent brain functions appears feasible. One should bear in mind that anatomical afferents
and efferents converge at the observed, activated nodes allowing the processing of
information gathered from different sources. This view of a task-related assembly of
cerebral networks comprising different critical nodes would apply to the so-called primary
areas and also to the higher order associative cortices. The view is also capable of
accounting for plastic topographical and temporal reorganization related to physiological
learning and to adaptation to pathological conditions. Finally, this view would explain the
relative paucity of corticofugal efferent and corticopetal afferent projections in the pres-
ence of a plethora of cortical-cortical connections, which can be estimated from the large
dimension of cerebral white matter compared to the small diameter of the pyramidal tract,
the posterior spinal columns, and the optic nerves.
REFERENCES
Alexander, M.P., Naeser, M.A. and Palumbo, C. (1990) Broca’s area aphasia: aphasia after lesions including the
frontal operculum. Neurology, 40, 353–362.
Alexander, G.E. and Moeller, J.R. (1994) Application of the subprofile model to functional imaging in neuro-
psychiatric disorders: A principal component approach to modeling brain function in disease. Human Brain
Mapping, 2, 79–94.
Amunts, K., Schmidt-Passos, F., Schleicher, A. and Zilles, K. (1997) Postnatal development of inter-
hemispheric asymmetry in the cytoarchitecture of human area 4. Anatomy and Embryology (Berlin),
196, 393–402.
Arnold, S., Schlaug, G., Niemann, H., Ebner, A., Lüders, H., Witte, O.W. and Seitz, R.J. (1996) Topography of
interictal metabolic depressions in mesiotemporal lobe epilepsy. Neurology, 46, 1422–1430.
Azari, N.P., Binkofski, F., Pettigrew, K.D., Freund, H.-J. and Seitz, R.J. (1996) Enhanced regional cerebral
metabolic interactions in thalamic circuitry predicts motor recovery in hemiparetic stroke. Human Brain
Mapping, 4, 240–253.
Azari, N.P., Arnold, S., Schlaug, G., Niemann, H., Ebner, A., Lüders, H., Witte, O.W. and Seitz, R.J.
(1999) Reorganized patterns of metabolic interactions in temporal lobe epilepsy. Neuropsychologia,
37, 625–636.
Bandettini, P.A., Kwong, K.K., Davis, T.L., Tootell, R.B.H., Wong, E.C., Fox, P.T., Belliveau, J.W., Weisskoff,
R.M. and Rosen, B.R. (1997) Characterization of cerebral blood oxygenation and flow changes during
prolonged brain activation. Human Brain Mapping, 5, 93–109.
Beauchamp, M.S., Haxby, J.V., Jennings, J.E. and DeYeo, E.A. (1999) An fMRI version of the Farnsworth-
Munsell 100Hue test reveals multiple color-sensitive areas in human ventral occipitotemporal cortex.
Beckers, G. and Hömberg, V. (1991) Impairment of visual perception and visual short term memory scanning by
transcranial magnetic stimulation of occipital cortex. Experimental Brain Research, 87, 421–432.
Binkofski, F., Seitz, R.J., Arnold, S., Claßen, J., Benecke, R. and Freund, H.-J. (1996) Thalamic metabolism and
integrity of the pyramidal tract determine motor recovery in stroke. Annals of Neurology, 39, 460–470.
Binkofski, F., Dohle, C., Posse, S., Stephan, K.M., Hefter, H., Seitz, R.J. and Freund, H.-J. (1998) Human
anterior intraparietal area subserves prehension. A combined lesion and functional MRI activation study.
Neurology, 50, 1253–1259.

Binkofski, F., Buccino, G., Dohle, C., Seitz, R.J. and Freund, H.-J. (1999a) Mirror agnosia and mirror ataxia
constitute different parietal lobe disorders. Annals of Neurology, 46, 51–61.
Binkofski, F., Buccino, G., Posse, S., Seitz, R.J., Rizzolatti, G. and Freund, H.-J. (1999b) A fronto-parietal
circuit for object manipulation in man: evidence from an fMRI study. European Journal of Neuroscience,
11, 3276–3286.
Binkofski, F., Fink, G.R., Wittsack, H.-J., Buccino, G., Seitz, R.J. and Freund, H.-J. (1999c) Lesion location
determines cortical activation patterns in stroke recovery. Neuroimage, 9, S726.
Binkofski, F., Kunesch, E., Dohle, C., Seitz, R.J., Freund, H.-J. (2001) Tactile Apraxia: unimodal apractic dis-
order of tactile object exploration associated with parietal lobe lesions. Brain, 124, 132–144.
Blomqvist, G., Seitz, R.J., Sjögren, I., Halldin, C., Stone-Elander, S., Widen, L., Solin, O. and Haaparanta, M.
(1994) Regional cerebral oxidative and total glucose consumption during rest and activation studied with
positron emission tomography. Acta Physiologica Scandinavica, 151, 29–43.
Bohm, C., Greitz, T., Seitz, R.J. and Eriksson, L. (1991) Specification and selection of regions of interest (ROIs)
in a computerized brain atlas. Journal of Cerebral Blood Flow and Metabolism, 11 (Suppl 1), A64–A68.
Bogousslavsky, J., Regli, F. and Assal, G. (1986) The syndrome of unilateral tuberothalamic artery territory
infarction. Stroke, 17, 434–441.
Bolay, H. and Dalkara, T. (1998) Mechanisms of motor dysfunction after transient MCA occlusion: persistent
transmission failure in cortical synapses is a major determinant. Stroke, 29, 1988–1994.
Brühl, C. and Witte, O.W. (1995) Cellular activity underlying altered brain metabolism during focal epileptic
activity. Annals of Neurology, 38, 414–420.
Büchel, C. and Friston, K.J. (1997) Modulation of connectivity in visual pathways by attention: cortical inter-
actions evaluated with structural equation modelling and fMRI. Cerebral Cortex, 7, 768–778.
Buchkremer-Ratzmann, I. and Witte, O.W. (1997) Extended brain disinhibition following small photothrombotic
lesions in rat frontal cortex. NeuroReport, 8, 519–522.
Buckner, R.L., Goodman, J., Burock, M., Rotte, M., Koutstaal, W., Schacter, D., Rosen, B. and Dale, A.M.
(1998) Functional-anatomic correlates of object priming in humans revealed by rapid presentation of
event-related fMRI. Neuron, 20, 285–296.
Buxton, R.B. and Frank, L.R. (1997) A model for the coupling between cerebral blood flow and oxygen
metabolism during neural stimulation. Journal of Cerebral Blood Flow and Metabolism, 17, 64–72.
Calamante, F., Thomas, D.L., Pell, G.S., Wiersma, J. and Turner, R. (1999) Measuring cerebral blood flow using
magnetic resonance imaging techniques. Journal of Cerebral Blood Flow and Metabolism, 9, 701–735.
Caselli, R.J. (1993) Ventrolateral and dorsomedial somatosensory association cortex damage produces distinct
somesthetic syndromes in humans. Neurology, 43, 762–771.
Chesselet, M.F. (2000) Mapping the basal ganglia. In: A.W. Toga and J.C. Mazziotta (eds), Brain Mapping: the
systems. Academic Press: San Diego, pp. 177–206.
Christensen, G.E., Rabbitt, R.D. and Miller, M.I. (1994) 3-D brain mapping using deformable neuroanatomy.
Physics in Medicine and Biology, 39, 609–618.
Classen, J., Kunesch, E., Binkofski, F., Hilperath, F., Schlaug, G., Seitz, R., Glickstein, M. and Freund, H.-J.
(1995) Subcortical origin of visuomotor apraxia. Brain, 118, 1365–1374.
Classen, J., Schnitzler, A., Binkofski, F., Werhahn, K.J., Kim, Y.-S., Kessler, K.R. and Benecke, R. (1997) The
motor syndrome associated with exaggerated inhibition within the primary motor cortex. Brain, 120, 605–619.
Classen, J., Knorr, U., Werhahn, K.J., Schlaug, G., Kunesch, E., Cohen, L.G., Seitz, R.J. and Benecke, R.
(1998a) Output mapping of human central motor representation: multiple modalities and multiple spatial
scales. Journal of Physiology (London), 512, 163–179.
Classen, J., Liepert, J., Wise, S.P., Hallett, M. and Cohen, L.G. (1998b) Rapid plasticity of human cortical
movement representation induced by practice. Journal of Neurophysiology, 79, 1117–1123.
Cremer, J.E., Cunninham, V.J. and Seville, M.P. (1983) Relationsphips between extraction and metabolism of
glucose, blood flow, and tissue blood volume in regions of rat brain. Journal of Cerebral Blood Flow and
Metabolism, 3, 291–302.
Creutzfeld, O. and Ojemann, G. (1989) Neuronal activity in the human lateral temporal lobe. I. Responses to
speech. Experimental Brain Research, 77, 451–475.
Dale, A.M. and Sereno, M.I. (1993) Improved localization of cortical activity by combining EEG and MEG with
MRI cortical surface reconstruction: a linear approach. Journal of Cognitive Neuroscience, 5, 162–176.
Deppe, M., Knecht, S., Papke, K., Lohmann, H., Fleischer, H., Heindel, W., Ringelstein, E.B. and Henningsen, H.
(2000) Assessment of hemispheric language lateralization: a comparison between fMRI and fTCD. Journal
of Cerebral Blood Flow and Metabolism, 20, 263–268.
de Peralta Menendez, R.G., Hauk, O., Andino, S.G., Vogt, H. and Michel, C. (1997) Linear inverse solutions with
optimal resolution kernels applied to electromagnetic tomography. Human Brain Mapping, 5, 454–467.
D’Esposito, M., Aguirre, G.K., Zarahn, E. and Ballard, D. (1997) Functional MRI studies of spatial and non-
spatial working memory. Cognitive Brain Research, 7, 1–13.
Dietrich, W.D., Ginsberg, M.D., Busto, R. and Smith, D.W. (1985) Metabolic alterations in rat somatosensory
cortex following unilateral vibrissal removal. Journal of Neuroscience, 5, 874–880.

Dominkus, M., Grisold, W. and Jelinek, V. (1990) Transcranial electrical motor evoked potentials as a prog-
nostic indicator for motor recovery in stroke patients. Journal of Neurology Neurosurgery and Psychiatry,
53, 745–748.
Drevets, W.C., Burton, H., Videen, T.O., Snyder, A.Z., Simpson, J.R.Jr. and Raichle, M.E. (1995) Blood flow
changes in human somatosensory cortex during anticipated stimulation. Nature (London), 373, 249–252.
Ebeling, U., Schmid, U.D., Ying, H. and Reulen, H.J. (1992) Safe surgery of lesions near the motor cortex using
intra-operative mapping techniques: a report on 50 patients. Acta Neurochirurgia, 119, 23–28.
Ebersole, J.S. (1997) Magnetoencephalography/magnetic source imaging in the assessment of patients with
epilepsy. Epilepsia, 38 (Suppl 4), S1–S5.
Erecinska, M. and Silver, I.A. (1989) ATP and Brain Function. Journal of Cerebral Blood Flow and Metabolism,
9, 2–19.
Evans, A.C., Beil, C., Marrett, S., Thompson, C.J. and Hakim, A. (1988) Anatomical-functional correlation
using an adjustable MRI-based region of interest atlas with positron emission tomography. Journal of
Cerebral Blood Flow and Metabolism, 8, 513–530.
Eysel, U.T. (1992) Cortical plasticity: remodelling receptive fields in sensory cortices. Current Biology, 2, 389–391.
Fetz, E.E. (1993) Cortical mechanisms controlling limb movement. Current Opinion in Neurobiology, 3, 932–939.
Fischl, B., Sereno, M.I., Tootell, R.B.H. and Dale, A.M. (1999) High-resolution intersubject averaging and
a coordinate system for the cortical surface. Human Brain Mapping, 8, 272–284.
Foerster, O. (1936) Motorische Felder und Bahnen. In: O. Bumke and O. Foerster (eds), Handbuch der Neuro-
logie, Vol. 6, Allgemeine Neurologie. Berlin: Julius Springer Verlag, pp. 1–357.
Fox, P.T. and Raichle, M.E. (1984) Stimulus rate dependence of regional cerebral blood flow in human striate
cortex, demonstrated by positron emission tomography. Journal of Neurophysiology, 51, 1109–1120.
Fox, P.T., Burton, H. and Raichle, M.E. (1987a) Mapping human somatosensory cortex with positron emission
tomography. Journal of Neurosurgery, 67, 34–43.
Fox, P.T., Miezin, F.M., Allman, J.M., van Essen, D.C. and Raichle, M.E. (1987b) Retinotopic organization of
human visual cortex mapped with positron emission tomography. Journal of Neuroscience, 7, 913–922.
Fox, P.T., Raichle, M.E., Mintun, M.A. and Dence, C. (1988) Nonoxidative glucose consumption during focal
physiologic neural activity. Science, (Washington), 241, 462–464.
Fox, P.T. (1997) The growth of human brain mapping. Human Brain Mapping, 5, 1–2.
Frackowiak, R.S.J. (1994) Functional mapping of verbal memory and language. Trends in Neuroscience, 17, 109–115.
Frahm, J., Bruhn, H., Merboldt, K.-D. and Hänicke, W. (1992) Dynamic MR imaging of human brain oxygena-
tion during rest and photic stimulation. Journal of Magnetic Resonance Imaging, 2, 501–505.
Friel, K.M. and Nudo, R.J. (1998) Recovery of motor function after focal cortical injury in primates: compensatory
movement patterns used during rehabilitative training. Somatosensory and Motor Research, 15, 173–189.
Frisk, V. and Milner, B. (1990) The role of the left hippocampal region in the acquisition and retention of story
content. Neuropsychologia, 28, 349–359.
Friston, K.J., Frith, C.D., Liddle, P.F. and Frackowiak, R.S.J. (1991) Investigating a network model of word
generation with positron emission tomography. Proceedings of the Royal Society of London, B, 244, 101–106.
Friston, K.J., Frith, C.D., Liddle, P.F. and Frackowiak, R.S.J. (1993) Functional connectivity: the principal-
component analysis of large (PET) data sets. Journal of Cerebral Blood Flow and Metabolism, 13, 5–14.
Friston, K.J. (1994) Functional and effective connectivity in neuroimaging: a synthesis. Human Brain Mapping,
2, 56–78.
Friston, K.J., Holmes, A.P., Worsley, K.J., Poline, J.-P., Frith, C.D. and Frackowiak, R.S.J. (1994) Statistical
parametric maps in functional imaging: a generalized linear approach. Human Brain Mapping, 2, 189–210.
Friston, K.J. (1997) Imaging cognitive anatomy. Trends in Cognitive Science, 1, 21–27.
Friston, K.J., Fletcher, P., Josephs, O., Holmes, A., Rugg, M.D. and Turner, R. (1998) Event-related fMRI:
Characterizing differential responses. Neuroimage, 7, 30–40.
Frostig, R.D., Lieke, E.E., Ts’o, D. and Grinvald, A. (1990) Cortical functional architecture and local coupling
between neuronal activity and the microcirculation revealed by in vivo high-resolution optical imaging of
intrinsic signals. Proceedings of the National Academy of Science, U.S.A., 87, 6082–6086.
Fukuyama, H., Ouchi, Y., Matsuzaki, S., Nagahama, Y., Yamauchi, H., Ogawa, M., Kimura, J. and Shibasaki, H.
(1997) Brain functional activity during gait in normal subjects: a SPECT study. Neuroscience Letters, 228,
183–186.
Gadian, D.G., Mishkin, M. and Vargha-Khadem, F. (1999) Early brain pathology and its relation to cognitive
impairment: the role of quantitative magnetic resonance techniques. In: P. Stefan, F. Andermann,
S.A. Shorvon and P. Chauvel (eds), Plasticity and epilepsy. (Advances in Neurology, Volume 81). New
York: Raven Press, pp. 307–315.
Gevins, A., Cutillo, B., DuRousseau, D., Le, J., Leong, H., Martin, N., Smith, M.E., Bressler, S., Bricket, P.,
McLaughlin, J., Barbero, N. and Laxer, K. (1994) Imaging the spatiotemporal dynamics of cognition with
high-resolution evoked potential methods. Human Brain Mapping, 1, 101–116.
George, J.S., Aine, C.J., Mosher, J.C., Schmidt, D.M., Ranken, D.M., Schlitt, H.A., Wood, C.C., Lewine, J.D.,
Sanders, J.A. and Belliveau, J.W. (1995) Mapping function in the human brain with magnetoencephalo-

graphy, anatomical magnetic resonance imaging, and functional magnetic resonance imaging. Journal of
Clinical Neurophysiology, 12, 406–431.
Gerloff, C., Grodd, W., Altenmüller, E., Kolb, R., Naegele, T., Klose, U., Voigt, K. and Dichgans, J. (1996)
Coregistration of EEG and fMRI in a simple motor task. Human Brain Mapping, 4, 199–209.
Geyer, S., Ledberg, A., Schleicher, A., Kinomura, S., Schormann, T., Bürgel, U., Klingberg, T., Larsson, J.,
Zilles, K. and Roland, P.E. (1996) Two different areas within the primary motor cortex of man. Nature
(London), 382, 805–807.
Gilman, S., Dauth, G.W., Frey, K.A. and Penney, J.B. (1987) Experimental hemiplegia in the monkey: basal
ganglia glucose activity during recovery. Annals of Neurology, 22, 370–376.
Grabowski, T.J., Frank, R.J., Brown, C.K., Damasio, H., Boles Ponto, L.L., Watkins, G.L. and Hichwa, R.D.
(1996) Reliability of PET activation across statistical methods, subject groups, and samples sizes. Human
Grafton, S.T., Mazziotta, J.C., Presty, S., Friston, K.J., Frackowiak, R.S.J. and Phelps, M.E. (1992) Functional
anatomy of human procedural learning determined with regional cerebral blood flow and PET. Journal of
Neuroscience, 12, 2542–2548.
Gray, C.M., Engel, A.K., König, P. and Singer, W. (1990) Stimulus-dependent neuronal oscillations in cat visual
cortex: receptive field properties and feature dependence. European Journal of Neuroscience, 2, 607–619.
Grinvald, A., Frostig, R.D., Siegel, R.M. and Bartfeld, E (1991) High-resolution optical imaging of functional
brain architecture in the awake monkey. Proceedings of the National Academy of Science, U.S.A., 88,
11559–11563.
Gulyas, B. (1997) Functional organization of human visual cortical areas. In: K.S. Rockland, J.H. Kaas and
A. Peters (eds), Cerebral Cortex, Vol 12, New York: Plenum Press, pp. 743–775.
Gur, R.C., Turetsky, B., Matsui, M., Yan, M, Bilker, W., Hughett, P. and Gur, R.E. (1999) Sex differences in
brain gray and white matter in healthy young adults: correlations with cognitive performance. Journal of
Haglund, M.M., Ojemann, G.A. and Hochman, D.W. (1992) Optical imaging of epileptiform and functional
activity in human cerebral cortex. Nature (London), 358, 668–671.
Hämäläinen, M., Hari, R., Ilmoniemi, R.J., Knuutila, J. and Lounasmaa, O.V. (1993) Magnetoencephalo-
graphy—theory, instrumentation, and applications to noninvasive studies of the working human brain.
Reviews of Modern Physics, 65, 413–497.
Hari, R. and Lounasmaa, O.V. (1989) Recording and interpretation of cerebral magnetic fields. Science,
(Washington), 244, 432–436.
Hasnain, M.K., Fox, P.T. and Woldorff, M.G. (1998) Intersubject variability of functional areas in the human
visual cortex. Human Brain Mapping, 6, 301–315.
Haxby, J.V., Horwitz, B., Ungerleider, L.G., Maisog, J.M., Pietrini, P. and Grady, C.L. (1994) The functional
organization of human extrastriate cortex: a PET-rCBF study of selective attention to faces and locations.
Heald, A., Bates, D., Cartlidge, N.E.F., French, J.M. and Miller, S. (1993) Longitudinal study of motor conduc-
tion time following stroke. 2. Central motor conduction measured within 72 h after stroke as a predictor of
functional outcome at 12 months. Brain, 116, 1371–1385.
Heinze, H.J., Mangun, G.R., Burchert, W., Hinrichs, H., Scholz, M., Münte, T.F., Gös, A., Scherg, M.,
Johannes, S., Hundeshagen, H., Gazzaniga, M.S. and Hillyard, S.A. (1994) Combined spatial and temporal
imaging of brain activity during visual selective attention in humans. Nature (London), 372, 543–546.
Heiss, W.D., Kessler, J., Thiel, A., Ghaemi, M. and Karbe, H. (1999) Differential capacity of left and right
hemispheric areas for compensation of poststroke aphasia. Annals of Neurology, 45, 430–438.
Herzog, H., Seitz, R.J., Tellmann, L. and Müller-Gärtner, H.-W. (1996) Quantification of regional cerebral blood
flow with 15O-butanol and positron emission tomography in humans. Journal of Cerebral Blood Flow and
Metabolism, 16, 645–649.
Hoge, R.D., Atkinson, J., Gill, B., Crelier, G.R., Marrett, S. and Pike, G.B. (1999) Linear coupling between
cerebral blood flow and oxygen consumption in activated human cortex. Proceedings of the National
Horwitz, B. (1994) Data analysis paradigms for metabolic-flow data: combining neural modeling and functional
neuroimaging. Human Brain Mapping, 2, 112–122.
Hubel, D.H. and Wiesel, T.N. (1963) Shape and arrangement of columns in cat’s striate cortex. Journal of
Physiology (London), 165, 559–568.
Hubel, D.H., Wiesel, T.N. and Stryker, M.P. (1978) Anatomical demonstration of orientation columns in
macaque monkey. Journal of Comparative Neurology, 177, 361–380.
Hunton, D.L., Miezin, F.M., Buckner, R.L., van Mier, H.I., Raichle, M.E. and Petersen, S.E. (1996) An assess-
ment of functional-anatomical variability in neuroimaging studies. Human Brain Mapping, 4, 122–139.
Ibánez, V., Deiber, M.P., Sadato, N., Toro, C., Grissom, J., Woods, R.P., Mazziotta, J.C., and Hallett, M. (1995)
Effects of stimulus rate of regional cerebral blood flow after median nerve stimulation. Brain, 118,
1339–1351.

Iadecola, C. (1993) Regulation of the cerebral microcirculation during neural activity: is nitric oxide the missing
link? Trends in Neuroscience, 16, 206–214.
Jenkins, W.M. and Merzenich, M.M. (1987) Reorganization of neocortical representations after brain injury:
a neurophysiological model of the bases of recovery from stroke. In: F.J. Seil, E. Herbert and B.M. Carlson
(eds), Neural regeneration. (Progress in Brain Research, Vol. 71). Amsterdam: Elsevier, pp. 249–266.
Jenkins, I.H., Brooks, D.J., Nixon, P.D., Frackowiak, R.S.J. and Passingham, R.E. (1994) Motor sequence
learning: a study with positron emission tomography. Journal of Neuroscience, 14, 3775–3790.
Jokeit, H., Seitz, R.J., Markowitsch, H.J., Neumann, N., Witte, O.W. and Ebner, A. (1997) Prefrontal asym-
metric interictal glucose hypometabolism and cognitive impairment in patients with temporal lobe epilepsy.
Brain, 120, 2283–2294.
Juliano, S.L., Hand, P.J. and Whitsel, B.L. (1981) Patterns of increased metabolic activity in somatosensory
cortex of monkeys macaca fascicularis, subjected to controlled cutaneous stimulation: a 2-deoxyglucose
study. Journal of Neurophysiology, 46, 1260–1284.
Juliano, S.L., Hand, P.J. and Whitsel, B.L. (1983) Patterns of metabolic activity in cytoarchitectural area SII and
surrounding cortical fields of the monkey. Journal of Neurophysiology, 50, 961–980.
Kaas, J.H. (1993) The functional organization of somatosensory cortex in primates. Anatomische Anziege, 175,
509–518.
Kadekaro, M., Crane, A.M. and Sokoloff, L. (1985) Differential effects of electrical stimulation of sciatic nerve
on metabolic activity in spinal cord and dorsal root ganglion in the rat. Proceedings of the National
Karni, A., Meyer, G., Jezzard, P., Adams, M.M., Turner, R. and Ungerleider, L.G. (1995) Functional MRI
evidence for adult motor cortex plasticity during motor skill learning. Nature (London), 377, 155–158.
Kastrup, A., Li, T.-Q., Glover, G.H., Krüger, G. and Moseley, M.E. (1999) Gender differences in cerebral blood
flow and oxygenation response during focal physiological neural activity. Journal of Cerebral Blood Flow
and Metabolism, 19, 1066–1071.
Kennedy, C., DesRosiers, M.H., Skurada, O., Shinohara, M., Reivich, M., Jehle, W. and Sokoloff, L. (1976)
Metabolic mapping of the primary visual system of the monkey by means of the autoradiographic
[14C]deoxyglucose technique. Proceedings of the National Academy of Science, U.S.A., 73, 4230–4234.
Kennedy, D.N., Lange, N., Makris, N., Bates, J., Meyer, J. and Caviness, Jr., V.S. (1998) Gyri of the human
neocortex: an MRI-based analysis of volume and variance. Cerebral Cortex, 8, 372–384.
Kertezs, A. and Ferro, J.M. (1984) Lesion size and location in ideomotor apraxia. Brain, 107, 921–933.
Kincses, W.E., Brain, C., Kaiser, S. and Elbert, T. (1999) Modeling extended sources of event-related potentials
using anatomical and physiological constraints. Human Brain Mapping, 8, 182–193.
Kislyakov, Y.Y. and Ivanov, K.P. (1986) O2 transport in cerebral microregions (mathematical simulation).
Journal of Biomechanical Enginineering, 108, 28–32.
Kleinschmidt, A., Merboldt, K.D., Hänicke, W., Steinmetz, H. and Frahm, J. (1994) Correlational imaging of
thalamocortical coupling in the primary visual pathway of the human brain. Journal of Cerebral Blood
Flow and Metabolism, 14, 952–957.
Kleinschmidt, A., Büchel, C., Zeki, S. and Frackowiak, R.S.J. (1998) Human brain activity during spontane-
ously reversing perception of ambiguous figures. Proceedings of the Royal Society London, B, 265,
2427–2433.
Knorr, U., Weder, B., Kleinschmidt, A., Wirrwar, A., Huang, Y., Herzog, H. and Seitz, R.J. (1993) Identification
of task-specific rCBF changes in individual subjects: validation and application for positron emission tomo-
graphy. Journal of Computer Assisted Tomography, 17, 517–528.
Korvenoja, A., Huttunen, J., Salli, E., Pohjonen, H., Martinkauppi, S., Palva, J.M., Lauronen, L., Virtanen, J.,
Ilmoniemi, R.J. and Aronen, H.J. (1999) Activation of multiple cortical areas in response to somatosensory
stimulation: combined magnetoencephalographic and functional magnetic resonance imaging. Human
Kosslyn, S.M., Alpert, N.M. and Thompson, W.L. (1995) Identifying objects at different levels of hierarchy:
a positron emission tomography study. Human Brain Mapping, 3, 107–132.
Kosslyn, S.M., Pascual-Leone, A., Felician, O., Camposano, S., Keenan, J.P., Thompson, W.L., Ganis, G.,
Sukel, K.E. and Alpert, N.M. (1999) The role of area 17 in visual imagery: convergent evidence from PET
and rTMS. Science (Washington), 284, 167–170.
Kossut, M., Hand, P.J., Greenberg, J. and Hand, C.L. (1988) Single vibrissal cortical column in SI cortex of rat
and its alterations in neonatal and adult vibrissa-deafferent animals: a quantitative 2DG study. Journal of
Neurophysiology, 60, 829–852.
Krakow, K., Woermann, F.G., Symms, M.R., Allen, P.J., Lemieux, L., Barker, G.J., Duncan, J.S. and Fish, D.R.
(1999) EEG-triggered functional MRI of interictal epileptiform activity in patients with partial seizures.
Brain, 122, 1679–1688.
Kristeva, R., Cheyne, D. and Deecke, L. (1991) Neuromagnetic fields accompanying unilateral and bilateral
voluntary movements: topography and analysis of cortical sources. Electroencephalography and Clinical

Kuschinsky, W., Suda, S. and Sokoloff, L. (1981) Local cerebral glucose utilization and blood flow during
metabolic acidosis. American Journal of Physiology, 241, H772–H777.
Lemon, R. (1988) The output map of the primate motor cortex. Trends in Neuroscience, 11, 501–506.
Logothetis, N.K., Pauls, J., Augath, M., Trinath, T. and Oeltermann, A. (2001) Neurophysiological investigation
of the basis of the fMRI signal. Nature, 412, 150–157.
Lueck, C.J., Zeki, S., Friston, K.J., Deiber, M.-P., Cope, P., Cunningham, V.J., Lammertsma, A.A., Kennard, C.
and Frackowiak, R.S.J. (1989) The colour centre in the cerebral cortex of man. Nature (London), 340, 386–389.
Madsen, P.L., Cruz, N.F., Sokoloff, L. and Dienel, G.A. (1999) Cerebral oxygen/glucose ratio is low during
sensory stimulation and rises above normal during recovery: excess glucose consumption during stimula-
tion is not accounted for by lactate efflux from or accumulation in brain tissue. Journal of Cerebral Blood
Magistretti, P.J. and Pellerin, L. (1999) Cellular mechanisms of brain energy metabolism and their relevance to
functional brain imaging. Philosophical Transactions of the Royal Society, London, series B, 354, 1155–1163.
Malonek, D. and Grinvald, A. (1996) Interactions between electrical activity and cortical microcirculation
revealed by imaging spectroscopy: implications for functional brain mapping. Science (Washington), 272,
551–554.
Marin, G., Guerin, C., Baillet, S., Garnero, L. and Meunier, G. (1998) Influence of skull anisotropy for the
forward and inverse problem in EEG: simulation studies using FEM on realistic head models. Human Brain
Mapping, 6, 250–269.
Mata, M., Fink, D.J., Gainer, H., Smith, C.B., Davidsen, L., Saviki, H., Schwartz, W.J. and Sokoloff, L. (1980)
Activity-dependent energy metabolism in rat posterior pituitary primarily reflects sodium pump activity.
Journal of Neurochemistry, 34, 213–215.
Maynard, E.M., Hatsopoulos, N.G., Ojakangas, C.L., Acuna, B.D., Sanes, J.N., Normann, R.A. and Donoghue,
J.P. (1999) Neuronal interactions improve cortical population coding of movement direction. Journal of
Mazziotta, J.C., Phelps, M.E., Plummer, D. and Kuhl, D.E. (1981) Quantitation in positron emission computed
tomography: 5. physical-anatomical effects. Journal of Computer Assisted Tomography, 5, 734–743.
McIntosh, A.R. and Gonzalez-Lima, F. (1994) Structural equation modeling and its application to network
analysis in functional brain imaging. Human Brain Mapping, 2, 23–44.
McIntosh, A.R., Grady, C.L., Ungerleider, L.G., Haxby, J.V., Rapoport, S.I. and Horwitz, B. (1994) Network
analysis of cortical visual pathways mapped with PET. Journal of Neuroscience, 14, 655–666.
Menon, R.S. and Kim, S.-G. (1999) Spatial and temporal limits in cognitive neuroimaging with fMRI. Trends in
Cognitive Science, 3, 207–216.
Mesulam, M.M. (1981) A cortical network for directed attention and unilateral neglect. Annals of Neurology, 10,
309–325.
Mesulam, M.M. (1990) Large-scale neurocognitive networks and distributed processing for attention, language,
and memory. Annals of Neurology, 28, 597–613.
Missimer, J., Knorr, U., Maguire, R.P., Herzog, H., Seitz, R.J., Tellman, L. and Leenders, K.L. (1999) On two
methods of statistical image analysis. Human Brain Mapping, 8, 245–258.
Mountcastle, V.B., Talbot, W.H., Sakata, H. and Hyvärinen, J. (1969) Cortical neuronal mechanisms in flutter-
vibration studied in unanaesthetized monkeys: Neural periodicity and frequency discrimination. Journal of
Müller, R.-A., Rothermal, R.D., Behen, M.E., Muzik, O., Mangner, Rj. and Chugani, H.T. (1998) Differential
patterns of language and motor reorganization following early left hemispheric lesion. Archives of Neuro-
logy, 55, 1113–1119.
Näätänen, R., Ilmoniemi, R.J. and Alho, K. (1994) Magnetoencephalography in studies of human cognitive brain
function. Trends in Neuroscience, 17, 389–395.
Neumann-Haefelin, T., Wittsack, H.-J., Wenserski, F., Siebler, M., Seitz, R.J., Mödder, U. and Freund, H.-J.
(1999) Diffusion- and perfusion-weighted MRI. The DWI/PWI mismatch region in acute stroke. Stroke, 30,
1591–1597.
Nudo, R.J., Wise, B.M., SiFurentes, F. and Milliken, G.W. (1996) Neural substrates for the effects of rehabilita-
tive training on motor recovery after ischemic infarct. Science (Washington), 272, 1791–1794.
Ojemann, G., Ojemann, J., Lettich, E. and Berger, M. (1989) Cortical language localization in left, dominant hemi-
sphere. An electrical stimulation mapping investigatin in 117 patients. Journal of Neurosurgery, 71, 316–326.
O’Sullivan, B.T., Roland, P.E. and Kawashima, R. (1994). A PET study of somatosensory discrimination in
man: Microgeometry versus macrogeometry. European Journal of Neuroscience, 6, 137–148.
Pascual-Leone, A., Grafman, J. and Hallett, M. (1994) Modulation of cortical motor output maps during devel-
opment of implicit and explicit knowledge. Science (Washington), 263, 1287–1289.
Pascual-Leone, A., Wassermann, E.M., Sadato, N. and Hallett, M. (1995) The role of reading activity on the
modulation of motor cortical outputs to the reading hand in Braille readers. Annals of Neurology, 38, 910–915.
Paus, T., Marrett, S., Worsley, K. and Evans, A. (1996) Imaging motor-to-sensory discharges in the human
brain: an experimental tool for the assessment of functional connectivity. Neuroimage, 4, 78–86.

Pendlebury, S.T., Blamire, A.M., Lee, M.A., Style, P. and Matthews, P.M. (1999) Axonal injury in the internal
capsule correlates with motor impairment after stroke. Stroke, 30, 956–962.
Penfield, W. and Boldrey, E. (1953) Somatic motor and sensory representation in the cerebral cortex of man as
studied by electrical stimulation. Brain, 60, 389–443.
Penfield, W. and Jasper, H. (1954) Epilepsy and the functional anatomy of the human brain. Little, Brown, Boston.
Peters, A. and Kara, D.A. (1987) The neuronal composition of area 17 of rat visual cortex. IV. The organization
of pyramidal cells. Journal of Comparative Neurology, 260, 573–590.
Petersen, S.E., Fox, P.T., Posner, M.I., Mintun, M. and Raichle, M.E. (1988) Positron emission tomographic
studies of the cortical anatomy of single-word processing. Nature (London), 331, 585–589.
Phillips, J.R., Johnson, K.O. and Hsiao, S.S. (1988) Spatial pattern representation and transformation in monkey
somatosensory cortex. Proceedings of the National Academy of Science, U.S.A., 85, 1317–1321.
Poeck, K., deBleser, R. and Graf v. Keyserlingk, D. (1984) Neurolinguistic status and localization of lesion in
aphasic patients with exclusively consonant-vowel recurring utterances. Brain, 107, 199–217.
Poline, J.B., Vandenberghe, R., Holmes, A.P., Friston, K.J. and Frackowiak, R.S.J. (1996) Reproducibility of
PET activation studies: lessons from a multi-center European experiment. EU concerted action on functional
imaging. Neuroimage, 4, 34–54.
Porro, C.A., Francescato, M.P., Cettolo, V., Diamond, M.E., Baraldi, P., Zuiani, C., Bazzocchi, M. and
di Prampero, P.E. (1996) Primary motor and sensory cortex activation during motor performance and motor
imagery: a functional magnetic resonance imaging study. Journal of Neuroscience, 16, 7688–7698.
Prichard, J.W. and Brass, L.M. (1992) New anatomical and functional imaging methods. Annals of Neurology,
32, 395–400.
Price, C.J., Moore, C.J. and Friston, K.J. (1997) Subtractions, conjunctions, and interactions in experimental
design of activation studies. Human Brain Mapping, 5, 264–272.
Rademacher, J., Caviness, V.S., Steinmetz, H. and Galaburda, A.M. (1993) Topographical variation of the
human primary cortices: implications for neuroimaging and brain mapping studies. Cerebral Cortex, 3,
313–329.
Raichle, M.E. (1987) Circulatory and metabolic correlates of brain function in normal humans. In: F. Plum (ed.),
Handbook of Physiology—Nervous System Volume V, Higher Functions of the brain. Bethesda, M.D.:
American Physiological Society, pp. 643–674.
Rao, S.M., Binder, J.R., Bandettini, P.A., Hammeke, T.A., Yetkin, F.Z., Jesmanowicz, A., Lisk, L.M.,
Morris, G.L., Mueller, W.M., Estkowski, L.D., Wong, E.C., Haughton, V.M. and Hyde, J.S. (1993)
Functional magnetic resonance imaging of complex human movements. Neurology, 43, 2311–2318.
Regard, M., Cook, N.D., Wieser, H.G. and Landis, T. (1994) The dynamics of cerebral dominance during uni-
lateral limbic seizures. Brain, 117, 91–104.
Ribary, U., Ioannides, A.A., Singh, K.D., Hasson, R., Bolton, J.P., Lado, F., Mogilner, A. and Llinas, R.R.
(1991) Magnetic field tomography of coherent thalamocortical 40-Hz oscillations in humans. Proceedings
of the National Academy of Science, U.S.A., 88, 11037–11041.
Rijntjes, M., Dettmers, C., Buchel, C., Kiebel, S., Frackowiak, R.S. and Weiller, C. (1999) A blueprint for move-
ment: functional and anatomical representations in the human motor system. Journal of Neuroscience, 19,
8043–8048.
Ringelstein, E.B., Binieck, R., Weiller, C., Ammeling, B., Nolte, P.N. and Thron, A. (1992) Type and extent of
hemispheric brain infarctions and clinical outcome in early and delayed middle cerebral artery recanaliza-
tion. Neurology, 42, 289–298.
Rizzolatti, G., Luppino, G. and Matelli, M. (1998) The organization of the cortical motor system: new concepts.
Electroencephalography and Clinical Neurophysiology, 106, 283–296.
Roland, P.E., Eriksson, L., Stone-Elander, S. and Widén, L. (1987) Does mental activity change the oxidative
metabolism of the brain? Journal of Neuroscience, 7, 2373–2389.
Roland, P.E. (1993) Brain activation. John Wiley, New York.
Roland, P.E. and Gulyas, B. (1994) Visual imagery and visual representation. Trends in Neuroscience, 17, 281–287.
Roland, P.E., Geyer, S., Amunts, K., Schormann, T., Schleicher, A., Malikovic, A. and Zilles, K. (1997) Cyto-
architectural maps of the human brain in standard anatomical space. Human Brain Mapping, 5, 222–227.
Romani, G.L. and Rossini, P. (1988) Neuromagnetic functional localization: principles, state of the art, and
perspectives. Brain Topography, 1, 5–21.
Rondot, P., DeRecondo, J. and Ribadeu Duma, J.L. (1977) Visuomotor ataxia. Brain, 100, 355–376.
Roth, M., Decety, J., Raybaudi, M., Massarelli, R., Delon-Martin, C., Segebarth, C., Morand, S., Gemignani, A.,
Décorps, M. and Jeannerod, M. (1996) Possible involvement of primary motor cortex in mentally simulated
movement: a functional magnetic resonance imaging study. NeuroReport, 7, 1280–1284.
Scherg, M. (1990) Fundamentals of dipole source potential analysis. In: F. Grandori, M. Hoke and G.L. Romani
(eds), Auditory evoked magnetic fields and electrical potentials. (Advances in Audiology, Vol. 6), Basel:
Karger, pp. 40–69.
Schieber, M.H. and Hibbard, L.S. (1993) How somatotopic is the motor cortex hand area? Science (Washington),
261, 489–492.

Schlaug, G., Knorr, U. and Seitz, R.J. (1994) Inter-subject variability of cerebral activations in acquiring a motor
skill. A study with positron emission tomography. Experimental Brain Research, 98, 523–534.
Schlaug, G., Jäncke, L., Huang, Y. and Steinmetz, H. (1995) In vivo evidence of structural brain asymmetry in
musicians. Science, Washington, 267, 699–701.
Schmahmann, J.D. (2000) Cerebellum and Brainstem. In: A.W. Toga and J.C. Mazziotta (eds), Brain Mapping:
the systems. San Diego: Academic Press, pp. 207–259.
Schneider, R. and Gautier, J.-C. (1994) Leg weakness due to stroke. Site of lesions, weakness patterns and
causes. Brain, 117, 347–354.
Schnitzler, A., Witte, O.W., Cheyne, D., Haid, G., Vrba, J. and Freund, H.J. (1995). Modulation of somato-
sensory evoked magnetic fields by sensory and motor interferences. NeuroReport, 6, 1653–1658.
Schnitzler, A., Volkmann, J., Enck, J., Frieling, T., Witte, O.W. and Freund, H.-J. (1999) Different cortical
organization of visceral and somatic sensation in humans. European Journal of Neuroscience, 11, 305–315.
Seitz, R.J., Bohm, C., Greitz, T., Roland, P.E., Eriksson, L., Blomkvist, G., Rosenkvist, G. and Nordell, B. (1990)
Accuracy and precision of the computerized brain atlas programme (CBA) for localization and quantifica-
tion in positron emission tomography. Journal of Cerebral Blood Flow and Metabolism, 10, 443–457.
Seitz, R.J., Roland, P.E., Bohm, C., Greitz, T. and Stone-Elander, S. (1991) Somatosensory discrimination of
shape: tactile exploration and cerebral localization. European Journal of Neuroscience, 3, 481–492.
Seitz, R.J. and Roland, P.E. (1992a) Vibratory stimulation increases and decreases the regional cerebral
blood flow and oxidative metabolism: a positron emission tomography (PET) study. Acta Neurologica
Scandinavica, 86, 60–67.
Seitz, R.J. and Roland, P.E. (1992b) Learning of finger movement sequences: a combined kinematic and
positron emission tomography study. European Journal of Neuroscience, 4, 154–165.
Seitz, R.J., Schlaug, G., Kleinschmidt, A., Knorr, U., Nebeling, B., Wirrwar, A., Steinmetz, H., Benecke, R. and
Freund, H.-J. (1994) Remote depressions of cerebral metabolism in hemiparetic stroke: topography and
relation to motor and somatosensory functions. Human Brain Mapping, 1, 81–100.
Seitz, R.J., Schlaug, G., Knorr, U., Steinmetz, H., Tellmann, L. and Herzog, H. (1996) Neurophysiology of the
human supplementary motor area: positron emission tomography. In: H.O. Lueders (ed.), Supplementary
sensorimotor area. (Advances in Neurology, Vol. 70). New York: Raven Press, pp. 167–175.
Seitz, R.J., Canavan, A.G.M., Yagüez, L., Herzog, H., Tellmann, L., Knorr, U., Huang, Y. and Hömberg, V.
(1997) Representations of graphomotor trajectories in the human parietal cortex: evidence for controlled
processing and automatic performance. European Journal of Neuroscience, 9, 378–389.
Seitz, R.J., Höflich, P., Binkofski, F., Tellmann, L., Herzog, H. and Freund, H.-J. (1998) Role of the premotor
cortex for recovery from middle cerebral artery infarction. Archives of Neurology, 55, 1081–1088.
Seitz, R.J. and Azari, N.P. (1999) Postlesional neuroplasticity in man. In: P. Stefan, F. Andermann, S.A. Shorvon
and P. Chauvel (eds), Plasticity and epilepsy. (Advances in Neurology, Volume 81). New York: Raven
Press, pp. 37–47.
Seitz, R.J., Azari, N.P., Knorr, U., Binkofski, F., Herzog, H, and Freund, H.-J. (1999) The role of diaschisis in
stroke recovery. Stroke, 30, 1844–1850.
Seitz, R.J., Volkmann, J. and Witte, O.W. (2000) Functional Mapping of the Human Brain. The Neuroscientist,
6, 75–76.
Seitz, R.J., Knorr, U., Azari, N.P. and Weder, B. (2001) Cerebral networks in sensorimotor disturbances. Brain
Research Bulletin, 54, 299–305.
Shulman, G.L., Fiez, J.A., Corbetta, M., Buckner, R.L., Miezin, F.M., Raichle, M.E. and Petersen, S.E. (1997)
Common blood flow changes across visual tasks. II. Decreases in cerebral cortex. Journal of Cognitive
Sitzer, M., Knorr, U. and Seitz, R.J. (1994) Cerebral hemodynamics during sensorimotor activation in humans.
Journal of Applied Physiology, 77, 2804–2811.
Sokoloff, L., Reivich, M., Kennedy, C., Des Rosiers, M.H., Patlak, C.S., Pettigrew, K.D., Sakurada, O. and
Shinohara, M. (1977) The [14C]deoxyglucose method for the measurement of local cerebral glucose
utilization: theory, procedure, and normal values in the conscious and anesthetized albino rat. Journal of
Neurochemistry, 28, 897–916.
Snyder, A.Z. (1991) Dipole source localization in the study of EP generators: a critique. Electroencephalography
and Clinical Neurophysiology, 80, 321–325.
Stehling, M.K., Turner, R. and Mansfield, P. (1991) Echo-planar imaging: magnetic resonance imaging in a
fraction of a second. Science, Washington, 254, 43–50.
Steinmetz, H., Fürst, G. and Freund, H.-J. (1989) Cerebral cortical localization: application and validation of the
proportional grid system in MR imaging. Journal of Computer Assisted Tomography, 13, 10–19.
Steinmetz, H., Huang, Y., Seitz, R.J., Knorr, U., Herzog, H., Hackländer, T., Kahn, T. and Freund, H.-J. (1992a)
Individual integration of positron emission tomography and high-resolution magnetic resonance imaging.
Journal of Cerebral Blood Flow and Metabolism, 12, 919–926.

Steinmetz, H., Jäncke, L., Kleinschmidt, A., Schlaug, G., Volkmann, J. and Huang, Y. (1992b) Sex but no hand
difference in the isthmus of the corpus callosum. Neurology, 42, 749–752.
Steinmetz, H., Herzog, A., Schlaug, G., Huang, Y. and Jäncke, L. (1995) Brain (a)symmetry in monogzygotic
twins. Cerebral Cortex, 5, 296–300.
Stephan, K.M., Fink, G.R., Passingham, R.E., Silberzweig, D., Ceballos-Baumann, A.O., Frith, C.D. and
Frackowiak, R.S.J. (1995a) Functional anatomy of the mental representation of upper extremity movements
in healthy subjects. Journal of Neurophysiology, 73, 373–386.
Stephan, K.M., Dabringhaus, A., Fink, G.R., Geyer, S., Passingham, R.E., Schormann, T., Knorr, U., Seitz, R.J.,
Roland, P.E., Frith, C.D., Frackowiak, R.S.J. and Zilles, K. (1995b) Motor imagery and motor performance:
PET activations compared to cytoarchitectonic data. Human Brain Mapping, Suppl 1, 301.
Stephan, K.M. and Frackowiak, R.S.J. (1996) Motor imagery—anatomical representation and electrophysiologi-
cal characteristics. Neurochemical Research, 21, 1105–1116
Stephan, K.M., Binkofski, F., Halsband, U., Dohle, C., Wunderlich, G., Schnitzler, A., Tass, P., Posse, S.,
Herzog, H., Sturm, V., Zilles, K., Seitz, R.J. and Zilles, K. (1999) The role of ventral medial wall motor
areas for bimanual coordination: a combined lesion and activation study. Brain, 122, 351–368.
Stepniewska, I., Preuss, T.M. and Kaas, J. (1993) Architectonics, somatotopic organization, and ipsilateral
cortical connections of the primary motor area (M1) of Owl monkey. Journal of Comparative Neurology,
330, 238–271.
Talairach, J., Szikla, G., Tournoux, P., Prossalentis, A., Bordas-Ferrer, M., Covello, L., Joacob, M., Mempel, A.,
Buser, P. and Bancaud, J. (1967) Atlas d’anatomie stéréotaxique du télencéphale. Paris: Masson.
Talairach, J. and Tournoux, P. (1988) Co-planar stereotaxic atlas of the human brain. Thieme, Stuttgart,
New York.
Thurfjell, L., Bohm, C., Bengtsson, E. (1995) CBA—an atlas based software tool used to facilitate the interpreta-
tion of neuroimaging data. Computer Methods and Programs in Biomedicine, 47, 51–71.
Toma, K., Honda, M., Hanakawa, T., Okada, T., Fukuyama, H., Ikeda, A., Nishizawa, S., Konishi, J. and
Shibasaki, H. (1999) Activities of the primary and supplementary motor areas increase in preparation and exe-
cution of voluntary muscle relaxation: an event-related fMRI study. Journal of Neuroscience, 19, 3527–3534.
Tong, D.C., Yenari, M.A., Albers, G.W., O’Brien, M., Marks, M.P. and Moseley, M.E. (1998) Correlation of
perfusion- and diffusion-weighted MRI with NIHSS score in acute (<6.5 hour) ischemic stroke. Neurology,
50, 864–870.
Tootell, R.B.H., Reppas, J.B., Kwong, K.K., Malach, R., Born, R.T., Brady, T.J., Rosen, B.R. and Belliveau, J.W.
(1995) Functional analysis of human MT and related visual cortical areas using magnetic resonance
imaging. Journal of Neuroscience, 15, 3215–323.
Tranel, D. (1991) Dissociated verbal and nonverbal retrieval and learning following left anterior temporal dam-
age. Brain and Cognition, 15, 187–200.
Ueki, M., Linn, F. and Hossmann, K.-A. (1988) Functional activation of cerebral blood flow and metabolism
before and after global ischemia of rat brain. Journal of Cerebral Blood Flow and Metabolism, 8, 486–494.
Ungerleider, L.G. and Haxby, J.V. (1994) “What” and “where” in the human brain. Current Opinion in Neuro-
biology, 4, 157–165.
Van Essen, D.C. (1985) Functional organization of primate visual cortex. In: A. Peters and E.G. Jones (eds),
Cerebral Cortex, Vol 3, Visual cortex. New York: Plenum Press, pp. 259–329.
Votaw, J.R., Henry, T.R., Shoup, T.M., Hoffman, J.M., Woodard, J.L. and Goodman, M.M. (1999) Butanol is
superior to water for performing positron emission tomography activation studies. Journal of Cerebral
Blood Flow and Metabolism, 19, 982–989.
Volkmann, J., Witte, O.W., Dammers, J., Arnold, S., Seitz, R.J., Müller-Gärtner, H.W. and Freund, H.-J. (1998)
Relationship between epileptogenic and symptomatogenic zone in epilepsia partialis continua: A MEG and
PET study. Journal of Neuroimaging, 8, 103–106.
von Giesen, H.J., Schlaug, G., Steinmetz, H., Benecke, R., Freund, H.-J. and Seitz, R.J. (1994) Cerebral network
underlying unilateral motor neglect: evidence from positron emission tomography. Journal of Neurological
Science, 125, 29–38.
Wassermann, E.M., Wang, B., Zeffiro, T.A., Sadato, N., Pascual-Leone, A., Toro, C. and Hallett, M. (1996)
Locating the motor cortex on the MRI with transcranial magnetic stimulation and PET. Neuroimage, 3, 1–9.
Watson, J.D.G., Myers, R., Frackowiak, R.S.J., Hajnal, J.V., Woods, R.P., Mazziotta, J.C., Shipp, S. and Zeki, S.
(1993) Area V5 of the human brain: evidence from a combined study using positron emission tomography
and magnetic resonance imaging. Cerebral Cortex, 3, 79–94.
Weiss, D.S. and Keller, A. (1994) Specific patterns of intrinsic connections between representations zones in the
rat motor cortex. Cerebral Cortex, 4, 205–214.
Welker, E., Rao, S.B., Dörfl, J., Melzer, P. and van der Loos, H. (1992) Plasticity in the barrel cortex of the adult
mouse: effects of chronic stimulation upon deoxyglucose uptake in the behaving animal. Journal of Neuro-
science, 12, 153–170.
Winstein, C.J. and Pohl, P.S. (1995) Effects of unilateral brain damage on the control of goal-directed hand
movements. Experimental Brain Research, 105, 163–174.

Wood, C.C., Cohen, D., Cuffin, B.N., Yarita, M. and Allison, T. (1985) Electrical sources in human somato-
sensory cortex: identification by combined magnetic and potential recordings. Science (Washington), 227,
1051–1053.
Worsley, K.J., Evans, A.C., Marrett, S. and Neelin, P. (1992) A three-dimensional statistical analysis for CBF
activation studies in human brain. Journal of Cerebral Blood Flow and Metabolism, 12, 900–918.
Wunderlich, G., Knorr, U., Kiwit, J.C.W., Herzog, H., Freund, H.-J. and Seitz, R.J. (1998) Precentral glioma
location determines the displacement of cortical hand representation. Neurosurgery, 42, 18–27.
Xerri, C., Merzenich, M.M., Peterson, B.E. and Jenkins, W. (1998) Plasticity of primary somatosensory cortex
paralleling sensorimotor skill recovery from stroke in adult monkeys. Journal of Neurophysiology, 79,
2119–2148,
Yarowsky, P., Kadekaro, M. and Sokoloff, L. (1983) Frequency-dependent activation of glucose utilization in
the superior cervical ganglion by electrical stimulation of cervical sympathic trunk. Proceedings of the
National Academy of Science, U.S.A., 80, 4179–4183.
Young, I.R., Bailes, D.R., Bult, M., Collins, A.G., Smith, D.T., McDonnell, M.J., Orr, J.S., Banks, L.M.,
Bydder, G.M., Greenspan, R.H. and Steiner, R.E. (1982) Initial clinical evaluation of a whole body nuclear
magnetic resonance (NMR) tomograph. Journal of Computer Assisted Tomography, 6, 1–18.
Yousry, T.A., Schmid, U.D., Jassoy, A.G., Schmidt, D., Eisner, W.E., Reulen, H.J., Reiser, M.F. and Lissner, J.
(1995) Topography of the cortical motor hand area: prospective study with functional MR imaging and
direct motor mapping at surgery. Radiology, 195, 23–29.
Zihl, J. and von Cramon, D. (1985) Visual field recovery from scotoma in patients with postgeniculate damage.
A review of 55 cases. Brain, 108, 335–365.
Zihl, J., von Cramon, D., Mai, N. and Schmid, C. (1991) Disturbance of movement vision after bilateral posterior
brain damage. Brain, 114, 2235–2252.

Part II
CORTICAL AREAS: CORRELATION

WITH CONNECTIVITY

6 Regional Dendritic Variation in Primate Cortical
Pyramidal Cells
Bob Jacobs1 and Arnold B. Scheibel2
1
Laboratory of Quantitative Neuromorphology, Department of Psychology,
The Colorado College, 14 E. Cache La Poudre, Colorado Springs, CO 80903
2
Department of Neurobiology, Brain Research Institute, University of California,
Los Angeles, CA 90024-1769
Correspondence to: Bob Jacobs, Ph.D. Laboratory of Quantitative Neuromorphology,
Department of Psychology, The Colorado College, 14 East Cache La Poudre, Colorado Springs,
CO 80903, USA
Tel: (719) 389-6594; FAX: (719) 389-6284; e-mail: bjacobs@coloradocollege.edu
This chapter reviews quantitative neuromorphological investigations of primate neocortex. In par-

ticular, we explore regional variation in the basal dendritic and spine systems of pyramidal neurones.
This synthesis indicates a relatively consistent, stepwise increase in dendritic extent and spine number
in a caudal-rostral direction. Cortical regions involved in the early stages (e.g. primary sensory areas)
of processing generally exhibit less complex dendritic/spine systems than those regions involved in
the latter stages of information processing (e.g. prefrontal cortex). This dendritic progression appears
to reflect significant differences in the nature of cortical processing, with spine-dense neurones at
hierarchically higher association levels integrating a broader range of synaptic input than those at
lower cortical levels. In concluding the chapter, we consider the characteristics of the receptive
dendritic membrane of individual neuronal elements (e.g. voltage-gated channels, input resistance,
voltage attenuation) and how such factors may relate to cortical computation.
KEYWORDS: cerebral cortex, dendrite, quantitative, regional variation, spine
1. INTRODUCTION
This chapter considers the degree to which neurones in different areas of the primate
cerebral cortex vary with regard to the extent and complexity of their dendritic ensembles.
We focus here on pyramidal neurones because they are the principal component of
neocortical circuitry. Moreover, we limit our discussion to the basal dendrite ensembles of
pyramidal cells, and thus follow the precedent set in most quantitative studies of cortical
neurones (Schlaug et al., 1993). Since the horizontal components of the pyramidal cell
ensemble, and notably the basal dendrites, provide the main receptive surface for axons of
intracortical origin (Globus and Scheibel, 1967a,b,c), there is a robust anatomical basis for
this choice. In addition, the enormous morphological variation of the stellate and non-
pyramidal cell contingents has always made these fascinating elements more problematic
for quantitative evaluation (see Prinz et al., 1997; Seldon, 1982).
In the last century, the classic qualitative studies of Golgi (1886), Cajal (1909), and
Lorente de Nó (1922), among others, provided the structural framework for a brilliant
111
112 Bob Jacobs and Arnold B. Scheibel
period of descriptive structuro-functional studies of the cerebral hemispheres. But as we

leave the decade of the brain and enter the 21st century, a more discriminating and compu-
tationally knowledgeable group of investigators poses questions that demand information
with higher levels of resolution. The intimate organization of dendrite membrane patches,
the structure, placement, and dimensions of individual dendrite spines, the length and
orientation of dendrite tips, and so on, become data necessary for a more profound under-
standing of the structuro-functional basis of neural computation.
In what follows, we attempt an overview of several quantitative approaches to analysis
of cortical dendrites as seen from an end-of-the-century point of view. We lean heavily
upon our own Golgi-based studies of human cortices with the following justifications:
(1) We are probably more aware of the shortcomings of our own work than of others’; and
(2) the field of quantitative human neural morphology is still very young with relatively
few published investigations. The major thrust of this overview is to provide evidence
relating dendritic length and spine number to cortical area; in general, the more complex
the computational functions of the area, the longer the dendrites and the more numerous
the spines. This relationship may be meaningful in terms of what is known about the
underlying physiology of dendrites and their spines, and may reflect significant regional
differences in the nature of cortical processing.
2. QUANTITATIVE TECHNIQUES AND METHODOLOGICAL

CONSIDERATIONS
Comparisons across neuromorphological studies are somewhat problematic because each

quantitative technique exhibits its own strengths and weaknesses (for review, see Uylings
et al., 1975, 1986). Moreover, all quantitative neuromorphological investigations face
formidable methodological constraints.
2.1. Quantitative Techniques

With considerable histological and technological advances over the last few decades, several
quantitative methodologies have emerged. At present, three morphological techniques are
commonly employed for dendritic quantification. The oldest and most widely used is the
semi-quantitative analysis of Sholl (1953, 1956), whereby a series of equidistant, concentric
rings are superimposed over a (typically traced) neurone, allowing the number of dendritic
intersections per ring to be counted (see Figure 6.1A, 6.1B). In the Eayrs’ (1955) concentric
circle variation of this technique, the researcher also quantifies the number of bifurcations and
terminal endings within each ring. As such, these techniques do not quantify dendritic seg-
ments in great detail, but do provide a first order approximation of the overall dendritic profile.
The second technique, a metric reconstruction, involves tracing the dendritic tree,
branch by branch, and measuring the length (and possibly diameter) of each segment
(see Figure 6.1C, 6.1D). This can be accomplished either indirectly by tracing cells with
a camera lucida and entering the coordinates on a digitizing tablet, or directly by quantify-
ing neurones through a microscope interfaced with a computer system (e.g. the Neuro-
lucida system, MicroBrightfield, Inc.). Although older versions of this technique were
somewhat limited, most recent versions can reconstruct accurately the entire dendritic
ensemble in 3-dimensional space, and can also provide estimates of spine number/density.

Regional Dendritic Variation
Figure 6.1. Illustrative depictions of common neuromorphological quantitative techniques. (A) Sample supragranular pyramidal cell (scale bar = 100 µm) with Sholl concentric
rings superimposed. (B) Results of a Sholl analysis within one individual comparing the basal dendritic systems in the trunk and hand-finger region of the somatosensory cortex.
Note the relatively more complex dendrites in the hand-finger region compared with the trunk region, as determined by the higher number of concentric ring crossings (adapted
from Scheibel et al., 1990). (C) A metric reconstruction of individual basal dendritic segments, which are traced here in a somatofugal pattern. (D) Results of such a metric
reconstruction illustrating a greater total dendritic length in prefrontal (BA10) compared with occipital (BA18) pyramidal neurones (adapted from Jacobs et al., 1997).
(E) Sample basal dendritic system (tangential view from the pial surface) with a polygon (hull) around dendritic tips (scale bar = 100 µm). This analysis provides a rough estim-
ate of dendritic field size. (F) A modified Sholl analysis is performed by superimposing concentric rings with 30° polar angle intervals, resulting in a polar plot (G) of dendritic tree
intersections as a function of direction from the soma. This polar plot provides a measure of dendritic tree orientation. In order to quantify the degree of dendritic bias, the
113
researchers calculate the sum of the angles (q) subtended between a circle (radius equal to the half-maximal value of the polar plot) and the polar plot. Cells with tightly
clustered dendritic trees have low q values; those with no bias have q value near 360° (E, F and G: adapted from Elston et al., 1998b).

Such analyses permit a fine-level, topographical depiction (e.g. length, number, and vol-
ume) of individual dendritic segments, and allow these segments to be integrated somato-
fugally (Van der Loos, 1959) within the 3-dimensional geometry of the dendritic array.
The third technique examines the dendritic field of the neurone (see Figure 6.1E), an
idea first investigated in a quantitative manner by Colonnier (1964). To examine the over-
all area of the dendritic field, researchers draw a polygon (hull) that joins the distal tips of
the outermost dendrites, and calculate the enclosed area. This technique provides a rela-
tively simple estimate of dendritic spread, but does not address overall complexity of
branching. To estimate such complexity, along with the clustering and orientation of the
dendritic tree, researchers can employ a modified Sholl analysis, which involves (1) com-
paring the overall number of intersections per concentric ring, and (2) examining the
distribution of intersections as a function of the radial position of the dendrite relative to
the soma, and depicting these in polar plots (see Figure 6.1F, 6.1G). This has been a par-
ticularly useful technique for investigating the geometry of the domains of dendritic trees
in different cortical regions.
2.2. Methodological Considerations

Extensive quantitative neuromorphological research has been conducted on non-human
organisms (e.g. Valverde, 1976; Juraska, 1982; Murphy and Magness, 1984; Braitenberg
and Schüz, 1991; Bannister and Larkman, 1995; Ishizuka et al., 1995), with appropriately
cautious heterospecific comparisons providing valuable insights into the characteristics
of cortical neuropil. Unfortunately, there have only been a handful of such studies in
non-human primates, and even fewer in humans. Moreover, when analyzing data from
human subjects, several special methodological issues constrain potential generaliza-
tions (Scheibel, 1988; Flood, 1993; Jacobs et al., 1993b, 1997).
Human subjects. Practical considerations in human research are numerous, including
the restrictions of retrospective analyses and the problems of relatively small sample sizes.
These are significant obstacles because correlative research on human dendritic systems
typically requires the broadest possible sociocultural, vocational, and avocational histories
on subjects (Scheibel et al., 1990), and because large sample sizes are required to over-
come the tremendous interindividual variation that characterizes human tissue (Ojemann
and Whitaker, 1978; Ojemann et al., 1989; Stensaas et al., 1974; Whitaker and Selnes,
1976). Moreover, it is impossible to determine in post-mortem tissue whether topograph-
ically identical cortical areas in different individuals share the same function. Fortunately,
for the purpose of regional comparisons, each subject can serve as his/her own control,
insofar as all areas of the cortex have been exposed to the same historical variables.
Histology. Even more limiting in human research are the autolytic consequences of
post-mortem delays (Williams et al., 1978; de Ruiter, 1983) and the restricted number of
histological techniques that can be employed on immersion-fixed tissue. By far the most
common stain for human brain tissue is the Golgi silver impregnation technique (see Figure
6.2), the relative merits of which are well documented (Scheibel and Scheibel, 1978;
Braak and Braak, 1985). There are, however, innumerable variations of Golgi impregna-
tions, each with its own idiosyncratic characteristics (Buell, 1982; Meller and Dennis,
1990), which further complicate cross-study comparisons. Recently, modern intracellular
injection techniques have focused attention on the limitations of such silver impregna-
tion methods. Although fluorescent dyes (e.g. Lucifer Yellow) have proven effective in

Regional Dendritic Variation 115
Figure 6.2. Photomicrographs of human supragranular pyramidal cells (A and B) and associated basal den-
dritic spines (C and D) stained with a modified rapid Golgi technique. Several individuals and Brodmann’s areas
(BA) are represented: (A) M32 (=32 year-old male), angular gyrus (BA39); (B) M23, prefrontal pole (BA10);
(C) F15 (=15 year-old female), somatosensory cortex (BA3-1-2); and (D) M14, inferior prefrontal pole (BA11).
For A and B, scale bars = 50 µm; for C and D, scale bars = 10 µm.
neuromorphological reconstructions in non-human animals (Trommald et al., 1995; Elston

et al., 1996), they remain somewhat problematic in human autopsy tissue (Ohm and Diekmann,
1994; Belichenko and Dahlström, 1995) due to factors such as dye leakage. Therefore,
despite the development of new histological techniques, the Golgi methods remain the
stain of choice in extensive quantitative studies of immersion fixed human tissue.
Quantification. Several factors determine what a particular quantitative and histological
technique will reveal about cortical neuropil. Here we outline four of these factors.
(1) Findings will be substantially affected if separate subpopulations of cells are quanti-
fied (e.g. layer II vs layer III pyramidal cells), even within the same region (Larkman,
1991; Matsubara et al., 1996). (2) Section thickness, which typically ranges from 50 µm to
200 µm in human Golgi preparations (e.g. Takashima et al., 1981; Koenderink et al.,
1994), can also substantially affect quantitative measures by producing varying degrees of
cut dendritic segments (Jacobs et al., 1997). (3) Quantitative studies conducted under dry
(as opposed to oil-immersion) objectives will result in a diminution of observed length
measurements (Uylings et al., 1986). (4) Spine counts, which underestimate actual num-
bers if they quantify only visible spines, will vary considerably with distance from the
soma (Globus and Scheibel, 1967a; Marin-Padilla, 1967), and between the basal and

apical dendrites of pyramidal neurones (Uemura, 1980). Given these constraints, among
others, it should be clear that quantitative measurements typically represent relative rather
than absolute values, and provide but a small window into the overall arrangement and
complexity of cortical neuropil.
3. INDIRECT INDICATORS OF REGIONAL VARIATION

IN CORTICAL NEUROPIL
The first quantitative study of cortical dendritic structure appears to have been Bok’s
(1936) examination of the relationship between the nuclear volume of the cell and the
number of dendritic branches. In terms of regional cortical variation, the first quantitative
documentation appears to be Sholl’s (1953) pioneering exploration of the dendritic
branching pattern in cat visual and motor cortices. In addition to finding a positive rela-
tionship between overall dendritic length and segment number, Sholl observed a greater
number of branches in the visual area compared with the motor area for both pyramidal
and stellate cells. Since Sholl’s initial observations, however, most quantitative neuro-
morphological investigations have not directly addressed regional dendritic variation.
Instead, they have focused primarily on one cortical area at a time, while exploring factors
such as hemispheric differences, cortical development, and aging. Limited inferences
related to cortical variation are nevertheless possible when findings from other types of
research (e.g. neuroimaging) are integrated.
3.1. Hemispheric Differences

The most general indicator of regional variation involves comparison of homologous
areas in the two cerebral hemispheres. Several investigations of this nature have been
conducted in humans, particularly on cortical areas involved in language. In an extensive
study of the human auditory cortex, Seldon (1981a,b, 1982) found that pyramidal cell
basal dendrites exhibited a larger tangential extent in the left hemisphere (LH) than in the
right hemisphere (RH). It was postulated that this dendritic advantage was related to
greater capacity for differential phonemic responses in the computationally specialized
LH. Similarly, slightly more complex basal dendritic systems have been documented in
classical Wernicke’s area over the RH homologue (Jacobs et al., 1993b). Findings in
classical Broca’s area have been mixed. Scheibel et al., (1985) found that higher order
dendritic segments tended to be more complex in the LH over the RH, presumably
because of the complexity of LH speech processing (cf. Jacobs et al., 1993a). In a very
limited study, however, Hayes and Lewis (1996) failed to observe interhemispheric differ-
ences of dendrites in magnopyramidal neurones in Broca’s area, but did suggest that LH
cells may be specialized to receive a more restricted complement of afferents. Given that
these quantitative morphological observations are consistent with documented interhemi-
spheric functional differences (Walker, 1980; Seldon, 1985; Zatorre et al., 1992), one can
suppose with some confidence that quantitative intrahemispheric differences should also
obtain across various cortical areas. It should be emphasized, however, that these dendritic
systems are extremely plastic. As such, each quantitative study provides but a synchronic
“snapshot” of the cortical neuropil, a snapshot that can be diachronically enriched by
observations of developmental and aging.

3.2. Development
Dramatic histological changes characterize pre- and post-natal cortical development,
including transient overproduction of neuropil, and selective elimination of excessive
connectivity (Marin-Padilla, 1970; Rakic et al., 1986; Mrzljak et al., 1990). Several quanti-
tative dendritic investigations have elucidated the timeline in primates for some of these
changes in individual cortical areas (Schadé and Van Groenigen, 1961; Schulz et al.,
1992; Koenderink et al., 1994; Koenderink and Uylings, 1995) but have not directly sug-
gested regional variability. In their cross-sectional, developmental human study, however,
Simonds and Scheibel (1989) inferred a gradual transition in dendritic primacy (i.e. relative
complexity of dendritic trees) from the RH to the LH, and from the orofacial motor cortex
to the motor speech region, intimating the adult pattern they had previously observed
(Scheibel et al., 1985). More recently, the Jacobs laboratory (unpublished data) has
compared basal dendritic systems across multiple cortical regions (Brodmann’s area, BA
3-1-2, BA4, BA18, and BA10) in human infants and adults. In infants, those regions that
mature earliest (BA3-1-2 and BA4) tend to exhibit more complex dendritic trees than
those that mature later (BA18, and especially BA10), a finding consistent with the fact
that primary cortical areas are initially more active metabolically than association regions
(Chugani et al., 1987). In the adult, the regional dendritic pattern tends to be reversed for
these four regions; that is, BA10 in the adult surpasses all other regions in dendritic
complexity (see below).
These morphological findings, coupled with measures of quantitative synaptogenesis
(Huttenlocher and Dabholkar, 1997) and with metabolic indicators (Chugani et al., 1987;
Jacobs et al., 1995), indicate a heterochronous path for cortical development. This path is
particularly clear in humans, where synaptic density peaks at approximately 3 months
postnatally in auditory cortex, at 8–12 months in striate cortex, and after 15 months in
frontal cortex (Huttenlocher et al., 1982; Huttenlocher and de Courten, 1987; Huttenlocher
and Dabholkar, 1997). Coincident with this synaptic proliferation, local cerebral metabolic
rates begin to increase between 1–2 years of age (Chugani et al., 1987). Dendritic growth
continues for several more years, even after neuronal and synaptic density decline (Conel,
1939–67; Schadé and Van Groenigen, 1961). With some regional variation, greatest
dendritic length in humans is probably reached between 8–10 years of age (Becker et al.,
1984; Semenova et al., 1989; Mrzljak et al., 1990; Jacobs and Scheibel, 1993). This dra-
matic growth in dendritic arborization coincides with a concomitant increase in the brain’s
metabolic demands (Mata et al., 1980; Nudo and Masterton, 1986). Thus, the metabolic
plateau in humans (at about 50% above adult levels) is achieved between 4–9 years of age
(Chugani et al., 1987). In terms of regional dendritic variability, the importance of this
heterochronous cortical development lies in the finding that some cortical areas in the
resting adult brain (e.g. prefrontal cortex) tend to exhibit higher rates of metabolism than
other cortical areas (Roland, 1984), in turn suggesting enhanced synaptic activity and
greater dendritic complexity in those more active regions.
3.3. Aging
Quantitative dendritic investigations have contributed greatly to our understanding of the
aging process. In one of the first (cross-sectional) studies to examine individual segment
length by digitizing camera lucida tracings, Cupp and Uemura (1980) suggested that

terminal basal segment growth may continue with age in some neurones in the frontal
cortex of rhesus monkeys, although some neurones may exhibit a distoproximal type of
degeneration. Neurones in different layers may be particularly susceptible to the aging
process. Nakamura et al. (1985), for example, noted that layer V basal dendrites appeared
to be more affected by the aging process than those in layer III. Using a Sholl analysis in
the parahippocampal gyrus, Buell and Coleman (1981) postulated that the normal aging
cortex appears to contain both regressing and proliferating (pyramidal) dendritic systems,
suggesting continued plasticity in the adult human brain. This age-related cortical plasti-
city appears wide-spread, insofar as the possibility of these two co-existing populations of
neurones has been reconfirmed in subsequent studies on granule cells of the dentate gyrus
(Flood et al., 1985) and supragranular pyramidal neurones in Wernicke’s area (Jacobs and
Scheibel, 1993).
In one of the first studies to examine multiple regions of the human cerebral cortex
(specifically, BA4, BA6, BA39, and BA10), Schierhorn (1981) documented age-related
decreases in spine density, which appeared to be consistent for all four areas, although
specific regional comparisons were not made. The only study to date that has specifically
compared age-related changes in dendritic/spine systems across multiple cortical regions
is Jacobs et al. (1997), which explored supragranular pyramidal neurones in secondary
visual (BA18) and prefrontal (BA10) regions in 26 human brains ranging from 14 to 106
years of age. Dendritic measures, particularly for spine systems, decreased substantially
from the youngest individuals to approximately 40 years of age, after which the measures
remained relatively stable. These losses appeared to be somewhat greater in BA10 than in
BA18, which is consistent with other research indicating that certain regions (e.g. frontal
lobes, selected association regions) may be particularly susceptible to aging (Kuhl et al.,
1982; Terry et al., 1987; Raz et al., 1993, 1997; Sullivan et al., 1995). More extensive
quantitative neuromorphological investigations are still required to evaluate age-related
regional loss in cortical neuropil (for review, see Coleman and Flood, 1987).
4. DIRECT INVESTIGATIONS OF REGIONAL VARIATION

IN CORTICAL NEUROPIL: NON-HUMAN PRIMATES
In recent years, an extensive series of studies by Elston and Rosa on monkey visual path-
ways has directly addressed the question of regional dendritic variation. Using Lucifer
Yellow injection techniques, a modified Sholl analysis, and dendritic field measures, these
investigations have examined basal dendritic complexity along the hierarchically arranged
occipitotemporal and occipitoparietal visual pathways, and in the frontal eye fields (see
Figure 6.3 for a summary).
4.1. The Occipitotemporal Visual Pathway

In one of their earliest comparative studies, Elston et al. (1996) examined four areas of the
adult Marmoset monkey brain (Callithrix jacchus): first (V1) and second (V2) visual
areas, the dorsolateral part (DL) and the fundus of the superior temporal (FST) region.
They found larger dendritic fields in the pyramidal cells of extrastriate regions, which are
involved in shape/color processing (DL) and motion/spatial analyses (FST; Boussaoud et al.,
1990), than in primary visual cortical areas (V1, V2) involved in the early stages of visual

Figure 6.3. Relative position of cortical tissue samples from the monkey, and representative tracings of supra-
granular pyramidal neurones (synthesized from Elston and Rosa, 1997, 1998a,b). Basal systems are drawn in
a plane tangential to the cortical layers and represent neurones in the 60th complexity percentile (scale
bar = 100 µm). These areas are arranged in a relative hierarchy, reflecting a general progression in dendritic
complexity along the occipitotemporal and occipitoparietal visual pathways, and in the frontal eye fields: pri-
mary visual cortex (V1), secondary visual cortex (V2), the fourth visual region (V4), a temporal lobe subdivision
(TEO), the middle temporal region (MT), the lateral intraparietal area (LIPv), the anterior bank of the superior
temporal sulcus in the parietal lobe (7a), and the frontal eye fields (FEF).
processing. This caudal-rostral progression in field size suggests a more extensive input
sampling by dendritic systems at higher levels, and corresponds with demonstrated size
increases in intrinsic axonal clusters (Yoshioka et al., 1992; Amir et al., 1993). This is
expected insofar as the area of axonal patches appears positively correlated with the basal
dendritic field of superficial pyramidal cells along this caudal-rostral visual gradient
(Lund et al., 1993).
In a similar, but more extensive investigation of adult Macaca fascicularis, Elston and
Rosa (1998b) explored the dendritic and spine characteristics of pyramidal neurones in
four visual regions: V1, V2, V4, and the occipitotemporal transitional zone (TEO). Several
morphological differences were observed both within and across cortical regions. Within
V1, cells located in cytochrome oxidase-rich blobs had more extensive dendritic fields
than those in interblob regions. Dendritic trees which were morphologically-oriented

(i.e. dendrites clustered in two diametrically opposing directions) and directionally-biased

(i.e. dendrites clustered in a particular direction) were more common in V1 and V2 than in
more rostral visual areas (V4 and TEO), where non-biased cells predominated. Finally,
not only did more rostral, higher visual areas (V4 and TEO) exhibit significantly larger
dendritic fields than primary regions (V1 and V2), but the number of spines along the basal
dendritic array roughly doubled at each successive stage in the pathway. These findings
suggest a stepwise progression in dendritic complexity, with the more rostrally located,
spine-dense neurones integrating a wider range of (non-visual) modulatory input than
the more caudally located, sparsely-spiny dendritic trees (Moran and Desmond, 1985;
Miyashita et al., 1993).
4.2. The Occipitoparietal Visual Pathway and Frontal Eye Fields

By extending their investigations to macaque temporo-parietal lobe visual areas (specific-
ally, the middle temporal area, MT; the lateral intraparietal area, LIPv; and the dorsal
superior temporal sulcus, 7a), Elston and Rosa (1997) have provided additional support
for a progressive hierarchy, although morphological differences were not as pronounced
as those in the occipitotemporal visual pathway. They documented a serial increase in
basal dendritic field territories and in branching complexity for pyramidal neurones in the
early stages of the occipitoparietal pathway (V1, V2 and MT), but not in the latter stages
(MT, LIPv and 7a). As in the occipitotemporal pathway, orientation and directional biases
were less marked in more rostral visual areas than in the primary visual regions. Finally,
by coupling (visible) spine estimates with basal (but not apical) dendritic extent, they
extrapolated a clear stepwise progression in the average spine counts from V1 (799 spines/
neurone) to 7a (2572 spines/neurone).
When pyramidal neurones in the two parietal regions (LIPv and 7a) were compared
with those in the frontal eye fields (FEF), clear differences emerged, with FEF neurones
being more complex in terms of basal dendritic field size and branching complexity
(Elston and Rosa, 1998a). Moreover, spine counts in FEF neurones were approximately
30% higher than in the parietal regions. FEF neurones had the largest and most complex
basal dendritic systems, with most dense spines, of any of the areas Elston and Rosa had
examined, presumably because FEF cells are less functionally compartmentalized, integ-
rating extensive polymodal input (Huerta et al., 1986). Incorporation of FEF neurones into
the visual pathways thus reveals an even more general caudal-rostral complexity gradient
for basal dendritic systems. Although progression along a proposed hierarchy may not be
paralleled exactly by concomitant increases in dendritic field area, especially if functional
or anatomical classification within that hierarchy is uncertain, other factors such as spine
density and dendritic orientation may contribute significantly to the ultimate structure of
hierarchically arranged neuropil.
5. DIRECT INVESTIGATIONS OF REGIONAL VARIATION

IN CORTICAL NEUROPIL: HUMANS
Much of the early quantitative work on regional variation in human dendritic systems was
performed in the Scheibel laboratory, and was inspired by the observation that certain func-
tional talents (e.g. eidetic imagery) are associated with particular anatomical arrangements

(Scheibel, 1988). In a pioneering undertaking to explore this potential structure-function

relationship, Scheibel et al. (1990) quantified basal dendritic systems by means of a Sholl
analysis in four cortical regions of the LH: somatosensory cortex (BA3-1-2—thoracic
region; BA3-1-2—finger region), prefrontal cortex (BA9), and the supramarginal gyrus
(BA40). They found partial support for a positive relationship between dendritic extent
and functional complexity insofar as the dendritic arbors in the two association cortices
(BA9 and BA40) and in the finger region of BA3-1-2 were typically more complex than
those in the thoracic region of BA3-1-2 (recall Figure 6.1B).
More recently, Schlaug et al. (1993) compared layer V basal dendritic systems in the
anterior (BA24b) and posterior (BA23b) cingulate gyrus by means of a Sholl analysis.
Consistent with the posterior-anterior gradient suggested by Elston and Rosa’s research,
the anterior cingulate exhibited greater dendritic complexity than the posterior cingulate.
This anterior advantage may reflect functional differences between the two regions as well
as differences in interconnectivity (Vogt et al., 1979), with the anterior cingulate pyra-
midal cells receiving diverse synaptic input of both an affective and a cognitive nature
(Devinsky et al., 1995). Importantly, the study also incorporated a quantitative measure
that has seldom been examined with regard to dendritic extent: cell packing density. The
anterior cingulate region was characterized by a lower cell packing density than the
posterior portion, thus indicating an inverse relationship between cell packing density and
dendritic arborization, at least in homotypical isocortex. Incorporating both measures in
future quantitative morphological research should provide valuable insights into regional
distributions of neuropil and interconnectivity patterns.
At present, the most extensive work on regional dendritic variation in humans has been
performed in the Jacobs laboratory, which has been exploring the morphological under-
pinnings of the functional cortical hierarchy proposed by Benson (1993, 1994). Benson’s
hierarchy draws heavily on the sensory-fugal gradients of cortical connectivity proposed
by Mesulam (1985), which have recently undergone considerable elaboration (Mesulam,
1998). In Benson’s over-simplified, but useful hierarchical schema, the cerebral cortex is
roughly classified into four divisions based on clinical/anatomical correlations: primary,
unimodal, heteromodal, and supramodal (see Table 6.1). These cortical types represent
progressively more complex levels of neural processing. Although these divisions and
their anatomical boundaries are far from absolute, they do provide an initial framework for
examining dendritic/spine systems vis-à-vis a functional hierarchy. To date, two studies
have been performed within this hierarchical schema to explore regional dendritic variation.
Table 6.1. Proposed functional hierarchy for human cerebral cortexa
Cortical divisions Function Sample areas
Primary cortex Transfer of sensory or motor impulses BA3-1-2, BA4

Unimodal association cortex Discrimination, categorization, and BA18, BA22,
integration of single modality BA44
information to form a unimodal percept
Heteromodal association cortex Formation and processing of complex BA6B, BA39
multimodal percepts
Supramodal association cortex Executive control of cognitive networks BA10, BA11
Note
a
Based on Benson (1993, 1994).

The first investigation, mentioned previously with regards to aging (Jacobs et al.,
1997), examined dendritic/spine differences between the secondary occipital area (BA18)
and the prefrontal cortex (BA10). BA18, which distributes functionally unique streams of
visual information (e.g. color, form, orientation) to other extrastriate areas (Burkhalter and
Van Essen, 1986; Gegenfurtner et al., 1996), represents typical unimodal cortex. BA10,
which is involved in several higher level integrative functions (e.g. drive, executive
control, planning; Stuss and Benson, 1984), represents the quintessential supramodal
region of the brain. As predicted, the basal dendrites and associated spines of supragranu-
lar pyramidal cells in BA10 were significantly more extensive than those in BA18—by
approximately 18% for total dendritic length, and by approximately 35% for spine number
(recall Figure 6.1D). Reinforcing the robust nature of this finding, this regional advantage
for BA10 was observed in all but a few of the 26 individuals examined. Recently, we have
also documented that the pyramidal cell packing density in layer III was 24% greater
in BA18 than in BA10 (unpublished data), a finding that further supports an inverse rela-
tionship between dendritic length and cell packing density in homotypical isocortex
(cf. Schlaug et al., 1993). Regardless of cellular density, the more complex dendritic array
in BA10 neurones appears to facilitate a broader sampling of afferent information, thereby
potentially increasing their integrative capacity. In contrast, the more limited basilar
dendritic systems in BA18 neurones may correspond with more discrete sampling of
afferent information (i.e. smaller receptive fields; cf. Rosa and Schmid, 1995), as would
be characteristic of information processing that is more unimodal than supramodal in
nature. These results appear consistent with the anatomical hierarchy proposed by Pandya
and Yeterian (1990), and provide initial support at the dendritic level for Benson’s functional
hierarchy.
The second investigation (Prather et al., 1997; Jacobs et al., 2001) is perhaps the most
extensive quantitative neuromorphological study to date. It examined the basal dendritic/
spine systems of supragranular pyramidal cells (N = 800) across eight regions of human
cerebral cortex. Tissue was removed from the lateral surface of the LH to represent
each level of Benson’s hierarchical functional schema: primary cortex (somatosensory,
BA3-1-2; motor, BA4), unimodal association cortex (Wernicke’s area, BA22; Broca’s
area, BA44), heteromodal association cortex (supplementary motor area, BA6b; angular
gyrus, BA39), and supramodal cortex (superior frontopolar zone, BA10; inferior fronto-
polar zone, BA11). Subsequently, primary and unimodal areas were grouped as “lower
integrative regions;” heteromodal and supramodal areas were grouped as “higher integrative
regions.”
Despite the considerable interindividual variation that typifies human tissue, there were
significant differences across Brodmann’s areas and between the higher and lower integ-
rative zones for all dendritic and spine measures (see Figure 6.4). Dendritic systems in
primary and unimodal regions were consistently less complex than heteromodal and supra-
modal areas. Nevertheless, the exact sequence of individual Brodmann’s areas depended
somewhat on what aspect of the dendritic tree was examined (e.g. total dendritic
length: BA3-1-2 < BA22 < BA4 < BA44 < BA11 < BA39 < BA6b < BA10; spine number:
BA3-1-2 < BA22 < BA4 < BA44 < BA6b < BA11 < BA39 < BA10). The range within these
rankings is substantial, with total dendritic length in BA10 being 31% greater than that in
BA3-1-2, and dendritic spine number being 69% greater.
In terms of individual areas, it is not surprising that BA6b, BA10, and BA39 exhibited
the most elaborate dendritic systems given the vast interconnections and functional

Figure 6.4. Sample tracings of human supragranular pyramidal cells and a bar graph of the dendritic spine
number (DSN) for eight Brodmann areas (BA), arranged from lowest (BA3-1-2) to highest (BA10) in terms of
overall complexity. Areas have also been grouped as Low (BA3-1-2, BA22, BA4 and BA44) and High (BA6,
BA11, BA39 and BA10) Integration regions, with the average DSN value for each grouping indicated by the
dotted lines. Note that DSN values represent only spines that were visible on the basal dendrites. In general,
DSN values in the High Integration regions are considerably higher than those in the Low Integration regions.
This hierarchy is roughly reflected in the individual tracings of neurones (synthesized from Prather et al., 1997;
Jacobs et al., 2001). Scale bars = 100 µm.
complexity of these cortical regions (for reviews, see Goldberg, 1985; Zilles, 1990;
Roland, 1993). It is unclear why the dendritic arbors of BA22 were at the low end of the
spectrum, although this may be due to this area’s close proximity to the primary auditory
cortex (see Jacobs and Scheibel, 1993). Conceivably, a section more posterior along the
superior temporal gyrus would be involved in synthesizing a greater proportion of poly-
modal information, especially given that the sensory speech region receives a wide sampling
of cortical and subcortical input (Pandya et al., 1969; Jones and Powell, 1970; Seldon,
1985). Thus, in some instances, there were minor exceptions to Benson’s functional hierarchy
(e.g. a primary area such as BA4 being slightly more complex than a predominantly—
though probably not exclusively—unimodal area such as BA22). Such exceptions should

be expected, given the vast interconnectivity of cortical areas, which do not readily
conform to strict hierarchical boundaries. On the whole, however, the results indicate that
the dendritic/spine systems of cortical areas involved in the initial stages of information
processing are not as complex as those involved later in the processing stream, and further
underscore that the processing demands placed on dendritic systems in various cortical
regions substantially influence their ultimate expression (Cajal, 1894; Hebb, 1949;
Diamond et al., 1964).
6. DENDRITIC INTEGRATION AND FUNCTIONAL IMPLICATIONS
Functional localization in the cerebral cortex has become one of the conceptual founda-
tions of neuroscience. It has traditionally been based on two streams of research activity:
(1) clinical and physiological dissections (e.g. Broca, 1861; Fritsch and Hitzig, 1870;
Woolsey, 1958), which established areal specificity for the various aspects of perception
and behaviour at a cortical level, and (2) cytoarchitectonic studies, which began to define
the anatomical extent and organization of such areas (e.g. Brodmann, 1909; Von
Economo, 1929). At the same time, the beginnings of studies at higher resolution, based
on visualization of dendritic ensembles and axonal patterns, became possible with the use
of the Golgi silver impregnation methods (Golgi, 1886; Cajal, 1909; Lorente de Nó,
1922). However, useful anatomical comparisons of cell structure and neuropil patterns
across areas have only become possible with the development of quantitative morpholo-
gical techniques (e.g. Bok, 1936; Sholl, 1953). A sufficient number of such analyses now
suggests that some patterns in neuropil may eventually be understood in terms of the
nature of the cortical function subsumed.
One such pattern appears to be the length and complexity of dendrite arrangements. We
had previously suggested a positive correlation between computational complexity and
dendritic extent (Scheibel et al., 1985). Recently, several studies have provided more rig-
orous support for this idea (Elston and Rosa, 1998b; Jacobs et al., 2001), which correlates
reasonably well with hierarchical conceptions of cortical organization. Essentially, as one
progresses from “first level” cortical input stages through intermediate levels of associ-
ation areas (unimodal and heteromodal) to the presumed hierarchically highest levels of
cortical associative activity (supramodal), there is a coincidental increment in basal den-
drite length and spine number. As indicated above, the increase in dendrite length in
humans over the entire sequence is of the order of one third, and that of total spine number
by about two thirds. Although fairly consistent, and significant in a quantitative sense,
these are by no means massive changes. On an entirely intuitive level, it might be difficult
to conceive that the robust qualitative differences assumed to exist between processing
mechanisms active in BA18 and BA10, for instance, can in large part be accounted for by
a 30% difference in dendritic extension. Certainly, factors other than dendrite extent must
be responsible for the dramatic inter-areal differences in function.
Phrasing the question differently, we might wonder whether the relatively modest
though consistent differences in basal dendrite dimensions among the cortical areas
studied are minor variants on a standard pattern, thereby providing only processing
changes of a quantitative nature, or whether these differences reflect much more funda-
mental computational mechanisms, thereby providing variations of a qualitative nature.
Studies by Tyc-Dumont and colleagues suggest that the latter may be true (Gogan and

Tyc-Dumont, 1989). In a series of communications, these researchers report significant

differences in the behaviour of individual dendrites. Each dendrite seems to have its own
electrical characteristics depending on a group of parameters, including input resistance,
voltage attenuation, and charge transfer effectiveness ratio (Gogan and Tyc-Dumont,
1989). These factors are computed for every dendritic site in the arborization to give the
electrical image of the tree.
Structural studies have emphasized the functional importance of the small caliber outer-
most branches of basal dendrite systems. In the enrichment paradigm, for example, it is
these peripheral processes in rats which develop in response to enhanced input and which
disappear when the animal is input-deprived (Connor et al., 1981). Similarly in the human
brain, it is the development of these smallest caliber (fifth and sixth order) branches in
Broca’s area of the left hemisphere that accompanies the development and maturation of
the language faculty (Simonds and Scheibel, 1989).
Some years ago, on the basis of a group of selective lesion experiments in rabbits,
Globus and Scheibel (1967a,b,c) emphasized the essentially modular nature of the den-
dritic domains of cortical pyramids. For any one cortical area, there was a high degree of
consonance in the extent of the horizontal dendritic components (basal and oblique
branches and the apical arch). Depending on the location of the cell body, however, the
apical shaft might be as short as 100 µm (layer II), or as long as 4000 µm (deep layer V).
Fibre terminals of intracortical derivation seemed almost exclusively related to the
horizontal branch systems of the cell (i.e. apical obliques and basal dendrites), whereas
extracortically derived afferents appeared to terminate predominantly on apical shafts.
The extent of the horizontal dendritic component of the cell could then be considered
a measure of its exposure to intracortically derived information, especially since the neo-
cortex communicates predominantly with itself (Braitenberg, 1978; Nieuwenhuys, 1994).
Progressive increase in the extent of the basal dendritic skirt might therefore be expected
to increase the exposure of the neurone to intracortical influences, a correlation that
intuitively meets assumptions about the needs of increasingly more complex associative
functions.
There have been suggestions that the fine, tapering, outermost branches of the dendritic
ensemble may assume a physiological importance out of proportion to the modest fraction
of the neuronal dendrite that they represent:
“Most distally from the soma, the extremely fine dendritic branches are practically
independent subunits where nonlinear synaptic interactions operate, and only the
results of these operations are transmitted to the soma with different efficiencies,
depending on the cable properties of the individual dendritic channel. [Thus]
each neuron is like a complex computer with many nonlinear coprocessors oper-
ating in parallel.” (Gogan and Tyc-Dumont, 1989, p. 129)
It is interesting to recall that earlier physiological studies of neurones led some to the
conclusion that, in the case of motoneurones at least, electrotonic considerations indicated
that distal dendrites contribute little to electrical events at the soma (Eccles, 1964). On the
basis of theoretical considerations, however, Rall (1974) suggested that remote dendrite
terminals might exercise significant electrical effects at the cell body. A number of
possible mechanisms for this have been advanced over the last thirty years. Among these
possibilities, Redman (1973) suggested that more current may be injected by distal

synapses than by proximal ones, thereby resulting in excitatory postsynaptic potentials of

approximately similar shape and size at the cell body.
An alternative suggestion by Rall (1974) is based on the observation that the stems of
dendritic spines are generally longest and thinnest at the peripheral portions of dendrites
and become increasingly short and stubby as the soma is approached (Jones and Powell,
1969). The long thin spine stem has a much higher resistance value and therefore a lower
current carrying capacity. Rall suggested that this attenuated spine stem might have value as
an impedance matching device vis-à-vis the underlying dendritic branch input. These most
peripheral spines might therefore be particularly effective in adjusting synaptic potency:
“The design principle involved here is to sacrifice maximum power in order to

gain flexibility and control. Adjustability of potency means either increase or
decrease relative to other synapses. Thus we think of delicate adjustments of the
relative weights (potency) of many different synapses to any given neurone. We
think of these changes as responsible for changes in dynamic patterns of activity
in assemblies of neurones organized with convergent and divergent connective
overlaps.” (Rall, 1974, p. 17)
Indeed, recent research suggests that the excitability of an entire dendrite may be regulated
by changes in distal spine density (Jaslove, 1992). It thus seems likely that synaptic
activity in the outermost segments of the dendrite tree not only affects neuronal activity,
but may, in fact, exert effects out of proportion to dendritic extent and geographic distance
from the soma.
Another aspect of terminal dendrite function that may enrich the role of this portion of
the dendrite tree is the possibility of interaction among dendrite terminal branches. In add-
ition to dendro-dendritic synapses such as those described in the olfactory bulb (Rall et al.,
1966), a number of investigators have described the presence of dendrite bundles in many
sites throughout the central nervous system including the cerebral cortex (Fleischhauer,
1974; Scheibel and Scheibel, 1970). Although the role of such structural complexes has
never been fully clarified, it has been suggested that distal dendrites in close apposition to
those from other neurones may in fact transmit information, thereby emphasizing again
the possibly special role of this portion of the nerve cell (Bras et al., 1987; Gogan and
Tyc-Dumont, 1989).
Finally, complementing Rall’s theoretical speculations are more recent insights into the
active characteristics of dendritic branches provided by high-speed fluorescence imaging
and dendritic patch clamping. It seems relatively clear that dendrites can no longer be
viewed as simply passive structures adhering to the principles of cable theory (for review,
see Johnston et al., 1996). Indeed, at least some cortical dendrites appear capable of
sustaining active propagation of electrical potentials through voltage-sensitive Na+ and
Ca2+ channels. These active characteristics would indeed boost the effect of distal synaptic
input, contributing significantly to synaptic integration locally and across the entire neuron.
7. CONCLUSION
Quantitative neuromorphological techniques have substantially enhanced our understand-

ing of the dendritic ensembles first described and categorized according to qualitative

observations (e.g. Ramón-Moliner, 1962). Indeed, as noted by Sholl (1953), the characteristic
dendritic patterns across various cortical regions can only be revealed through quantitative
investigations. We conclude this chapter by noting that there is a very strong likelihood
that these inter-areal variations in basal dendritic dimensions revealed by quantitative neu-
romorphological investigations of the cerebral cortex reflect significant differences in the
nature of cortical processing. Many other factors are undoubtedly involved in determining
the range of computational strategems as one moves from first level sensory representations
to the highest associational levels. Some of these are considered elsewhere in this volume.
Nonetheless, the characteristics of the receptive dendritic membrane of individual neuronal
elements and their variations along the length of the dendritic shaft are bound to represent
central issues in our developing knowledge of cortical computation.
ACKNOWLEDGEMENTS
We would like to thank the following individuals for their suggestions on preliminary
versions of this manuscript: Jesse Jacobs, Elisa Kapler, Jennifer Ransom and Sharon Sann.
REFERENCES
Amir, Y., Harel, M. and Malach, R. (1993) Cortical hierarchy reflected in the organization of intrinsic connec-
tions in macaque monkey visual cortex. Journal of Comparative Neurology, 334, 19–64.
Bannister, N.J. and Larkman, A.U. (1995) Dendritic morphology of CA1 pyramidal neurons from the rat hippo-
campus: I. Branching patterns. Journal of Comparative Neurology, 360, 150–160.
Becker, L.E., Armstrong, D.L., Chan, F. and Wood, M.M. (1984) Dendritic development in human occipital
cortical neurons. Developmental Brain Research, 13, 117–124.
Belichenko, P.V. and Dahlström, A. (1995) Studies of the 3-dimensional architecture of dendritic spines and
varicosities in human cortex by confocal laser scanning microscopy and Lucifer Yellow microinjections.
Journal of Neuroscience Methods, 57, 55–61.
Benson, D.F. (1993) Progressive frontal dysfunction. Dementia, 4, 149–153.
Benson, D.F. (1994) The Neurology of Thinking. New York: Oxford Press.
Bok, S.T. (1936) The branching of the dendrites in the cerebral cortex. Proceedings of the Academy of Sciences
(Amsterdam), 39, 1209–1218.
Boussaoud, D., Ungerleider, L.G. and Desimone, R. (1990) Pathways for motion analysis: Cortical connections
of the medial superior temporal and fundus of the superior temporal visual areas in the macaque. Journal of
Braak, H. and Braak, E. (1985) Golgi preparations as a tool in neuropathology with particular reference to
investigations of the human telencephalic cortex. Progress in Neurobiology, 25, 93–139.
Braitenberg, V. (1978) Cortical architectonics: General and areal. In: M.A.B. Brazier and H. Petsche (eds),
Architectonics of the Cerebral Cortex, New York: Raven Press, pp. 443–465.
Braitenberg, V. and Schüz, A. (1991) Anatomy of the Cortex: Statistics and Geometry. Berlin: Springer.
Bras, H., Destombes, J., Gogan, P. and Tyc-Dumont, S. (1987) The dendrites of single brain-stem motoneurons
intracellularly labelled with horseradish peroxidase in the cat. An ultrastructural analysis of the synaptic
covering and the microenvironment. Neuroscience, 22, 971–981.
Broca, P. (1861) Remarques sur la siège de la faculté du language articulé, suivies d’une observation d’aphemie
(perte de la parole). Bulletin de Societé d’Anatomie (Paris), 36, 330–357.
Brodmann, K. (1909) Vergleichende Lokalisationlehre der Grosshirnrinde in ihren Prinzipien dargestellt auf
Grund des Zellenbaues. Leipzig: J.A. Barth.
Buell, S.J. and Coleman, P.D. (1981) Quantitative evidence for selective dendritic growth in normal human
aging but not in senile dementia. Brain Research, 214, 23–41.
Buell, S.J. (1982) Golgi-Cox and rapid Golgi methods as applied to autopsied human brain tissue: Widely
disparate results. Journal of Neuropathology and Experimental Neurolology, 41, 500–507.
Burkhalter, A. and Van Essen, D.C. (1986) Processing of color, form and disparity information in visual areas
VP and V2 of ventral extrastriate cortex in the macaque monkey. Journal of Neuroscience, 6, 2327–2351.

Cajal, S. Ramón y (1894) Les Nouvelles Ideés sur la Structure du Système Nerveux chez L’homme et chez les
Vertébrés. Trans. by L. Azoulay. Paris: Reinwald.
Cajal, S. Ramón y (1909, 1911) Histologie du Système Nerveux de L’homme et des Vertébrés, 2 Vols. Trans. by
L. Azoulay. Paris: Maloine.
Chugani, H.T., Phelps, M.E. and Mazziotta, J.C. (1987) Positron emission tomography study of human brain
functional development. Annals of Neurology, 22, 487–497.
Coleman, P.D. and Flood, D.G. (1987) Neuron numbers and dendritic extent in normal aging and Alzheimer’s
disease. Neurobiology of Aging, 8, 521–545.
Colonnier, M. (1964) The tangential organization of the visual cortex. Journal of Anatomy (London), 98, 327–344.
Conel, J. (1939–67) The Postnatal Development of the Human Cerebral Cortex (Vols. 1–6). Cambridge,
Massachusetts: Harvard University Press.
Connor, J.R., Melone, J., Yuen, A. and Diamond, M.C. (1981) Terminal segment lengths in dendrites in aged
rats: An experimentally induced response. Experimental Neurology, 73, 827–830.
Cupp, C.J. and Uemura, E. (1980) Age-related changes in prefrontal cortex of Macaca mulatta: Quantitative
analysis of dendritic branching patterns. Experimental Neurology, 69, 143–163.
de Ruiter, J.P. (1983) The influence of post-mortem fixation delay on the reliability of the Golgi silver impregna-
tion. Brain Research, 266, 143–147.
Devinsky, O., Morrell, M.J. and Vogt, B.A. (1995) Contributions of anterior cingulate cortex to behaviour.
Brain, 118, 279–306.
Diamond, M.C., Krech, D. and Rosenzweig, M.R. (1964) The effects of an enriched environment on the histo-
logy of the rat cerebral cortex. Journal of Comparative Neurology, 123, 111–119.
Eayrs, J.T. (1955) The cerebral cortex of normal and hypothyroid rats. Acta Anatomica, 25, 160–183.
Eccles, J.C. (1964) The Physiology of Synapses. Berlin: Springer.
Elston, G.N. and Rosa, M.G.P. (1997) The occipitoparietal pathway of the macaque monkey: Comparison
of pyramidal cell morphology in layer III of functionally related cortical visual areas. Cerebral Cortex, 7,
432–452.
Elston, G.N. and Rosa, M.G.P. (1998a) Complex dendritic fields of pyramidal cells in the frontal eye field of the
macaque monkey: Comparison with parietal areas 7a and LIP. NeuroReport, 9, 127–131.
Elston, G.N. and Rosa, M.G.P. (1998b) Morphological variation of layer III pyramidal neurones in the occipito-
temporal pathway of the macaque monkey visual cortex. Cerebral Cortex, 8, 278–294.
Elston, G.N., Rosa, M.G.P. and Calford, M.B. (1996) Comparison of dendritic fields of layer III pyramidal
neurons in striate and extrastriate visual areas of the marmoset: A Lucifer Yellow intracellular injection
study. Cerebral Cortex, 6, 807–813.
Fleischhauer, K. (1974) On different patterns of dendrite bundling in the cerebral cortex of the rat. Zeitschrift für
Anatomie und Entwicklungsgeschichte, 143, 115–126.
Flood, D.G. (1993) Critical issues in the analysis of dendritic extent in aging humans, primates, and rodents.
Neurobiology of Aging, 14, 649–654.
Flood, D.G., Buell, S.J., Defiore, C.H., Horwitz, G.J and Coleman, P.D. (1985) Age-related dendritic growth in
dentate gyrus of human brain is followed by regression in the ‘oldest old.’ Brain Research, 345, 366–368.
Fritsch, G. and Hitzig, E. (1870) Über die elektrische Erregbarkeit des Grosshirns. Archiv für Anatomie und
Physiologie (Leipzig), 37, 300–332.
Gegenfurtner, K.R., Kiper, D.C. and Fenstemaker, S.B. (1996) Processing of color, form, and motion in macaque
area V2. Visual Neuroscience, 13, 161–172.
Globus, A. and Scheibel, A.B. (1967a) Pattern and field in cortical structure: The rabbit. Journal of Comparative
Neurology, 131, 155–172.
Globus, A. and Scheibel, A.B. (1967b) Synaptic loci on visual cortical neurons of the rabbit: The specific afferent
radiation. Experimental Neurology, 18, 116–131.
Globus, A. and Scheibel, A.B. (1967c) Synaptic loci on parietal cortical neurons: Termination of corpus callo-
sum fibers. Science (Washington), 156, 1127–1129.
Gogan, P. and Tyc-Dumont, S. (1989) How do dendrites process neural information? News in Physiological
Sciences (American Physiological Society), 4, 127–130.
Goldberg, G. (1985) Supplementary motor area structure and function: Review and hypothesis. Behavioral and
Brain Sciences, 8, 567–616.
Golgi, C. (1886) Studi Sulla Fina Anatomia degli Organi Centrali del Sistema Nervoso. Milano: Hoepli.
Hayes, T.L. and Lewis, D.A. (1996) Magnopyramidal neurons in the anterior motor speech region. Archives of
Neurology, 53, 1277–1283.
Hebb, D. (1949) The Organization of Behavior. New York: Wiley.
Huerta, M.F., Krubitzer, L.A. and Kaas, J.H. (1986) Frontal eye field as defined by intracortical microstimula-
tion in squirrel monkeys, owl monkeys, and macaque monkeys: I. Subcortical connections. Journal of
Huttenlocher, P.R. and Dabholkar, A.S. (1997) Regional differences in synaptogenesis in human cerebral cortex.

Huttenlocher, P.R. and de Courten, C. (1987) The development of synapses in the striate cortex of man. Human
Neurobiology, 6, 1–9.
Huttenlocher, P.R., de Courten, C., Garey, L.J. and Van Der Loos, H. (1982) Synaptogenesis in human visual
cortex: Evidence for synapse elimination during normal development. Neuroscience Letters, 33, 247–252.
Ishizuka, N., Maxwell Cowan, W. and Amaral, D.G. (1995) A quantitative analysis of the dendritic organization
of pyramidal cells in the rat hippocampus. Journal of Comparative Neurology, 362, 17–45.
Jacobs, B. and Scheibel, A.B. (1993) A quantitative dendritic analysis of Wernicke’s area in humans. I. Lifespan
changes. Journal of Comparative Neurology, 327, 83–96.
Jacobs, B., Batal, H.A., Lynch, B., Ojemann, G., Ojemann, L. and Scheibel, A.B. (1993a) Quantitative dendritic
and spine analyses of speech cortices: A case study. Brain and Language, 44, 239–253.
Jacobs, B., Schall, M. and Scheibel, A.B. (1993b) A quantitative dendritic analysis of Wernicke’s area in
humans. II. Gender, Hemispheric, and Environmental Factors. Journal of Comparative Neurology, 327,
97–111.
Jacobs, B., Chugani, H.T., Allada, V., Chen, S., Phelps, M.E., Pollack, D.B. and Raleigh, M.J. (1995). Develop-
mental changes in brain metabolism in sedated rhesus macaques and vervet monkeys revealed by positron
emission tomography. Cerebral Cortex, 5, 222–233.
Jacobs, B., Driscoll, L. and Schall, M. (1997) Life-span dendritic and spine changes in areas 10 and 18 of human
cortex: A quantitative Golgi Study. Journal of Comparative Neurology, 386, 661–680.
Jacobs, B., Schall, M., Prather, M., Kapler, E., Driscoll, L., Baca, S., Jacobs, J., Ford, K., Wainwright, M. and
Treml, M. (2001) Regional dendritic and spine variation in human cerebral cortex: A quantitative Golgi
study. Cerebral Cortex, 11, 558–571.
Jaslove, S.W. (1992) The integrative properties of spiny distal dendrites. Neuroscience, 47, 495–519.
Johnston, D., Magee, J.C., Colbert, C.M. and Christie, B.R. (1996) Active properties of neuronal dendrites.
Annual Review of Neuroscience, 19, 165–186.
Jones, E.G. and Powell, T.P.S. (1969) Morphological variations in the dendritic spines of the neocortex. Journal
of Cell Science, 5, 509–529.
Jones, E.G. and Powell, T.P.S. (1970) An anatomical study of converging sensory pathways with the cerebral
cortex of the monkey. Brain, 93, 793–820.
Juraska, J.M. (1982) The development of pyramidal neurons after eye opening in the visual cortex of hooded
rats: A quantitative study. Journal of Comparative Neurology, 212, 208–213.
Koenderink, M.J.Th., Uylings, H.B.M. and Mrzljak, L. (1994) Postnatal maturation of the layer III pyramidal
neurons in the human prefrontal cortex: A quantitative Golgi analysis. Brain Research, 653, 173–182.
Koenderink, M.J.Th. and Uylings, H.B.M. (1995) Postnatal maturation of layer V pyramidal neurons in the
human prefrontal cortex: A quantitative Golgi analysis. Brain Research, 678, 233–243.
Kuhl, D.E., Metter, E.J., Riege, W.H. and Phelps, M.E. (1982) Effects of human aging on patterns of local
cerebral glucose utilization determined by the [18F] fluorodeoxyglucose method. Journal of Cerebral Blood
Larkman, A.U. (1991) Dendritic morphology of pyramidal neurones of the visual cortex of the rat: III. Spine
distributions. Journal of Comparative Neurology, 306, 332–343.
Lorente de Nó, R. (1922) La corteza cerebral del raton. Primera contribucion: La corteza acustica. Trabajos del
Instituto Cajal de Investigaciones Biologicas (Madrid), 20, 41–78.
Lund, J.S., Yoshioka, T. and Levitt, J.B. (1993) Comparison of intrinsic connectivity in different areas of
macaque monkey cerebral cortex. Cerebral Cortex, 3, 148–162.
Marin-Padilla, M. (1967) Number and distribution of the apical dendritic spines of the layer V pyramidal cells in
man. Journal of Comparative Neurology, 131, 475–490.
Marin-Padilla, M. (1970) Prenatal and early postnatal ontogenesis of the human motor cortex: A Golgi study. I.
The sequential development of the cortical layers. Brain Research, 23, 167–183.
Mata, M., Fink, D.J., Gainer, H., Smith, C.B., Davidsen, L., Savaki, H. et al. (1980) Activity-dependent energy
metabolism in rat posterior pituitary reflects sodium pump activity. Journal of Neurochemistry, 34, 213–215.
Matsubara, J.A., Chase, R. and Thejomayen, M. (1996) Comparative morphology of three types of projection-
identified pyramidal neurons in the superficial layers of cat visual cortex. Journal of Comparative Neuro-
logy, 366, 93–108.
Meller, S.T. and Dennis, B.J. (1990) A multiple Golgi analysis of the periaqueductal gray in the rabbit. Journal
of Comparative Neurology, 302, 66–86.
Mesulam, M.-M. (1985) Principles of Behavioral Neurology. Philadelphia: F.A. Davis.
Mesulam, M.-M. (1998) From sensation to cognition. Brain, 121, 1013–1052.
Miyashita, Y., Okuno, H. and Hasegawa, I. (1993) Tuning and association: Neuronal memory mechanisms of
complex visual forms in monkey temporal cortex. Biomedical Research, 14, 89–94.
Moran, J. and Desmone, R. (1985) Selective attention gates visual processing in the extrastriate cortex. Science
(Washington), 229, 782–784.
Mrzljak, L., Uylings, H.B.M., van Eden, C.G. and Judás, M. (1990) Neuronal development in human prenatal
and postnatal stages. Progress in Brain Research, 85, 185–222.

Murphy, E.H. and Magness, R. (1984) Development of the rabbit visual cortex: A quantitative Golgi analysis.
Nakamura, S., Akiguchi, I., Kameyama, M. and Mizuno, N. (1985) Age-related changes of pyramidal cell basal
dendrites in layers III and V of human motor cortex: A quantitative Golgi study. Acta Neuropathologica
(Berlin), 65, 281–284.
Nieuwenhuys, R. (1994) The neocortex. Anatomy and Embryology, 190, 307–337.
Nudo, R.J. and Masterton, R.B. (1986) Stimulation-induced (14C)2-deoxyglucose labeling of synaptic activity in
the central auditory system. Journal of Comparative Neurology, 245, 553–565.
Ohm, T.G. and Diekmann, S. (1994) The use of Lucifer Yellow and Mini-Ruby for intracellular staining in fixed
brain tissue: Methodological considerations evaluated in rat and human autopsy brains. Journal of Neuro-
science Methods, 55, 105–110.
Ojemann, G.A. and Whitaker, H.A. (1978) Language localization and variability. Brain and Language, 6, 239–260.
Ojemann, G.A., Ojemann, J., Lettich, E. R.E.E.G.T. and Berger, M. (1989) Cortical language localization in left,
dominant hemisphere: An electrical stimulation mapping investigation in 117 patients. Journal of Neuro-
surgery, 71, 316–326.
Pandya, D.N. and Yeterian, E.H. (1990) Prefrontal cortex in relation to other cortical areas in rhesus monkey:
Architecture and connections. In: H.B.M. Uylings, C.G. Van Eden, J.P.C. De Bruin, M.A. Cornere and
M.G.P. Feenstra (eds), The Prefrontal Cortex. Its Structure, function and pathology (Progress in Brain
Research, Vol. 85) Amsterdam: Elsevier, pp. 63–94.
Pandya, D.N., Hallett, M. and Mukherjee, S.K. (1969) Intra- and interhemispheric connections of neocortical
auditory system in rhesus monkey. Brain Research, 14, 49–65.
Prather, M., Treml, M., Driscoll, L., Schall, M. and Jacobs, B. (1997) Regional variation in dendritic and spine
complexity: A quantitative Golgi analysis of human cerebral cortex. Society for Neuroscience Abstracts, 23,
87.13.
Prinz, M., Prinz, B. and Schulz, E. (1997) The growth of non-pyramidal neurons in the primary motor cortex of
man: A Golgi study. Histology and Histopathology, 12, 895–900.
Rakic, P., Bourgeois, J.-P., Eckenhoff, M.F., Zecevic, N. and Goldman-Rakic, P.S. (1986) Concurrent over-
production of synapses in diverse regions of the primate cerebral cortex. Science (Washington), 232, 232–235.
Rall, W., Shepherd, G.M., Reese, T.S. and Brightman, M.W. (1966). Dendrodendritic synaptic pathway for
inhibition in the olfactory bulb. Experimental Neurology, 14, 44–56.
Rall, W. (1974) Dendritic spines, synaptic potency and neuronal plasticity. In Cellular Mechanisms Subserving
Changes in Neuronal Activity, (eds) C.D. Woody, K.A. Brown, T.J. Crow and J.D. Knispel, Brain Information
Service Research Report #3: University of California, Los Angeles, pp. 13–21.
Ramón-Moliner, E. (1962) An attempt at classifying nerve cells on the basis of their dendritic patterns. Journal
Raz, N., Torres, I.J., Spencer, W.D. and Acker, J.D. (1993) Pathoclysis in aging human cerebral cortex: Evidence
from in vivo MRI morphometry. Psychobiology, 21, 151–160.
Raz, N., Gunning, F.M., Head, D., Dupuis, J.H., McQuain, J., Briggs, S.D., Laken, W.J., Thornton, A.E. and
Acker, J.D. (1997) Selective aging of the human cerebral cortex observed in vivo: Differential vulnerability
of the prefrontal gray matter. Cerebral Cortex, 7, 268–282.
Redman, S.J. (1973) The attenuation of passively propogating dendritic potentials in a motoneurone cable
model. Journal of Physiology (London), 234, 637–664.
Roland, P.E. (1984) Metabolic measurements of the working frontal cortex in man. Trends in Neuroscience, 7,
430–435.
Roland, P.E. (1993) Brain Activation. New York: John Wiley & Sons.
Rosa, M.G.P. and Schmid, L.M. (1995) Visual areas in the dorsal and medial extrastriate cortices of the marmo-
set. Journal of Comparative Neurology, 359, 272–299.
Schadé, J.P. and Van Groenigen, W.B. (1961) Structural organization of the human cerebral cortex. 1. Matura-
tion of the middle frontal gyrus. Acta Anatomica, 47, 74–111.
Scheibel, M.E. and Scheibel, A.B. (1970) Organization of spinal motoneuron dendrites in bundles. Experimental
Neurology, 28, 106–112.
Scheibel, M.E. and Scheibel, A.B. (1978) The methods of Golgi. In: R.T. Robinson (ed.), Neuroanatomical
Research Techniques, New York: Academic Press, pp. 89–114.
Scheibel, A.B., Paul, L.A., Fried, I., Forsythe, A.B., Tomiyasu, U., Wechsler, A., Kao, A. and Slotnick, J. (1985)
Dendritic organization of the anterior speech area. Experimental Neurology, 87, 109–117.
Scheibel, A.B. (1988) Dendritic correlates of human cortical function. Archives Italiennes de Biologie, 126, 347–357.
Scheibel, A.B., Conrad, T., Perdue, S., Tomiyasu, U. and Wechsler, A. (1990) A quantitative study of dendrite
complexity in selected areas of the human cerebral cortex. Brain and Cognition, 12, 85–101.
Schierhorn, H. (1981) Strukturwandel neokortikaler Pyramidenneurone des Menschen während des 5. bis 9.
Dezenniums. Psychiatrie, Neurologie und medizinische Psychologie, 33, 664–673.
Schlaug, G., Armstrong, E., Schleicher, A. and Zilles, K. (1993) Layer V pyramidal cells in the adult human
cingulate cortex: A quantitative Golgi-study. Anatomy and Embryology, 187, 515–522.

Schulz, E., Renner, J. and Meyer, U. (1992) Quantitative Analyse von Lamina III-Pyramiden-Neuronen im
Parietalcortex des Neugeborenen. Journal für Hirnforschung (Berlin), 6, 661–672.
Seldon, H.L. (1981a) Structure of human auditory cortex. I. Cytoarchitectonics and dendritic distributions. Brain
Research, 229, 277–294.
Seldon, H.L. (1981b) Structure of human auditory cortex. II. Axon distributions and morphological correlates of
speech perception. Brain Research, 229, 295–310.
Seldon, H.L. (1982) Structure of human auditory cortex. III. Statistical analysis of dendritic trees. Brain
Research, 249, 211–221.
Seldon, H.L. (1985) The anatomy of speech perception: Human auditory cortex. In: A. Peters and E.G. Jones
(eds), Cerebral Cortex (Vol. 4), New York: Plenum Press, pp. 273–327.
Semenova, L.K., Vasil’eva, V.A., Tsekhmistrenko, T.A. and Shumeiko, N.S. (1989) Characteristics of the
ensemble organization of the human cerebral cortex from birth to 20 years of age. Arkhiv Anatomii,
Gistologii I Embriologii, 97, 15–24.
Sholl, D.A. (1953) Dendritic organization in the neurones of the visual and motor cortices of the cat. Journal of
Anatomy, 87, 387–406.
Sholl, D.A. (1956) The Organization of the Cerebral Cortex. New York: Wiley.
Simonds, R.J. and Scheibel, A.B. (1989) The postnatal development of the motor speech area: A preliminary
study. Brain and Language, 37, 42–58.
cortex in man. Journal of Neurosurgery, 40, 747–755.
Stuss, D.T. and Benson, D.F. (1984) Neuropsychological studies of the frontal lobes. Psychological Bulletin, 95,
3–28.
Sullivan, E.V., Marsh, L., Mathalon, D.H., Lim, K.O. and Pfefferbaum, A. (1995) Age-related decline in MRI
volumes of temporal lobe gray matter but not hippocampus. Neurobiology of Aging, 16, 591–606.
Takashima, S., Becker, L.E., Armstrong, D.L. and Chan, F. (1981) Abnormal neuronal development in the visual
cortex of the human fetus and infant with Down’s syndrome. A quantitative and qualitative Golgi study.
Brain Research, 225, 1–21.
Terry, R.D., DeTeresa, R. and Hansen, L.A. (1987) Neocortical cell counts in normal human adult aging. Annals
of Neurology, 21, 530–539.
Trommald, M., Jensen, V. and Andersen, P. (1995) Analysis of dendritic spines in rat CA1 pyramidal cells intra-
cellularly filled with a fluorescent dye. Journal of Comparative Neurology, 353, 260–274.
Uemura, E. (1980) Age-related changes in prefrontal cortex of Macaca mulatta: Synaptic density. Experimental
Neurology, 69, 164–172.
Uylings, H.B.M., Smit, G.J. and Veltman, W.A.M. (1975) Ordering methods in quantitative analysis of branch-
ing structures of dendritic trees. In: G.W. Kreutzberg (ed.), Advances in Neurology (Vol. 12), New York:
Raven Press, pp. 247–254.
Uylings, H.B.M., Ruiz-Marcos, A. and van Pelt, J. (1986) The metric analysis of three-dimensional dendritic tree
patterns: A methodological review. Journal of Neuroscience Methods, 18, 127–151.
Valverde, F. (1976) Aspects of cortical organization related to the geometry of neurons with intra-cortical axons.
Journal of Neurocytology, 5, 509–528.
Van der Loos, H. (1959) Dendro-dendritic Connections in the Cerebral Cortex. [in Dutch] Doctoral dissertation,
University of Amsterdam. Haarlem: Stam.
Vogt, B.A., Rosene, D.L. and Pandya, D.N. (1979) Thalamic and cortical afferents differentiate anterior from
posterior cingulate cortex in the monkey. Science (Washington), 204, 205–207.
Von Economo, C.F. (1929) The Cytoarchitectonics of the Human Cerebral Cortex. London: Oxford Medical
Publications.
Walker, S.F. (1980) Lateralization of function in the vertebrate brain: A review. British Journal of Psychology,
71, 329–367.
Whitaker, H.A. and Selnes, O.A. (1976) Anatomic variations in the cortex: Individual differences and the problem
of the localization of language functions. Annals of the New York Academy of the Sciences, 280, 844–854.
Williams, R.S., Ferrante, R.J. and Caviness, Jr., V.S. (1978) The Golgi rapid method in clinical neuropathology:
Morphological consequences of suboptimal fixation. Journal of Neuropathology and Experimental Neuro-
logy, 37, 13–33.
Woolsey, C.N. (1958) Organization of somatic sensory and motor areas of the cerebral cortex. In: H.F. Harlow
and C.N. Woolsey (eds), Biological and Biochemical Bases of Behavior, Madison: University of Wisconsin,
pp. 63–81.
Yoshioka, T., Levitt, J.B. and Lund, J.S. (1992) Intrinsic lattice connections of macaque monkey visual cortical
area V4. Journal of Neuroscience, 12, 2785–2802.
Zatorre, R.J., Evans, A.C., Meyer, E. and Gjedde, A. (1992) Lateralization of phonetic and pitch discrimination
in speech processing. Science (Washington), 256, 846–849.
Zilles, K. (1990) Cortex. In: G. Paxinos (ed.), The Human Nervous System, San Diego: Academic Press,
pp. 757–802.

7 Intrinsic Connections in Mammalian Cerebral Cortex
Jonathan B. Levitt1 and Jennifer S. Lund2
1
Corresponding author. Department of Biology,
City College of the City University of New York,
138th Street & Convent Avenue, New York NY 10031, USA
2
Department of Ophthalmology, Moran Eye Center, University of Utah,
50 North Medical Drive, Salt Lake City, UT 84132, USA
Tel: +1-212-650-8539; FAX: +1-212-650-8585; e-mail: jbl@sci.ccny.cuny.edu
The superficial layers of all areas of mammalian cerebral cortex are characterised by an extensive
network of axons running parallel to the cortical surface. These projections, furnished primarily by
pyramidal neurones (but in certain species also by inhibitory basket neurones) span several milli-
meters of cortex, and make synaptic contacts in terminal fields that are generally patchy, or clustered
into discrete zones of high terminal density separated by zones of low density. These are referred to
as intrinsic connections, as they interconnect loci within a single cortical area. This chapter reviews
the basic structure of this connectional lattice: its laminar specificity, sources and targets, spatial
arrangement, and relation to functional maps. This description will centre on primary visual cortex
in cat and monkey as these have been most intensively studied, but we will also discuss the cortices
of other species where data are available, and will touch upon extrastriate visual areas, as well as
auditory, somatosensory, motor and prefrontal cortices. It is our purpose to show which features of
this connectional lattice are common to all areas of mammalian cerebral cortex, thus illustrating
what makes this a fundamental architectural feature of the mammalian cerebral cortex.
KEYWORDS: circuitry, horizontal connections, neuroanatomy, pyramidal neurone, visual cortex
1. INTRODUCTION
The columnar structure of the cerebral cortex has now been appreciated for well over 40
years as a fundamental architectural feature of cerebral cortex organization (Mountcastle,
1957; Powell and Mountcastle, 1959; Hubel and Wiesel, 1962). Cells situated in a column,
perpendicular to the cortical surface, share a number of important functional properties such
as receptive field position or stimulus selectivity. While we now appreciate the limitations
of this scheme (in that cells situated in different layers are in fact known to differ to some
degree in their connections and receptive field properties), it is still clear that in many
important respects neurones located in a single column from pia to white matter share key
receptive field properties. This columnar organisation is mirrored by the anatomical organ-
isation of cerebral cortex: in any area one chooses to investigate, there is a dense vertical focus
of connections linking cells at different depths within a single column of cortex. This has
been appreciated for even longer than the physiological columnar organisation: golgi
studies have shown that connections running perpendicular to the cortical surface, linking
133
134 Jonathan B. Levitt and Jennifer S. Lund
the different layers, are more prominent than those running parallel to the cortical surface
(Cajal, 1922; O’Leary, 1941; Lund and Boothe, 1975). Connections running horizontally,
parallel to the cortical surface, are difficult to trace with the Golgi technique. In such studies,
axons are generally traced no more than a few hundred microns from the soma. However, it was
clear, even from these relatively crude techniques, that there did exist neural circuits to mediate
communication among cells at different locations across the cortex (i.e. in different columns).
Increasingly sophisticated techniques have been used to study these horizontal connections
in the cerebral cortex: These include fibre degeneration, autoradiographic demonstration
of transport of radioactive amino acids, transport of neural tracer substances, in vitro intra-
cellular labeling and electrophysiological recording of individual neurones, and optical
imaging to name but a few. In this chapter, we will review the key features of these intrinsic
horizontal connections, i.e. circuits linking different columns within a single cortical area
(as opposed to those circuits making vertical connections within a column—e.g. Lund,
1973; Blasdel et al., 1985; Yoshioka et al., 1994). This chapter is not intended as an
exhaustive review of intrinsic connections. In the visual cortex alone there is now an
enormous literature concerning the organisation of cortical connections linking cells both
at different depths within a column, or in different columns (for recent reviews of this
topic see Fitzpatrick, 1996; Callaway, 1998b). Despite obvious differences among differ-
ent cortical areas in cytoarchitecture, functional properties, extrinsic connections, etc., it
has also long been appreciated that certain neuronal components and architectural features
are common to all areas of cerebral cortex (Cajal, 1909; Rockel et al., 1980; Tyler et al.,
1998). Our purpose here is to review the basic features of this connectional system to
discern what is common to all areas of cerebral cortex. We will show that horizontal
connections intrinsic to any area are a fundamental feature of the organization of the
cerebral cortex. We will emphasise primary visual cortex (V1) in cat and monkey, which
has been studied most intensively, but we will also discuss differences among other species
and cortical areas to elucidate common threads.
2. BASIC FEATURES OF INTRINSIC CONNECTIONS
Here we review the evidence that the basic features of intrinsic connections are similar in
nearly all cortical areas of all mammalian species, differing only quantitatively rather than
qualitatively. Later in this chapter we will provide a rationale for both the similarities and
the exceptions.
2.1. Degeneration Studies

The first studies to describe in any detail, the pattern of connections within any single area of
the cerebral cortex used degeneration techniques. Small lesions (surgical slits or electrolytic
lesions) are placed in the cortex, and the pattern of degenerating fibres is revealed, thus
indicating the extent and topography of projections from the lesion site. Studies in monkey
visual (Fisken et al., 1973), motor (Gatter et al., 1978), and somatosensory cortex (Shanks
et al., 1978; Vogt and Pandya, 1978), showed that degenerating fibres spread from the
lesion site in a similar pattern in all these areas, suggesting that intrinsic cortical connections
were essentially similar in all cortical areas. Degenerating fibres were most dense within
a millimeter of the lesion, but could span several millimeters of cortex. The density of

Intrinsic Connections of the Cortex 135
degenerating fibres appeared to fall off with distance from the lesion, but with no obvious
substructure to the label pattern. However, the total field of cortex containing degenerating
fibres was not always isotropic, for example forming an ellipse elongated in the anteroposterior
axis in the motor cortex.
The 1975 study of the primate striate cortex by Fisken et al. was the most detailed of
these, and provided a fuller characterisation of intrinsic connections using the degeneration
technique. They quantified and described more carefully a number of features of intrinsic
connections not previously appreciated. One important factor was the use of smaller
lesions. Fibre degeneration superficial and deep to lesions occupied roughly the same
width as the lesion itself, confirming the columnar vertical projection focus suggested by
the earlier Golgi studies. However, these authors also reported a laminar specificity in the
spread of fibre degeneration: horizontal connections were essentially absent from layer 4,
were light in layers 1 and 6, and were most prominent in lower layers 2/3 (extending into
the uppermost part of layer 4) and layer 5. They quantified the distribution of degenerating
fibres across the cortex, and found that fibre density was greatest within 1 mm of the lesion
(roughly 60% of all such fibres they observed). The distribution appeared essentially
continuous, not obviously patchy or clustered, but there were suggestions that the degen-
erating fibres and terminals furthest from the lesion did segregate into discrete zones
(see their Figure 25). They noted intense fine degeneration within 200 µm of the lesion
site, and moderate terminal and fibre degeneration up to 2–3 mm from the lesion. Fisken
et al. also examined the ultrastructure of degenerating profiles to determine their synaptic
relationships, and noted that the vast majority (roughly 90%) of degenerating synaptic
profiles were of the asymmetric type, and were found contacting spines; degenerating
symmetric profiles were much rarer (although as they conceded it is more difficult to be
confident that a degenerating axon terminal has a symmetric type of synapse), contacting both
pyramidal and nonpyramidal cells. Intrinsic connecions thus appeared to be a connectional
network primarily linking excitatory pyramidal neurones to one another.
In these early papers, what was emphasised was the limited extent of degeneration
following a small lesion to the cerebral cortex, but they are significant for revealing that
horizontal connections within a given cortical area could indeed span several millimeters
(up to 5–6 mm total in visual cortex), thus linking neurones at physically offset points across
the cortex. However, certain of their conclusions can be questioned on a number of technical
grounds. Firstly, the lack of any obvious fine structure to the projections might simply
have reflected the uncertainty of whether it was fibres originating from the lesion site rather
than simply those passing through the lesion site that were being revealed. A number of the
lesions in these studies were quite large, which also might have masked specificity of
intrinsic connections. Secondly, it is possible that with different survival times, the maximum
extent of intrinsic connections would have been greater. Nonetheless, they are noteworthy
for being the first demonstration of the phenomenon of long-range clustered projections
parallel to the cortical surface. Indeed, it is remarkable, how many details of the organisa-
tion of intrinsic connections were elucidated using these relatively crude techniques.
2.2. Tracer Studies

Such connections are now best revealed using sensitive neuronal tracers: The tracers of
choice are the Choleratoxin B subunit (CTb), biocytin, Phaseolus vulgaris leucoagglutinin
(PHA-L), biotinylated dextran amine (BDA), fluorescent latex beads, fluorescent dyes such

as Nuclear Yellow or Fast Blue, or wheatgerm agglutinin-horseradish peroxidase (WGA-HRP).

These offer several advantages over older techniques, namely the ability to label tissue
without destroying it, the ability to confine injections to small cortical loci (of the same size
as single columns), and their much superior sensitivity. The first studies to use tracer tech-
niques to study the organisation of intrinsic connections were those of Rockland and Lund
(1982, 1983) and Gilbert and Wiesel (1979, 1983). These studies in cat, tree shrew, and
monkey demonstrated, in agreement with the earlier degeneration studies, that a restricted
locus of cells (or even single cells) in the visual cortex furnished projections spanning
several millimeters across the cortical surface, and whose axons ramified into terminal
clusters at discrete locations (clusters or patches) across the cortex. They showed that these
projections were most prominent in the supragranular layers, though they were to be found
in layer 5 as well, and terminal clusters were typically 200–300 µm in diameter, spaced
350–600 µm apart (depending on the particular species). Furthermore, the coexistence of
labeled cells and terminals in these patches suggested that such connections were reciprocal.
Figures 7.1 and 7.2 illustrate the basic features of this connectional lattice. The photo-
micrograph of Figure 7.1 shows the result of making a small injection of the neuronal
Figure 7.1. Biocytin injection site in macaque primary visual cortex (injection core marked by arrowhead).
Note strong vertical focus of projections, and fibers running laterally past area marked by open arrow before
ending in a terminal cluster. Scale bar = 200 µm.

tracer biocytin into the primary visual cortex of a macaque monkey. Biocytin is primarily
taken up by cell bodies at the injection site and transported in the orthograde direction to
label axon terminals; in a less reliable way it also labels cells in the retrograde direction
which project to the injection site. In this injection, centred in the lower part of layer 3, the
dense columnar focus of projections is apparent. Interlaminar axonal projections pass
through layer 4 with very little lateral spread, whereas laterally-spreading fibres are appar-
ent in the superficial layers as well as in layer 5. Axons in the superficial layers run paral-
lel to the cortical surface without ramifying until they reach a particular target location on
the cortex where they ramify profusely, making terminal endings. A similar terminal
cluster in layer 5 is found in vertical registration with that in the superficial layers. This
photomicrograph shows only the terminal cluster nearest to the injection site. For a better
appreciation of the overall connectivity pattern, the cortex is best viewed tangential to the
pial surface.
The photomicrographs of Figure 7.2 show the results of making injections of the bidir-
ectional tracer CTb (choleratoxin B subunit). This tangential section was taken through
layer 3, shown in Figure 7.1 to be the site of the most prominent lateral projections. The low-
power view shows the injection site surrounded by a number of distinct clusters of ortho-
gradely-labeled terminals and retrogradely-labeled cell bodies. Apart from the projection
field being patchy and spanning several millimeters of cortex, this figure also shows two
other common features of such projections, their anisotropy and their reciprocity. Generally,
connections do not spread out isotropically, instead extending further from the injection
site along one axis. Furthermore, most (although not all) patches of terminals are also
coincident with retrogradely labeled cells. This indicates that pyramidal cells in the super-
ficial layers generally receive inputs from those cortical columns to which they project.
The higher power view of Figure 7.2B shows clusters of labeled pyramidal neurones
coincident with labeled fibres making terminal endings.
Although these patches of terminal label are round or oval-shaped in most cortical
areas, in certain areas the pattern can be rather different. The photomicrograph in Figure
7.3 shows the biocytin-labeled projections in macaque dorsolateral prefrontal cortex. Here
it is very common to find elongated terminal fields; one such zone is indicated by the
arrow below the injection core.
Figure 7.4 shows complete serial reconstructions of the pattern of orthograde terminal
label resulting from six separate injections into macaque cerebral cortex (four into primary
visual cortex, one into extrastriate visual area V4, and one into dorsolateral prefrontal
cortex). These complete reconstructions again illustrate the clustered and anisotropic
spread of projections from a given cortical locus, and indicate how different the overall
topography can be in different cortical regions.
We now review the basic features of intrinsic circuits, and their generality to all cortical
areas.
2.2.1. Laminar specificity

In nearly all species and cortical areas where it has been studied, the same laminar pattern
of lateral projections is observed, i.e. laterally spreading fibers are most prominent in
layers 2/3, but are also to be observed in layer 5, with a more columnar focus within
layer 4. This has been found to be true in primary visual cortex of cat (Gilbert and Wiesel,
1979; Martin and Whitteridge, 1984; Gabbot et al., 1987), macaque (Blasdel et al., 1985;

Figure 7.2. Tangential view through layer 3 of intrinsic connections in macaque primary visual cortex
surrounding a choleratoxin B subunit (CTb) injection site (indicated by dashed white circle) viewed at low (A)
and high (B) magnifications. Note the clusters of labeled cells and terminals surrounding the injection site. Scale
bar = 500 µm (A), 100 µm (B).

Figure 7.3. Tangential view through layer 3 of a biocytin injection site in macaque dorsolateral prefrontal
cortex (area 9). The arrow indicates an elongated band of terminal label. Scale bar = 200 µm. From Levitt et al.
(1993) with permission.
Yoshioka et al., 1994), tree shrew (Rockland and Lund, 1982; Rockland et al., 1982;
Fitzpatrick, 1996) and human (Burkhalter and Bernardo, 1989). The same laminar pattern is
found in macaque extrastriate visual areas V2 and V4 (Yoshioka et al., 1992; Kritzer et al.,
1992; Levitt et al., 1994), dorsolateral prefrontal cortex (areas 9/46: Levitt et al., 1993;
Kritzer and Goldman-Rakic, 1995), motor cortex (area 4: Gatter et al., 1978), and somato-
sensory cortex (areas 3,1,2: Shanks et al., 1978; Vogt and Pandya, 1978), as well as cat
area 18 (Matsubara et al., 1985, 1987) and primary auditory cortex (Wallace et al., 1991).
It may be assumed that the band of horizontal projections in layer 2/3 corresponds to the
outer band of Baillarger seen in myelin preparations; however, since this tier of projec-
tions can extend down into layer 4, it may be innaccurate to simply define the outer band
of Baillarger as being in a single cortical layer.
We note however, that while lateral projections do appear generally more restricted
within layer 4, they are by no means entirely absent. Studies in primate V1 have clearly
shown projections running laterally within upper layer 4 for several hundred microns
(Anderson et al., 1993; Yoshioka et al., 1994). Projections within the uppermost part of
layer 4, layer 4B (the stria of Gennari), can extend several millimeters. The significance of
anatomical links between columns in the deeper portion of layer 4, in the topographically-
precise input layers, remains to be determined.
2.2.2. Total extent and anisotropy of the intrinsic connectional field

A fundamental characteristic of intrinsic projections is that projections from a particular
cortical locus span several millimeters of cortex, a distance much greater than the dendritic

Figure 7.4. Reconstructions from serial sections of orthograde label pattern in the superficial layers resulting
from biocytin injections into different areas of macaque cerebral cortex. All panels show views tangential to the
cortical surface. (A) Four separate small iontophoretic injections (fringed stippled circles) into visual area V1.
Terminal patches (solid black zones) are plotted in relation to cytochrome-oxidase blobs (dashed lines).
(B) Large pressure injection into visual area V4. Surrounding the large injection core (central solid black region)
is a dense radiation of labeled fibres (hatched area) and terminal patches (solid black zones). (C) Injection (central
open circle) into prefrontal area 9. Zones of highest terminal density are black, lower density stippled. Scale
bars = 1 mm. From Lund et al. (1993) with permission.
field of neurones at the source site, or of the arborisation of thalamic inputs to that site.
This system therefore serves to interconnect neurones in cortical columns at different parts
of the map within that cortical area. Generally, these patchy projection patterns stop at
areal borders, forming a different pattern in the adjoining area. However, in certain cases
an injection close to an areal border leads to a continuous pattern of labelling, which spreads
into the neighbouring area, ignoring the border. This was observed in the dorsolateral
prefrontal cortex, in which projections spread freely across the border between areas 9 and
46 (Levitt et al., 1993; Pucak et al., 1996).
The shape of the overall projection field was not well characterised in most of the earlier
studies, but it is now appreciated that projections from a given cortical locus are not

isotropic, but instead can extend further along one axis than another, not necessarily
always anteroposterior. Matsubara et al. (1987) noted, in agreement with the earlier
degeneration studies, that the majority of projections in cat area 18 were found within
1.4 mm of an injection site, but that isolated terminal projections could be found up to
3.4 mm away. The overall spread of label was elongated in the anteroposterior axis.
Kisvárday and Eysel (1992) came to the identical conclusion in cat area 17; they showed
that the overall connectional field resulting from a restricted tracer injection measured
6.5 mm in the anteroposterior axis, but only 3.5 mm mediolaterally. Similar measures
have been reported in macaque V1: both Malach et al. (1993) and Yoshioka et al. (1994)
reported that connections spread up to 3.7 mm from an injection site, but with the greatest
spread of label running mediolaterally, parallel to the V1/V2 border and orthogonal to the
ocular dominance bands. In the visual cortex, the general rule thus seems to be that the
long axis parallels the area 17/18 border. The ratio of long to short axes of the connec-
tional field ranged from 1.5–1.8, closely similar to the ratios observed in cat visual
cortex, and such anisotropy is also observed in primate motor cortex (Gatter et al., 1978)
and cat auditory cortex (Matsubara and Phillips, 1988; Wallace et al., 1991). It appears
that this anisotropic spread of connections relates to the sensory map in each area, such
that (at least in cat and monkey) a roughly circular region of the sensory periphery is
monosynaptically linked—although this preferential spread of connections along one
axis is also found in the prefontal cortex (Levitt et al., 1993; Kritzer and Goldman-Rakic,
1995), where there is as yet no clearly-defined “map”.
Optical imaging studies are also consistent with the anatomically-measured extent and
anisotropy of connections. Albowitz and Kuhnt (1993) found the spread of voltage-sensitive
dye signals following focal electrical stimulation of guinea pig V1 slices to span up to
3.6 mm, while Grinvald et al. (1994) recording from macaque V1 found that visual stimuli
smaller than 1 degree in diameter activated a region of cortex 2.7 × 1.5 mm in extent, with
the greatest spread of activation parallel to the V1 border.
Recent electrophysiological results seem to confirm the dramatic extent of these mono-
synaptic projections. Bringuier et al. (1999) used in vivo intracellular recording methods in
cat V1 to demonstrate the spatial extent over which subthreshold depolarising inputs could
be measured in response to visual stimuli. They found that this subthreshold field could
span up to 20 degrees in total diameter (i.e. 10 degrees from the receptive field center). Such
extensive inputs could well result from some other input spanning more space, such as
cortical feedback. However, this study showed a linear relationship between response
latency and stimulus distance from the receptive field; this property does not seem required
of a feedback circuit, and the most parsimonious explanation is therefore that intrinsic
connections mediate these responses. The increasing response latencies with distance from
the receptive field reflect simply the conduction delays across the cortex.
Similarly, in slices of macaque dorsolateral prefrontal cortex, Gonzalez-Burgos et al.
(2000) were able to measure postsynaptic currents evoked by low-intensity electrical
stimulation at sites located up to 2200 microns lateral to the recorded cell. This is surely an
underestimate of the total extent over which monosynaptic activation can be activated, since
the slice preparation would almost certainly cut off some more remote inputs, or would
cause the investigators to miss weaker ones. But what was also noteworthy was that mono-
synaptic EPSCs mediated by even the most remote connections had amplitudes similar or
larger than short-distance connections, suggesting that the excitatory input provided by
long-distance intrinsic connections is still quite robust.

2.2.3. Patchiness of the projection

Perhaps the defining characteristic of intrinsic connections is their discontinuous nature.
As illustrated in Figures 7.1–7.4 above, projections to or from a given cortical locus are
not uniformly arranged around that point, simply decreasing in density with increasing
distance from the injection. Rather, neurones providing input to one cortical location are
clustered into a number of discrete patches. Similarly, projections from a given point
terminate in a number of discrete zones of high terminal density interspersed with zones
relatively free of such terminals. Clustered intrinsic connections have been found in every
cortical area examined. A by-no-means-exhaustive list would include:
• V1 of cat (Gilbert and Wiesel, 1983; Gabbott et al., 1987; Kisvárday and Eysel,
1992), macaque and squirrel monkey (Rockland and Lund, 1983; Malach et al., 1993;
Yoshioka et al., 1996), human (Burkhalter and Bernardo, 1989), ferret (Durack and
Katz, 1996; Ruthazer and Stryker, 1996), tree shrew (Rockland et al., 1982; Rockland
and Lund, 1982; Fitzpatrick, 1996), and rodent (though of a somewhat different
topography: Burkhalter, 1989)
• extrastriate visual areas V2, V4, 7a, and MT in macaque (Rockland, 1985b; Yoshioka
et al., 1992; Amir et al., 1993; Levitt et al., 1994), squirrel monkey (Malach et al.,
1994), owl monkey (Malach et al., 1997), and cat (Matsubara et al., 1985, 1987)
• dorsolateral prefrontal cortex of the macaque (Levitt et al., 1993; Kritzer and
Goldman-Rakic, 1995)
• somatosensory cortex of macaque (Jones et al., 1978; Juliano et al., 1990), cat
(Juliano et al., 1989; Sonty and Juliano, 1997), and ferret (Juliano et al., 1996)
• auditory cortex of cat (Matsubara and Phillips, 1988; Ojima et al., 1991; Wallace
et al., 1991) and ferret (Wallace and Bajwa, 1991)
• motor cortex of macaque (Jones et al., 1978; Huntley and Jones, 1991) and cat
(Landry et al., 1980; Keller, 1993)
As noted above, these clusters of cells or terminals are generally round or oval-shaped,
but in some cortical areas they can be distinctly elongated and stripe-like as in macaque
prefrontal cortex or owl monkey middle temporal area. We have preliminary evidence that
intrinsic projections within layer 4B of macaque V1 also take this form (Asi et al., 1996).
Depending on the cortical area, these elongated patches average between 130–300 µm in
width of the narrower axis, with a mean centre-to-centre spacing of 230–500 µm.
Why are the projections patchy? Rockland and Lund (1982) suggested early on that some
cells in V1 might simply not furnish local connections. Mitchison and Crick (1982) sug-
gested instead that the patchiness or substructure of the projection might reflect constraints
on connections set by the functional map, for example that cells would be linked if their
receptive fields had similar orientation preferences and were linked across the map in a par-
ticular direction in visual space. Although it is still not entirely clear, why intrinsic projec-
tions take the form they do, this suggestion is probably closer to the truth, since wherever
one injects a tracer into a given cortical area, a patchy projection obtains—although patch size
increases somewhat, the cortex never “fills in” despite the largest tracer injections (Yoshioka
et al., 1992; Amir et al., 1993). We will return to this point toward the end of this chapter.
Each terminal patch appears rather stereotyped, apparently containing roughly the same
number of synaptic boutons. For example, Kisvárday and Eysel (1992) showed in cat V1

that each labeled layer-3 pyramidal cell furnishes axonal projections to 4–8 (or more)
patches, each patch containing 80 synaptic boutons from that cell. Furthermore, these
connections are mainly but not strictly reciprocal, i.e. one can find labeled cells not
surrounded by labeled terminals and vice versa (Boyd and Matsubara, 1991; Amir et al.,
1993; Levitt et al., 1994).
2.2.4. Cells furnishing and targeted by intrinsic connections

In both visual and prefrontal cortices, long-range intrinsic projections are furnished primarily
by spine-bearing pyramidal neurones, and their main synaptic targets are the distal dendritic
spines of other pyramidal neurones (Fisken et al., 1975; Rockland, 1985a; Kisvárday et al.,
1986; LeVay, 1988; McGuire et al., 1991; Melchitzky et al., 1998). In agreement with the
earlier degeneration study of Fisken et al., the later studies using tracers to label projections
to be examined ultrastructurally, showed that cells labeled from an injection received syn-
aptic contacts characteristic of excitatory neurones, while synaptic terminals labeled from
an injection mainly made asymmetric contacts onto dendritic spines. It is clear, however,
from all these studies that a significant proportion of synaptic contacts are also made onto
inhibitory neurones. In fact, one study (Keller and Asanuma, 1993) analysed the synaptic
contacts made by 3 cells in cat motor cortex onto other cells within a few hundred microns,
and found one to have most of its outputs to a non-pyramidal (presumably inhibitory) neurone.
In cat and monkey V1, roughly 20% of the supragranular cells are GABAergic (Gabbott
and Somogyi, 1986; Hendry et al., 1987). As pointed out by McGuire et al. (1991), this is
essentially the same proportion as intrinsic pyramidal neurone synapses made onto smooth
stellate cell dendrites, while the proportion of synapses onto spines (75%) is nearly the
same as the overall proportion of axospinous type I synapses in the neuropil of the supra-
granular layers (Beaulieu and Colonnier, 1985). Thus projections seem not to target
particular cell types, but rather to end randomly on all neurones within a terminal cluster
according to the proportion of neurones found there. However, the possibility still remains
that synapses at different distances from the cell of origin may exhibit different patterns of
specificity or topography. This remains to be resolved.
Furthermore, it should not be assumed that inhibitory GABAergic neurones do not
furnish such long-range projections. While their contribution may be less prominent, they
do exist. In cat V1, Albus et al. (1991) retrogradely labeled cells, and then labeled the tissue
for GABA, searching for cells that were GABAergic as well as lateral projectors. In layer 3,
70% of the double-labeled cells were found within 1 mm of the injection site, while 30%
were found between 1–2.5 mm from the injection. Double-labeled cells were not so clearly
clustered as excitatory neurones; nearly half the labeled cells were scattered and not in
clusters. The relatively small numbers and different topography implied that the rule for
inhibitory circuits may be different from that for excitatory ones. Kisvárday et al. (1993)
used a more sensitive method to show that large GABAergic basket cells in layer 3 can pro-
ject across a region of cortex spanning 2.3 × 2.2 mm, while those in layer 5 can span
3.8 × 1.7 mm. They further showed that these basket cells contact both other basket cells as
well as pyramidal neurones, thus mediating direct inhibition and facilitation via disinhibi-
tion. Kritzer et al. (1992) similarly demonstrated the existence of GABAergic projections
parallel to the cortical surface in macaque visual areas V1, V2, and V4. The functional
relevance of these differences in proportion and spatial extent between excitatory and
inhibitory projections within cortex remains enigmatic, particularly since some GABAergic

neurones in cat and monkey send axons into the white matter (e.g. Lund and Wu, 1997),
and in rat at least can also project to other cortical areas (McDonald and Burkhalter, 1993).
2.2.5. Strength and reliability of intrinsic synapses

While a great deal is now known about the cellular biophysics of pyramid to pyramid or
pyramid to interneurone interactions (for recent reviews see Thomson and Deuchars,
1994, 1997), this is once again a topic whose complete treatment is beyond the scope of
this chapter. Broadly, studies on the synaptic physiology of intrinsic connections have
shown that synaptic effects due to activation of intrinsic connections can be detected. In
visual or prefrontal areas, electrical stimulation of intracortical pathways leads to mono-
synaptic EPSPs in recorded neurones, while stronger stimuli evoke disynaptic IPSPs as
well (Hirsch and Gilbert, 1991; Gonzalez-Burgos et al., 2000). Generally, the EPSPs are
weak, but if the cell is depolarised, these EPSPs evoked by stimulation of intrinsic pathways
can elicit spike activity in the cell. Furthermore, the strength of these intrinsic pathways
is not fixed, but appears to be modifiable by the parameters of the stimulation (Hirsch and
Gilbert, 1993). Bringuier et al. (1999) have directly shown in cat V1 that visual stimuli
beyond the classical receptive field (which translates to activation of locations lateral to
the recording site) can also evoke subthreshold excitation in cortical neurones.
A fundamental feature of corticocortical synapses between neurones laterally offset from
one another is that they are weaker and more variable than thalamocortical synapses or
corticocortical synapses between neurones vertically offset within the same column (Strat-
ford et al., 1996; Yoshimura et al., 2000). Nonetheless it appears that intracortical inputs
may well be strong enough to provide a large part of the excitation to cortical cells in vivo.
Yoshimura et al. also showed that EPSPs evoked by simultaneous activation of vertical
and lateral inputs summated linearly when the postsynaptic cell was at resting potential,
but supralinearly when the postsynaptic cell was depolarised. Furthermore, it appears that
there are important differences in how presynaptic activity in pyramidal neurones leads to
activity in other pyramids versus interneurones. For example, the synaptic interaction
between two pyramidal cells offset laterally from one another shows paired-pulse depres-
sion—the postsynaptic response to the second of two presynaptic spikes is usually smaller
(Thomson and West, 1993; Yoshimura et al., 2000), while pyramid–pyramid interactions
between cells vertically aligned sometimes showed synaptic facilitation in this paradigm.
However, pyramid-interneurone connections showed pronounced paired-pulse facilitation
(Thomson et al., 1993). Thus, with increasing activity levels (and greater depolarisation),
pyramidal neurones seem more sensitive to inputs from within that same cortical area, and
preferentially able to recruit inhibitory interneurone activity; this may serve as one form of
gain-control in the cortex.
2.2.6. Conduction velocity

To understand the function of these laterally-running cortical circuits, it is also of course
critical to know how quickly signals can propagate along them. There have now been
a number of measurements of the conduction velocity of intrinsic connections, using
a number of different techniques and in different species. These studies all come to a
remarkably consistent conclusion, that horizontal projections within cortex conduct
impulse activity rather slowly, in the range of 0.1–0.2 m/s.

Tanifuji et al. (1994) used optical imaging techniques in slice preparations of young rat
primary visual cortex. They measured the rate at which the optical signals from voltage-
sensitive dyes spread across the cortex following electrical stimulation of the white matter.
They reported a mean value in layer 3 of 0.06 ± 0.035 m/s (whereas activity spread in the
columnar axis with a mean velocity of 0.2 ± 0.051 m/s). Albowitz and Kuhnt (1993) used
closely similar techniques in slice preparations of guinea pig V1; their data suggest similar
conduction velocity values of 0.06–0.12 m/s. Grinvald et al. (1994) measured the rate of
spread of voltage-sensitive dye signals in in vivo recordings from macaque monkey V1,
using visual stimuli to activate cortical neurones; they showed activity to spread across the
cortex at a rate of 0.09–0.25 m/s.
Gonzalez-Burgos et al. (2000) used whole-cell patch clamp recordings in an in vitro
preparation of macaque dorsolateral prefrontal cortex. They measured postsynaptic cur-
rents evoked by low-intensity electrical stimulation of the slice. Their data indicate that
monosynaptic connections within cortex are conveyed across the cortex at approximately
0.14 m/s. Bringuier et al. (1999) also recorded intracellularly from single neurones in cat
V1 in vivo, but using visual stimuli to drive activity. They found that the latency of
subthreshold responses increased with distance of the stimulus in the visual field from
the receptive field centre. They arrived at an apparent conduction velocity of 0.1–0.2 m/s
(since they did not directly measure conduction velocity, but simply translated spread of
activation in degrees of visual angle to millimeters of cortex, by assuming a cortical
magnification factor of 1 mm/deg which is reasonable for cat V1 at the retinal eccentri-
cities they studied).
These studies thus all concur that for neuronal signals to propagate 2–3 mm laterally
from a given site on the cortex requires roughly 10–20 ms. The calibre of the axons
furnishing these projections has been reported as roughly 1–3 µm in diameter in cat V1
(Kisvárday and Eysel, 1992), and probably finer in the primate. Such slow conduction
velocities ought to result from axons having diameters much finer than this (Miller, 1996;
Swadlow, 2000), so it remains puzzling why propagation rate is so slow in these axons.
2.2.7. Specificity of connections and functional correlates

A central question about intrinsic circuits concerns their functional role and what
governs their detailed topography—why do projections from a given site on the cortex
target particular other sites? While many questions remain, it is now clear that intrinsic
connections relate to the functional map in each area, such that neurones with similar
functional properties are connected. This relationship has been most thoroughly examined
in the visual cortex, with respect to the arrangement of orientation columns. Rockland
et al. (1982) first attempted to correlate the local transport of HRP with the 2-deoxyglucose
pattern evoked in tree shrew V1 in response to a single stimulus orientation. They noted the
similarities of the patterns, but reached no firm conclusion. Gilbert and Wiesel (1989) sub-
sequently repeated the experiment in cat V1, injecting fluorescent latex beads into a site of
known orientation preference. They found that retrograde label was essentially confined to
regions with the same orientation preference as the injection site. Cross-correlation studies
are consistent with this result: one observes peaks in cross-correlograms between single
cells separated by up to 4.2 mm if they have similar orientation preferences and spatially
overlapping receptive fields (Ts’o et al., 1986; Schwarz and Bolz, 1991). This relationship
has been subsequently confirmed and refined by combining transport of more sensitive

tracers (such as biocytin) with more sophisticated optical imaging techniques that allow
the entire orientation map to be specified. In this way, it has been shown that intrinsic con-
nections are closely related to the orientation map; regions of similar orientation prefer-
ence are more likely to be interconnected. This has now been demonstrated repeatedly in
V1 of cat (Kisvárday et al., 1994, 1997), macaque (Malach et al., 1993), and tree shrew
(Bosking et al., 1997), as well as V2 of squirrel monkey (Malach et al., 1994) and cat
(Matsubara et al., 1985, 1987) although Matsubara et al.’s studies suggested connections
were predominantly between regions preferring orthogonal orientations.
This relationship needs to be qualified, however. Firstly, as Mitchison and Crick (1982)
originally proposed, axis in space is also relevant: neurones seem more likely to be inter-
connected not only if they share orientation preference, but also if the connections’ axis in
visual space is parallel to the preferred orientation so as to link cells responding to an
elongated contour (Fitzpatrick, 1996; Bosking et al., 1997; Schmidt et al., 1997). Indeed,
electrophysiological experiments have revealed response facilitation among such axially
aligned cells with similar orientation preferences (Nelson and Frost, 1985). Secondly, the
correlation is not perfect. Very local connections (within about 500 µm of an injection site)
are much less specific than long-range ones (Malach et al., 1993; Bosking et al., 1997;
Toth et al., 1997), and even the more specific long-range projections seem to show some
proportion (as much as a third) which do not obey the strict “like-to-like” connectivity
rule. Of course connections within visual cortex may also relate to some other parameter
such as ocular dominance (Katz et al., 1989; Yoshioka et al., 1996), or the distribution of
particular afferent populations (indicated by staining for cytochrome oxidase: Livingstone
and Hubel, 1984; Yoshioka et al., 1996; Yabuta and Callaway, 1998).
In other cortical areas, the rule that intrinsic connections link neurones with similar
properties is not so obvious. In auditory cortex, intrinsic connections appear to relate to
the tonotopic map. Several studies have shown that projections in A1 spread preferentially
in the dorsoventral direction (Reale et al., 1983; Matsubara and Phillips, 1988; Wallace et al.,
1991; Wallace and Bajwa, 1991) such that cells with similar characteristic frequencies are
interconnected. However, as in visual cortex, connections do not conform strictly to this
rule, but instead seem to relate to the particular axis of isofrequency contours (patches being
more numerous dorsoposterior to an injection when isofrequency contours run obliquely);
many cells or terminal patches are found in regions of higher frequency. Furthermore,
connections do not bear any strict relationship to the distribution of binaural properties.
The rules governing connections in sensorimotor cortex are similarly equivocal. Juliano
et al. (1990) made tracer injections into physiologically-characterized sites in monkey
somatosensory cortex and then compared the resulting label pattern with the activity
pattern in cortex elicited by that stimulus (revealed by 2DG mapping). They found ortho-
and retrograde label largely confined to regions with similar response properties. However,
in the barrel field of mouse somatosensory cortex, cells representing one entire whisker
row are more strongly interconnected than cells representing different whisker rows
(Bernardo et al., 1990), and in motor cortex, connections can link neurones representing
portions of the forelimb representation as far apart as the digits and the shoulder (Huntley
and Jones, 1991). Despite the patchy organization, there are still no clear functional
correlates of local connections in somatosensory cortex. The main problem (as noted by
Juliano et al., 1990) is that there still does not seem to be a generally-agreed set of para-
meters to define a functional column in somatosensory cortex. Correlating anatomical
circuits with functional units therefore remains problematic.

Any parameter which is mapped regularly across the cortex could potentially serve to
guide intrinsic connections (for example direction preference in certain visual cortices:
Swindale et al., 1987; Weliky et al., 1996). A key role of future work will be to determine
which parameter, or constellation of parameters, determines the detailed spatial topo-
graphy of intrinsic connections in cortex.
The functional role of these connections remains unclear. In visual cortex, for example,
it is now known that neurones’ orientation and direction selectivity is much more dynamic
than previously appreciated: Tuning can be dramatically modified by the presence outside
the classical receptive field of stimuli which themselves evoke no responses (Gilbert and
Wiesel, 1990; Sillito et al., 1995; Levitt and Lund, 1997). Intrinsic circuits could serve to
link neurones at different retinotopic locations, thus mediating these effects, which could
be one manifestation of a response gain-control mechanism in the cortex. Theoretical
studies also suggest that intrinsic cortical circuits may play an important role in generation
of basic receptive field properties, either by amplifying weakly-tuned thalamic afferent
signals (Somers et al., 1995) or even by generating strong tuning themselves (Adorjan et al.,
1999). It has also been suggested that patchy connections may serve to maximize the
amount of information available to neurones in the cortex (Malach, 1994) or to enhance
signal transmission through cortex (Schüz, 1994). These issues are far from decided, but
their resolution will clearly be necessary for a complete understanding of how circuitry leads
to the elaboration of functional properties in cerebral cortex.
2.2.8. Development and refinement

Why do intrinsic cortical circuits develop this way? Although a detailed discussion of this
issue is beyond the scope of this chapter, it is now clear that, like all other neural circuits,
intrinsic cortical conections are not generated with their final topography or synaptic
complement, but require some period to refine. Although certain basic features of intrinsic
connections may be present even prenatally (as in macaque monkey V1: Callaway,
1998a), these generally require a substantial postnatal maturation or refinement period
before intrinsic circuits attain their final adult state, in terms of both inter- and intra-
laminar specificity (as in cat or ferret visual or somatosensory cortices: Callaway and
Katz, 1990; Juliano et al., 1996; Durack and Katz, 1996; Ruthazer and Stryker, 1996;
Sonty and Juliano, 1997). There is also a huge overproduction and refinement of synapses
during the early postnatal period (Lund et al., 1977; Rakic et al., 1986); thus, the synapses
in these connectional fields must undergo modification. It is also now known that these
circuits’ final layout may be affected by experience; for example, Löwel and Singer (1992)
showed that rearing cats with a strabismus modified the pattern of local connections in
V1 relative to ocular dominance columns, essentially restricting intracortical projections
to sites of similar ocular dominance as the injection site. Such data indicate that refinement
is not merely an activity-dependent process, but that patterned activity serves to sculpt
these circuits. While many details of the mechanisms by which this refinement take place
remain to be elucidated, laterally-running glial processes can be found in cortex
(Albus and Luebke, 1992; Juliano et al., 1996) which could serve as a scaffolding for
developing lateral projections just as they seem to do for vertical interlaminar projections.
While much is now known, a great deal remains to be explored regarding the relative
timing of maturation of intrinsic circuits, extrinsic circuits, and receptive field properties
of cortical neurones.

3. SCALING OF INTRINSIC CONNECTIONS
We conclude by attempting to answer what may determine the size of these patchy con-
nectional fields in all cortical areas. One striking feature of this connectional lattice is that
not only does the extent of cortex monosynaptically linked vary across areas, but the size
of individual cell or terminal clusters also varies across areas. Admittedly, both of these
characteristics might simply reflect differences in the maps in each area (e.g. cortical
magnification factor) such that a “column” of cells necessary to process a given amount of
visual field, or range of auditory frequencies, or extent on the body surface has a different
dimension. Nevertheless, it has also been noted that there are differences among areas
in the dimensions of the pyramidal neurones that furnish and receive these projections.
However, different studies use different techniques, tracers, survival times, injection sizes,
etc. to study intrinsic connections, making comparisons problematic. Both Lund et al.
(1993) and Amir et al. (1993) used consistent techniques to label intrinsic connections in
several different areas of macaque cerebral cortex. They both reported consistent dif-
ferences among areas in mean patch size and spacing: in visual cortex, these parameters,
as well as the total extent of the labeled field, increase the further one gets from primary
visual cortex into higher visual areas. While this may also be true of other sensory systems,
these have been almost exclusively studied in primary sensory areas. Proof awaits future
experiments in non-visual cortical areas. Based on these criteria, primary motor cortex
appears to resemble a higher sensory area, rather than a primary one, but this also remains
equivocal. Lund et al. then compared these dimensions to those of the basal dendritic
fields of pyramidal neurones in each area, labeled by Golgi impregnations. Figure 7.5
illustrates their finding, that the size and spacing of this connectional latice is highly
correlated with the dimensions of basal dendritic fields, despite a two-fold range in den-
dritic field size.
This finding is consistent with an earlier statement by Landry et al. (1980), who noted
in motor cortex that “The tangential expansion of this local field corresponds to that of the
basal dendritic domain of pyramidal tract neurons.” What seems to be the case is not that
the dimensions are strictly commensurate in any area, but rather that they covary across
areas.
There is an interesting consequence of this arrangement. Rockel et al. (1980)
reported that in all cortical areas (with the exception of V1), a column of fixed diameter
contains the same number of cells through the depth of the cortex. Furthermore, bouton
density (interbouton interval) along intrinsically-projecting axons remains essentially
constant all along these processes (Amir et al., 1993; Yabuta and Callaway, 1998); axons
simply ramify more in the vicinity of a terminal cluster. Thus, areas whose connectional
lattice has larger terminal patches are likely to be contacting an increasing number of
cells. The functional relevance of this is completely unknown. Perhaps intrinsic
“modulatory” influences are greater in cortical areas with larger patch sizes for this
reason.
Finally, we note that patchy intrinsic connections are also found not only in the
eutherian species described above, but also in V1 of the marsupial quokka (Tyler
et al., 1998). Given that patchy superficial layer pyramidal neurone projections are
found in species that diverged at least 135 million years ago, this further suggests that
this is indeed an ancient and fundamental architectural element in the organisation of
neocortex.

Figure 7.5. (A) Plot of the relationship between mean diameter of terminal patches or narrow dimension of
terminal bands (measured from orthograde transport of biocytin) and the mean diameter of basal dendritic field
of individual layer 2/3 pyramidal neurones in different areas of macaque cerebral cortex (measured from Golgi
impregnations). Error bars indicate standard deviations. Each point summarises data from one cortical area:
macaque visual (V1, V2, V4), somatosensory (3b, 1, 2), motor (4), and prefrontal (9, 46: indicated by asterisk)
areas. Also included are measures from V1 of cat and tree shrew (TS). (B) Plot of the relationship between mean
terminal patch centre-to-centre spacing and mean terminal patch diameter. Dashed lines are regression lines
through macaque data (A: r = 0.779, p < 0.05; B: r = 0.989, p < 0.001). From Lund et al. (1993) with permission.

4. SUMMARY
We have reviewed a number of anatomical and physiological studies, and highlight the
following as central defining characteristics of intrinsic connections in all areas of cerebral
cortex:
• Connections parallel to the cortical surface are most prominent in layers 2/3 and 5,
less so in layers 1, 4, 6.
• Connections spread several millimeters from any locus, spanning a total extent up to
6–8 mm depending on the particular cortical area. The overall connectional field is
generally anisotropic, forming an elongated ellipse, apparently related to the functional
map in that area.
• These projections are patchy rather than diffuse; axons ramify and make extensive
terminal branches only at particular discrete locations across the cortex (the exception
being very local, within 300 µm of a projection source, where connections are dense
and relatively uniform). Intervening regions are relatively free of terminals. Single
neurones’ projections are also patchy. Individual terminal patches are generally
round/oval, but these can appear more stripelike (for example in visual area MT,
prefrontal areas 9/46, or in layer 4B of V1).
• The size and spacing of these terminal patches differs between areas, and seems to
scale to the diameter of the basal dendritic field of pyramidal neurones in each area.
• These connections are furnished by excitatory, spine-bearing pyramidal neurones
(though in some species like the cat, certain inhibitory cells—the basket cells—can
make medium-range inhibitory connections). Intrinsic connections contact mainly
other excitatory neurones, but a smaller proportion of synaptic contacts are on to
inhibitory interneurones.
• Synapses made by horizontal intrinsic corticocortical connections evoke EPSPs and
IPSPs of smaller magnitude and lower reliability than thalamocortical or vertical
(columnar) corticocortical synapses. However their strength appears to be dynamically
modifiable.
• The average conduction velocity of intrinsic connections is 0.1–0.2 m/s.
• Intrinsic connections relate to the functional map in each area linking mainly, but not
exclusively, neurones of like kind.
REFERENCES
Adorjan, P., Levitt, J.B., Lund, J.S. and Obermayer, K. (1999) A model for the intracortical origin of orientation
preference and tuning in macaque striate cortex. Visual Neuroscience, 16, 303–318.
Albowitz, B. and Kuhnt, U. (1993) The contribution of intracortical connections to horizontal spread of activity
in the neocortex as revealed by voltage sensitive dyes and a fast optical recording method. European
Albus, K., Wahle, P., Luebke, J. and Matute, C. (1991) The contribution of GABAergic neurons to horizontal
intrinsic connections in upper layers of the cat’s striate cortex. Experimental Brain Research, 85, 235–239.
Albus, K. and Luebke, J. (1992) Widespread lateral processes of glial cells in the immature striate cortex of the
cat. Developmental Brain Research, 68, 278–281.
Anderson, J.C., Martin, K.A.C. and Whitteridge, D. (1993) Form, function, and intracortical projections of
neurons in the striate cortex of the monkey Macacus nemestrinus. Cerebral Cortex, 3, 412–420.

Asi, H., Levitt, J.B. and Lund, J.S. (1996) In macaque V1 lateral connections in layer 4B have a different
topography than in layers 2/3. Society for Neuroscience Abstracts, 22, 1608.
Beaulieu, C. and Colonnier, M. (1985) A laminar analysis of the number of round-symmetrical and flat-
asymmetrical synapses on spines, dendritic trunks, and cell bodies in area 17 of the cat. Journal of Compara-
tive Neurology, 231, 180–189.
Bernardo, K.L., McCasland, J.S. and Woolsey, T.A. (1990) Local axonal trajectories in mouse barrel cortex.
Blasdel, G.G., Lund, J.S. and Fitzpatrick, D. (1985) Intrinsic connections of macaque striate cortex: axonal
projections of cells outside lamina 4C. Journal of Neuroscience, 5, 3350–3369.
Bosking, W.H., Zhang, Y., Schofield, B. and Fitzpatrick, D. (1997) Orientation selectivity and the arrangement
of horizontal connections in tree shrew striate cortex. Journal of Neuroscience, 17, 2112–2127.
Boyd, J. and Matsubara, J. (1991) Intrinsic connection in cat visual cortex: a combined anterograde and retro-
grade tracing study. Brain Research, 560, 207–215.
Bringuier, V., Chavane, F., Glaeser, L. and Fregnac, Y. (1999) Horizontal propagation of visual activity in the
synaptic integration field of area 17 neurons. Science (Washington), 283, 695–699.
Burkhalter, A. (1989) Intrinsic connections of rat primary visual cortex: laminar organization of axonal
projections. Journal of Comparative Neurology, 279, 171–186.
Burkhalter, A. and Bernardo, K.L. (1989) Organization of corticocortical connections in human visual cortex.
Proceedings of the National Academy of Science U.S.A., 86, 1071–1075.
Cajal, S. Ramón y (1909–1911) Histologie du système nerveux de l’homme et des vertébrés. Vols. I and II. Paris,
Maloine. Reprinted Madrid, Consejo Superior de Investigaciones Cientificas, 1952.
Cajal, S. Ramón y (1922) Studien über die Sehrinde der Katze. Journal of Psychology and Neurology, 29, 161–181.
Callaway, E.M. (1998a) Prenatal development of layer-specific local circuits in primary visual cortex of the
macaque monkey. Journal of Neuroscience, 18, 1505–1527.
Callaway, E.M. (1998b) Local circuits in primary visual cortex of the macaque monkey. Annual Review of Neuro-
science, 21, 47–74.
Callaway, E.M. and Katz, L.C. (1990) Emergence and refinement of clustered horizontal connections in cat striate
Durack, J.C. and Katz, L.C. (1996) Development of horizontal projections in layer 2/3 of ferret visual cortex.
Fisken, R.A., Garey, L.J. and Powell, T.P.S. (1973) Patterns of degeneration after intrinsic lesions of the visual
cortex (area 17) of the monkey. Brain Research, 4, 369–374.
Fisken, R.A., Garey, L.J. and Powell, T.P.S. (1975) The intrinsic, association and commissural connections of area
17 of the visual cortex. Philosophical Transactions of the Royal Society of London, Series B, 272, 487–536.
Fitzpatrick, D. (1996) The functional organization of local circuits in visual cortex: insights from the study of
tree shrew striate cortex. Cerebral Cortex, 6, 329–341.
Gabbott, P.L.A. and Somogyi, P. (1986) Quantitative distribution of GABA-immunoreactive neurons in the
visual cortex (area 17) of the cat. Experimental Brain Research, 61, 323–331.
Gabbott, P.L.A., Martin, K.A.C. and Whitteridge, D. (1987) Connections between pyramidal neurons in layer 5
of the cat visual cortex (area 17). Journal of Comparative Neurology, 259, 364–381.
Gatter, K.C., Sloper, J.J. and Powell, T.P.S. (1978) The intrinsic connections of the cortex of area 4 of the monkey.
Brain, 101, 513–541.
Gilbert, C.D. and Wiesel, T.N. (1979) Morphology and intracortical projections of functionally characterised
neurones in the cat visual cortex. Nature (London), 280, 120–125.
science, 3, 1116–1133.
Gilbert, C.D. and Wiesel, T.N. (1989) Columnar specificity of intrinsic horizontal and corticocortical connections
in cat visual cortex. Journal of Neuroscience, 9, 2432–2442.
Gilbert, C.D. and Wiesel, T.N. (1990) The influence of contextual stimuli on the orientation selectivity of cells
in primary visual cortex of the cat. Vision Research, 30, 1689–1701.
Gonzalez-Burgos, G., Barrionuevo, G. and Lewis, D.A. (2000) Horizontal synaptic connections in monkey
prefrontal cortex: an in vitro electrophysiological study. Cerebral Cortex, 10, 82–92.
Grinvald, A., Lieke, E.E., Frostig, R.D. and Hildesheim, R. (1994) Cortical point-spread function and long-range
lateral interactions revealed by real-time optical imaging of macaque monkey primary visual cortex.
Hendry, S.H.C., Schwark, H.D., Jones, E.G. and Jan, J. (1987) Numbers and proportions of GABA-
immunoreactive neurons in different areas of monkey cerebral cortex. Journal of Neuroscience, 7,
1503–1519.
Hirsch, J.A. and Gilbert, C.D. (1991) Synaptic physiology of horizontal connections in the cat’s visual cortex.
Hirsch, J.A. and Gilbert, C.D. (1993) Long-term changes in synaptic strength along specific intrinsic pathways in
the cat visual cortex. Journal of Physiology (London), 461, 247–262.

Hubel, D.H. and Wiesel, T.N. (1962) Receptive fields, binocular interaction, and functional architecture in the
cat’s visual cortex. Journal of Physiology (London), 160, 106–154.
Huntley, G.W. and Jones, E.G. (1991) Relationship of intrinsic connections to forelimb movement representa-
tions in monkey motor cortex: a correlative anatomic and physiological study. Journal of Neurophysiology,
66, 390–413.
Jones, E.G., Coulter, J.D. and Hendry, S.H.C. (1978) Intracortical connections of architectonic fields in the
somatic sensory, motor and parietal cortex of monkeys. Journal of Comparative Neurology, 181, 291–348.
Juliano, S.L., Whitsel, B.L., Tommerdahl, M. and Cheema, S.S. (1989) Determinants of patchy metabolic label-
ing in the somatosensory cortex of cats: a possible role for intrinsic inhibitory circuitry. Journal of Neuro-
science, 9, 1–12.
Juliano, S.L., Friedman, D.P. and Eslin, D.E. (1990) Corticocortical connections predict patches of stimulus-
evoked metabolic activity in monkey somatosensory cortex. Journal of Comparative Neurology, 298,
23–39.
Juliano, S.L., Palmer, S.L., Sonty, R.V., Noctor, S. and Hill, G.F. (1996) Development of local connections in
ferret somatosensory cortex. Journal of Comparative Neurology, 374, 259–277.
Katz, L.C., Gilbert, C.D. and Wiesel, T.N. (1989) Local circuits and ocular dominance columns in monkey
Keller, A. (1993) Patterns of intrinsic connections between motor representation zones in the cat motor cortex.
NeuroReport, 4, 515–518.
Keller, A. and Asanuma, H. (1993) Synaptic relationships involving local axon collaterals of pyramidal neurons
in the cat motor cortex. Journal of Comparative Neurology, 336, 229–242.
Kisvárday, Z.F., Martin, K.A.C., Freund, T.F., Magloczky, Z., Whitteridge, D. and Somogyi, P. (1986) Syn-
aptic targets of HRP-filled layer III pyramidal cells in the cat striate cortex. Experimental Brain Research,
64, 541–552.
Kisvárday, Z.F. and Eysel, U.T. (1992) Cellular organization of reciprocal patchy networks in layer III of cat
visual cortex (area 17). Neuroscience, 46, 275–286.
Kisvárday, Z.F., Beaulieu, C. and Eysel, U.T. (1993) Network of GABAergic large basket cells in cat visual
corex (area 18): implications for lateral disinhibition. Journal of Comparative Neurology, 327, 398–415.
Kisvárday, Z.F., Kim, D.S., Eysel, U.T. and Bonhoeffer, T. (1994) Relationship between lateral inhibitory con-
nections and the topography of the orientation map in cat visual cortex. European Journal of Neuroscience,
6, 1619–1632.
Kisvárday, Z.F., Toth, E., Rausch, M. and Eysel, U.T. (1997) Orientation-specific relationship between populations
of excitatory and inhibitory lateral connections in the visual cortex of the cat. Cerebral Cortex, 7, 605–618.
Kritzer, M.F., Cowey, A. and Somogyi, P. (1992) Patterns of inter- and intralaminar GABAergic connections
distinguish striate (V1) and extrastriate (V2, V4) visual cortices and their functionally specialized sub-
divisions in the rhesus monkey. Journal of Neuroscience, 12, 4545–4564.
Kritzer, M.F. and Goldman-Rakic, P.S. (1995) Intrinsic circuit organization of the major layers and sub-
layers of the dorsolateral prefrontal cortex in the rhesus monkey. Journal of Comparative Neurology,
359, 131–143.
Landry, P., Labelle, A. and Deschenes, M. (1980) Intracortical distribution of axonal collaterals of pyramidal
tract cells in the cat motor cortex. Brain Research, 191, 327–336.
LeVay, S. (1988) Patchy intrinsic projections in visual cortex, area 18, of the cat: morphological and immuno-
cytochemical evidence for an excitatory function. Journal of Comparative Neurology, 269, 265–274.
Levitt, J.B., Lewis, D.A., Yoshioka, T. and Lund, J.S. (1993) Topography of pyramidal neuron intrinsic connections
in macaque monkey prefrontal cortex (areas 9 and 46). Journal of Comparative Neurology, 338, 360–376.
Levitt, J.B., Yoshioka, T. and Lund, J.S. (1994) Intrinsic cortical connections in macaque visual area V2: evidence
for interaction between different functional streams. Journal of Comparative Neurology, 342, 551–570.
Levitt, J.B. and Lund, J.S. (1997) Contrast dependence of contextual effects in primate visual cortex. Nature
(London), 387, 73–76.
Livingstone, M.S. and Hubel, D.H. (1984) Specificity of intrinsic connections in primate primary visual cortex.
Löwel, S. and Singer, W. (1992) Selection of intrinsic horizontal connections in the visual cortex by correlated
neuronal activity. Science (Washington), 255, 209–212.
Lund, J.S. (1973) Organization of neurons in the visual cortex, area 17 of the monkey (Macaca mulatta). Journal
Lund, J.S. and Boothe, R.G. (1975) Interlaminar connections and pyramidal neuron organization in the visual
cortex, area 17, of the macaque monkey. Journal of Comparative Neurology, 159, 305–334.
Lund, J.S., Boothe, R.G. and Lund, R.D. (1977) Development of neurons in the visual cortex of the monkey
(Macaca nemestrina): A Golgi study from fetal day 127 to postnatal maturity. Journal of Comparative
Neurology, 176, 149–188.

Lund, J.S. and Wu, C.Q. (1997) Local circuit neurons of macaque monkey striate cortex: IV. Neurons of laminae
1–3A. Journal of Comparative Neurology, 384, 109–126.
Malach, R., Amir, Y., Harel, M. and Grinvald, A. (1993) Relationship between intrinsic connections and func-
tional architecture revealed by optical imaging and in vivo targeted biocytin injections in primate striate
cortex. Proceedings of the National Academy of Science U.S.A., 90, 10469–10473.
Malach, R. (1994) Cortical columns as devices for maximizing neuronal diversity. Trends in Neuroscience, 17,
101–104.
Malach, R., Tootell, R.B. and Malonek, D. (1994) Relationship between orientation domains, cytochrome
oxidase stripes, and intrinsic horizontal connections in squirrel monkey area V2. Cerebral Cortex, 4, 151–165.
Malach, R., Schirman, T.D., Harel, M., Tootell, R.B. and Malonek, D. (1997) Organization of intrinsic connec-
tions in owl monkey area MT. Cerebral Cortex, 7, 386–393.
Martin, K.A.C. and Whitteridge, D. (1984) Form, function, and intracortical projections of spiny neurones in the
Matsubara, J., Cynader, M., Swindale, N.V. and Stryker, M.P. (1985) Intrinsic projections within visual cortex:
evidence for orientation-specific local connections. Proceedings of the National Academy of Science U.S.A.,
82, 935–939.
Matsubara, J.A., Cynader, M.S. and Swindale, N.V. (1987) Anatomical properties and physiological correlates
of the intrinsic connections in cat area 18. Journal of Neuroscience, 7, 1428–1446.
Matsubara, J.A. and Phillips, D.P. (1988) Intracortical connections and their physiological correlates in the
primary auditory cortex (AI) of the cat. Journal of Comparative Neurology, 268, 38–48.
McDonald, C.T. and Burkhalter, A. (1993) Organization of long-range inhibitory connections within rat visual
McGuire, B.A., Gilbert, C.D., Rivlin, P.K. and Wiesel, T.N. (1991) Targets of horizontal connections in
macaque primary visual cortex. Journal of Comparative Neurology, 305, 370–392.
Melchitzky, D.S., Sesack, S.R., Pucak, M.L. and Lewis, D.A. (1998) Synaptic targets of pyramidal neurons
providing intrinsic horizontal connections in monkey prefrontal cortex. Journal of Comparative Neurology,
390, 211–224.
Miller, R. (1996) Axonal conduction time and human cerebral laterality. Harwood Academic Publishers,
Amsterdam.
Mitchison, G. and Crick, F. (1982) Long axons within the striate cortex: their distribution, orientation, and
patterns of connection. Proceedings of the National Academy of Science, U.S.A., 79, 3661–3665.
Mountcastle, V.B. (1957) Modality and topographic properties of single neurons of cat’s somatic sensory cortex.
Journal of Neurophysiology, 20, 408–434.
Nelson, J.I. and Frost, B.J. (1985) Intracortical facilitation among co-oriented, co-axially aligned simple cells in
cat striate cortex. Experimental Brain Research, 61, 54–61.
Ojima, H., Honda, C.N. and Jones, E.G. (1991) Patterns of axon collateralization of identified supragranular
pyramidal neurons in the cat auditory cortex. Cerebral Cortex, 1, 80–94.
O’Leary, J.L. (1941) Structure of the area striata of the cat. Journal of Comparative Neurology, 75, 131–161.
Powell, T.P.S. and Mountcastle, V.B. (1959) Some aspects of the functional organization of the cortex of the
postcentral gyrus of the monkey: a correlation of findings obtained in a single unit analysis with cyto-
architecture. Johns Hopkins Hospital Bulletin, 105, 133–162.
Pucak, M.L., Levitt, J.B., Lund, J.S. and Lewis, D.A. (1996) Patterns of intrinsic and associational circuitry in
monkey prefrontal cortex. Journal of Comparative Neurology, 376, 614–630.
Rakic, P., Bourgeois, J.-P., Zecevic, N. and Goldman-Rakic, P.S. (1986) Concurrent overproduction of synapses
in diverse regions of the primate cerebral cortex. Science (Washington), 232, 232–235.
Reale, R.A., Brugge, J.F. and Feng, J.Z. (1983) Geometry and orientation of neuronal processes in cat primary
auditory cortex (AI) related to characteristic-frequency maps. Proceedings of the National Academy of
Sciences U.S.A., 80, 5449–5453.
Rockel, A.J., Hiorns, R.W. and Powell, T.P.S. (1980) The basic uniformity in structure of the neocortex. Brain,
103, 221–244.
Rockland, K.S. and Lund, J.S. (1982) Widespread periodic intrinsic connections in the tree shrew visual cortex.
Science (Washington), 215, 1532–1534.
Rockland, K.S., Lund, J.S. and Humphrey, A.L. (1982) Anatomical binding of intrinsic connections in striate
cortex of tree shrews (Tupaia glis). Journal of Comparative Neurology, 209, 41–58.
Rockland, K.S. and Lund, J.S. (1983) Intrinsic laminar lattice connections in primate visual cortex. Journal of
Rockland, K.S. (1985a) Intrinsically projecting pyramidal neurons of monkey striate cortex: an EM-HRP study.
Society for Neuroscience Abstracts, 11, 17.
Rockland, K.S. (1985b) A reticular pattern of intrinsic connections in primate area V2 (area 18). Journal of
Ruthazer, E.S. and Stryker, M.P. (1996) The role of activity in the development of long-range horizontal
connections in area 17 of the ferret. Journal of Neuroscience, 16, 7253–7269.

Schmidt, K.E., Goebel, R., Löwel, S. and Singer, W. (1997) The perceptual grouping criterion of colinearity
is reflected by anisotropies of connections in the primary visual cortex. European Journal of Neuroscience,
9, 1083–1089.
Schüz, A. (1994) Patchiness as a means to get a message across. Trends in Neuroscience, 17, 365.
Schwarz, C. and Bolz, J. (1991) Functional specificty of a long-range horizontal connections in cat visual cortex:
a cross-correlation study. Journal of Neuroscience, 11, 2995–3007.
Shanks, M.F., Pearson, R.C.A. and Powell, T.P.S. (1978) The intrinsic connections of the primary somatic
sensory cortex of the monkey. Proceedings of the Royal Society of London (Biology), 200, 95–101.
Sillito, A.M., Grieve, K.L., Jones, H.E., Cudeiro, J. and Davis, J. (1995) Visual cortical mechanisms detecting
focal orientation discontinuities. Nature (London), 378, 492–496.
Somers, D.C., Nelson, S.B. and Sur, M. (1995) An emergent model of orientation selectivity in cat visual cortical
simple cells. Journal of Neuroscience, 15, 5448–5465.
Sonty, R.V. and Juliano, S.L. (1997) Development of intrinsic connections in cat somatosensory cortex. Journal
Stratford, K.J., Tarczy-Hornoch, K., Martin, K.A.C., Bannister, N.J. and Jack, J.J.B. (1996) Excitatory synaptic
inputs to spiny stellate cells in cat visual cortex. Nature (London), 382, 258–261.
Swadlow, H.A. (2000) Information flow along neocortical axons. In: R. Miller (ed.), Time and the brain.
Conceptual Advances in Brain Research series, Harwood Academic Publishers, Amsterdam, pp. 131–150.
Swindale, N.V., Matsubara, J.A. and Cynader, M.S. (1987) Surface organization of orientation and direction
selectivity in cat area 18. Journal of Neuroscience, 7, 1414–1427.
Tanifuji, M., Sugiyama, T. and Murase, K. (1994) Horizontal propagation of excitation in rat visual cortical
slices revealed by optical imaging. Science (Washington), 266, 1057–1059.
Thomson, A.M. and West, D.C. (1993) Fluctuations in pyramid-pyramid excitatory postsynaptic potentials
modified by presynaptic firing pattern and postsynaptic membrane potential using paired intracellular
recording in rat neocortex. Neuroscience, 54, 329–346.
Thomson, A.M., Deuchars, J. and West, D.C. (1993) Single axon excitatory postsynaptic potentials in neocor-
tical interneurons exhibit pronounced paired pulse facilitation. Neuroscience, 54, 347–360.
Thomson, A.M. and Deuchars, J. (1994) Temporal and spatial properties of local circuits in neocortex. Trends in
Thomson, A.M. and Deuchars, J. (1997) Synaptic interactions in neocortical local circuits: dual intracellular
recordings in vitro. Cerebral Cortex, 7, 510–522.
Toth, L.J., Kim, D.-S., Rao, S.C. and Sur, M. (1997) Integration of local inputs in visual cortex. Cerebral Cortex,
7, 703–710.
Ts’o, D.Y., Gilbert, C.D. and Wiesel, T.N. (1986) Relationships between horizontal interactions and func-
tional architecture in cat striate cortex as revealed by cross-correlation analysis. Journal of Neuroscience,
6, 1160–1170.
Tyler, C.J., Dunlop, S.A., Lund, R.D., Harman, A.M., Dann, J.F., Beazley, L.D. and Lund, J.S. (1998) Anatom-
ical comparison of the macaque and marsupial visual cortex: common features that may reflect retention of
essential cortical elements. Journal of Comparative Neurology, 400, 449–468.
Vogt, B.A. and Pandya, D.N. (1978) Cortico-cortical connections of somatic sensory cortex (areas 3,1, and 2) in
the rhesus monkey. Journal of Comparative Neurology, 177, 179–191.
Wallace, M.N. and Bajwa, S. (1991) Patchy intrinsic connections of the ferret primary auditory cortex. Neuro-
report, 2, 417–420.
Wallace, M.N., Kitzes, L.M. and Jones, E.G. (1991) Intrinsic inter- and intralaminar connections and their rela-
tionship to the tonotopic map in cat primary auditory cortex. Experimental Brain Research, 86, 527–544.
Weliky, M., Bosking, W.H. and Fitzpatrick, D. (1996) A systematic map of direction preference in primary
visual cortex. Nature (London), 379, 725–728.
Yabuta, N.H. and Callaway, E.M. (1998) Cytochrome-oxidase blobs and intrinsic horizontal connections of
layer 2/3 pyramidal neurons in primate V1. Visual Neuroscience, 15, 1007–1027.
Yoshimura, Y., Sato, H., Imamura, K. and Watanabe, Y. (2000) Properties of horizontal and vertical inputs to
pyramidal cells in the superficial layers of the cat visual cortex. Journal of Neuroscience, 20, 1931–1940.
Yoshioka, T., Levitt, J.B. and Lund, J.S. (1992) Intrinsic lattice connections of macaque monkey visual cortical
area V4. Journal of Neuroscience, 12, 2785–2802.
Yoshioka, T., Levitt, J.B. and Lund, J.S. (1994) Independence and merger of thalamocortical channels within
macaque monkey primary visual cortex: anatomy of interlaminar projections. Visual Neuroscience, 11,
467–489.
Yoshioka, T., Blasdel, G.G., Levitt, J.B. and Lund, J.S. (1996) Relation between patterns of intrinsic lateral
connectivity, ocular dominance, and cytochrome oxidase-reactive regions in macaque monkey striate
cortex. Cerebral Cortex, 6, 297–310.

8 Thalamic Systems and the Diversity of Cortical Areas
Catherine G. Cusick
Department of Structural and Cellular Biology and Neuroscience Program,
Tulane University School of Medicine, 1430 Tulane Avenue, New Orleans
Louisiana, U.S.A. 70112
e-mail: cusick@tulane.edu
The thalamus of higher primates may be organized into sets of nuclei that reflect the organization of
different cortical sensory systems (visual, somatosensory, auditory) into complex distributed hier-
archies. For the visual system, there appear to be three levels of thalamic nuclear organization: a first
level nucleus corresponding to the lateral geniculate nucleus, a second level nucleus corresponding
to the inferior pulvinar complex, and a third, high order group comprising the dorsal pulvinar nuclei.
Thalamic nuclei can be defined using criteria similar to those used for cortical areas, especially
chemoarchitecture, cortical interconnections, and topographic maps. First level, primary sensory
thalamic nuclei share many common chemoarchitectonic features. However, other thalamocortical
and corticothalamic sensory systems appear not to be so extensively developed as those for vision.
For example, whereas the somatosensory and auditory thalamocortical relations support first level and
high order nuclei, the suggested second level nuclei are not so easily identified nor neurochemically
differentiated as the visual pulvinar.
KEYWORDS: auditory system, lateral geniculate nucleus, primate, somatosensory system, thalamic
nuclei, visual system
1. INTRODUCTION
Research in the last several decades on the organization of the cerebral cortex into separ-
ate highly interconnected functional areas and systems has revealed principles that have
parallels in thalamic organization. The focus of this chapter will be to highlight aspects of
thalamic organization that provide for partially segregated, or parallel processing, within
systems, and to explore some hints that thalamic nuclei respect but also may coordinate
activity within the well-described cortical hierarchies. The emphasis will be on visual
thalamocortical systems in macaque monkeys, as the interconnections and distributed
parallel processing are best understood within this model. It should be noted that while the
macaque visual thalamus has become one of the better understood models, the hierarchical
organization exhibited within this system may not be typical of most others, whose organ-
ization may be more strongly parallel. It should be noted that much work on microcircuitry
and physiology has been conducted on other thalamocortical systems and species (see
Jones, 1985; Sherman and Koch, 1998; McCormick and Huguenard, 1992; Ramcharan et al.,
2000). Given the many specializations in the organization of cerebral cortex of different
mammals, reliance on a single well-studied model is attractive for extracting general rules
155
156 Catherine G. Cusick
of thalamocortical organization. To maintain an awareness of possible limits of this

approach, other systems will be contrasted with the macaque visual thalamocortical model
to illustrate the range of different organizational strategies that may exist.
1.1. Overview of Cortical Organization

Cortical systems are traditionally regarded as broadly divided into sensory, motor and
“high order” (supramodal) systems. Cortical systems consist of functional units termed
areas that are interconnected with each other in specific patterns that are both distributed
and hierarchical (for reviews see Felleman and Van Essen, 1991; Kaas and Garraghty,
1991; Casagrande, 1994). In monkeys, the first visual area (V1) receives the bulk of the
projections from the primary relay nucleus, the lateral geniculate (LGN). V1 sends inputs
to the immediately adjacent cortex, V2, and to several other retinotopically organized
areas, e.g. the middle temporal (MT), dorsomedial (DM), and medial (M) visual areas,
that may be considered to be at the same level in the cortical hierarchy. Several further
processing levels exist and there are in turn multiple areas at each level (Felleman and Van
Essen, 1991). Interconnections among cortical areas diverge into different dorsal and
ventral streams that are directed toward the posterior parietal cortex and inferior temporal
lobe and are concerned with visually guided actions and perception, respectively
(Ungerleider and Mishkin, 1982; Milner and Goodale, 1993). The headwaters, as it were,
of these partially segregated streams are already identifiable within separate layers and
modules of V1 and different modules of V2. To an extent, the dorsal and ventral cortical
streams are traceable back to the LGN (Casagrande, 1994; Van Essen and Gallant, 1994).
1.2. Changing Views of Thalamocortical Relations

Views on how the thalamic inputs relate to the conceptualization of cortical organization
into systems, hierarchies, areas, modules and layers have changed dramatically in recent
years. The early studies on thalamocortical relations developed the concept that each
cortical area received information from one thalamic nucleus (Rose and Woolsey, 1949).
It soon became clear that individual thalamic nuclei might target more than one cortical
area within the same functional system. This was particularly true for the cat visual
system, in which the lateral geniculate nucleus was recognized to project to areas 17, 18,
and 19 of visual cortex (Garey et al., 1968; Heath and Jones, 1970). In monkeys, however,
V1 was thought to be the only target of projections from the LGN, and extrastriate visual
cortex was thought to receive inputs only from the pulvinar, although the pulvinar was
known to project to most of the individual areas of extrastriate cortex (Garey and Powell,
1971).
Tracing studies utilizing more sensitive methods suggest that many if not most cortical
areas in monkeys receive projections from more that one thalamic nucleus (Friedman and
Murray, 1986). Thus, the extrastriate areas MT and V4 receive major projections from the
pulvinar, but also minor inputs from the LGN (Cusick et al., 1993; Stepniewska et al.,
1999). Taking a further example from the somatosensory system, the primary area 3b
receives connections from the major relay nucleus, the ventroposterior (VP) complex,
including its component divisions ventroposterior proper (VP), ventroposterior inferior
(VPI) and ventroposterior superior (VPS) (see below), as well as from the adjacent
anterior pulvinar nucleus (Cusick and Gould, 1990).

Thalamic Systems and Cortical Areas 157
Since interconnections between thalamus and cortex do not follow a “single nucleus to
single area” rule, or vice versa, the question is raised about the determinants of specificity
of thalamocortical relations. To gain an approximate answer to this question will require
cellular and developmental approaches, and obviously also a full understanding of the
degree of specificity exhibited in the adult brain.
2. CONCEPTS OF THALAMIC ORGANIZATION
2.1. Organization of Thalamic Nuclei into Systems

Within the thalamus, there have been considered to be two broad sorts of cortical projec-
tion nuclei, specific (terminating in layers III/IV) and modulatory (terminating in layers I
and VI; (Herkenham, 1980; Sherman and Guillery, 1998). Modulatory nuclei have topo-
graphic projections with widespread regions of the cerebral hemisphere, and in many
cases target the basal ganglia, via thalamostriatal projections, as well as the cerebral cortex
(Jones, 1985). The specific nuclei form groups that are analogous to cortical systems. The
traditional anatomical parcellation of the thalamus into medial, lateral, and anterior
groups, provided by the position of the internal medullary lamina, is meaningful in terms
of the topography of thalamocortical relations, but the “thalamic system” spoken of here
is the set of nuclei that interconnects with particular cortical systems. Thalamic systems
are proposed to consist of primary nuclei, second order nuclei and “high order” nuclei.
Although somewhat similar to the schema of Guillery (1995), the term high order is
reserved here for non-modality specific (“supramodal”) nuclei, and the non-primary thalamic
nucleus within a modality is termed second order.
Cortical systems exhibit several hierarchically organized levels of processing, with esti-
mates for the visual system at over 10 levels (Felleman and Van Essen, 1991). By contrast,
thalamic systems can be identified at this time as having perhaps 2–3 levels within each
system. For the visual system, the primary nucleus is ofcourse the LGN, the second level
nucleus, the inferior pulvinar complex, and high order nuclei would include other portions
of the pulvinar, especially the dorsal pulvinar complex.
2.2. What is a Thalamic Nucleus?

Because the collections of neurones within the thalamus sometimes have complicated
three-dimensional shapes, the definitions of nuclei have been difficult, and there are many
examples in which different nomenclatures have been proposed in different species for
apparently homologous structures. Similar definitions of thalamic nuclei can be developed
as have long existed for cortical areas (Van Essen and Zeki, 1978; Kaas, 1990; Krubitzer,
1995). First, single cortical areas are characterized by uniform cyto- and myeloarchitec-
ture. Incorporating recent evidence that modularity may exist within a cortical area, such
modular structure would be a feature of the cortical area (e.g. V2; Rosa and Krubitzer,
1999). Second, for “lower” or “early” cortical areas, there is a topographic representation
of the entire contralateral sensory receptor sheet or motor space. Third, there are topo-
graphic patterns of connections with the similar sets of cortical and subcortical structures,
allowing for differences in strength of connections for different portions of the functional

map. Fourth, cortical areas contain neurones which have similar physiological properties
and which may be clustered into a columnar organization that has a structural correlate to
architectonic modules (e.g. cytochrome oxidase domains in V2; myelin densities in area
3b of monkeys [Jain et al., 1998]).
A thalamic nucleus has a parallel cytoarchitectural integrity: similar types of neurones
are disposed in clusters or layers, and with particular relations to the encapsulating fibre
lamellae within the thalamus. For the primary nuclei, there are orderly topographic maps
of the sensory or motor space. The mapping rules for thalamic nuclei are not as straight-
forward as they are for cortex, onto which sensory representations are laid out as two-
dimensional sheets.
3. THE MACAQUE LGN AS A MODEL OF A PRIMARY RELAY NUCLEUS
3.1. Layers and Maps

Representations in thalamus are best understood for the LGN, in which the visual hemi-
field is mapped onto separate layers that can be modelled as a series of discs folded around
the hilus of the LGN (Kaas et al., 1978). In primates, these layers consist of three different
types, magno-, parvo- and koniocellular (M, P, and K) layers (Casagrande, 1994). There
are 2 sets of M layers, one for each eye, and the 2 sets of P layers are variably divided in
different primate species into further leaflets according to response types (Kaas et al.,
1978). Each LGN contains a map of the contralateral visual hemifield, with maps in the
different layers in visuotopic register, such that a point in visual space is represented by
a projection line that traverses all of the layers (Malpeli and Baker, 1975; Malpeli et al.,
1996; Erwin et al., 1999). Each of the separate layers of the LGN then contains a complete
map of the contralateral visual hemifield, and the aligned layers could be viewed in the
aggregate as several stacked maps of visual space, each concerning a different type of
information channel (M, P, or K) and receiving inputs from one of the two eyes.
The map of visual space in the LGN can also be demonstrated by its pattern of anatom-
ical connections with the striate cortex. A focal projection of a neuroanatomical tracer
reveals retrograde labeled neurones that follow a radial path or projection line through all
of the LGN laminae. Although broader in mediolateral extent, feedback axons from the
cortex target the same columnar region (Shatz and Rakic, 1981; Robson, 1983, 1984), and
the terminals form fine beaded projections termed “elongate” (Type I) (Robson, 1983;
Rockland, 1996).
A primary thalamic nucleus can thus be viewed as containing a single aggregate map
that contains multiple maps sorted by physiological response types. The subdivision of
thalamic nuclei into clusters or layers is analogous to the partitioning of cortical areas into
separate modules. The different LGN layers exhibit different laminar projections. The
principal (P and M) layers have projections to separate sublayers of IVc (Blasdel and
Lund, 1983; Blasdel and Fitzpatrick, 1984), and the koniocellular (K) layers, a term
including both the superficial and interlaminar cell groups, have projections that are
predominantly to the CO dense blobs within the supragranular layers, as well as to layer I
(Lachica and Casagrande, 1992; Hendry and Yoshioka, 1994). The principal layers have
been referred to as contributing a “lemniscal” (secure, fast conducting pathway from the
discrete receptive fields on the sensory periphery to the cortex) thalamocortical pathway

and the K layers as participating in a “non-lemniscal” (slower conducting projections with

broader receptive fields) pathway (Casagrande, 1994).
3.2. Layers and Neurochemical Signatures

The LGN layers have neurochemical signatures that have certain commonalities with other
specific thalamic nuclei (Table 8.1 and Figure 8.1). Principal layers have metabolically
highly active relay neurones that stain intensely with Nissl stains, and for cytochrome
oxidase. In addition to containing calcium binding proteins (e.g. Ca-calmodulin kinases)
that are ubiquitously found in neurones, the principal layer neurones contain parvalbumin,
a calcium binding protein whose functional significance in thalamus is not well known.
The K relay neurones also contain parvalbumin, but in addition they express calbindin,
a calcium binding protein not normally found in the principal layers, at least in normal
Table 8.1. Compartments (layers/modules) and projections of selected thalamic nuclei
“Compartment” Cat-301/WFA CO Parvalbumin Calbindin AChE Cortical

targets
LGN M layers + ++ ++ ++ V1 layer IV

P layers + + + layer IV
K layers ++ layer I
VP CO rich zones +(variable) ++ ++ Areas 3b/1
layer IV
CO poor zones ++ layer I
PI complex PIm + ++ ++ ++ MT
PIc + + ++ + DLc
Notes
Plus symbols indicate relative density of staining for the different neurochemical markers. Compartments
indicated in bold are thought to be lemniscal; plain text indicates non-lemniscal type of chemoarchitecture
and cortical projection. Not all cortical projections are listed for the individual nuclei. For further explanation
see text.
Figure 8.1. Neurochemical stains demarcate the laminar organization of the macaque LGN into principal
(magnocellular and parvocellular) and small-celled (koniocellular) layers. Immunocytochemistry for Cat-301
antibody and binding for the lectin Wisteria floribunda agglutinin (WFA) localize more densely to the magno-
cellular layers than the parvocellular layers. Calbindin immunocytochemistry stains cells in the thin koniocellular
layers.

adult monkeys (e.g. Gutierrez and Cusick, 1994). Thus, in calbindin-immunoreactive

sections of the LGN, there are numerous calbindin-positive fibres in the nucleus, many of
these apparently of retinal origin, but calbindin neurones only occupy the K layers adja-
cent to the optic tract and the interlaminar zones. The K neurones also express a high level
of the alpha-subunit of the CaMKII-kinase complex, a subunit not found in the P and M
layers (Hendry and Yoshioka, 1994).
The principal layers of the LGN also show different neurochemical staining patterns
from each other. The M layers stain somewhat more intensely for cytochrome oxidase and
Nissl than the P layers. The M neurones are decorated with specific extracellular matrix
components (ECM) that are relatively lacking in P neurones. These can be demonstrated
by binding of the Cat-301 antibody to protein cores of ECM (Hockfield et al., 1983;
Hendry et al., 1984), or by binding of the lectin Wisteria floribunda agglutinin (WFA) to
ECM sugar residues (Preuss et al., 1998). The magnocellular layers also show stronger
staining for non-phosphorylated neurofilament protein with the SMI-32 antibody (Gutierrez
et al., 1995), and for acetylcholinesterase (McDonald et al., 1993), than do the parvocellular
layers. Although at the present time there is not a specific marker with higher affinity for
the parvocellular layers, the three layer types obviously have distinct neurochemical
signatures.
3.3. Parallels in Neurochemical Organization of Primary Somatosensory,

Auditory, and Motor Thalamic Nuclei
Certain of the neurochemical characteristics of the LGN layers have parallels in other
thalamic nuclei. In the ventroposterior complex, one compartment is cytochrome oxidase-
poor and calbindin-rich, consisting of two zones termed the ventroposterior inferior
nucleus (VPI) and the CO-poor matrix. The CO-poor matrix occurs in patches in the more
dorsal part of the ventroposterior nucleus (VP) proper. The rest of VP proper (Figure 8.2), by
contrast, is a zone with dense clusters of cytochrome oxidase stain that often correspond to
important subregions of the somatosensory map, separated by thin fibre lamellae or septa
(Jones et al., 1986; Cusick and Gould, 1990; Rausell et al., 1991; Krubitzer and Kaas, 1992).
These CO patches are parvalbumin-rich and calbindin-poor, receive dense inputs from
the medial lemniscus, and project densely to layer IV of the first somatosensory cortex
area 3b. The CO-poor matrix on the other hand is the target of patchy foci of projections
from the spinothalamic tract and contains many smaller neurones that project to supra-
granular layers of the anterior parietal cortex (Rausell and Jones, 1991; Rausell et al.,
1992a,b). The “lemniscal” CO patches thus appear to be analogous to the LGN principal
layers, whereas the “non-lemniscal” CO-poor matrix, including VPI, appears to be analog-
ous to the LGN koniocellular layers. Interestingly, the positions of entry of the retinal
afferents to the LGN and the somatosensory inputs to the ventroposterior complex, the
S layers and VPI respectively, are both calbindin-rich compartments.
A further division of the VP patch compartment into two subregions is suggested by the
clustering of inputs from slowly adapting (SA) and rapidly adapting (RA) fibers (Dykes et al.,
1981; Kaas et al., 1984). These somatosensory responses resemble, in part, the sustained
and transient properties of the P and M layers (Sherman et al., 1976; Schiller and Malpeli,
1978), respectively. Interestingly, the pattern of binding for WFA and Cat-301 antibody
(Figure 8.2) suggests that the CO dense (“patch”) compartment may contain functional
subdivisions. Based on somatotopic mapping and receptor distributions, the most ventral

Figure 8.2. Binding for the lectin WFA in squirrel monkey thalamus delimits thalamic nuclei chemoarchitec-
tonically and shows their lamellar/modular organization as well. Rostral-to-caudal series of coronal sections.
Abbreviations of thalamic nuclei: AD, anterodorsal nucleus; AM, anteromedial; AV, anteroventral; CL, central
lateral; CM, centromedian; Hl, lateral habenula; Hm, medial habenula; LD, lateral dorsal; Lim, limitans; LP,
lateral posterior; mc, magnocellular layers; MD, mediodorsal; MDpc, parvocellular division of mediodorsal;
MDmc, magnocellular division of MD; MDmf, multiform division of MD; LGN, lateral geniculate; MGN,
medial geniculate; Pa, anterior pulvinar; Pc, paracentral; pc, parvocellular layers; Pf, parafascicular; PIc, central
division of the inferior pulvinar complex; PIl, lateral division of PI complex; PIm, medial division of PI
complex; PIp, posterior division of PI complex PL, lateral pulvinar; PM, medial pulvinar; Ret, reticular; VAmc,
magnocellular division of ventral anterior; VApc, parvocellular division of ventral anterior; VLa, anterior divi-
sion of ventrolateral nucleus; VLd, dorsal division of ventrolateral nucleus; VLp, posterior division of VL; VLx,
division “x” of VL; VM, ventral medial; VPI, ventroposterior inferior; VPL, ventroposterior lateral; VPM,
ventroposterior medial; VPMpc, parvocellular division of VPM; VPS, ventroposterior superior nucleus. From
Preuss et al., 1998.
portion of the “hand subnucleus” of VP would be expected to have a heavy representation

of SA inputs from the finger tip pads (Kaas et al., 1984), and this portion of VP is rela-
tively devoid of WFA and Cat-301 stain but has dense CO (Preuss et al., 1998). Thus, it is
possible that WFA and Cat-301 binding might discriminate RA and SA cell clusters in the
VP proper, in a similar manner to the pattern shown in M and P layers of the LGN. This
point deserves further investigation.

It should be noted, however, that other neurochemical patterns are found in VP which
differ strongly from those of the LGN: for example, whereas LGN principal layers are
AChE rich (McDonald et al., 1993), presumably related to the cholinergic input from the
parabrachial nucleus in the brainstem (Hu et al., 1989a,b), the patch compartment of VP is
AChE poor (Hirai and Jones, 1989).
Parallel parvalbumin- and calbindin-rich compartments have been identified in the
medial geniculate nucleus of monkeys (Hashikawa et al., 1991, 1995; Molinari et al.,
1995). The ventral nucleus contains dense clusters of parvalbumin neurones that project
onto layer IV of A1, which in turn has an especially dense plexus of parvalbumin fibers.
Calbindin neurones concentrate in the caudal part of the posterodorsal and in the magno-
cellular divisions of the medial geniculate complex, and project to layer I of A1. The ventral
division of the medial geniculate complex projects both to a core zone of three primary-
like koniocellular fields, and adjacent belt areas receive from all divisions of the medial
geniculate (Kaas and Hackett, 1998). Thus, similar to the somatosensory thalamocortical
system, there is evidently more parallelism in the projections from the primary auditory
nucleus than there is for the geniculocortical projections.
Figure 8.3. Chemoarchitecture of the ventroposterior complex revealed by cytochrome oxidase (CO) histo-
chemistry and WFA binding. Section A is adjacent to B, and C to D. Note that some regions of intense CO
stain correspond to zones with dense WFA binding, but the WFA binding pattern is more restricted, suggestive
of further compartmentation within the ventroposterior complex. From Preuss et al., 1998.

For motor thalamus, the pattern of neurochemical stain has recently been described in
several investigations and there is general agreement that the pattern of AChE stain is
highly correlated with presumably separate functional zones of cerebellar and pallidal
input (Hirai and Jones, 1993; Stepniewska et al., 1994a,b; Sakai et al., 2000). Similar
correlations may be provided by WFA binding, which shows marked differences between
different divisions of the motor thalamus (Figure 8.3; Preuss et al., 1998). In summary, it
has been suggested that lemniscal and non-lemniscal categories may extend to compart-
ments within each of the major thalamic relay nuclei (Jones, 1998a,b).
4. THE INFERIOR PULVINAR COMPLEX: A “SECOND LEVEL”

THALAMIC NUCLEUS
The organization of visual thalamic cell groups in the monkey pulvinar has been extensively
investigated recently by means of neurochemical markers correlated with patterns of
cortical connections. In primates, the expansion of extrastriate visual cortex and differenti-
ation into multiple areas and levels has been accompanied by the enormous enlargement
of the pulvinar nucleus of the thalamus. Striate cortex and “early” extrastriate visual areas
that are retinotopically organized, such as V2, V4, and MT, have retinotopically organized
connections within approximately the ventral half of the pulvinar, including portions of
the traditional medial, lateral, and inferior pulvinar nuclei (Figure 8.5). This retinotopi-
cally organized zone includes several distinct neurochemical units that also overlap the
traditional boundaries. In order to simplify the language used to describe these architec-
tonic-connectional units, it was proposed that the entire zone encompassed by the striate
cortex connections be termed the inferior pulvinar (PI) complex (Gutierrez et al., 1995;
Gutierrez and Cusick, 1997). Five neurochemically distinct regions have been distin-
guished within the PI complex, and similar patterns of chemoarchitecture have been
identified in several species of monkeys (Gray et al., 1999), in chimpanzees (Cola et al.,
2000), and in humans (Cola et al., 1999).
The organization of the PI complex into different neurochemical subdivisions has been
proposed to be analogous to the organization of the LGN into separate layers, with the PI
complex as a whole having the status of a thalamic nucleus (Gutierrez et al., 1995) (see
Figure 8.4). Thus, although the newly defined PI complex encompasses the traditional PI,
a small ventromedial corner of PM, and the adjacent parts of the PL “nuclei”, the architec-
tonic and especially the connectional patterns argue for a larger, single nucleus. The status
of the separate subdivisions as “layers” rather than as separate nuclei is supported by the
patterns of visuotopic organization revealed by large injections within striate cortex that
show continuous or nearly continuous projection of the center of gaze representation
across the different subdivisions (Gutierrez and Cusick, 1997), and evidence, from double
labeling studies, that individual neurones project to topographically similar representa-
tions in different visual areas (Kennedy and Bullier, 1985; Kaske et al., 1991). This concept
clearly needs further investigation, as the results of small injections do not give the impres-
sion of wedge-like bands crossing all of the subdivisions as they do for the layers of the
LGN. The present evidence of differential connections of the V4 and MT regions (Adams
et al., 2000; Cusick et al., 1993; Gray et al., 1999; Kaske et al., 1991) can be interpreted
either as support for the subdivisions being separate nuclei or layers of a single nucleus.

Figure 8.4. Summary of divisions of the pulvinar according to traditional Nissl architecture (A, A′) compared
to chemoarchitecture (B, B′), in this example for AChE histochemistry. A′ and B′ are adjacent sections; the
arrows point to the same blood vessel. Note the divisions drawn in A follow the positions of fibre bundles and
the brachium of the superior colliculus (bsc) on the Nissl section in A′. The histochemical divisions of the
inferior pulvinar complex are indicated by thick lines, and the divisions of the dorsal pulvinar by thinner lines.
Bsc, brachium of the superior colliculus; L-S, lateral-shell of the inferior pulvinar (PI) complex; -P, posterior
division of the PI complex; PIc, central division of the PI complex; PIm, medial division of the PI complex; PIl,
lateral division of the PI complex; PLd, dorsal Lateral Pulvinar; PMl, lateral division of PM; PMm, medial
division of PM; PMm-c, medial-central PM. Modified from Gutierrez et al. (2000).
4.1. Lamellae and Maps in the PI Complex

The PI complex can be conceptualized as a set of nested lamellae, enclosing a central core,
something in the manner of the layers of an onion. The core zone consists of the medial
subdivision PIm, and the medially and laterally adjacent posterior (PIp) and central (PIc)
divisions (Lin and Kaas, 1979; Cusick et al., 1993; Gutierrez et al., 1995). Based on a uni-
form architectonic appearance, direct continuity with each other behind PIm, and strong
connections of both zones with the superior colliculus, it seems likely that PIp and PIc are
in fact a single unit that envelops the caudal pole of PIm (Stepniewska and Kaas, 1997;
Gray et al., 1999; Stepniewska et al., 1999). There is broad agreement about the existence
and defining characteristics of PIm, PIc, and PIp (even though the latter two might be
considered as one; Stepniewska and Kaas, 1997; Adams et al., 2000; Gray et al., 1999),
but the organization of the remainder of the PI complex, its larger, more lateral part, has

been an area of some disagreement. This zone has been termed the lateral division of the
PI complex, or PIl, since it is a composite of the lateral part of the traditional PI and the
inferior part of the traditional PL (Gutierrez et al., 1995; Gray et al., 1999). PIl contains
a coherent set of topographically organized projections from the upper and lower visual
quadrants, wherein the lower quadrant extends above the brachium of the superior collicu-
lus (bsc), outside of the limits of the traditional PI. It should also be noted that this concept
makes minimal assumptions about the significance of fibres of the bsc, which do not form
a border for any of the agreed medial three subdivisions.
4.2. Modulatory and Driving Inputs

The connections of the PI complex with V1 thus form a guide to the unifying of the dif-
ferent architectonic components into a single zone. The PI subdivisions or “lamellae” of
the PI complex, each have connections with multiple extrastriate visual areas and vice versa
(Lin and Kaas, 1979; Gutierrez and Cusick, 1997; Beck and Kaas, 1998b; Gray et al.,
1999; Stepniewska et al., 1999; Adams et al., 2000). However, in the pulvinar there is
a new degree of corticothalamic complexity that characterizes the non-primary nuclei in
general. That is, the pulvinar receives two different types of corticothalamic fibres. The
first type (termed elongated, E or Type I) is similar to the feedback axons provided by
striate cortex to the LGN and consists of fine axons with a broadly distributed set of fine
terminals. These appear to extend across the subdivisions of the PI complex in a manner
analogous to feedback axons coursing through the projection lines of the LGN, and pro-
vide sprays of branches in specific locations (Rockland, 1994, 1996). A reasonable but
currently untested hypothesis is that these axon sprays target retinotopically similar repre-
sentations within individual PI subdivisions. Furthermore, it is likely that the E-type axons
arise from cortical layer VI and target zones containing neurones that project to the same
cortical region. Interestingly, Type I axons have been shown to contain growth associated
protein 43 (GAP-43), suggesting a role in plasticity of corticothalamic circuits (Bickford,
1999).
The second type of corticothalamic axon to the PI complex provides dense foci of large
round terminals (R or Type II terminals) to more discrete regions of the pulvinar (Rockland,
1996, 1998). Morphologically, the terminals resemble those of the specific sensory inputs
(e.g. retinal fibres, medial lemniscus fibres) to primary nuclei (Guillery and Colonnier,
1970; Mason et al., 1984; Ralston and Ralston, 1994) and it has been hypothesized that
these R type cortical terminals within second level nuclei essentially substitute for the
“primary” type of inputs within principal sensory relay nuclei (Guillery, 1995). Thus, these
axons have been termed feedforward, since they are large, presumably glutamatergic and
target proximal portions of dendrites on their targeted neurones, and are proposed to form
a driving influence on the pulvinar. They appear to arise from large neurones in layer V,
some of which also send a collateral branch to other targets such as the superior colliculus
(Rockland, 1998).
Some Type II terminal fields are found in locations that appear not to contain pulvinar
neurones projecting to the same cortical site. Most evidence for this is indirect, combining,
for example, the dense foci of R type terminations of V1 axons within PIm (Feig and
Harting, 1998; Gutierrez and Cusick, 1997) with the absence of retrograde cell labeling
from V1 within PIm (Adams et al., 2000). This segregated pattern of microcircuitry

implies that forward-going pathways between sets of cortical areas can be not only direct,
corticocortical, but also indirect, by way of a corticothalamocortical loop through the
pulvinar (Guillery, 1995; Feig and Harting, 1998). It should be remembered that the con-
tributions of pulvinocortical circuits to cortical hierarchies could also be either modulatory
(to supragranular layers and layer I) or driving (to layer III/IV) and that individual axons
do show one or both types of terminations within different cortical fields (Rockland et al.,
1999). Thus, a projection can serve different functions in different areas, and since a sub-
stantial number of pulvinar neurones send collateral projections to retinotopically-matched
locations in different extrastriate areas (Rockland et al., 1999), the pulvinocortical circuit
can serve not only to transfer information downstream, but may also coordinate and perhaps
co-activate similar topographic and functional sites between areas (Nowak et al., 1999).
4.3. The PI Complex as the Striate and Tectal Recipient “Visual” Pulvinar
Previous concepts of the role of pulvinar in visual processing have emphasized, for non-
primate taxa, that the tectal recipient zone of pulvinar provides a “second visual system”
route to extrastriate visual cortex (Diamond, 1976). Primates, by contrast, emphasize the
role of the geniculocortical pathway in driving the pulvinocortical circuit. Thus, the striate
cortex sends a strong feedforward connection to PI subdivisions, including PIl (the largest
and its outermost sector, PIl-s) and PIm. Interestingly, V1-to-pulvinar projections are not-
ably weak in PIc and have not been reported for PIp (Gutierrez and Cusick, 1997), the two
divisions that receive a dense projection from the superior colliculus (Stepniewska et al.,
1999). Thus, the geniculostriate and collicular inputs are predominantly segregated within
different subdivisions of the PI complex, and considered together, these two projection
systems outline the complex as a whole. Based on the relative sizes of the striate- and colli-
cular-recipient zones, it appears that the PI complex is dominated by cortical inputs, and
indeed removal of striate cortex inputs produces a profound decrease in visual driving in
the pulvinar, while collicular ablation produces minimal effects (Bender, 1983).
4.4. M- and P-Stream Segregation in the PI Complex?

The PI complex thus appears most heavily influenced by its cortical inputs. Since the
“early” cortical visual areas show evidence of parcellation according to their relative
influences of M and P layers of the lateral geniculate nucleus (Maunsell et al., 1990;
Casagrande, 1994; Van Essen and Gallant, 1994), the question arises whether that partial
segregation is retained or integrated within the PI complex. Cortical connections of PIm,
especially its strong interconnections with the motion sensitive area MT (Figure 8.6) as
well as with area DM (Cusick et al., 1993), suggest that PIm is predominantly concerned
with the M stream. Its connections with V1 and V2 may be closely associated with
M stream as well, but this has not been determined. PIm generally lacks connections with
the V4 region (Adams et al., 2000; Cusick et al., 1993), whose inputs may emphasize the
P and K pathways (Van Essen and Gallant, 1994) but which does show significant M contri-
bution (Ferrera et al., 1992). Histochemically, PIm has several features in common with
the magnocellular layers of the LGN: These include conspicuously dense staining for CO,
AChE, parvalbumin, non-phosphorylated neurofilament protein, Cat-301, and WFA bind-
ing (Cusick et al., 1993; Adams et al., 2000; Gray et al., 1999; Gutierrez et al., 1995;
Lysakowski et al., 1986; Stepniewska and Kaas, 1997). Interestingly, in macaques, all of

Figure 8.5. Connections of V1 with the inferior pulvinar complex in a macaque. Band-like injections of two
different tracers were placed near the vertical meridian representation in V1. The tracer deposit in lower field
(wheat germ agglutinin conjugated to HRP) is shown in gray, and for upper field (35S-labeled methionine),
in black. The upper field injections curved upward toward the representation of the horizontal meridian. Thick
lines indicate the upper border of the PI complex. PI subdivisions are outlined. Very thin outlines surround the
zones of label. Zones of label transported from the lower field representation are indicated by minus signs. Stars
denote the estimated centre-of-gaze representation in the PI complex, where the two labels touch. Serial 25 µm
thick sections are numbered from caudal to rostral. Modified from Gutierrez and Cusick (1997).

these chemoarchitectonic methods show non-uniform staining, with small patches of

about the same calibre as the foci of projections from V1 (Gutierrez and Cusick, 1997).
PIm is characterized by very light staining for calbindin, and taking into account all of its
chemoarchitectonic features described to date, appears a candidate for a metabolically
very active “lemniscal” pathway through the inferior pulvinar complex (Cusick et al., 1993),
albeit with its driving inputs arising from layer V in the striate cortex (Lund et al., 1975;
Rockland, 1996).
By contrast, evidence suggests that PIl is concerned mainly with parvocellular but
possibly also with koniocellular LGN layers, considering its interconnections with V2 and
with V4, and the general segregation of MT connections always from the zones of V4 and
MT input into the PIl-s region (Gray et al., 1999; Adams et al., 2000). PIl has neurochemical
staining characteristics more or less intermediate in density levels between PIm and PIc.
From the dense neuropil staining for calbindin found within PIc (Gray et al., 1999;
Stepniewska and Kaas, 1997), one might speculate not only a neurochemical but also
a functional parallel with the K pathway (Behan et al., 1992; Gutierrez and Cusick, 1994).
This supposition is circumstantial but attractive because of the dense inputs to PIc from
the superior colliculus (Stepniewska et al., 1999), coupled with the dense inputs from the
superficial layers of the superior colliculus with the koniocellular layers of the LGN
(Harting et al., 1978). Interestingly, connections of PIc are dense, specifically with an area
of cortex that is in the DLr region directly caudal to MT (Gray et al., 1999), and with DM,
which is densely interconnected with that same cortical location (Beck and Kaas, 1998a,b).
The functional contributions of these cortical areas and of PIc to visual processing,
however, are not well understood. The PIc, similar to the koniocellular layers of the LGN
(Hendry and Reid, 2000), may be a non-leminiscal type of pathway through the PI com-
plex, involving the superior colliculus rather than the striate cortex as the main driving
influence.
4.5. Second Level Nuclei for Other Systems?

This possible further elaboration of sensory channels within a second level nucleus, the PI
complex, does not have an obvious correspondence within other sectors of sensory and
motor thalamus, chiefly because the pulvinar is so large and well-differentiated in pri-
mates as compared to those other systems. Some comparisons with the somatosensory
system are worth noting, however, as the thalamocortical somatosensory system appears
to emphasize more parallel than hierarchical structure. The primary, ventroposterior,
nucleus connects with several different somatosensory areas, including areas 3a (Huffman
and Krubitzer, 2001), 3b, 1 and 2 (Pons and Kaas, 1985; Cusick and Gould, 1990; see
Kaas and Garraghty, 1991 for review) and has clustered neurochemical organization
reminiscent of the separate LGN layers. In view of the multiple architectonic areas that are
targeted by the cutaneous “core” relay nucleus, it is perhaps not surprising that the defini-
tion of the ventroposterior has been somewhat problematic. A “shell” region dorsal to the
ventroposterior nucleus has been termed a separate nucleus, the ventroposterior superior
or VPS (Cusick et al., 1985; Pons and Kaas, 1985; Cusick and Gould, 1990). VPS conveys
information about deep body receptors predominantly to cortical areas 3a and 2, but there
are minor connections from VPS to the cutaneous core areas 3b and 1 (Pons and Kaas,
1985; Cusick and Gould, 1990). Thus, other possible interpretations are that VP and VPS
together constitute a single thalamic nucleus rather than being separate nuclei, or that VP

and VPS represent an intermediate form in which different modalities are represented in
thalamic structures that are only partially differentiated into separate nuclei. The hierarch-
ical patterns of functional cortical connections between the anterior parietal fields, at least,
argue in favour of the view that area 3b is the primary somatosensory area for cutaneous
somatosensory processing and against the view that VPS is a primary relay nucleus targeting
area 3b.
The anterior pulvinar (Pa) may be considered a second order nucleus, and high order,
multisensory thalamic processing would involve the LP/dorsal pulvinar complex (Gutierrez
et al., 2000). The anterior pulvinar has been proposed on the basis of architecture, relative
position to VP, and spinothalamic inputs (Rausell et al., 1992a) to correspond to the
medial division of the posterior nucleus (Pom) described in other groups of mammals
(Krubitzer and Kaas, 1992). Analogous to corticothalamic fibres within the LGN and
inferior pulvinar complex, cortical fibres exhibit both type E and type R terminations
within “medial pulvinar” (here termed Pa; Darian-Smith et al., 1999). Pa shows somato-
topically organized connections with area 3b of anterior parietal cortex (squirrel monkey;
Cusick and Gould, 1990) as well as with area 2 (macaque; Pons and Kaas, 1985), and
appears to have widespread connections with most somatosensory areas. The nucleus is
relatively pale in cytochrome oxidase stains and dense for calbindin, and thus has connec-
tional and neurochemical features of non-lemniscal thalamocortical systems. It should be
noted that the type of “laminar” histochemical heterogeneity characteristic of the PI
complex has not been found for the anterior pulvinar, and Pa appears not to contain cell
groups with lemniscal neurochemical features, such as found in PIm.
In the auditory system, a candidate second level nucleus is not quite so clear, but, on the
basis of cortical connections with areas outside of the koniocellular fields, it might include
such diverse cell groups as the magnocellular medial geniculate nucleus (Molinari et al.,
1995), a portion of the suprageniculate/limitans complex that caps the medial geniculate,
and the ventral portion of the medial pulvinar nucleus PMmv (Hackett et al., 1998;
Gutierrez et al., 2000). More dorsal portions of PM, e.g. PMl (Gutierrez et al., 2000) may
receive the auditory contributions to thalamic multisensory processing (Hackett et al.,
1998; Romanski et al., 1997).
5. A “THIRD LEVEL” THALAMIC COMPONENT OF THE VISUAL,

SOMATOSENSORY AND AUDITORY SYSTEMS: THE DORSAL
PULVINAR COMPLEX
A “high order” or third level of thalamic nuclei for vision is represented by the dorsal
pulvinar complex, recently determined to contain three broad histochemical zones. The
dorsolateral corner is sharply outlined in chemoarchitectonic stains and termed PLd
(Lysakowski et al., 1986; Gutierrez et al., 2000). Darkly-stained for AChE and calbindin-
pale, PLd occupies the dorsal sector of the traditional PL and is continuous rostrally with
the lateral posterior nucleus LP. PLd has connections with “high order” visual cortex of
the inferior parietal lobule, and with dorsolateral prefrontal cortex in the location of the
frontal eye fields (Asanuma et al., 1985; Gutierrez et al., 2000). The traditional PM
nucleus of the dorsal pulvinar has two distinct neurochemical divisions. The more lateral
division, PMl, stains moderately for AChE, and the more medial division, PMm, is pale
for AChE, although both divisions show heterogeneity of histochemical staining patterns

Table 8.2. Neurochemical characteristics and selected cortical connec-

tions of dorsal pulvinar subdivisions
Histochemical marker PLd PMl PMm PMm-c
AChE ++ +, patchy +, patchy +

Parvalbumin ++ +, patchy +, patchy +
Calbindin – + + –
Selected Connections
DPF ++ ++ + –
IPL ++ +++ ++ –
STG – + ++ ++
Notes
Pluses indicate relative intensity of neurochemical stain or density of connections.
Minuses indicate low levels of staining or absence of connections. DPF, dorsolateral
prefrontal cortex; IPL, inferior parietal lobule; STG, superior temporal gyrus. For
other abbreviations, see legend to Figure 8.4.
suggestive of a further modular organization. Similar patterns of histochemical staining in

the dorsal pulvinar have been identified in humans (Cola et al., 1999).
The dorsal pulvinar connects with visual cortical areas whose neurones have large
receptive fields, and which as a whole show little retinotopic organization. Thus, the
organization of dorsal pulvinar is more difficult to determine based on sensory topo-
graphy, but the patterns of connections with different cortical areas are nevertheless
informative. The traditional medial pulvinar nucleus PM has been shown to have connec-
tions with many if not most high order regions of the cerebral hemisphere, including
visual regions (inferotemporal cortex, superior temporal polysensory region and inferior
parietal lobule: Asanuma et al., 1985; Yeterian and Pandya, 1989; Baleydier and Morel,
1992; Baizer et al., 1993), auditory regions (superior temporal gyrus: Hackett et al., 1998;
Gutierrez et al., 2000) and somatosensory regions (insula: Mufson and Mesulam, 1984).
Connections also exist with cingulate and prefrontal cortex, and thus PM has associations
with a hemisphere-wide system or network proposed to be concerned with directed spatial
attention (see Gutierrez et al., 2000 for review).
The connections of PM with functionally different cortical areas appear to follow rules that
parallel those for connections between the same cortical areas (Figure 8.6). In other words,
cortical areas that do not have direct corticocortical links interconnect with different parts
of PM, so that there is partial segregation with somatosensory connections rostrally, auditory
cortex connections at a rostro-intermediate level, and visual connections located caudally
within PM (Pons and Kaas, 1985; Hackett et al., 1998; Grieve et al., 2000; Gutierrez et al.,
2000). Of particular interest are the connections at the mid-rostral levels, where visual
cortex (inferior parietal lobule, IPL) and auditory cortex (middle and caudal superior
temporal gyrus, STG) both connect. The visual and auditory connections often lie directly
adjacent to each other, in a tightly interlocking pattern, with almost no evidence of direct
overlap (Gutierrez et al., 2000). This is true of both the corticocortical and corticothalamic
relations of the IPL and STG (Seltzer et al., 1996; Gutierrez et al., 2000). It should be
pointed out that the connection patterns of PM have not been worked out in terms of cortico-
thalamic and thalamocortical microcircuitry. While vision-related connections from IT
cortex (and others) have been reported to be of the R-type (Rockland, 1996), it is not fully
known which of the many cortical areas that interconnect with PM provide R-type, driving

projections. With respect to the model proposed by Guillery (1995; see Sherman and
Koch, 1998), one might expect that PM interconnections with highest cortical levels
would be particularly informative.
6. FUNCTIONAL AND COMPARATIVE IMPLICATIONS

OF THALAMIC SYSTEMS
Synchronization of neuronal activity within cortical networks has been observed and
extensively characterized. In the visual cortex, oscillatory activity has been proposed as
a possible coding mechanism for sensory processing as well as a means of achieving
“binding”, the simultaneous perception of different stimulus attributes (Frien et al., 1994;
Tallon-Baudry et al., 1997; Castelo-Branco et al., 1998; Frien and Eckhorn, 2000; but
see Tovee and Rolls, 1992; Ghose and Freeman, 1997). Electrophysiological studies
have suggested that thalamocortical and corticothalamic activity contributes to gamma-
oscillations, which correlate with sensory processing, as well as to delta activity, a com-
ponent of slow wave sleep (Nunez et al., 1992; Bekisz and Wrobel, 1999; Blumenfeld and
Figure 8.6. Summary of selected connections of the inferior pulvinar complex of squirrel monkeys (A) and the
dorsal pulvinar of macaques (B) Part A shows the strongest connections identified following injections into area
MT are with PIm whereas the caudal division of the dorsolateral area (DLc) connects most strongly with PIc.
Although these two cortical areas both connect with PIl and with the dorsal pulvinar, their densest connections
support the concept of segregation of function between the different PI subdivisions. Based on its strong connec-
tions with area MT, PIm appears associated with M-stream. Part B shows dorsal pulvinar in macaques connect-
ing with dorsolateral prefrontal, inferior parietal, and superior temporal gyrus cortex in macaques. Each area has
separate connections with individual divisions. Overlapping connections are revealed by pairs of injections
within prefrontal and posterior parietal cortex of the same hemisphere (lines going to the same arrowheads).
Injections within the superior temporal gyrus and posterior parietal cortex, however, show only non-overlapping
connections. A is from Gray et al., 1999 and B from Gutierrez et al., 2000.

McCormick, 2000). Embedded in each specific relay nucleus there appears a set of parv-
albumin neurones that project onto layer IV, and a set of “non-specific” callbindin cells that
project to layer I (Jones 1998a,b). It is tempting to associate the different histochemical
compartments of the thalamus with different oscillatory behaviors. One might speculate
that the differential expression of calcium binding proteins in the thalamus indicates
different intracellular requirements for calcium buffering and may provide molecular sig-
natures for neurones involved in different thalamocortical oscillatory behaviors.
In support of this, it should be noted that the neurochemical “signatures” present in the
lateral geniculate nucleus appear to be modified under different functional states. For
example, calbindin, which is usually expressed in a sparse, small-celled component of the
VP complex and the LGN, is upregulated in response to peripheral deafferentation
(Rausell et al., 1992b; Gutierrez and Cusick, 1994). Numerous species differences exist in
localization of calcium binding proteins within the thalamus (Leuba and Saini, 1997;
Fortin et al., 1998). Nevertheless, a broad pattern exists which allows the patterns of neuro-
chemical markers to be useful in delineating functional units of thalamus (Morel et al.,
1997; Munkle et al., 1999, 2000; Blomqvist et al., 2000) and in states of long term injury
(Woods et al., 2000). Calcium conductances are thought to play important roles in rhyth-
mic thalamocortical activity (Pedroarena and Llinas, 1997). Differential requirements for
intracellular calcium buffering may underlie the different distributions of calcium binding
proteins in discrete thalamic populations.
For first and second order cortical areas, the thalamic connections are well positioned to
play a role in synchronization of cortical networks. Driving influences to this early portion
of the cortical network appear to emanate from the first order cortical areas (Felleman and
Van Essen, 1991). There are examples of collateralized thalamocortical connections that
target pairs of early areas, for example, finger tip representations of anterior parietal fields
in monkeys (Cusick et al., 1985) and retinotopically matched sites in V1 and V2 (Kennedy
and Bullier, 1985), and V1 and V4 (Lysakowski et al., 1988). Of particular interest are
observations that certain visual responses in the superficial layers of V2 are strongly syn-
chronized to those of V1 (Nowak et al., 1999). It has been proposed that the hierarchical
patterns of cortical connectivity may be matched by patterns of feedforward and feedback
connections through the thalamus (Sherman and Koch, 1998). Alternatively, the role of
the thalamus may be to activate or enhance selectively the activity of matched cortical
sites. In this regard, it has been suggested that the LP/pulvinar complex in the cat strength-
ens and propagates oscillatory synchronization of cortical networks (Molotchnikoff and
Shumikhina, 1996). The microcircuitry of cortico-thalamo-cortical loops with regard to
driving and modulatory inputs will lend only partial insights to their functional role;
sequential or coordinated activation, whether or not by collateralization of thalamocortical
axons, needs to be directly assessed.
As a final point, the macaque visual system provides a well-characterized model system
for addressing some of these questions, but there are some qualifications. Patterns of
visual corticocortical and thalamocortical connections suggest distributed hierarchical
processing (Felleman and Van Essen, 1991), but this is not the only way that a system can
be organized. Generalizations to other thalamic systems, such as somatosensory and audi-
tory (Kaas and Garraghty, 1991; Kaas and Hackett, 1998), can only be made with caution.
These systems are organized with more strongly parallel aspects, i.e. a first level nucleus
that connects with a primary cortical area as well as with adjacent second-level or “belt”
areas. It may be that the second-level nucleus, the inferior pulvinar complex of the visual

thalamic system, uniquely arose in evolution as a primate specialization, in concert with

the differentiation of extrastriate cortex into a system with only sparse connections with
the primary nucleus.
ACKNOWLEDGEMENTS
I thank Drs. Jon Kaas and Todd Preuss for comments on an earlier version of the manu-
script. The research from this laboratory was supported by NIH Grant EY08906.
REFERENCES
Adams, M.M., Hof, P.R., Gattass, R., Webster, M.J. and Ungerleider, L.G. (2000) Visual cortical projections and
chemoarchitecture of macaque monkey pulvinar. Journal of Comparative Neurology, 419, 377–393.
Asanuma, C., Andersen, R.A. and Cowan, W.M. (1985) The thalamic relations of the caudal inferior parietal
lobule and the lateral prefrontal cortex in monkeys: divergent cortical projections from cell clusters in the
medial pulvinar nucleus. Journal of Comparative Neurology, 241, 357–381.
Baizer, J.S., Desimone, R. and Ungerleider, L.G. (1993) Comparison of subcortical connections of inferior
temporal and posterior parietal cortex in monkeys. Visual Neuroscience, 10, 59–72.
Baleydier, C. and Morel, A. (1992) Segregated thalamocortical pathways to inferior parietal and inferotemporal
cortex in macaque monkey. Visual Neuroscience, 8, 391–405.
Beck, P.D. and Kaas, J.H. (1998a) Cortical connections of the dorsomedial visual area in prosimian primates.
Beck, P.D. and Kaas, J.H. (1998b) Thalamic connections of the dorsomedial visual area in primates. Journal of
Behan, M., Jourdain, A. and Bray, G.M. (1992) Calcium binding protein (calbindin D28k) immunoreactivity in
the hamster superior colliculus: ultrastructure and lack of co-localization with GABA. Experimental Brain
Research, 89, 115–124.
Bekisz, M. and Wrobel, A. (1999) Coupling of beta and gamma activity in corticothalamic system of cats attending
to visual stimuli. NeuroReport, 10, 3589–3594.
Bender, D.B. (1983) Visual activation of neurons in the primate pulvinar depends on cortex but not colliculus.
Bickford, M.E. (1999) Growth-associated protein 43 is located in type I corticothalamic terminals in the cat
visual thalamus. Journal of Neuroscience (Online), 19, RC22.
Blasdel, G.G. and Lund, J.S. (1983) Termination of afferent axons in macaque striate cortex. Journal of Neuro-
science, 3, 1389–1413.
Blasdel, G.G. and Fitzpatrick, D. (1984) Physiological organization of layer 4 in macaque striate cortex. Journal
of Neuroscience, 4, 880–895.
Blomqvist, A., Zhang, E.T. and Craig, A.D. (2000) Cytoarchitectonic and immunohistochemical characterization
of a specific pain and temperature relay, the posterior portion of the ventral medial nucleus, in the human
thalamus. Brain, 123, 601–619.
Blumenfeld, H. and McCormick, D.A. (2000) Corticothalamic inputs control the pattern of activity generated in
thalamocortical networks. Journal of Neuroscience, 20, 5153–5162.
Casagrande, V.A. (1994) A third parallel visual pathway to primate area V1. Trends in Neuroscience, 17, 305–310.
Castelo-Branco, M., Neuenschwander, S. and Singer, W. (1998) Synchronization of visual responses between the
cortex, lateral geniculate nucleus, and retina in the anesthetized cat. Journal of Neuroscience, 18, 6395–6410.
Cola, M.G., Gray, D.N., Seltzer, B. and Cusick, C.G. (1999) Human thalamus: neurochemical mapping of
inferior pulvinar complex. NeuroReport, 10, 3733–3738.
Cola, M.G., Preuss, T.M., Seltzer, B. and Cusick, C.G. (2000) Neurochemical organization of chimpanzee
inferior pulvinar complex. Society for Neuroscience Abstracts, 26, 1203.
Cusick, C.G., Steindler, D.A. and Kaas, J.H. (1985) Corticocortical and collateral thalamocortical connections of
postcentral somatosensory cortical areas in squirrel monkeys: a double-labeling study with radiolabeled
wheatgerm agglutinin and wheatgerm agglutinin conjugated to horseradish peroxidase. Somatosensory
Research, 3, 1–31.
Cusick, C.G. and Gould, H.J.3d. (1990) Connections between area 3b of the somatosensory cortex and sub-
divisions of the ventroposterior nuclear complex and the anterior pulvinar nucleus in squirrel monkeys.

Cusick, C.G., Scripter, J.L., Darensbourg, J.G. and Weber, J.T. (1993) Chemoarchitectonic subdivisions of the
visual pulvinar in monkeys and their connectional relations with the middle temporal and rostral dorso-
lateral visual areas, MT and DLr. Journal of Comparative Neurology, 336, 1–30.
Darian-Smith, C., Tan, A. and Edwards, S. (1999) Comparing thalamocortical and corticothalamic microstruc-
ture and spatial reciprocity in the macaque ventral posterolateral nucleus (VPLc) and medial pulvinar.
Diamond, I.T. (1976) Organization of the visual cortex: comparative anatomical and behavioral studies. Feder-
ation Proceedings, 35, 60–67.
Dykes, R.W., Sur, M., Merzenich, M.M., Kaas, J.H. and Nelson, R.J. (1981) Regional segregation of neurons
responding to quickly adapting, slowly adapting, deep and Pacinian receptors within thalamic ventroposter-
ior lateral and ventroposterior inferior nuclei in the squirrel monkey (Saimiri sciureus). Neuroscience, 6,
1687–1692.
Erwin, E., Baker, F.H., Busen, W.F. and Malpeli, J.G. (1999) Relationship between laminar topology and
retinotopy in the rhesus lateral geniculate nucleus: results from a functional atlas. Journal of Comparative
Neurology, 407, 92–102.
Feig, S. and Harting, J.K. (1998) Corticocortical communication via the thalamus: ultrastructural studies of cortico-
thalamic projections from area 17 to the lateral posterior nucleus of the cat and inferior pulvinar nucleus of
the owl monkey. Journal of Comparative Neurology, 395, 281–295.
Felleman, D.J. and Van Essen, D.C. (1991) Distributed hierarchical processing in the primate cerebral cortex.
Ferrera, V.P., Nealey, T.A. and Maunsell, J.H. (1992) Mixed parvocellular and magnocellular geniculate signals
in visual area V4. Nature (London), 358, 756–761.
Fortin, M., Asselin, M.C., Gould, P.V. and Parent, A. (1998) Calretinin-immunoreactive neurons in the human
thalamus. Neuroscience, 84, 537–548.
Friedman, D.P. and Murray, E.A. (1986) Thalamic connectivity of the second somatosensory area and neighbor-
ing somatosensory fields of the lateral sulcus of the macaque. Journal of Comparative Neurology, 252,
348–373.
Frien, A., Eckhorn, R., Bauer, R., Woelbern, T. and Kehr, H. (1994) Stimulus-specific fast oscillations at zero
phase between visual areas V1 and V2 of awake monkey. NeuroReport, 5, 2273–2277.
Frien, A. and Eckhorn, R. (2000) Functional coupling shows stronger stimulus dependency for fast oscillations
than for low-frequency components in striate cortex of awake monkey. European Journal of Neuroscience,
12, 1466–1478.
Garey, L.J., Jones, E.G. and Powell, T.P. (1968) Interrelationships of striate and extrastriate cortex with
the primary relay sites of the visual pathway. Journal of Neurology, Neurosurgery & Psychiatry, 31,
135–157.
Garey, L.J. and Powell, T.P. (1971) An experimental study of the termination of the lateral geniculo-cortical
pathway in the cat and monkey. Proceedings of the Royal Society of London—Series B: Biological
Sciences, 179, 41–63.
Ghose, G.M. and Freeman, R.D. (1997) Intracortical connections are not required for oscillatory activity in the
visual cortex [corrected and republished in Vis Neurosci 1997 Nov–Dec;14(6):963R-979R]. Visual Neuro-
science, 14, 963–979.
Gray, D., Gutierrez, C. and Cusick, C.G. (1999) Neurochemical organization of inferior pulvinar complex in
squirrel monkeys and macaques revealed by acetylcholinesterase histochemistry, calbindin and Cat-301
immunostaining, and Wisteria floribunda agglutinin binding. Journal of Comparative Neurology, 409,
452–468.
Grieve, K.L., Acuña, C. and Cudeiro, J. (2000) The primate pulvinar nuclei: vision and action. Trends in Neuro-
science, 23, 35–39.
Guillery, R.W. and Colonnier, M. (1970) Synaptic patterns in the dorsal lateral geniculate nucleus of the monkey.
Zeitschrift Für Zellforschung und Mikroskopische Anatomie, 103, 90–108.
Guillery, R.W. (1995) Anatomical evidence concerning the role of the thalamus in corticocortical communica-
tion: a brief review. Journal of Anatomy, 187, 583–592.
Gutierrez, C. and Cusick, C.G. (1994) Effects of chronic monocular enucleation on calcium binding proteins
calbindin-D28k and parvalbumin in the lateral geniculate nucleus of adult rhesus monkeys. Brain Research,
651, 300–310.
Gutierrez, C., Yaun, A. and Cusick, C.G. (1995) Neurochemical subdivisions of the inferior pulvinar in macaque
monkeys. Journal of Comparative Neurology, 363, 545–562.
Gutierrez, C. and Cusick, C.G. (1997) Area V1 in macaque monkeys projects to multiple histochemically
defined subdivisions of the inferior pulvinar complex. Brain Research, 765, 349–356.
Gutierrez, C., Cola, M.G., Seltzer, B. and Cusick, C. (2000) Neurochemical and connectional organization of the
dorsal pulvinar complex in monkeys. Journal of Comparative Neurology, 419, 61–86.
Hackett, T.A., Stepniewska, I. and Kaas, J.H. (1998) Thalamocortical connections of the parabelt auditory cortex
in macaque monkeys. Journal of Comparative Neurology, 400, 271–286.

Harting, J.K., Casagrande, V.A. and Weber, J.T. (1978) The projection of the primate superior colliculus upon
the dorsal lateral geniculate nucleus: autoradiographic demonstration of interlaminar distribution of tecto-
geniculate axons. Brain Research, 150, 593–599.
Hashikawa, T., Rausell, E., Molinari, M. and Jones, E.G. (1991) Parvalbumin- and calbindin-containing neurons
in the monkey medial geniculate complex: differential distribution and cortical layer specific projections.
Hashikawa, T., Molinari, M., Rausell, E. and Jones, E.G. (1995) Patchy and laminar terminations of medial
geniculate axons in monkey auditory cortex. Journal of Comparative Neurology, 362, 195–208.
Heath, C.J. and Jones, E.G. (1970) Connexions of area 19 and the lateral suprasylvian area of the visual cortex of
the cat. Brain Research, 19, 302–305.
Hendry, S.H., Hockfield, S., Jones, E.G. and McKay, R. (1984) Monoclonal antibody that identifies subsets of
neurones in the central visual system of monkey and cat. Nature (London), 307, 267–269.
Hendry, S.H. and Yoshioka, T. (1994) A neurochemically distinct third channel in the macaque dorsal lateral
geniculate nucleus. Science (Washington), 264, 575–577.
Hendry, S.H. and Reid, R.C. (2000) The koniocellular pathway in primate vision. Annual Review of Neuroscience,
23, 127–153.
Herkenham, M. (1980) Laminar organization of thalamic projections to the rat neocortex. Science (Washington),
207, 532–535.
Hirai, T. and Jones, E.G. (1989) A new parcellation of the human thalamus on the basis of histochemical staining.
Brain Research—Brain Research Reviews, 14, 1–34.
Hirai, T. and Jones, E.G. (1993) Comparative anatomical study of ventrolateral thalamic mass in humans and
monkeys. Stereotactic & Functional Neurosurgery, 60, 6–16.
Hockfield, S., McKay, R.D., Hendry, S.H. and Jones, E.G. (1983) A surface antigen that identifies ocular domin-
ance columns in the visual cortex and laminar features of the lateral geniculate nucleus. Cold Spring Harbor
Symposia on Quantitative Biology, 48, 877–889.
Hu, B., Steriade, M. and Deschenes, M. (1989a) The cellular mechanism of thalamic pontogeniculo-occipital
waves. Neuroscience, 31, 25–35.
Hu, B., Steriade, M. and Deschenes, M. (1989b) The effects of brainstem peribrachial stimulation on neurons of
the lateral geniculate nucleus. Neuroscience, 31, 13–24.
Huffman, K.J. and Krubitzer, L.A. (2001) Thalamo-cortical connections of areas 3a and M1 in marmoset monkeys.
Jain, N., Catania, K.C. and Kaas, J.H. (1998) A histologically visible representation of the fingers and palm in
primate area 3b and its immutability following long-term deafferentations. Cerebral Cortex, 8, 227–236.
Jones, E.G. (1985) The Thalamus. New York: Plenum Press.
Jones, E.G., Hendry, S.H. and Brandon, C. (1986) Cytochrome oxidase staining reveals functional organization
of monkey somatosensory thalamus. Experimental Brain Research, 62, 438–442.
Jones, E.G. (1998a) A new view of specific and nonspecific thalamocortical connections. Advances in Neurology,
77, 49–71; discussion 72–43.
Jones, E.G. (1998b) Viewpoint: the core and matrix of thalamic organization. Neuroscience, 85, 331–345.
Kaas, J.H., Huerta, M.F., Weber, J.T. and Harting, J.K. (1978) Patterns of retinal terminations and laminar
organization of the lateral geniculate nucleus of primates. Journal of Comparative Neurology, 182,
517–553.
Kaas, J.H., Nelson, R.J., Sur, M., Dykes, R.W. and Merzenich, M.M. (1984) The somatotopic organization of the
ventroposterior thalamus of the squirrel monkey, Saimiri sciureus. Journal of Comparative Neurology, 226,
111–140.
Kaas, J.H. (1990) How sensory cortex is subdivided in mammals: implications for studies of prefrontal cortex.
In: H.B.M. Uylings, C.G. Van Eden, J.P.C. De Bruin, M.A. Corner and M.G.P. Feenstra (eds), The Pre-
frontal cortex: its structure, function and pathology (Progress in Brain Research, Volume 85), Elsevier
Science, Amsterdam, pp. 3–10; (also discussion pp. 10–11).
Kaas, J.H. and Garraghty, P.E. (1991) Hierarchical, parallel, and serial arrangements of sensory cortical areas:
connection patterns and functional aspects. Current Opinion in Neurobiology, 1, 248–251.
Kaas, J.H. and Hackett, T.A. (1998) Subdivisions of auditory cortex and levels of processing in primates.
Audiology & Neuro-Otology, 3, 73–85.
Kaske, A., Dick, A. and Creutzfeldt, O.D. (1991) The local domain for divergence of subcortical afferents to the
striate and extrastriate visual cortex in the common marmoset (Callithrix jacchus): a multiple labelling
study. Experimental Brain Research, 84, 254–265.
Kennedy, H. and Bullier, J. (1985) A double-labeling investigation of the afferent connectivity to cortical areas
V1 and V2 of the macaque monkey. Journal of Neuroscience, 5, 2815–2830.
Krubitzer, L.A. and Kaas, J.H. (1992) The somatosensory thalamus of monkeys: cortical connections and
a redefinition of nuclei in marmosets. Journal of Comparative Neurology, 319, 123–140.
Krubitzer, L. (1995) The organization of neocortex in mammals: are species differences really so different?
Trends in Neuroscience, 18, 408–417.

Lachica, E.A. and Casagrande, V.A. (1992) Direct W-like geniculate projections to the cytochrome
oxidase (CO) blobs in primate visual cortex: axon morphology. Journal of Comparative Neurology,
319, 141–158.
Leuba, G. and Saini, K. (1997) Colocalization of parvalbumin, calretinin and calbindin D-28k in human cortical
and subcortical visual structures. Journal of Chemical Neuroanatomy, 13, 41–52.
Lin, C.S. and Kaas, J.H. (1979) The inferior pulvinar complex in owl monkeys: architectonic subdivisions and
patterns of input from the superior colliculus and subdivisions of visual cortex. Journal of Comparative
Neurology, 187, 655–678.
Lund, J.S., Lund, R.D., Hendrickson, A.E., Bunt, A.H. and Fuchs, A.F. (1975) The origin of efferent pathways
from the primary visual cortex, area 17, of the macaque monkey as shown by retrograde transport of
horseradish peroxidase. Journal of Comparative Neurology, 164, 287–303.
Lysakowski, A., Standage, G.P. and Benevento, L.A. (1986) Histochemical and architectonic differentiation of
zones of pretectal and collicular inputs to the pulvinar and dorsal lateral geniculate nuclei in the macaque.
Lysakowski, A., Standage, G.P. and Benevento, L.A. (1988) An investigation of collateral projections of the
dorsal lateral geniculate nucleus and other subcortical structures to cortical areas V1 and V4 in the macaque
monkey: a double label retrograde tracer study. Experimental Brain Research, 69, 651–661.
Malpeli, J.G. and Baker, F.H. (1975) The representation of the visual field in the lateral geniculate nucleus of
Macaca mulatta. Journal of Comparative Neurology, 161, 569–594.
Malpeli, J.G., Lee, D. and Baker, F.H. (1996) Laminar and retinotopic organization of the macaque lateral
geniculate nucleus: magnocellular and parvocellular magnification functions. Journal of Comparative
Neurology, 375, 363–377.
Mason, C.A., Guillery, R.W. and Rosner, M.C. (1984) Patterns of synaptic contact upon individually labeled
large cells of the dorsal lateral geniculate nucleus of the cat. Neuroscience, 11, 319–329.
Maunsell, J.H., Nealey, T.A. and DePriest, D.D. (1990) Magnocellular and parvocellular contributions to
responses in the middle temporal visual area (MT) of the macaque monkey. Journal of Neuroscience, 10,
3323–3334.
McCormick, D.A. and Huguenard, J.R. (1992) A model of the electrophysiological properties of thalamocortical
relay neurons. Journal of Neurophysiology, 68, 1384–1400.
McDonald, C.T., McGuinness, E.R. and Allman, J.M. (1993) Laminar organization of acetylcholinesterase and
cytochrome oxidase in the lateral geniculate nucleus of prosimians. Neuroscience, 54, 1091–1101.
Milner, A.D. and Goodale, M.A. (1993) Visual pathways to perception and action. Progress in Brain Research,
95, 17–337.
Molinari, M., Dell’Anna, M.E., Rausell, E., Leggio, M.G., Hashikawa, T. and Jones, E.G. (1995) Auditory
thalamocortical pathways defined in monkeys by calcium-binding protein immunoreactivity. Journal of
Molotchnikoff, S. and Shumikhina, S. (1996) The lateral posterior-pulvinar complex modulation of stimulus-
dependent oscillations in the cat visual cortex. Vision Research, 36, 2037–2046.
Morel, A., Magnin, M. and Jeanmonod, D. (1997) Multiarchitectonic and stereotactic atlas of the human
thalamus [published erratum appears in J Comp Neurol 1998 Feb 22;391(4):545]. Journal of Comparative
Neurology, 387, 588–630.
Mufson, E.J. and Mesulam, M.M. (1984) Thalamic connections of the insula in the rhesus monkey and comments
on the paralimbic connectivity of the medial pulvinar nucleus. Journal of Comparative Neurology, 227,
109–120.
Munkle, M.C., Waldvogel, H.J. and Faull, R.L. (1999) Calcium-binding protein immunoreactivity delineates the
intralaminar nuclei of the thalamus in the human brain. Neuroscience, 90, 485–491.
Munkle, M.C., Waldvogel, H.J. and Faull, R.L. (2000) The distribution of calbindin, calretinin and parvalbumin
immunoreactivity in the human thalamus. Journal of Chemical Neuroanatomy, 19, 155–173.
Nowak, L.G., Munk, M.H., James, A.C., Girard, P. and Bullier, J. (1999) Cross-correlation study of the
temporal interactions between areas V1 and V2 of the macaque monkey. Journal of Neurophysiology, 81,
1057–1074.
Nunez, A., Amzica, F. and Steriade, M. (1992) Intrinsic and synaptically generated delta (1–4 Hz) rhythms in
dorsal lateral geniculate neurons and their modulation by light-induced fast (30–70 Hz) events. Neuro-
science, 51, 269–284.
Pedroarena, C. and Llinas, R. (1997) Dendritic calcium conductances generate high-frequency oscillation in
thalamocortical neurons. Proceedings of the National Academy of Sciences, U.S.A., 94, 724–728.
Pons, T.P. and Kaas, J.H. (1985) Connections of area 2 of somatosensory cortex with the anterior pulvinar and
subdivisions of the ventroposterior complex in macaque monkeys. Journal of Comparative Neurology, 240,
16–36.
Preuss, T.M., Gray, D. and Cusick, C.G. (1998) Subdivisions of the motor and somatosensory thalamus of
primates revealed with Wisteria floribunda agglutinin histochemistry. Somatosensory & Motor Research,
15, 211–219.

Ralston III, H.J. and Ralston, D.D. (1994) Medial lemniscal and spinal projections to the macaque thalamus: an
electron microscopic study of differing GABAergic circuitry serving thalamic somatosensory mechanisms.
Ramcharan, E.J., Gnadt, J.W. and Sherman, S.M. (2000) Burst and tonic firing in thalamic cells of unanesthetized,
behaving monkeys. Visual Neuroscience, 17, 55–62.
Rausell, E. and Jones, E.G. (1991) Chemically distinct compartments of the thalamic VPM nucleus in monkeys
relay principal and spinal trigeminal pathways to different layers of the somatosensory cortex. Journal of
Rausell, E., Bae, C.S., Viñuela, A., Huntley, G.W. and Jones, E.G. (1992a) Calbindin and parvalbumin cells in
monkey VPL thalamic nucleus: distribution, laminar cortical projections, and relations to spinothalamic
terminations. Journal of Neuroscience, 12, 4088–4111.
Rausell, E., Cusick, C.G., Taub, E. and Jones, E.G. (1992b) Chronic deafferentation in monkeys differentially
affects nociceptive and nonnociceptive pathways distinguished by specific calcium-binding proteins and
down-regulates gamma-aminobutyric acid type A receptors at thalamic levels. Proceedings of the National
Academy of Sciences, U.S.A., 89, 2571–2575.
Robson, J.A. (1983) The morphology of corticofugal axons to the dorsal lateral geniculate nucleus in the cat.
Robson, J.A. (1984) Reconstructions of corticogeniculate axons in the cat. Journal of Comparative Neurology,
225, 193–200.
Rockland, K.S. (1994) Further evidence for two types of corticopulvinar neurons. NeuroReport, 5, 1865–1868.
Rockland, K.S. (1996) Two types of corticopulvinar terminations: round (type 2) and elongate (type 1). Journal
Rockland, K.S. (1998) Convergence and branching patterns of round, type 2 corticopulvinar axons. Journal of
Rockland, K.S., Andresen, J., Cowie, R.J. and Robinson, D.L. (1999) Single axon analysis of pulvinocortical
connections to several visual areas in the macaque. Journal of Comparative Neurology, 406, 221–250.
Romanski, L.M., Giguere, M., Bates, J.F. and Goldman-Rakic, P.S. (1997) Topographic organization of medial
pulvinar connections with the prefrontal cortex in the rhesus monkey. Journal of Comparative Neurology,
379, 313–332.
Rosa, M.G. and Krubitzer, L.A. (1999) The evolution of visual cortex: where is V2? Trends in Neuroscience, 22,
242–248.
Rose, J.E. and Woolsey, C.N. (1949) Organization of the mammalian thalamus and its relationships to the
cerebral cortex. Electroencephalography and Clinical Neurophysiology, 1, 391–403.
Sakai, S.T., Stepniewska, I., Qi, H.X. and Kaas, J.H. (2000) Pallidal and cerebellar afferents to pre-supplementary
motor area thalamocortical neurons in the owl monkey: a multiple labeling study. Journal of Comparative
Neurology, 417, 164–180.
Schiller, P.H. and Malpeli, J.G. (1978) Functional specificity of lateral geniculate nucleus laminae of the rhesus
monkey. Journal of Neurophysiology, 41, 788–797.
Seltzer, B., Cola, M.G., Gutierrez, C., Massee, M., Weldon, C. and Cusick, C.G. (1996) Overlapping and non-
overlapping cortical projections to cortex of the superior temporal sulcus in the rhesus monkey: double
anterograde tracer studies. Journal of Comparative Neurology, 370, 173–190.
Shatz, C.J. and Rakic, P. (1981) The genesis of efferent connections from the visual cortex of the fetal rhesus
Sherman, S.M., Wilson, J.R., Kaas, J.H. and Webb, S.V. (1976) X- and Y-cells in the dorsal lateral geniculate
nucleus of the owl monkey (Aotus trivirgatus). Science (Washington), 192, 475–477.
Sherman, S.M. and Guillery, R.W. (1998) On the actions that one nerve cell can have on another: disting-
uishing “drivers” from “modulators”. Proceedings of the National Academy of Science, U.S.A., 95,
7121–7126.
Sherman, S.M. and Koch, C. (1998) Thalamus. In: G.M. Shepherd, (ed.), The Synaptic Organization of the
Brain, 4th edition. Oxford University Press: New York, pp. 289–328.
Stepniewska, I., Preuss, T.M. and Kaas, J.H. (1994a) Architectonic subdivisions of the motor thalamus of owl
monkeys: Nissl, acetylcholinesterase and cytochrome oxidase patterns. Journal of Comparative Neurology,
349, 536–557.
Stepniewska, I., Preuss, T.M. and Kaas, J.H. (1994b) Thalamic connections of the primary motor cortex
(M1) of owl monkeys [published erratum appears in J Comp Neurol 1995 Feb 27;353(1):160]. Journal
of Comparative Neurology, 349, 558–582.Stepniewska, I. and Kaas, J.H. (1997) Architectonic sub-
divisions of the inferior pulvinar in New World and Old World monkeys. Visual Neuroscience, 14,
1043–1060.
Stepniewska, I., Qi, H.X. and Kaas, J.H. (1999) Do superior colliculus projection zones in the inferior pulvinar
project to MT in primates? European Journal of Neuroscience, 11, 469–480.
Tallon-Baudry, C., Bertrand, O., Delpuech, C. and Permier, J. (1997) Oscillatory gamma-band (30–70 Hz)
activity induced by a visual search task in humans. Journal of Neuroscience, 17, 722–734.

Tovee, M.J. and Rolls, E.T. (1992) Oscillatory activity is not evident in the primate temporal visual cortex with
static stimuli. NeuroReport, 3, 369–372.
Ungerleider, L.G. and Mishkin, M. (1982). Two cortical visual systems. In: D.J. Ingle, M.A. Goodale and
R.J.W. Mansfield (eds), Analysis of Visual Behavior. Cambridge, Massachusetts: MIT Press, pp. 549–586.
Van Essen, D.C. and Zeki, S.M. (1978) The topographic organization of rhesus monkey prestriate cortex.
Journal of Physiology, 277, 193–226.
Van Essen, D.C. and Gallant, J.L. (1994) Neural mechanisms of form and motion processing in the primate
visual system. Neuron, 13, 1–10.
Woods, T.M., Cusick, C.G., Pons, T.P., Taub, E. and Jones, E.G. (2000) Progressive transneuronal changes in
the brainstem and thalamus after long-term dorsal rhizotomies in adult macaque monkeys. Journal of Neuro-
science, 20, 3884–3899.
Yeterian, E.H. and Pandya, D.N. (1989) Thalamic connections of the cortex of the superior temporal sulcus in
the rhesus monkey. Journal of Comparative Neurology, 282, 80–97.

9 Cortical Areas and Patterns of Cortico-Cortical
Connections
Jon H. Kaas
Department of Psychology, Vanderbilt University
Correspondence to: Jon H. Kaas, Vanderbilt University, Department of Psychology,
111 21st Avenue South, 301 Wilson Hall, Nashville, TN 37240
Tel: 615-322-6029; FAX: 615-343-4342; e-mail: jon.kaas@vanderbilt.edu
Traditionally, cortical areas have been defined by differences in histological appearance, first by
cytoarchitectonic features and then by myeloarchitectonic and chemoarchitectonic features. How-
ever, architectonic distinctions are often subtle, and interpretations of how the cortex is divided into
areas by various investigators have varied. Other methods such as recording or evoking patterns of
sensory and motor representation, evaluating response properties of neurones, and tracing patterns
of connections have been used to delimit or help delimit cortical areas. Many researchers now recog-
nize that the strengths of each method are additive, and that areas are best identified when results
from several methods agree. However, relatively few areas have been established as valid by the
congruence of conclusions based on different procedures, and the reality of many of the traditional
areas, conceived as functionally distinct “organs of the brain”, is in doubt. Nevertheless, the cortico-
cortical connections which have been demonstrated for some well-established areas suggest that
many and perhaps all valid areas have unique, identifying patterns of connections. Even so, connec-
tion patterns such as those between the two cerebral hemispheres do not always reflect areal bound-
aries; and because connections can be distributed in modules within an area and can be similar in
adjacent areas, identifying areas by connection patterns alone can be difficult. The relationship
between architectonic characteristics and connection patterns is uncertain, but at least some of the
architectonic features of areas may emerge as a result of their cortico-cortical connections.
KEYWORDS: cytoarchitecture, development, neocortex, representation
1. INTRODUCTION
The neocortex is a sheet of tissue that varies greatly in surface area, although only somewhat
in thickness, across mammalian species (Kaas, 2000). Neurones throughout this variable
sheet are similarly arranged in layers, and much of the processing is done within local
circuits of neurones that are highly interconnected in vertical arrays across the thickness of
cortex, wherever they are located. Thus, some aspects of neocortical structure and function
are relatively uniform across this sheet. Nevertheless, we have known from the 1782 report
of Gennari (see Gross, 1997) that this sheet is not completely uniform in structure, and
from the time of Broca (1861) at least, that it is not completely uniform in function. Dif-
ferent regions of the neocortex have long been recognized as mediating different functions,
and early neuroanatomists such as Brodmann (1909) laboured to identify and delimit such
functional regions or “organs of the brain” by the few methods available at the time.
179
180 Jon H. Kaas
Brodmann, and others at that time, stained brain sections by the Nissl method for cell
bodies and used regional differences in the detailed appearance of various sections of
cortex to reveal the locations of functional subdivisions of the brain. Brodmann in particular
studied the brains of many different species of mammals and developed elaborate and
complete theories or proposals of how their cortex is subdivided into areas. In brief,
Brodmann proposed that humans, with a large sheet of neocortex, have many areas, nearly
50, while small-brained mammals with little neocortex, such as hedgehogs, have fewer
than 15. Even so, hedgehogs and other small-brained mammals shared some areas with
humans, such as area 17 (or primary visual cortex), and other areas in such mammals were
“composites”, such as 5 plus 7, that had differentiated into separate areas in humans and
other large-brained mammals. Thus, Brodmann presented a broad, comprehensive theory
of brain organization and evolution. According to this theory, the neocortex is divided into
a number of areas that varies across mammalian taxa, thereby accounting at least partly for
variations in behaviour and ability. Humans and other large-brain mammals have more
areas, some areas are shared across species from a common ancestor, and some are new.
New areas emerged by differentiating from old areas.
The basic features of Brodmann’s proposal, as restated and simplified here, are those
that most investigators would hold today. It is in the specifics of the proposal that contem-
poraries of Brodmann and investigators ever since have come to differ. What is at issue is
not whether the cortex can be divided into areas, or if areas vary in number across species,
but exactly where each area is and exactly how they vary in number. Identifying areas in
such a reliable and compelling manner that nearly all investigators will agree, has proven
to be an extremely difficult task. It is a task so widely recognized as difficult, that few
today would attempt to replicate Brodmann’s effort to identify all the areas of neocortex,
and to do so in a range of species. Such uncertainty about areas clearly complicates the
present discussion of the relationship of cortico-cortical connections to cortical areas, but
fortunately some cortical areas have been well defined, and some conclusions about the
relationship of these areas to connection patterns are possible. First, we discuss the issue
of defining cortical areas.
2. WHAT IS A CORTICAL AREA?
The concept of the cortical area has been fundamental to the great progress that has been
made in understanding how brains function. Instead of a large sheet of tissue with
uniformly distributed functions and the equipotentiality of all sectors, the premise that
neocortex is divided into a number of areas of differing functions is now universally
accepted. We generally consider the cortical areas as “organs of the brain”, much as
Brodmann (1909) did. While this concept has been widely accepted and has been
extremely useful, there is little agreement about how to define cortical areas, their sizes
and numbers, how they vary across species, emerge in development, and even if they always
have sharp boundaries or sometimes gradually merge with each other. In a practical sense,
we may have difficulties distinguishing areas from parts of areas, from the larger modules
or columns, and from constellations of areas that function together.
Architectonic methods for identifying areas are based on the longstanding premise that
structure reflects function. If a cortical area is specialized for performing a function or set
of functions, the functional distinctiveness should be reflected in the structural design, and

Cortical Areas and Cortico-Cortical Connections 181
thus in the histological appearance of the area. In principle, all cortical areas can be identi-
fied by differences in cytoarchitecture, and all of the areas distinguished by Brodmann
could be valid. In practice, the delineation of cortical areas by cytoarchitecture alone has
often proven to be unreliable. A given investigator may subdivide the cortex of the same
brain in different ways at different times, and different investigators examining the same
brain or brains of the same species come to various conclusions (see Lashley and Clark,
1946 for a classical critical review). Of course, we have benefited since the time of
Brodmann by the introduction of a number of additional architectonic methods, including
myeloarchitectonic and chemoarchitectonic procedures, as well as by the recent introduc-
tion of semiautomatic, quantitative methods of architectonic analysis (see Schleicher et al.,
1999). These methods and procedures sometimes reveal additional boundaries that were
not obvious before, and they greatly extend the usefulness of the architectonic approach,
but they do not address the more basic issue of determining what the architectonic distinc-
tions mean. Are they marking borders between areas, or borders within areas (such as
between the monocular and binocular segments of the primary visual cortex) or are they
artifacts produced by fissures, blood vessels, or tissue folding and processing? Architec-
tonic differences are often not very impressive, even with newer methods, and we often do
not know what they mean. However, we can address these problems by also using other
methods to subdivide the cortex, and see if the conclusions agree, and obtain information
about the meaning of architectonic boundaries. Without other sources of information,
architectonic areas remain only hypothesized areas, and no more than that. One person’s
proposal is as good as the next. Brodmann’s maps have survived over others in current
textbooks, not because they have been shown to be valid, but because we have been slow
to develop and verify other proposals. Perhaps we need the illusion that we understand
how all parts of the neocortex are divided into areas and how this is done in many species.
Only Brodmann’s maps provide this scope and range. However, more recent proposals
(e.g. Felleman and Van Essen, 1991), although less extensive, are based on a range of
methods, and in many respects they bear little resemblance to those of Brodmann and
other early investigators. Even so, many of the newly proposed areas in these current
theories are still based on too little evidence, and the proposals contain areas as question-
able as many of the areas of Brodmann.
Elsewhere, I have argued (e.g. Kaas, 1982, 1989, 1997), as have others (e.g. Van Essen,
1985), that cortical areas are most reliably defined by the congruence of several different
types of evidence. For example, if a cortical area were to perform a distinct function or set
of functions, it would seem necessary that its pattern of axonal inputs and outputs reflect
those functions. Furthermore, if the computational procedures within two areas are very
similar, their separate functions may depend more on differences in inputs and outputs
rather than internal structure. In practice, connection patterns are commonly used to help
define cortical areas. Reliable methods for revealing such connections have only emerged
over the last 30 or so years, but many excellent methods are now available (for example,
see Angelucci et al., 1996). In addition, the internal organizations of areas should reflect
their functions. Sensory areas systematically represent sensory surfaces, and these repre-
sentations can be revealed by recording from arrays of locations across areas. Adjoining
areas are distinguished by having different patterns of representation. Motor areas can be
similarly revealed by electrically stimulating many sites in cortex while monitoring the
types of movements. Neurones within different areas should have different response
properties, at least at the population level, and these can be revealed by single neurone

182 Jon H. Kaas
recordings or can be visualized with functional imaging methods. These methods not only
help define areas, but suggest functions of areas that can be tested further by deactivating
regions of cortex during behaviour, or by perturbing the functions of areas in other ways,
such as by stimulating the cortex electrically or magnetically.
Proposed cortical areas, and boundaries of areas, are most likely to be valid when
conclusions based on a number of different approaches agree. Each method by itself has
its own problems of interpretation, as well as its strengths. For instance, architectonic
methods may best reveal the precise borders and full extent of areas, but they indicate little
about the functional significance of borders. (Admittedly, stains for metabolic enzymes
say something about sustained levels of neural activity, large pyramidal neurones suggest
long projections, etc.). Thus, an architectonic border may be between functional parts of
an area as well as between areas. Connections are often distributed in a patchy fashion, so
that patches can be misinterpreted as projections to separate areas, especially as adjoining
areas usually have similar connections (e.g. Young et al., 1995; Young, this volume).
In addition, projections to adjacent areas can be interpreted as if they are to the same area.
In studies of neurone response properties, it is difficult to collect a large enough sample of
single neurone recordings from two or more adjacent areas to characterize their contrasting
properties and identify a border between them. Functional imaging studies may confuse
constellations of functionally-related areas with a single area, and maps of sensory and
motor representations may include parts of adjoining areas with similar representations
and matched borders.
Defining cortical areas will be an even more difficult task if, as some suppose, borders
are sometimes or often not sharp, and areas gradually merge with each other. Brodmann
(1909) allowed for gradual borders, and “transition” zones between areas are sometimes
described in current studies.
As an additional complication, the broad goal is not only to define areas in a single
species, but to identify those areas that are homologous, and those that are not, in different
species. Without correct interpretations of homology, efforts to compare neocortex and
cortical function across species lose much of their meaning. Yet, this task is extremely
difficult given that homologous areas can vary in appearance, connections, neural proper-
ties and any other attribute (see Preuss and Kaas, 1999).
3. HOW DO ARCHITECTONIC AREAS RELATE TO PATTERNS

OF CORTICAL CONNECTIONS?
If cortical areas are most reliably identified by the agreement of different types of evidence,
and two major sources of evidence are architectonic distinctiveness and connectional
distinctiveness, then all correctly identified architectonic areas should be connectionally
unique. This premise can best be evaluated by considering the connections of cortical
areas that have been well defined. In the visual system of primates, all or nearly all inves-
tigators agree on the existence and architectonic boundaries of three visual areas (see
Kaas, 1997), the classical area 17 or primary visual cortex (V1), the second visual area
(V2), and the middle temporal visual area (MT). Area 17 is one of the most distinctive of
all cortical areas, and almost any histological procedure will identify it. Nevertheless,
misidentifications have occurred. For example, Brodmann mistook the less-developed
medial monocular portion of area 17 for area 18 in squirrels, leading to the preservation of

this error in the common portrayal of an “area 18” medial to area 17 in rats. However, the
boundaries of area 17 of most mammals are not in question today.
The second visual area, V2, has been more difficult to delimit (see Allman and Kaas,
1974). Its inner border with area 17, ofcourse, has been obvious, but its outer border has
not. Brodmann’s (1909) area 18 in New World marmosets closely corresponds in depicted
width with present-day estimates of V2, and it is tempting to refer to V2 as area 18, as
many investigators do. However, Brodmann’s area 18 in Old World monkeys was more
than twice as wide as present estimates of V2, so it is obvious that Brodmann did not
consistently define the same region of cortex as area 18, even in primates. Today, the
architectonic field corresponding to V2 in monkeys is rather easily identified by its
conspicuous banding pattern. This is best seen in “surface views” of V2 in sections cut
parallel to the surface of flattened cortex and processed for cytochrome oxidase or myelin
(e.g. Livingstone and Hubel, 1982; Tootell et al., 1985; Krubitzer and Kaas, 1989). V2
has alternating dark and light bands across the width of the area in both preparations, and
the full extent of V2 is obvious. This architectonic definition of V2 might be termed the
“redefined area 18”, since V2 is commonly referred to as area 18. This V2 also can be
distinguished by cytoarchitecture (see Allman and Kaas, 1974), but the distinctions in
Nissl-stained brain sections are not obvious enough for reliable identification.
The middle temporal visual area, MT, is a visual area first identified in New World
monkeys, when microelectrode recordings revealed it as a systematic representation of the
contralateral visual hemifield, that was co-extensive with a densely myelinated oval of
cortex (Allman and Kaas, 1971). Again, cytoarchitectonic distinctions were also apparent,
but they were too subtle to serve as a practical way of identifying MT. Subsequently, MT
has been identified by its myeloarchitecture and other characteristics in a broad range of
primate species, including humans (see Sereno, 1998). MT probably exists in all primates,
but it has not been identified with certainty in any non-primate mammal (see Kaas, 1997).
Thus, MT may be an area that emerged with the evolution of the first primates. Given the
distinctiveness of MT in brain sections stained for myelin, it seems surprising that the area
had not been delimited in earlier myeloarchitectonic studies of cortex. Brodmann and
other early investigators proposed no area like MT in shape and location.
The ease and reliability of identifying V1, V2, and MT by architectonic criteria, and
their wide acceptance as valid visual areas, makes them ideal for addressing the question
of how connection patterns relate to architectonic fields (although not necessarily the
classical fields of Brodmann). As early as 1965 (Myers, 1965; Kuypers et al., 1965),
lesion studies of connectivity using anterograde degeneration techniques revealed that V1
projects to the general region of cortex in the upper temporal lobe of macaque monkeys,
that was later identified as MT (Allman and Kaas, 1971). The existence of these projec-
tions was further established in subsequent investigations, but it was not until the report of
Spatz (1977) that the connections were specifically attributed to MT. Spatz (1977) demon-
strated that MT is reciprocally and topographically interconnected with area 17. The rela-
tionship between V1 and MT in terms of connections is now well established (for review,
see Weller and Kaas, 1983; Preuss et al., 1993). Injections of tracers in central V1 label
central regions of MT, and those along the outer margin of V1 label locations along the
outer margin of MT (Figure 9.1). Thus, the two representations are retinotopically
matched in their overall interconnection patterns. However, projections to MT are patchy
and more extensive than would conform to strictly-matched retinotopic locations, and the
projections from MT back to V1 are also more broadly distributed than expected from

184 Jon H. Kaas
V2
MT
V1
40˚ 20˚ 5˚
5˚ 20˚ 40˚
LF LF
HM HM HM HM
UF UF
VM
VM
V2
HM
Figure 9.1. The connection patterns of V1 and V2 with MT in primates. The primary visual area, V1 or area 17,
the second visual area, V2, and the middle temporal visual area, MT, are visual areas that are architectonically
distinct and easily delimited in primates. Both V1 and V2 interconnect with MT in topographic patterns that closely
match the retinotopic organizations of the three areas. The clear relationship between connection patterns and
these architectonically distinct fields suggests that most or all valid architectonic fields have their own identifying
patterns of cortico-cortical connections. Dots represent injection sites of tracers in V1 and V2. Visual hemifield
coordinates are indicated in V1, V2, and MT. The zero horizontal meridian (HM) bisects V1 and MT and forms
the outer border of V2. The zero vertical meridian forms the outer border of V1 and MT.
precisely-matched retinotopic connections (Krubitzer and Kaas, 1990a). Nevertheless, the

overall retinotopic pattern is clear. The inputs to MT respect MT boundaries, and conclu-
sions about the location and extent of MT from connectional studies agree with those from
architectonic studies.
MT can also be defined by connections with other visual areas. Although not so exten-
sively investigated, a topographic pattern of connections from V2 can be used to define
and delimit MT (see Stepniewska and Kaas, 1996). Perhaps as an even more dramatic
example of how connection patterns define cortical areas, injections in a visual area just
ventral to MT (FSTD, the dorsal area of the fundus of the superior temporal sulcus) label
large portions of MT in a patchy pattern, but none of the cortex immediately surrounding
MT; injections placed just slightly more ventral in the sulcus (in FSTV, the ventral area of
the fundus of the superior temporal sulcus) will label cortex in a ring-like area around MT
(MTC, the so-called MT crescent) but not MT (Figure 9.2; Kaas and Morel, 1993). The
specificity of these connection patterns, ofcourse, also helps to establish the two injected
regions, FSTD and FSTV, as separate visual areas, and the MTC crescent-shaped ring or
belt around MT as an additional visual area (see Kaas, 1997). Injections into another pro-
posed visual area, the dorsomedial area (DM) also demonstrate interconnections that
identify MT (Beck and Kaas, 1999; Krubitzer and Kaas, 1993).
In summary, connection patterns with any and all of at least four visual areas (V1, V2,
FSTD and DM) identify and delimit the same visual area, MT, as defined by myeloarchi-
tecture (or the dense expression of cytochrome oxidase; Tootell et al., 1985). The connec-
tions with each of these areas reach all parts of MT, and stop or change in density and
pattern sharply at the MT border. Thus, at least some long cortico-cortical connection
patterns precisely and completely conform to the limits of an easily and reliably identified
architectonic area. Subcortical connections do so as well. Most notably, a subdivision of

MST
MT
MTC
FSTD
FSTV
Figure 9.2. MT is also distinguished by its strong connections with adjoining visual area, FSTD, and its lack of
connections with a slightly more distant visual area, FSTV. (These visual areas are named after their more
ventral or more dorsal locations in the fundus of the superior temporal sulcus of macaque monkeys). MT is also
distinguished by connections with the medial superior temporal area, MST. Gray circles represent injection sites
of tracers in visual areas, while arrowheads mark axon terminations. The projections are broadly distributed, but
they may be roughly topographic between areas. This possibility has not yet been adequately explored.
the inferior pulvinar complex is densely interconnected with MT but not with other sub-
divisions of visual cortex (Lin and Kaas, 1980; Stepniewska et al., 1999). Many other
examples of connections conforming to architectonic boundaries exist, but evidence is
often less obvious. In the visual cortex, the connection patterns between V2 and other
visual areas (Stepniewska and Kaas, 1996) conform to V2 as now defined architectonically.
In the somatosensory cortex of primates, the connections of area 3b with areas 3a and 1, as
well as other subdivisions of somatosensory cortex, provide other examples of architec-
tonic areas with distinct patterns of connections (e.g. Krubitzer and Kaas, 1990b). The
results from these and other areas strongly suggest that any valid architectonic area
conforms to a unique and systematic pattern of cortico-cortical connections.
4. DO ALL LONG CORTICO-CORTICAL CONNECTIONS RESPECT

ARCHITECTONIC BOUNDARIES?
There are instances when patterns of connections may appear to fail to reflect architec-
tonic boundaries, where closer inspection shows that they do. Cortical areas often have
connections that are quite similar to those of the areas adjoining them (Young et al., 1995;

186 Jon H. Kaas
Young, this volume). For example, area 3b of monkeys projects to both area 1 and area 2
(see Pons and Kaas, 1986). Since areas 1 and 2 adjoin each other, the two projection
patterns may appear as one that fails to distinguish the two fields. However, the projec-
tion density to the two areas is quite different, with the terminations in area 2 being much
less dense than terminations in area 1. In addition, if injections are placed in different
locations across the width of area 3b, different topographic patterns of terminations in
areas 1 and 2 are revealed. Thus, some of the connections of adjoining areas may be with
the same fields, but they are likely to differ in density, topographic pattern, and even
laminar distribution.
In other instances, a lack of differences in connections for adjoining areas may suggest
that the so-called areas are parts of a larger area. If connections of two or more adjoining
areas are nearly identical, as the connections of areas 3b, 1 and 2 of somatosensory cortex
of cats seem to be (see Scannell et al., 1995), then one might question the validity of the
claim for separate fields. Microelectrode mapping data and architectonic results suggest
that all three proposed fields, 3b, 1 and 2, of somatosensory cortex of cats (Hassler and
Muhs-Clement, 1964) actually constitute a single field, S1 (Felleman et al., 1983). This S1
of cats appears to be homologous with area 3b of primates, and all of this larger field in
cats deserves to be called area 3b (Kaas, 1983). All parts of S1 in cats are similar in archi-
tectonic appearance. As a single field, S1 should have a single pattern of connections with
other fields, and the near identity of connections of “3b”, “1” and “2” of cats suggest that
they are parts of a single field.
However, in at least some instances, cortical connection patterns do not reflect architec-
tonic borders of valid areas very well. The best example may be the callosal connections
of area 17 in primates (Figure 9.3; see Cusick et al., 1984; Cusick and Kaas, 1986; Kaas,
1995). Callosal connections may be of several types (Figure 9.4), and they may be of
mixed types for the same area. Area 17 of Old World monkeys is almost devoid of
Callosal connections
Monkeys Prosimian Galagos
V2 V1 V2 V1
Figure 9.3. Callosal connections of the V1/V2 border region of primates. The connections are sharply con-
fined to the border region of V1 and larger extents of V2 in monkeys, but they extend well into V1 of galagos.
Thus, these connections do not always sharply mark the border of V1 with V2. However, connections are more
dense along the border.

1. Homotopic
A A'
B B'
C C'
2. Heterotopic but homoareal
A D
B E
C F
3. Heteroareal, but either

homotopic or heterotopic
A'
A B'
Figure 9.4. Callosal connections are of several types, and individual areas may have more than one type.
Connections between representations of visual hemifields in the two cerebral hemispheres may match precisely
along the two borders corresponding to the same zero vertical meridian (1. Homotopic), or neurones within one
representation may project to the border of the other (2. Heterotopic). One area may project in both patterns to
another area in the opposite hemisphere (3. Heteroareal). (See Kaas, 1995).
interhemispheric connections, with only a few projection cells and terminations along the
fringe of the outer border. In contrast, callosal connections are densely distributed along
the inner border of V2 and well into V2 (although not fully across V2 or even uniformly
within the inner half). In New World monkeys, callosal connections are also dense in V2,
but they barely invade V1. In Old World monkeys, the callosal connections of V1 appear
to be even more restricted to the border. In prosimian galagos and probably other pro-
simian primates, however, the distribution of callosal connections in V2 does not stop
abruptly at the V1 border, but instead continues well into V1 where it is patchy and it
gradually drops off in density. Thus, if the callosal connection pattern was used as the only
criterion for subdividing cortex into areas, and the compelling architectonic evidence
ignored, then one would place an approximate V1/V2 border somewhere within V1 of
galagos, and wrongly conclude that the border is not sharp, but rather constitutes a broad
transition zone.
There are probably other instances where all parts a valid architectonic field are not
connected in the same way with other regions of cortex. The portions of V1 and other
visual areas that represent central vision may connect with more areas than the portions
representing peripheral vision (see Casagrande and Kaas, 1994). Areas are often divided
into sets of modules with different connections, and studies of connections may involve
one set more than the other. The blob and non-blob regions of V1 in monkeys, for

188 Jon H. Kaas
example, have different connections with extrastriate cortex (see Casagrande and Kaas,
1994). A single focused injection in any cortical area typically labels a patchy distribution
of terminations and projection neurones in a number of other areas, and separate patches
of connections may be mistaken for separate visual areas. Nevertheless, the overwhelming
conclusion should be that studies of connection patterns provide a powerful way of identi-
fying cortical areas. When connectional and architectonic approaches identify the same
subdivisions of cortex, then the evidence that the proposed cortical area is a valid area is
quite compelling. Such evidence is further strengthened when results from microelectrode
recordings and maps, images of brain activity patterns, and ablation-behaviour studies
produce congruent results.
5. ARE ARCHITECTONIC DIFFERENCES PRODUCED

BY CONNECTIONAL DIFFERENCES?
Little is really known of how architectonic distinctions between fields are created. One
possibility is that the connection patterns develop as the result of some sort of competitive
process, and that differences in connections produce different activity patterns, gene
expression, and ultimately histological structures (see O’Leary, 1989; Killackey, 1990).
This is obviously true in at least trivial ways. The large Meynert cells that characterize V1
of monkeys are those that conduct impulses rapidly over long distances to MT and subcor-
tical structures. If the target fields of such neurones are ablated, these neurones would
degenerate or would not develop to such conspicuous sizes (see Cowan et al., 1984). If a
major activating input into a cortical area is removed, less of the metabolic enzymes such
as cytochrome oxidase may be expressed (Wong-Riley, 1994), and less of the inhibitory
neurotransmitter, GABA, and receptors for that transmitter, may be present (see Jones,
1993). Thus, the cyto- and chemoarchitecture of areas may be altered during and after
development by changes in their connections.
Some have emphasized how little most of neocortex varies in architectonic appearance
from one region to another (e.g. Zeki, 1978). The argument can be made that the local
circuit computations made by groups of neurones at any location in the cortex is much the
same, and that the only important difference is that each area has its unique pattern
of inputs and outputs. According to this argument, areas look much alike because the
neurones within them are doing basically the same things, but with different inputs and
outputs and with different consequences. The variable functions of areas of cortex thus
stem from differences in connections rather than from other morphological specializa-
tions. Again, this does not seem to be completely the case, if only because some areas such
as V1 are so remarkably distinct in their internal structure. However, when architectonic
differences between fields are minor and hard to demonstrate, they may have been created
in development largely or completely by connectional differences. Furthermore, if adjoin-
ing fields are performing nearly identical transformations, but on different inputs, there
may be few or no significant architectonic distinctions between the fields.
Despite this, it seems extremely likely that at least some cortical areas are differentiated
by the uneven distribution of chemical markers before connections form. Chemical gradi-
ents have been demonstrated in the developing neocortex well before thalamo-cortical and
cortico-cortical connections form (Rubenstein et al., 1998), and in mutant mice that fail to
develop thalamo-cortical connections (Nakagawa et al., 1999; Miyashita-Lin et al., 1999).

Through thresholding mechanisms these gradients could delimit at least a few primary
areas, such as V1. Although many of the architectonic features of V1 may depend on the
presence of normal inputs and other connections (see Rakic, 1991), some of the identify-
ing features of the area may emerge independently. Other cortical areas, according to this
scenario, might emerge as chemically specified primary areas form projections that
compete for cortical space. Thus, a few basic cortical areas might differentiate in part as
a consequence of early positional effects, and other areas may differentiate later as a con-
sequence of differences in connectivity.
Another possibility is that all cortical areas are created by positional effects on gene
expression, and that at least some distinctive architectonic features emerge in each of these
areas as a result of differences in gene expression. Usually, this possibility is related to
the concept of a “protomap” of cortex in which cortical areas are predetermined even
before cortex develops within a spatially differentiated sheet of germinal cells (Rakic,
1988). Cells for each area in the protomap would be labeled in some way so that after they
migrate to cortex, they would attract proper connections and contribute to the unique
structure of their area. While such a possibility has its attractive elements, it would seem
improbable in complex brains, such as the human brain with well over 50 areas, that all
such areas depend on positional effects on gene expression in the germinal layer that gives
rise to cortex.
In conclusion, cortical areas are distinguished by differences in connection patterns, and
connection differences have at least some structural and histological consequences. If con-
nections are experimentally altered early in development, cortical architecture may not
develop in a normal manner. This suggests that at least for those areas that are not very
distinct in architecture, many or all of those features that are distinct emerge as a con-
sequence of their connection patterns and the impact of those connections on neural activ-
ity during development. Nevertheless, many or most of the architectonic features of
cortical areas, especially for the most distinct primary areas, may depend on positional
effects on the regional development of cortex or on the regional distribution of germinal
cells that generate cortex. More experimental studies of the development of cortical areas
are needed.
ACKNOWLEDGEMENTS
This review was prepared while the author was a Fellow at the Center for Advanced Study
in the Behavioral Sciences at Stanford University. Financial support at the Center was
provided by the John D. and Catherine T. MacArthur Foundation. Helpful comments were
provided by Todd Preuss.
REFERENCES
Allman, J.M. and Kaas, J.H. (1971) A representation of the visual field in the caudal third of the middle temporal
gyrus of the owl monkey (Aotus trivirgatus). Brain Research, 31, 85–105.
Allman, J.M. and Kaas, J.H. (1974) The organization of the second visual area (V-II) in the owl monkey:
A second-order transformation of the visual hemifield. Brain Research, 76, 247–265.
Angelucci, A., Clascá, F. and Sur, M. (1996) Anterograde axonal tracing with subunit B of cholera toxin,
a highly sensitive immunohistochemical protocol for revealing fine axonal morphology in adult and
neonatal brains. Journal of Neuroscience Methods, 65, 101–102.

190 Jon H. Kaas
Beck, P.D. and Kaas, J.H. (1999) Cortical connections of the dorsomedial visual area in Old World macaque
monkeys. Journal of Comparative Neurology, 407, 487–502.
Broca, P. (1861) Remarks on the rest of the faculty of articulate language. In: G. von Bonin, trans. (1960), Some
papers on the cerebral cortex. C.C. Thomas: Springfield, IL, pp. 99–72.
Casagrande, V.A. and Kaas, J.H. (1994) The afferent, intrinsic, and efferent connections of primary visual cortex
in primates. In: A. Peters and K. Rockland (eds), Cerebral Cortex, Vol. 10, Primary Visual Cortex in Primates.
New York: Plenum Press, pp. 201–259.
Cowan, W.M., Fawatt, J.W., O’Leary, D.D.M. and Stanfield, B.B. (1984) Regressive events in neurogenesis.
Cusick, C.G., Gould, H.J. and Kaas, J.H. (1984) Interhemispheric connections of visual cortex of owl monkeys
(Aotus trivirgatus), marmosets (Callithrix jacchus), and galagos (Galago crassicaudatus). Journal of
Cusick, C.G. and Kaas, J.H. (1986) Interhemispheric connections of cortical sensory and motor maps in
primates, In: F. Lepore, M. Dtito and H.H. Jasper (eds), Two Hemispheres—One Brain, New York: Alan
R. Liss, Inc., pp. 83–102.
Felleman, D.J., Wall, J.T., Cusick, C.G. and Kaas, J.H. (1983) The representation of the body surface in S1 of
cats. Journal of Neuroscience, 3, 1648–1669.
Felleman, D.J. and Van Essen, D.C. (1991) Distributed hierarchical processing in primate cerebral cortex. Cerebral
Cortex, 1, 1–47.
Gross, C.G. (1997) From Imhotep to Hubel and Wiesel: the story of visual cortex. In: K.S. Rockland, J.H. Kaas
and A. Peters (eds), Cerebral Cortex, Volume 12, Extrastriate Cortex in Primates. New York: Plenum
Press, pp. 1–58.
Hassler, R. and Muhs-Clement, K. (1964) Architectonischer aufbau des sensorimotorischen and parietal en
cortex der katze. Journal für Hirnforschung, 6, 377–420.
Jones, E.G. (1993) GABAergic neurons and their role in cortical plasticity in primates. Cerebral Cortex, 3,
361–372.
Kaas, J.H. (1982) The segregation of function in the nervous system: Why do the sensory systems have so many
subdivisions? Contributions to Sensory Physiology, 7, 202–240.
Kaas, J.H. (1983) What, if anything, is S-1? The organization of the “first somatosensory area” of cortex. Physio-
logical Reviews, 63, 206–231.
Kaas, J.H. (1989) Why does the brain have so many visual areas? Journal of Cognitive Neuroscience, 1, 121–135.
Kaas, J.H. and Morel, A. (1993) Connections of visual areas in the upper temporal lobe of owl monkeys: The MT
crescent and dorsal and ventral subdivisions of FST. Journal of Neuroscience, 13, 534–546.
Kaas, J.H. (1995) The organization of callosal connections in primates. In: A.G. Reeves and D.W. Roberts (eds),
Epilepsy and the corpus callosum II. New York: Plenum Press, pp. 15–27.
Kaas, J.H. (1997) Theories of visual cortex organization in primates. In: K.S. Rockland, J.H. Kaas and A. Peters
(eds), Cerebral Cortex, Volume 12, Extrastriate Cortex in Primates. New York: Plenum Press, pp. 91–125.
Kaas, J.H. (2000) Why is brain size so important: Design problems and solutions as neocortex gets bigger or
smaller. Brain and Mind, 1, 7–23.
Killackey, H.P. (1990) Neocortical expansion: An attempt toward relating phylogeny and ontogeny. Journal of
Cognitive Neuroscience, 2, 1–17.
Krubitzer, L.A. and Kaas, J.H. (1989) Cortical integration of parallel pathways in the visual system of primates.
Krubitzer, L.A. and Kaas, J.H. (1990a) Cortical connections of MT in four species of primates: Areal, modular,
and retinotopic patterns. Visual Neuroscience, 5, 165–204.
Krubitzer, L.A. and Kaas, J.H. (1990b) The organization and connections of somatosensory cortex in marmosets.
Krubitzer, L.A. and Kaas, J.H. (1993) The dorsomedial visual area of owl monkeys: Connections, myeloarchi-
tecture, and homologies in other primates. Journal of Comparative Neurology, 334, 497–528.
Kuypers, H.G.J.M., Szwarcbart, M.K., Mishkin, M. and Rosvald, H.E. (1965) Occipito-temporal corticocortical
connections in the rhesus monkey. Experimental Neurology, 11, 245–262.
Lashley, K.S. and Clark, G. (1946) The cytoarchitecture of the cerebral cortex of Ateles: A critical examination of
architectonic studies. Journal of Comparative Neurology, 85, 223–306.
Lin, C.-S. and Kaas, J.H. (1980) Projections from the medial nucleus of the inferior pulvinar complex to the
middle temporal area of visual cortex. Neuroscience, 5, 2219–2228.
Livingstone, M.S. and Hubel, D.H. (1982) Thalamic inputs to cytochrome oxidase-rich regions in monkey visual
cortex. Proceedings of the National Academy of Sciences, U.S.A., 79, 6098–6101.
Miyashita-Lin, E.M., Hevener, R., Wassarman, K.M., Martinez, S. and Rubenstein, J. (1999) Neocortical region-
alization in the absence of thalamic innervation. Science (Washington), 285, 906–909.
Myers, R.E. (1965) Organization of visual pathways. In: E.G. Ehlinger (ed.), Functions of the Corpus Callosum.
London: Churchill, p. 133.

Nakagawa, Y., Johnson, J.E. and O’Leary, D.D.M. (1999) Graded and areal expression patterns of regulatory
genes and cadherins in embryonic neocortex independent of thalamocortical input. Journal of Neuro-
science, 19, 10877–10885.
O’Leary, D.D.M. (1989) Do cortical areas emerge from a protocortex. Trends in Neuroscience, 12, 401–406.
Pons, T.P. and Kaas, J.H. (1986) Corticocortical connections of areas 2, 1, and 5 of somatosensory cortex in
macaque monkeys: A correlative anatomical and electrophysiological study. Journal of Comparative Neuro-
logy, 248, 313–375.
Preuss, T.M., Beck, P.D. and Kaas, J.H. (1993) Areal, modular, and connectional organization of visual cortex in
a prosimian primate, the slow loris (Nycticebus coucong). Brain Behavior Evolution, 42, 237–251.
Preuss, T.M. and Kaas, J.H. (1999) Human brain evolution. In: M.J. Zigmond, F.E. Bloom, S.C. Landis, J.L. Roberts
and L.R. Squire (eds), Fundamental Neuroscience. San Diego: Academic Press, pp. 1283–1311.
Rakic, P. (1988) Specification of cerebral cortical areas. Science (Washington), 241, 170–176.
Rakic, P. (1991) Experimental manipulation of cerebral cortical areas in primates. Philosophical Transactions of
the Royal Society of London, Series B, 331, 291–294.
Rubenstein, J.L., Shimamura, K., Martinez, S. and Puelles, L. (1998) Regionalization of the prosencephalic
neural plate. Annual Review of Neuroscience, 21, 445–477.
Scannell, J.W., Blakemore, C. and Young, M.P. (1995) Analysis of connectivity in the cat cerebral cortex.
Schleicher, A., Amunts, K., Geyer, S., Morosan, P. and Zilles, K. (1999) Observer-independent method for
microstructural parcellation of cerebral cortex: A quantitative approach to cytoarchitectonics. Neuroimage,
9, 165–177.
Sereno, M.I. (1998) Brain mapping in animals and humans. Current Opinion in Neurobiology, 8, 188–194.
Spatz, W.B. (1977) Topographically organized reciprocal connections between area 17 and MT (visual area of
superior temporal sulcus) in the marmoset, Callithrix jacchus. Experimental Brain Research, 27, 559–572.
Stepniewska, I. and Kaas, J.H. (1996) Topographic patterns of V2 cortical connections in macaque monkeys.
Stepniewska, I., Qi, H.W. and Kaas, J.H. (1999) Do superior colliculus projection genes in the inferior pulvinar
project to MT in primates? Journal of European Neuroscience, 11, 856–866.
Tootell, R.B.H., Silverman, M.S., DeValois, R.L. and Jacobs, G.H. (1982) Functional organization of the second
cortical area of primates. Science (Washington), 220, 737–739.
Tootell, R.B.H., Hamilton, S.L. and Silverman, M.S. (1985) Topography of cytochrome oxidase activity in owl
monkey cortex. Journal of Neuroscience, 5, 2786–2800.
Van Essen, D.C. (1985) Functional organization of primate visual cortex. In: A. Peters and E.G. Jones (eds),
Cerebral Cortex, Volume 3, Visual Cortex. New York: Plenum Press, pp. 259–329.
Weller, R.E. and Kaas, J.H. (1983) Retinotopic patterns of connections of area 17 with visual areas V-II and MT
in macaque monkeys. Journal of Comparative Neurology, 220, 253–279.
Wong-Riley, M.T.T. (1994) Primate visual cortex. Dynamic metabolic organization and plasticity revealed by
cytochrome oxidase. In: A. Peters and K.S. Rockland (eds), Cerebral Cortex, Volume 10, Primary Visual
Cortex in Primates. New York: Plenum Press, pp. 191–200.
Young, M.P., Scannell, J.W. and Burns, G. (1995) The analysis of cortical connectivity. Austin: Landes.
Zeki, S.M. (1978) Uniformity and diversity of structure and function in rhesus monkey prestriate cortex. Journal
of Physiology (London), 277, 273–290.

Part III
CONSTANCY AND VARIATION

ACROSS SPECIES

10 The Cerebral Cortex of Mammals:
Diversity within Unity
Facundo Valverde, Juan A. De Carlos and
Laura López-Mascaraque
Laboratorio de Neuroanatomía Comparada, Instituto Cajal (CSIC),
Avenida del Doctor Arce 37,28002 Madrid, Spain
Correspondence: Prof. F. Valverde, Instituto Cajal (CSIC)
Avenida del Doctor Arce 37, 28002 Madrid, Spain
FAX: 91 585 47 54; e-mail: fvalverde@cajal.csic.es
In this chapter, we will review our observations on the structure of the cerebral cortex obtained with
the Golgi method. A comparative approach based on the study of several mammalian species will
place emphasis on differences and similarities of intrinsic neuronal connectivity. The chapter first
explores fundamental characteristics differentiating neocortex from allocortex, following a description
of cortical cell arrangement which has been considered as a model of neocortical organization: the
module concept. We review the different types of intrinsic neocortical cells and the morphology of
specific cortical afferent fibres.
The cerebral cortex of mammals contains an ample variety of cells which, from a general point of
view, can be classified into two principle types: pyramidal and non-pyramidal cells. From a morpho-
logical point of view, pyramidal cells are relatively homogeneous, having long axons projecting
outside the cortex. Non-pyramidal cells or intrinsic neurones have been classified according to their
dendritic and axonal characteristics. Their axons remain inside the cortical territory in which they
are located. Cells with unspecific axonal arborizations have been found in all cortical layers, with
the exception of layer I, and in all mammals thus far examined. Their name indicates the lack of
specialized axonal arborizations, in contrast to other cell varieties having uniquely elaborated axonal
morphologies.
The chapter further explores the morphology and distribution of specific thalamo-cortical fibres
and the varieties of intrinsic cells they contact. Special interest was dedicated to spiny stellate cells
with recurving (ascending) axons, which represent the principal relay of thalamic afferent fibres in
the visual cortex of primates. The varieties of spiny stellate cells and their proportions depend on the
animal and a comparison with other cortical areas show interesting and contrasting differences.
We conclude that, from a general point of view, the neocortex appears uniform at a rather funda-
mental level of organization, even though significant differences were found among certain varieties
of cells, some of which may be unique to a given species. Some of these differences can be minimal
in closely-related species, but they appear substantial when the comparison is made between distant
animals.
KEYWORDS: comparative neuroanatomy, Golgi method, intrinsic neurones, neocortex, visual cortex
1. INTRODUCTION
The cerebral cortex of mammals is an extremely complex laminated structure on which

sensory organs are mapped in specific cortical domains, known as primary sensory areas.
195
196 Facundo Valverde, Juan A. De Carlos and Laura López-Mascaraque
Figure 10.1. Nissl-stained sections of the auditory cortex in five representative mammals reproduced at the
same magnification. (A) hedgehog, (B) rat, (C) cat, (D) monkey, (E) man. The arrangement in vertical columns
is clearly visible in the cat, monkey and man. (F) shows a small sector of the auditory cortex of man showing
giant pyramidal cell. Scale bar at the left of F equals 200 µm; scale bar inside F equals 100 µm.

The Cerebral Cortex of Mammals 197
From olfaction to vision, each sensory modality has its own territory in the cortex which
remains relatively constant from lower mammals to humans. Beyond the primary cortical
areas, the cerebral cortex contains multiple secondary somatic and motor areas in which
afferent inputs are finally elaborated into complex behavioural responses. The organiza-
tion of this dominating structure, which achieves, from a human perspective, its highest
complexity in man, is the result of a slow evolutionary process that supposedly began
about 100 million years ago, evolving to reach a final prototype in primitive Insectivora
(Valverde, 1983).
The history of cortical discoveries is a passionate story of the human endeavour to
unveil the organization of this remarkable structure (Lorente de Nó, 1949; Jones, 1984a).
In 1878 Bevan Lewis proposed the six-layer scheme, which has remained since then, for
naming the different cortical layers (Figure 10.1). Based on this plan, the cerebral cortex
was then subdivided into two separate types: isocortex and allocortex, terms originally
coined by Brodmann (1909) and Cécile and Oskar Vogt (1919). The isocortex or neo-
cortex represents that part of the cerebral mantle in which a six-layered stratification can
always be recognized. The allocortex consists of the archicortex (hippocampus, including
the dentate gyrus) and paleocortex (olfactory cortex proper), which exhibit simple laminar
structures. Both cortical types are more or less clearly separated from each other by
a number of intervening para-cortical areas, which have been a matter of interest for the
study of certain evolutionary trends (Sanides, 1970).
The cerebral cortex contains nerve cells and fibres arranged in parallel layers. Each
layer has its own individuality, as given by its specific cell varieties and input and
output connections; but the various layers do not operate in isolation. The elements which
characterize them are linked to the constituents of the remaining layers. This is the philo-
sophy underlying the concepts of module operation, envisaged as spatial modules of
vertically-linked cell groups surrounding cortico-cortical and specific cortical afferents
(Szentágothai, 1975, 1978, 1979; Eccles, 1984). Fundamental aspects of cortical organiza-
tion are based on these anatomical entities, which will provide us a reference frame for the
study of those structural features that either remain constant or vary across mammalian
species. In this chapter, we will review our observations on the structure of the cerebral
cortex, obtained with the Golgi method in a large collection of more than 2000 brains,
from Insectivora to Primates. Due to the nature of our previous work, we will concentrate
on specific aspects of cortical organization from a comparative point of view, with emphasis
on study in the primary visual cortex, area 17.
2. PHYLOGENETIC APPROACH TO CORTICAL ORGANIZATION

ALLOCORTEX VERSUS NEOCORTEX
There are fundamental characteristics which differentiate the neocortex from the allocortex.
Apart from differences in pyramidal cells, which have a typical shape in the neocortex
different from that in the allocortex, in neocortex and mainly in primary sensory areas,
thalamocortical afferent fibres end in middle cortical layers. In the allocortex, the major
afferent fibre system is formed by a strong input extending into the plexiform, or first
layer; e.g. the lateral olfactory tract in the primary olfactory cortex, or perforant bundles
in the hippocampal and dentate gyri (Valverde, 1998). The dependence on afferent fibres
in allocortex is so strong that, for instance, unilateral olfactory bulb ablation in the rat

Figure 10.2. Transverse section through the somato-sensory cortical area in the hedgehog. In basal Insecti-
vores, the fundamental plan of neocortical organization can hardly be recognized. The cortex shows a thick layer
I, with densely populated layer II containing large polymorphic cells (e.g. cells l–t) with dendrites richly covered
by spines. An exuberant number of dendrites penetrate layer I where they receive afferent input. The granular
layer IV cannot be recognized, instead other pyramidal or pyramid-like cells (e.g. cells a–c) and polymorphic
cells with smooth dendrites, and various axonal morphology (e.g. cells d–k) populate the middle cortical layers
(III–IV). Layers V and VI appear well developed with large pyramidal cells and long apical dendrites reaching
layer I (e.g. cells u–w). When axons could be traced, they were numbered followed by the letter of their parent
cell (e.g. 1a). Camera lucida drawing. Golgi method. Hedgehog 30 days old. (From Valverde, 1986. Reprinted
with permission from Elsevier Science).

causes the death of almost all layer II pyramidal cells in the olfactory cortex (Heimer and
Kalil, 1978; Friedman and Price, 1986; López-Mascaraque and Price, 1997). Comparative
anatomical studies have shown that these structural features remain constant throughout
species, and can, therefore, be considered as essential for the information processing
typical of allocortical formations. On the other hand, primary sensory neocortical areas
receive their afferent inputs through relays made in different thalamic nuclei, ascending
vertically from the underlying white matter, which arborize mainly on middle (layers III
and IV) and upper (layer I) cortical levels.
In spite of these strong differences between the allocortex and isocortex, the study of
primitive Insectivora reveals that both types of cortical organization might have been
derived from a common prototype. In the hypothetical primordial vertebrate ancestor, the
end-brain developed almost exclusively as an olfactory brain, and successive differenti-
ations are thought to have taken place by the invasion of diverse somatic systems into this
primitive anlage. The hedgehog, Erinaceus europaeus, once called a “survivor of the
Paleocene”, is considered to be one of the most direct descendants of the primitive placen-
tals from which mammals originated (Simpson, 1945; Romer, 1966). The brain of the
hedgehog shows a pattern of predominantly olfactory structures with a small neocortex,
resembling the architecture of surrounding allocortical formations. The main cytoarchitec-
tonic features of this initial neocortex (Figures 10.1, 10.2) show an extremely wide layer I
(300 µm) with strong thalamic input, an accentuated layer II with extraverted pyramidal
cells, and an absence of layer IV, so that layers III and V have no distinct boundaries
(Sanides, 1970; De Carlos, 1986; Valverde and Facal-Valverde, 1986; Valverde et al.,
1986; Glezer et al., 1988).
3. THE MODEL OF NEOCORTICAL ORGANIZATION
Understanding the cerebral cortex has been dominated by the concept of columnar organ-
ization. This term was first used to designate the intrinsic neuronal connectivity within
a vertical cylinder, or column, of cortical tissue which has a central axis formed by specific
thalamo-cortical afferent fibres (Lorente de Nó, 1949). For several years, this elementary
unit was envisaged as a functional concept, because it explained earlier results obtained by
neurophysiology in the somatosensory (Mountcastle, 1957), auditory (Woolsey, 1960)
and visual (Hubel and Wiesel, 1962) primary areas; namely, that cells having similar func-
tional properties appear to be arranged along the vertical axis of the cortex from the pia to
the white matter. In the primary somatosensory (Mountcastle, 1979) and visual cortices
(Hubel and Wiesel, 1977), functional columns are defined in terms of receptive field prop-
erties, while in the primary auditory cortex, they are interpreted in terms of best frequency
responses (Abeles and Goldstein, 1970; Imig et al., 1982).
From a strictly anatomical point of view, vertical arrangements in the cerebral cortex
are not obvious in all the species examined. In this regard, we compared transverse sections
of the primary auditory cortex stained with the Nissl method in a representative series of
mammals (hedgehog, rat, cat, monkey and man). This comparison is shown in Figure
10.1, where we represent the auditory cortices of these five different species at the same
magnification. For instance, the thickness of the auditory cortex of the hedgehog and rat
appears to be the same, but the layer distribution is different. The hedgehog does not show
a clear granular layer IV; in addition, the thickness of its layer I is almost double that of

layer I of the rat. These distinctive features also occur in the brain of the dolphin, which
may constitute an interesting link between marine and terrestrial mammals (De Carlos, 1986).
The first anatomical evidence of an organization that could be correlated with physio-
logical columns was obtained by Hubel and Wiesel (1969) in the visual cortex of the
macaque monkey. Using silver impregnation techniques, they showed the existence of
two independent and overlapping systems of bands coincident with the system of ocular
dominance bands. The presence of comparable arrangements was subsequently demon-
strated in almost every cortical area. Newly developed methods for the tracing of pathways
and the use of enzymatic reactions and radioactive tracers provided evidence that the
cerebral cortex, as far as the sensory thalamic input is concerned, is organized into regu-
larly spaced, periodic subdivisions readily attributable to the spatial distribution of afferent
fibres in the somatosensory, auditory and visual cortices of various animal species. This
demonstrates that some of these thalamic distributions have specific patterns of cortical
representation, such as the barrel field in the somatosensory cortex of rodents (Welker,
1976; Woolsey and Van der Loos, 1970), the banding pattern of the ocular dominance
system of the visual cortex in the monkey (Le Vay et al., 1975) and the radial arrangement
of cell chains with the same characteristic frequencies in the auditory cortex (Abeles and
Goldstein, 1970; Brugge and Reale, 1985). This pattern of orderly partitions is not unique
to cortical primary sensory areas: there is now evidence that cortico-cortical (callosal and
association) projections also branch into alternating, vertically-oriented patches segre-
gated from thalamo-cortical afferents (Jones, 1984b; Wise and Jones, 1976; Záborsky and
Wolf, 1982).
4. INTRINSIC CORTICAL CELLS—ORIGIN, AREAL AND INTER-SPECIES

VARIATION
The cerebral cortex of mammals contains an immense variety of cell types. However,
from a general point of view, cortical neurones can be classified into two principal types:
pyramidal cells and non-pyramidal cells. Pyramidal cells represent the fundamental type
of cell: they have long axons projecting outside the cortex. Non-pyramidal cells, or
intrinsic cells, have axons that remain inside the cortical area where they are located, and
do not project extracortically. There is a general consensus indicating that pyramidal cells
represent the essence of cortical architecture, while intrinsic neurones add its “flavour”.
Despite considerable efforts during recent years, identification of the diverse types of
intrinsic neocortical cells is not yet complete. In addition, recent studies have shown that
cortical cells have a variety of neurotransmitters which can be used as selective markers
for different populations of cells (see reviews in DeFelipe and Fariñas, 1992; Nieuwenhuys,
1994).
Pyramidal cells, and some intrinsic cells, have different developmental origins. Cortical
projection neurones (pyramidal cells) originate from the neocortical ventricular zone, ascend-
ing to the surface by following the processes of radial glial cells (Rakic, 1972, 1985;
Valverde et al., 1989; Misson et al., 1991; Gould et al., 1999). In contrast, recent evidence
has shown that the ventricular and subventricular zones of some transient basal structures,
which develop in early embryonic stages—the ganglionic eminences—give rise to another
population of cortical neurones. These cells, cross the cortico-striatal boundary to enter
the cortical neuroepithelium using a tangential migration path (De Carlos et al., 1996).

They follow different routes: preplate (marginal zone or future layer I) and intermediate
zone (future white matter), and have been characterized as GABAergic (Anderson et al.,
1997a, 1999; Tamamaki et al., 1997; Zhu et al., 1999). These neurones are not transitory,
but become diluted in the thickness of the cerebral cortex during development, contrib-
uting there to the intracortical circuitry.
A new area of research is rapidly growing related especially to study of the diverse
genes controlling the processes of areal specification and connectivity (Anderson et al.,
1997b; Donoghue and Rakic, 1999; Rubenstein and Rakic, 1999). Study of the depend-
ence on a number of developmental genes is rapidly evolving and will provide new
advances concerning the possibility that differences in cellular origins might explain
which features remain constant throughout species, or to what degree cell types or their
ramification patterns differ between species (Karten, 1997). Another field of interest per-
tains to recent immunocytochemical studies of specific subpopulations of intrinsic cells
that have been shown to be immunoreactive for different proteins. Their differences in
laminar distribution and their relations with specific populations of pyramidal cells reflect
diverse patterns of cortical circuitry that have been correlated with functional subdivisions
of the neocortex (Kritzer et al., 1992; Fujita and Fujita, 1996; Elston et al., 1999).
In spite of these new advances, the Golgi method has remained one of the most elegant
procedures for studying the morphology of neurones. The fact that the various Golgi
methods selectively impregnate only a small proportion of neurones, but often stain them
in their entirety, has made them the methods of choice for studying individual neurones
for almost a century. Successful impregnation of brain tissue with this method provides
a complete picture of neuronal morphology, including all dendritic branches, axonal
arbors and finest terminal ramifications. It still remains true that most classes of intrinsic
neocortical cells, along with their inter-species and inter-areal varieties, have been
described based on Golgi preparations. In 1979, we proposed a classification of intrinsic
neurones, that takes into account dendritic, as well as axonal characteristics (Fairén and
Valverde, 1979). A short review follows:
(a) Cells with recurving axonal arcades. These cells either have smooth dendrites or bear
a very small number of spines. They have been found in all cortical layers of the
mouse area 17, with the exception of layer I (Figure 10.3, cells 16–18; Figure 10.4, b).
The axon emerges from the upper pole of the cell body, or from the base of a dendrite,
ascends vertically and forms thick recurving arcades of horizontal and oblique
branches. These cells are entirely comparable to others present in the monkey visual
(Lund, 1973) and somato-sensory (Jones, 1975) cortices.
(b) Cells with ascending axons. These are multipolar cells whose dendrites range from
smooth (spine-free) to sparsely spiny, even though in adult animals dendrites have
been found so richly endowed with spines that their classification as sparsely spiny
would seem inadequate (Figure 10.3, cells 8–15; Figure 10.4, e). The axon emerges
from the upper pole of the cell body and forms a straight ascending stem reaching
layer I, where it arborizes. Cells of this variety that are located in upper cortical layers
II and III have long horizontal axonal trajectories into layer I, while those located in
layers IV to VI have axons with thin descending collaterals forming a relatively dense
local arborization, and a second group of axonal branches contributing sparsely to
layer I. The more deeply located cells of this variety correspond to the type described
by Cajal (1911) as Martinotti cells.

202
Facundo Valverde, Juan A. De Carlos and Laura López-Mascaraque
Figure 10.3. Axonal patterns from various intrinsic cortical cells arranged according to the depth of their cells bodies below the pial surface collected from diverse neocortical
areas of the mouse. The first row (1–7) comprises axons ramifying above the cell body. The second row (8–15) shows ascending axons with a vertical principal branch emitting
collaterals at different levels. The first three examples in the third row (16–18) represent examples of axons with recurving and long descending collaterals ramifying above and
below their cells bodies. Examples of axonal complexes in the third row 19 to 21 represent types with diversely oriented axonal collaterals. Examples 22 and 23 correspond to
axons of pyramid-like cells with developed intra-cortically and forming strong recurrent arcs. The example 24 corresponds to a typical pyramidal cell with collaterals distributed
in layers IV and V. The age and depth from the surface (in microns) appears in each example. Camera lucida drawings. Golgi method. (From Valverde, 1976.)

Figure 10.4. Transverse section through the visual cortex of the mouse. Varieties of cells with intracortical
axons (intrinsic cells b and e) and pyramidal cells (cells a, c, d and f). The figure shows the contrasting differ-
ences between one cell with recurving axonal arcades (b), and one cell with ascending axon reaching layer I (e).
Typical superficial (c), middle (d) and deep (a, f) pyramidal cells were also reproduced. In the cell labelled f, the
possible sites of synaptic contacts provided by the axon of cell e were labelled s. When axons could be traced,
they were numbered followed by the letter of their parent cell (e.g. 1b). Camera lucida drawing. Golgi method.
Mouse 21 days old. (From Fairén and Valverde, 1979).
(c) Cells with specialized axonal arborizations. Included in this category are various
kinds of cells having smooth or sparsely spiny dendrites and uniquely elaborated axonal
arborizations, each with a peculiar morphology that suggests the name it bears. All of
them have in common that they are inhibitory interneurones.
Basket cells (Figure 10.5) have been considered to be large multipolar cells with
smooth dendrites and long horizontal axon collaterals forming pericellular nests
(baskets) around pyramidal cell bodies in layers III and V. First described by Cajal
(1911) in the human motor and visual cortices, several varieties have been
described in more recent times. However, not only does their existence remain
a matter of considerable debate, but strong evidence for their presence outside the
sensory and motor areas, or in species other than primates and carnivores, is also
still lacking (see reviews in Jones and Hendry, 1984 and Fairén et al., 1984).

Figure 10.5. Several terminal axonal branches from a large basket cell (not shown) outlining an unstained
large pyramidal cell of layer V in the human motor cortex. Camera lucida drawing made by F. Valverde from
a Golgi preparation made by S. Ramón y Cajal. (From the Cajal Museum.)
Recent studies using parvalbumin and GABA immunoreactivity have confirmed

the convergence of multiple axonal terminals around pyramidal and non-pyramidal
cells; these are considered to belong to large basket cells (Somogyi et al., 1983;
Hendry et al., 1989; Kisvárday et al., 1993), providing powerful inhibitory influ-
ences upon the cells they contact.
Chandelier cells, so named for their axonal morphology with terminal branches
resembling the vertical “candles” of a lamp fixture (Figure 10.6, a), were first
described by Szentágothai and Arbib (1974) in the gyrus cinguli of the cat. They
appear similar to type 4 cells of Jones (1975). Somogyi (1977) showed for the first
time, in the visual cortex of the rat, that the axon terminals composing the charac-
teristic “candles”, form symmetric synaptic junctions with the initial axonal
segments of pyramidal cells. Subsequent studies confirmed their existence in the
visual cortex of the cat (Fairén and Valverde, 1980). Since then, numerous observ-
ations have shown that chandelier cells are by no means specific for the visual
cortex, for they have been found not only in different neo- and paleo-cortical areas,
but also in practically all mammals thus far studied (Somogyi et al., 1982; Fairén
and Valverde, 1980; Valverde, 1983; Fairén et al., 1984; Peters, 1984; Kisvárday et al.,

Figure 10.6. Transverse section through the visual cortex of the cat. The figure is a composite drawing from
several adjacent sections. One specific afferent fibre (entering section from right) labelled 1, ramifies in layer IV.
Large stellate cells with markedly spinous dendrites were labelled b, c and d, representing examples of cells
receiving direct synaptic contact from specific cortical afferents. The figure also shows one cell with smooth
dendrites (e), identified as a “clewed cell”, and one example of chandelier cell (a) with characteristic axonal
terminals (2a) devoted to contact with the initial axonal segments of pyramidal cells (not stained), and one
descending axon (1a). Axons were numbered followed by the letter of their parent cell (e.g. 2b). Camera lucida
drawing. Golgi method. Cat 1 month old. (From Fairén and Valverde, 1979.)

1986; De Carlos et al., 1985, 1986, 1987; Marín-Padilla, 1987). Since chandelier
cells perform a strong inhibitory action on the axon initial segment of pyramidal
cells, they have been considered to have a decisive influence on pyramidal cell
output. Chandelier cells probably use GABA as a neurotransmitter, as it has been
reported that chandelier axon terminals in contact with the initial axon segment of
pyramidal cells are glutamic acid decarboxylase (GAD)-positive (Peters et al., 1982),
and some of them have been shown to contain parvalbumin (DeFelipe et al., 1989).
Clewed cells, also known as neurogliaform, spiderweb or clutch cells (Figure 10.6, e;
Figure10.7, g), were first described in sublayer IVc in the primary visual cortex,
area 17, of the monkey as being characteristically small cells with beaded, smooth
dendrites and one axon that soon resolves into numerous, densely interwoven
collaterals, forming a plexus of strictly local nature (Valverde, 1971). This cell
corresponds to the type 5 cell of Jones (1975) and a comparable cell variety has
also been described in the monkey area 18 (Valverde, 1978). Further studies have
demonstrated that the main targets of clewed cells in layer IV of the visual cortex in
the monkey are spiny stellate cells, and the presence of GABA in these cells suggests
that they provide a strong inhibitory input to these stellate cells (Kisvárday et al.,
1986).
Cells with vertical axonal bundles, also known as “double bouquet cells” were
described by Cajal (1911) as one cell variety having axonal and dendritic trees
arranged in vertical bundles. Although the vertical arrangement of the dendrites is
not clearly evident in some cases, the most distinctive feature is the distribution of
their axons running perpendicular to the surface (Figure 10.8, a, b), extending
through several cortical layers and forming synaptic contacts with dendritic spines
of pyramidal and non-pyramidal cells. These cells were described in layer II and
the upper part of layer III of area 18 in the monkey (Valverde, 1978). Since then,
they have been found and described in several neocortical areas in man and other
primates, in the cat and in rodents (Somogyi and Cowey, 1984). These cells most
probably use GABA as their primary transmitter and, as some recent studies have
shown, they also contain several neuropeptides (DeFelipe et al., 1990) and calcium-
binding protein (DeFelipe and Jones, 1992). Using immunocytochemical tech-
niques, double bouquet cells have been found to be particularly abundant in the
monkey area 18, probably being related to different processing of visual stimuli
(DeFelipe et al., 1999).
(d) Non-pyramidal cells with spiny dendrites. These are multipolar cells with dendrites
covered by spines, as in pyramidal cells, but lacking the typical apical dendrite. It is
clear that they do not constitute a uniform group: subgroups can be defined according
to cell size, dendritic orientation and axonal distribution, but in all cases their axons
are considered as generalized, due to the lack of specific terminal arborizations (as
found for non-pyramidal neurones with no or few spines). Some non-pyramidal cells
with spiny dendrites have axons projecting to the contralateral hemisphere via the
corpus callosum (Valverde, 1986). These cells are abundant in practically all cortical
layers, but most notably in layers III and IV, receiving specific afferent fibres. Due to
the importance of these cells in the context of cortical afferent fibres, we will consider
their detailed description in the next section.

Figure 10.7. Transverse section through the visual cortex, area 17 of the monkey. The drawing is a composite
illustration of varieties of cells recorded from several adjacent sections. Specific cortical afferents (geniculo-cortical
fibers) ascend from the white matter forming elaborate terminal ramifications in sublayer IVc (1, 2). Medium-sized
conventional pyramidal cells in layer III (a) and sublayer IVb (f) have axons (1a, 1f) descending to lower layers.
Small stellate cells with spinous dendrites (h, i, j) show ascending, recurrent axons (1h, 1i, 1j) ending with localized
axonal plexuses in sublayer IVa. In this sublayer, there is another variety of stellate cells with spinous dendrites (b, c)
running horizontally. Layer V also contains similar varieties of small pyramidal cells (l, m) with ascending recurv-
ing axons (1l, 2m). Two stellate cells with spinous dendrites (d, e) and one “clewed cell” (g) can also be observed.
Camera lucida drawing. Golgi method. Adult monkey Macaca rhesus (From Fairén and Valverde, 1979).

Figure 10.8. Transverse section through the visual cortex, area 18 of the monkey. The image shows cells with
vertical axonal bundles and examples of pyramidal cells. Cells with vertical axonal bundles or “double bouquet
cells” (a, b), so named by S. Ramón y Cajal, display characteristic vertical axonal trees arranged in tightly
packed vertical bundles spanning several cortical layers. They appear to be typical for area 18. Layer III appears
subdivided into IIIa and IIIb; together with layer IV, they contain large pyramidal cells (c, d, e and f) with
descending axons and collateral branches running for long distances. Collateral branches were numbered
consecutively followed by the letter of their parent cell (e.g. 3a). Inset drawing shows the region from which the
drawing was obtained. Camera lucida drawing. Golgi method. Adult monkey Macaca rhesus (From Valverde,
1978. Reprinted with permission from Springer-Verlag).

5. THE TARGETS OF CORTICAL AFFERENT FIBRES
Target cells for thalamo-cortical fibres appear to vary widely in different mammals and in
different cortical areas, and a common pattern is far from the rule. Numerous observa-
tions, based on studies carried out with the electron microscope alone, or after appropri-
ately placed lesions in the thalamic principal relay nuclei, indicate that practically all
synapses identified from thalamic terminals are of the asymmetric (presumed excitatory)
type (Peters and Feldman, 1976; Peters et al., 1979; White, 1979). Hence, it has been pro-
posed that nearly all cell bodies and dendrites located in the domain covered by thalamic
afferent fibres in the middle cortical layers, and capable of forming asymmetric synapses,
can receive thalamic input (Peters and Feldman, 1977).
The use of procedures combining Golgi techniques with electron microscopy, have
shown that the majority of dendritic spines receiving thalamo-cortical synapses belong to
dendrites of pyramidal cells whose cell bodies lie in layers III, IV and V, along with the
varieties of intrinsic cells located in the middle cortical layers that constitute the popula-
tion of spiny stellate cells (Somogyi, 1978; Peters et al., 1979; Davis and Sterling, 1979;
White, 1979; Hornung and Garey, 1981; Freund et al., 1985). As indicated, stellate cells
do not constitute a uniform population: not only are there significant differences in dend-
ritic and axonal morphology among various animal species and in different neocortical
areas, but their synaptic relations also appear different. Stellate cells have been classified
as: lacking spines (smooth stellate cells); having few spines (sparsely spiny stellate cells);
or having a high density of spines (spiny stellate cells). However, none of these types
constitute a uniform group, and other criteria, such as the form and distribution of the
dendritic tree, cell size, and, principally, the axonal pattern, have been considered to give
a comprehensive classification (Valverde, 1976; Feldman and Peters, 1978; Fairén and
Valverde, 1979; Fairén et al., 1984; Peters and Jones, 1984).
In the mouse, somato-sensory cortex corresponding to the barrel field, spiny stellate
cells are particularly abundant (Woolsey et al., 1975). They have dendrites covered by
numerous spines, no trace of an apical dendrite, and the patterns of dendritic orientation
depend on whether the cell body is located in the barrel wall or in the barrel hollow.
Except for these highly characteristic elements in this particular neocortical area, the
majority of postsynaptic elements (about 83%) are represented by dendritic spines belong-
ing to apical and oblique dendrites of pyramidal cells located in layers V and VI, and to
the basal dendrites of layer III pyramidal neurones (Peters and Feldman, 1977). There is
a small proportion of postsynaptic elements (about 17%) represented by the shafts of
smooth or sparsely spiny stellate cells residing in layer IV (Valverde, 1968; Peters et al.,
1976).
Spiny stellate cells have been a matter of considerable interest since Kelly and Van
Essen (1974) showed in the visual cortex of the cat that some cells, recovered by intra-
cellular dye injections after identifying their functional characteristics, belong to this
category. These spiny stellate cells (Figure 10.6, b–d) were similar to those described in
Golgi preparations in the same cortical area (Cajal, 1911, 1921; O’Leary, 1941; LeVay,
1973; Fairén and Valverde, 1979; Lund et al., 1979; Peters and Regidor, 1981; Meyer and
Ferres-Torres, 1984). It was clearly established that these cells receive direct contact from
thalamic terminals (Hornung and Garey, 1981; Martin and Whitteridge, 1984; Freund et al.,
1985; Figure 10.6, 1) and some of them, especially those cells located at the 17/18 border,
have axons projecting to the contralateral hemisphere (Sanides, 1979; Innocenti, 1979;
Hornung and Garey, 1980; Meyer and Albus, 1981). The fact that these cells project to

the white matter had already been mentioned by Cajal (1921), and confirmed later after
intracellular injections with horseradish peroxidase (Gilbert and Wiesel, 1979; Lin et al.,
1979).
In primates, target cells for thalamic axons are represented by intrinsic cells which do
not project outside the cortex. The visual cortex of the monkey (the striate area or area 17)
has been studied intensively since the demonstration of the system of ocular and
orientation columns (Hubel and Wiesel, 1969, 1972). In the visual cortex of primates,
layer IV has been subdivided into three tiers, labelled as IVa, IVb and IVc (Figure 10.7).
Sublayer IVc contains a population of spiny stellate cells with recurving (ascending)
axons, as well as varieties of cells with smooth dendrites. These spiny stellate cells are
the most distinctive element found in this sublayer, which probably has no counter-part
in non-primate species (Valverde, 1971). They have been a matter of considerable inter-
est in the analysis of the anatomical and functional organization of the visual cortex. The
axons of these cells originate at the lower pole of the cell body, descend for a short
distance and then turn upwards, forming single or several characteristic loops in ascend-
ing bundles that reach layer III and sub-layers IVa and IVb, where they develop into
distinct elongated plexuses of terminal fibres (Figure 10.7, h, i). Their general morpho-
logy has often been considered akin to pyramidal cells with truncated or absent apical
dendrites (Lund, 1984). They have even been considered as the result of evolutionary
transformations of pyramidal cells (Nieuwenhuys, 1994) probably related to changes in
the distribution of cortical afferents, which presumably shifted from layer I in primitive
mammals to predominate in middle cortical layers. Thus, in the study of the forms of
different cells with spiny dendrites, including pyramidal cells, one gets the impression
that all of them may share a common phylogenetic origin and that a continuum can be
traced from lower forms to the primate brain. It is possible that this shift modified the
intrinsic neocortical organization.
In spite of these differences, the visual cortex of the cat and monkey show strong simi-
larities that have been considered in a detailed study, using the Golgi method, by Lund and
collaborators (Lund et al., 1979). However, layer IVb in primates, corresponding to the
stria of Gennari, which does not receive direct input from the thalamic lateral geniculate
body, has no counterpart in cat area 17. Retrograde labelling with HRP revealed the exist-
ence of a reciprocal projection from the visual area of the superior temporal sulcus (STS)
to the striate area in Callithrix, ending in layer IVb (Spatz, 1977; Rockland and Pandya,
1981). The tangential spread of some axonal collaterals in the stria of Gennari, derived
from descending axons of large pyramidal and stellate cells, extends for up to several
millimeters (Colonnier and Sas, 1978; Fisken et al., 1975; Valverde, 1985).
There is a second and major type of target cell for thalamic axons, corresponding to
neurones with smooth dendrites and beaded axons. The most abundant ones show a very
limited axonal field formed by densely interwoven axonal collaterals and recurving
dendrites, giving the ensemble the appearance of a ball of yarn (Figure 10.7, g). They
were named “clewed cells” when observed for the first time (Valverde, 1971), and have
been considered previously as having specialized axonal arborizations. These cells have
been a matter of considerable interest, not only for their specific axonal patterns, but
also because they form symmetrical synaptic contacts and appear, therefore, to be inhibit-
ory interneurones. These cells are most probably identical with the small basket (clutch)
cells described in the visual cortex of the cat (Kisvárday et al., 1985) and monkey
(Kisvárday et al., 1986). Their synaptic relations with spiny stellate cells suggest an

interesting type of local interaction (Kisvárday et al., 1986), and they have been
considered as important key pieces in conceptual models of neocortical operation
(Eccles, 1981, 1984).
In view of the importance of spiny stellate cells as common targets for thalamic afferent
fibres, several studies have been made which show that their proportion varies depending
on the animal and cortical area. In the somato-sensory area of the mouse, about one-half of
the cells in layer IV have dendrites covered by spines (Woolsey et al., 1975). Spiny stellate
cells in the visual cortex of the rat (Golgi method) comprise around 11% (Feldman and
Peters, 1978). In the visual cortex of the cat, non-pyramidal cells account for 60–80% of
the cells in layer IV, where spiny stellate cells are found to be as common as cells with
smooth dendrites (Garey, 1971; Winfield and Powell, 1976). In the visual cortex of the
monkey, exclusively spiny stellate cells and cells with smooth dendrites form the popula-
tion of target cells in sublayer IVc, where the spiny stellate variety constitutes about 95%
(Mates and Lund, 1983). Spiny stellate cells have not been found in the visual cortex of
the mouse and they are virtually absent in the rabbit neocortex. Non-pyramidal spiny cells
have been detected only occasionally in layers III and IV of the auditory cortex of the
rabbit, where 87% of all impregnated neurones (Golgi-Cox method) are pyramidal cells
(McMullen et al., 1984). In our previous studies of the neocortex in the hedgehog, we
could not detect any typical spiny stellate cells (Valverde, 1983).
6. COMPARISON WITH OTHER CORTICAL AREAS
We have been interested in the comparison of area 17 (Figure 10.7) with area 18 (Figure
10.8) and other areas of the temporal lobe in the monkey brain (Valverde, 1978). Spiny
stellate cells are apparently absent from layer IV in area 18 (Lund et al., 1981); instead,
this layer contains small pyramidal cells with recurrent axons similar to those found in the
upper part of layer V of the visual cortex. Their axons form strong recurrent arcades
ascending to sublayer IIIb in area 18, apparently devoted to contacting large pyramidal
cells. In the cortex of the superior temporal sulcus, identical small pyramidal cells with
strong recurving axons have been found. Both layers III and IV also contain small stellate
cells with smooth, beaded dendrites having ascending axonal branches, as well as variet-
ies of large multipolar and bipolar cells, which are particularly abundant in these and
neighbouring auditory areas. One specific type of cell was that found by Cajal (1900) only
in the human auditory cortex. This type of cell is characterized by its large soma size
(40–60 µm), triangular or fusiform morphology, and presence in all cortical layers except
layer I. Thus far, we have found this type of giant cell (Figure 10.1F) only in the auditory
cortex of human specimens (De Carlos, 1986).
There is evidence that in certain cortical areas with a well-developed layer IV, thalamic
afferents are not distributed in this layer. In several of the parietal and temporal fields of
the monkey (e.g. areas 5, 7, second and third temporal), thalamic terminals are mainly
distributed upon pyramidal cells of layer IIIB suggesting endings upon basal dendrites and
proximal portions of their apical dendrites (Jones and Burton, 1976). This points to a dif-
ferent organization of the intrinsic circuitry of this layer, including different chemical
characteristics for synaptic input (see review in DeFelipe and Fariñas, 1992). In this case,
targets for thalamic fibres are represented by those cells with smooth dendrites and by
typical pyramidal cells.

7. IS THERE A BASIC PLAN OF NEOCORTICAL ORGANIZATION?

CONTRASTING DIFFERENCES OF TARGET NONSPECIFICITY
As one ascends the phylogenetic scale, the morphology of cortical neurones becomes less
uniform; more specific neuronal types appear, and the number of non-pyramidal cells
increases (Lorente de Nó, 1949). The neocortex, in general, does not show a uniform struc-
ture; instead, it contains a number of diverse cortical areas which are distinguished by
different anatomical and functional characteristics. However, from a general point of view,
the neocortex appears functionally uniform at a rather fundamental level of organization.
As just reviewed, layers III and IV are considered the principal level of termination for
specific cortical afferent fibers. From here, impulses are relayed to layers II and III, which
are the source of long and short association fibres and intrinsic descending connections
with layers V and VI, which contain the majority of cells projecting subcortically, and
a number of cells with ascending axons. However, the study of cell varieties and the mode
in which they intervene in intracortical microcircuits in different animals, clearly shows
the existence of important variations, some of which may be unique for a given animal.
Significant differences have been found among certain varieties: cells with ascending,
recurving axons, apparently unique to the primate brain; spiny stellate cells typical of the
somato-sensory cortex of rodents; spiny stellate cells with axons projecting to the white
matter in the visual cortex of carnivores; and extraverted pyramidal cells in the neocortex
of insectivores which have no equivalent in other animals. Important differences were
found in relation to the targets for thalamic afferent fibers, confirming that spiny stellate
cells are by no means a common target for thalamic inputs.
At a more detailed level, the existence of differences, representing adaptations to
unique cortical architectures, between the most closely-related animals have long been
suspected, even though their morphological characterizations still remain elusive. How-
ever, recent studies have shown, for the first time, significant differences in the cortical
architecture of the visual cortex between humans and apes, the animals most closely
related to humans (Preuss et al., 1999).
The stage of neocortical organization attained in primates, including man, may have
been accomplished, not by replication of a basic cortical module already present in lower
mammals (Glezer et al., 1988), but by reshaping dendritic and axonal arbors of various
categories of neocortical cells, developing into different intracortical connectivities. These
differences can be minimal in closely-related species, but they appear very substantial
when the comparison is made between distant animals.
ACKNOWLEDGEMENTS
Supported by Ministerio de Educación y Ciencia, Grant number PB96-0813 and Consejería

de Educación y Cultura de la Comunidad de Madrid, Grant number 08.5/0037/1998.
REFERENCES
Abeles, M. and Goldstein, M.H. (1970) Functional architecture in cat primary auditory cortex. Columnar
organization and organization according to depth. Journal of Neurophysiology, 33, 172–187.
Anderson, S.A., Eisenstat, D.D., Shi, L. and Rubenstein, J.L.R. (1997a) Interneuron migration from basal
forebrain to neocortex: dependence on Dlx genes. Science (Washington), 278, 474–476.

Anderson, S.A., Qiu, M.-S., Bulfone, A., Eisenstat, D.D., Meneses, J., Pedersen, R. and Rubenstein, J.L.R.
(1997b) Mutation of the homeobox genes Dlx-1 and Dlx-2 disrupt the striatal subventricular zone and
differentiation of late-born striatal neurons. Neuron, 19, 27–37.
Anderson, S.A., Mione, M., Yun, K. and Rubenstein, J.L.R. (1999) Differential origins of neocortical projection
and local circuit neurons: role of Dlx genes in neocortical interneurogenesis. Cerebral Cortex, 9, 646–654.
Brugge, J.F. and Reale, R.A. (1985) Auditory cortex. In: A. Peters and E.G. Jones (eds), Cerebral Cortex,
Association and Auditory Cortices, Vol. 4, New York and London: Plenum Press, pp. 229–271.
Cajal, S.R. (1900) Estudios sobre la corteza cerebral humana: III. Corteza acústica. Revista Trimestral Micrográ-
fica, tomo V, 129–183.
Cajal, S.R. (1911) Histologie du Systéme Nerveux de l’Homme et des Vertébrés. Vol. 2. Paris: Maloine
(Reimpress 1955, Madrid: Instituto Cajal, CSIC).
Cajal, S.R. (1921) Textura de la corteza cerebral del gato. Trabajos del Laboratorio de Investigaciones
Biológicas, 19, 113–144.
Colonnier, M. and Sas, E. (1978) An anterograde degeneration study of the tangential spread of axons in
cortical areas 17 and 18 of the squirrel monkey (Saimiri sciureus). Journal of Comparative Neurology, 179,
245–262.
Davis, T.L. and Sterling, P. (1979) Microcircuitry of cat visual cortex: Classification of neurons in layer IV of
area 17, and identification of the patterns of lateral geniculate input. Journal of Comparative Neurology,
188, 599–628.
De Carlos, J.A., López-Mascaraque, L. and Valverde, F. (1985) Development, morphology and topography of
chandelier cells in the auditory cortex of the cat. Developmental Brain Research, 22, 293–300.
De Carlos, J.A. (1986) Estructura y conexiones de la corteza auditiva en mamíferos. Doctoral Thesis. Univer-
sidad Autónoma de Madrid.
De Carlos, J.A., López-Mascaraque, L., Ramón y Cajal-Agüeras, S. and Valverde, F. (1987) Chandelier cells in
the auditory cortex of monkey and man: a Golgi study. Experimental Brain Research, 66, 295–392.
De Carlos, J.A., López-Mascaraque, L. and Valverde, F. (1996) Dynamics of cell migration from the lateral
ganglionic eminence in the rat. Journal of Neuroscience, 16, 6146–6156.
DeFelipe, J., Hendry, S.H.C. and Jones, E.G. (1989) Visualization of chandelier cells axons by parvalbumin
immunoreactivity in monkey cerebral cortex. Proceedings of the National Academy of Sciences, U.S.A., 86,
2093–2097.
DeFelipe, J., Hendry, S.H.C., Hashikawa, T., Molinari, M. and Jones, E.G. (1990) A microcolumnar structure of
monkey cerebral cortex revealed by immunocytochemical studies of double bouquet cell axons. Neuro-
science, 37, 655–673.
DeFelipe, J. and Fariñas, I. (1992) The pyramidal neuron of the cerebral cortex: morphological and chemical
characteristics of the synaptic inputs. Progress in Neurobiology, 39, 563–607.
DeFelipe, J. and Jones, E.G. (1992) High resolution light and electron microscopic immunocytochemistry of
colocalized GABA and calbindin D-28k in somata and double bouquet cell axons of monkey somato-
sensory cortex. European Journal of Neuroscience, 4, 46–60.
DeFelipe, J., González-Albo, M.C., Del Rio, M.R. and Elston, G.N. (1999) Distribution and patterns of connec-
tivity of interneurons containing calbindin, calretinin, and parvalbumin in visual areas of the occipital and
temporal lobes of the macaque monkey. Journal of Comparative Neurology, 412, 515–526.
Donoghue, M.J. and Rakic, P. (1999) Molecular evidence for the early specification of presumptive functional
domains in the embryonic primate cerebral cortex. Journal of Neuroscience, 19, 5967–5979.
Eccles, J.C. (1981) The modular operation of the cerebral neocortex considered as the material basis of mental
events. Neuroscience, 6, 1839–1856.
Eccles, J.C. (1984) The cerebral cortex. A theory of its operation. In: E.G. Jones and A. Peters (eds), Cerebral
Cortex. Functional Properties of Cortical Cells, Vol. 2, New York and London: Plenum Press, pp. 1–36.
Elston, G.N., Tweedale, R. and Rosa, M.G.P. (1999) Cortical integration in the visual system of the macaque
monkey: large scale morphological differences of pyramidal neurones in the occipital, parietal and temporal
lobes. Proceedings of the Royal Society, London, series B, 266, 1367–1374.
Fairén, A. and Valverde, F. (1979) Specific thalamo-cortical afferents and their presumptive targets in the visual
cortex. A Golgi study. In: M. Cuénod, G.W. Kreutzberg and F.E. Bloom (eds), Development and Specificity
of Neurons, (Progress in Brain Research, Vol. 51), Amsterdam: Elsevier, pp. 419–438.
Fairén, A. and Valverde, F. (1980) A specialized type of neuron in the visual cortex of cat: A Golgi and electron
microscope study of chandelier cells. Journal of Comparative Neurology, 194, 761–779.
Fairén, A., DeFelipe, J. and Regidor, J. (1984) Nonpyramidal neurons. General account. In: A. Peters and E.G.
Jones (eds), Cerebral Cortex, Cellular Components of the Cerebral Cortex, Vol. 1, New York and London:
Plenum Press, pp. 201–253.
Feldman, M. and Peters, A. (1978) The forms of non-pyramidal neurons in the visual cortex of the rat. Journal of
Fisken, R.A., Garey, L.J. and Powell, T.P.S. (1975) The intrinsic, association and commissural connections
of area 17 of the visual cortex. Philosophical Transactions of the Royal Society, London, series B, 272,
487–536.

Freund, T.F., Martin, K.A.C. and Whitteridge, D. (1985) Innervation of cat visual areas 17 and 18 by physiolo-
gically identified X- and Y-type thalamic afferents. I. Arborization patterns and quantitative distribution of
postsynaptic elements. Journal of Comparative Neurology, 242, 263–274.
Friedman, B. and Price, J.L. (1986) Age-dependent cell death in the olfactory cortex: Lack of transneuronal
degeneration in neonates. Journal of Comparative Neurology, 246, 20–31.
Fujita, I. and Fujita, T. (1996) Intrinsic connections in the macaque inferior temporal cortex. Journal of Compar-
Garey, L.J. (1971) A light and electron microscopic study of the visual cortex of the cat and monkey. Proceed-
ings of the Royal Society, London, series B, 179, 21–40.
Gilbert, C.D. and Wiesel, T.N. (1979) Morphology of intracortical projections of functionally characterised
neurons in the cat visual cortex. Nature (London), 280, 120–125.
Glezer, I.I., Jacobs, M.S. and Morgane, P.J. (1988) Implications of the “initial brain” concept for brain evolution
in Cetacea. Behavioral and Brain Sciences, 11, 75–116.
Gould, E., Reeves, A.J., Graziano, M.S.A. and Gross, C.G. (1999) Neurogenesis in the neocortex of adult
primates. Science (Washington), 286, 548–552.
Heimer, L. and Kalil, R. (1978) Rapid transneuronal degeneration and death of cortical neurons following
removal of olfactory bulb in adult rats. Journal of Comparative Neurology, 178, 559–610.
Hendry, S.H.C., Jones, E.G., Emson, P.C., Lawson, D.E.M., Heizmann, C.W. and Streit, P. (1989) Two classes
of cortical GABA neurons defined by differential calcium binding protein immunoreactivities. Experi-
mental Brain Research, 76, 467–472.
Hornung, J.P. and Garey, L.J. (1980) A direct pathway from thalamus to visual callosal neurons in the cat.
Hornung, J.P. and Garey, L.J. (1981) The thalamic projection to cat visual cortex: Ultrastructure of
neurons identified by Golgi impregnation or retrograde horseradish peroxidase transport. Neuroscience, 6,
1053–1068.
Hubel, D.H. and Wiesel, T.N. (1962) Receptive fields, binocular interaction and functional architecture in the
Hubel, D.H. and Wiesel, T.N. (1969) Anatomical demonstration of columns in the monkey striate cortex.
Nature (London), 221, 747–750.
Hubel, D.H. and Wiesel, T.N. (1972) Laminar and columnar distribution of geniculo-cortical fibers in the
macaque monkey. Journal of Comparative Neurology, 146, 421–450.
Hubel, D.H. and Wiesel, T.N. (1977) Functional architecture of macaque monkey visual cortex. Proceedings of
the Royal Society, London, series B, 198, 1–59.
Imig, T.J., Reale, R.A. and Brugge, J.F. (1982) The auditory cortex. Patterns of corticocortical projections
related to physiological maps in the cat. In: C.N. Woolsey (ed.), Cortical Sensory Organization, Vol. 3,
Multiple Auditory Areas, New Jersey: Humana Press, pp. 1–41.
Innocenti, G.M. (1979) Adult and neonatal characteristics of the callosal zone at the boundary between areas 17
and 18 in the cat. In: I.S. Russell, M.W. van Hof and G. Berlucchi (eds), Structure and Function of Cerebral
Commissures, London: MacMillan Press, pp. 244–258.
Jones, E.G. (1975) Varieties and distribution of non-pyramidal cells in the somatic sensory cortex of the squirrel
Jones, E.G. and Burton, H. (1976) Areal differences in the laminar distribution of thalamic afferents in cortical
fields of the insular, parietal and temporal regions of primates. Journal of Comparative Neurology, 168,
197–248.
Jones, E.G. and Hendry, S.H.C. (1984) Basket cells. In: A. Peters and E.G. Jones (eds), Cerebral Cortex, Vol. 1.
Cellular Components of the Cerebral Cortex, New York and London: Plenum Press, pp. 309–336.
Jones, E.G. (1984a) History of Cortical Cytology. In: A. Peters and E.G. Jones (eds), Cerebral Cortex, Vol. 1.
Cellular Components of the Cerebral Cortex, New York and London: Plenum Press, pp. 1–32.
Jones, E.G. (1984b) Laminar distribution of cortical efferent cells. In: A. Peters and E.G. Jones (eds), Cerebral Cortex,
Vol. 1. Cellular Components of the Cerebral Cortex, New York and London: Plenum Press, pp. 521–553.
Karten, H.J. (1997) Evolutionary developmental biology meets the brain: The origins of mammalian cortex.
Proceeding of the National Academy of Sciences, U.S.A., 94, 2800–2804.
Kelly, J.P. and Van Essen, D.C. (1974) Cell structure and function in the visual cortex of the cat. Journal of
Kisvárday, Z.F., Martin, K.A.C., Whitteridge, D. and Somogyi, P. (1985) Synaptic connections of intracellularly
filled clutch cells: A type of small basket cell in the visual cortex of the cat. Journal of Comparative Neuro-
logy, 241, 111–137.
Kisvárday, Z.F., Cowey, A. and Somogyi, P. (1986) Synaptic relationships of a type of GABA-immunoreactive
neuron (clutch cell), spiny stellate cells and lateral geniculate nucleus afferents in layer IVc of the monkey
striate cortex. Neuroscience, 19, 741–761.
Kisvárday, Z.F., Beaulieu, C. and Eysel, U.T. (1993) Network of GABAergic large basket cells in cat visual
cortex (area 18): implication for lateral inhibition. Journal of Comparative Neurology, 327, 398–415.

Kritzer, M., Cowey, A. and Somogyi, P. (1992) Patterns of inter- and intralaminar GABAergic connections
distinguish striate (V1) and extrastriate (V2, V4) visual cortices and their functionally specialized sub-
divisions in the rhesus monkey. Journal of Neuroscience, 12, 4545–4564.
Le Vay, S. (1973) Synaptic patterns in the visual cortex of the cat and monkey: Electronmicroscoscopy of
Golgi preparations. Journal of Comparative Neurology, 150, 53–86.
Le Vay, S., Hubel, D.H. and Wiesel, T.N. (1975) The pattern of ocular dominance columns in macaque visual
cortex revealed by a reduced silver stain. Journal of Comparative Neurology, 159, 559–576.
Lewis, W.B. (1878) On the comparative structure of the cortex cerebri. Brain, 1, 79–86.
Lin, C.S., Friedlander, M.J. and Sherman, S.M. (1979) Morphology of physiologically identified neurons in the
visual cortex of the cat. Brain Research, 172, 344–348.
López-Mascaraque, L. and Price, J.L. (1997) Protein synthesis inhibitors delay transneuronal death in the
piriform cortex of young adult rats. Neuroscience, 79, 463–475.
Lorente de Nó, R. (1949) Cerebral cortex: Architecture, intracortical connections, motor projections. In Fulton’s
Physiology of the Nervous System, London: Oxford University Press, pp. 288–313.
Lund, J.S. (1973) Organization of neurons in the visual cortex, area 17, of the monkey (Macaca mulatta).
Lund, J.S., Henry, G.H., Macqueen, C.L. and Harvey, A.R. (1979) Anatomical organization of the primary
visual cortex (area 17) of the cat. A comparison with area 17 of the macaque monkey. Journal of Comparat-
ive Neurology, 184, 599–618.
Lund, J.S., Hendrickson, A.E., Ogren, M.P. and Tobin, E.A. (1981) Anatomical organization of primate visual
cortex area VII. Journal of Comparative Neurology, 202, 19–45.
Lund, J.S. (1984) Spiny stellate neurons. In A. Peters and E.G. Jones (eds), Cerebral Cortex. Cellular Com-
ponents of the Cerebral Cortex, Vol. 1, New York and London: Plenum Press, pp. 255–308.
Marín-Padilla, M. (1987) The chandelier cell of the human visual cortex: A Golgi study. Journal of Comparative
Neurology, 256, 61–70.
Martin, K.A.C. and Whitteridge, D. (1984) Form, function and intracortical projections of spiny neurons in the
Mates, S.L. and Lund, J.S. (1983) Developmental changes in the relationship between type 2 synapses and spiny
neurons in the monkey visual cortex. Journal of Comparative Neurology, 221, 98–105.
McMullen, T.N., Glaser, E.M. and Tagamets, M. (1984) Morphometry of spine-free nonpyramidal neurons in
rabbit auditory cortex. Journal of Comparative Neurology, 222, 383–395.
Meyer, G. and Albus, K. (1981) Spiny stellates as cells of origin of association fibers from area 17 to area 18 in
the cat’s neocortex. Brain Research, 210, 335–341.
Meyer, G. and Ferres-Torres, R. (1984) Postnatal maturation of nonpyramidal neurons in the visual cortex of the
cat. Journal of Comparative Neurology, 228, 226–244.
Misson, J.P., Austin, C.P., Takahashi, T., Cepko, C.L. and Caviness, Jr., V.S. (1991) The alignment of
migrating neural cells in relation to the murine neopallial radial glial fiber system. Cerebral Cortex, 1,
221–229.
Mountcastle, V. (1957) Modality and topographic properties of single neurons of cat’s somatic sensory cortex.
Mountcastle, V. (1979) An organizing principle for cerebral function: the unit module and the distributed system.
In: F.O. Schmidt and F.G. Worden (eds), The Neurosciences. Fourth Study Program, Cambridge,
Massachusetts: MIT Press, pp. 21–42.
Nieuwenhuys, R. (1994) The neocortex. An overview of its evolutionary development, structural organization
and synaptology. Anatomy and Embryology, 190, 307–337.
O’Leary, J.L. (1941) Structure of the area striata of the cat. Journal of Comparative Neurology, 75, 131–164.
Peters, A. and Feldman, M.L. (1976) The projection of the lateral geniculate nucleus to area 17 of the rat cerebral
cortex. I. General description. Journal of Neurocytology, 5, 63–84.
Peters, A., Feldman, M.L. and Saldanha, J. (1976) The projection of the lateral geniculate nucleus to area 17 of
the rat cerebral cortex. II. Terminations upon neuronal perikarya and dendritic shafts. Journal of Neuro-
cytology, 5, 85–107.
Peters, A. and Feldman, M.L. (1977) The projection of the lateral geniculate nucleus to area 17 of the rat cerebral
cortex. IV. Termination upon spiny dendrites. Journal of Neurocytology, 6, 669–689.
Peters, A., Proskauer, C.C., Feldman, M.L. and Kimerer, L. (1979) The projection of the lateral geniculate
nucleus to area 17 of the rat cerebral cortex. V. Degenerating axon terminals synapsing with Golgi impreg-
nated neurons. Journal of Neurocytology, 8, 331–357.
Peters, A. and Regidor, J. (1981) A reassessment of the forms of nonpyramidal neurons in area 17 of cat visual
cortex. Journal of Comparative Neurology, 203, 685–716.
Peters, A., Proskauer, C.C. and Ribak, C.E. (1982) Chandelier cells in rat visual cortex. Journal of Comparative
Neurology, 206, 397–416.
Peters, A. (1984) Chandelier cells. In: A. Peters and E.G. Jones (eds), Cerebral Cortex, Cellular Components of
the Cerebral Cortex, Vol. 1, New York and London: Plenum Press, pp. 361–380.

Peters, A. and Jones, E.G. (1984) Classification of cortical neurons. In: A. Peters and E.G. Jones (eds), Cerebral Cor-
tex. Cellular Components of the Cerebral Cortex, Vol. 1, New York and London: Plenum Press, pp. 107–121.
Preuss, T.M., Qi, H. and Kaas, J.H. (1999) Distinctive compartmental organization of human primary visual
cortex. Proceedings of the National Academy of Science, U.S.A., 96, 11601–11606.
Rakic, P. (1972) Mode of cell migration to the superficial layers of fetal monkey neocortex. Journal of Compar-
Rakic, P. (1985) Contact regulation of neuronal migration. In: G.M. Edelman and J.P. Thiery (eds), The Cell in
Contact, New York: Wiley, pp. 67–91.
Rockland, K.S. and Pandya, D.N. (1981) Cortical connections of the occipital lobe in the rhesus monkey: inter-
connections between areas 17, 18, 19 and the superior temporal sulcus. Brain Research, 212, 249–270.
Romer, A.S. (1966) Vertebrate Paleontology. Chicago: The University of Chicago Press.
Rubenstein, J.L.R. and Rakic, P. (1999) Genetic control of cortical development. Cerebral cortex, 9, 521–523.
Sanides, D. (1979) Commissural connections of the visual cortex of the cat. In: I.S. Russell, M.W. van Hof
and G. Berlucchi (eds), Structure and Function of Cerebral Commissures, London: MacMillan Press,
pp. 236–243.
Sanides, F. (1970) Functional architecture of motor and sensory cortices in primates in the light of a new concept
of neocortex evolution. In: C.R. Noback and W. Montagna (eds), The Primate Brain, New York: Appleton
Century Crofts, pp. 137–208.
Simpson, G.G. (1945) The principles of classification and a classification of mammals. Bulletin of the American
Museum of Natural History, 85, 1–350.
Somogyi, P. (1977) A specific ‘axo-axonal’ interneuron in the visual cortex of the rat. Brain Research, 136,
345–350.
Somogyi, P. (1978) The study of Golgi stained cells and of experimental degeneration under the electron micro-
scope: A direct method for the identification in the visual cortex of three successive links in a neuron chain.
Somogyi, P., Freund, T. F. and Cowey, A. (1982) The axo-axonic interneuron in the cerebral cortex of the rat, cat
and monkey. Neuroscience, 7, 2577–2609.
Somogyi, P., Kisvárday, Z.F., Martin, K.A.C. and Whitteridge, D. (1983) Synaptic connections of morphologic-
ally identified and physiologically characterized large basket cells in the striate cortex of cat. Neuroscience,
10, 261–294.
Somogyi, P. and Cowey, A. (1984) Double bouquet cells. In: A. Peters and E.G. Jones (eds), Cerebral Cortex,
Cellular Components of the Cerebral Cortex, Vol. 1, New York and London: Plenum Press, pp. 337–360.
Spatz, W.B. (1977) Topographically organized reciprocal connections between areas 17 and MT (visual area of
superior temporal sulcus) in the marmoset Callithrix jacchus. Experimental Brain Research, 27, 559–572.
Szentágothai, J. (1975) The ‘module concept’ in cerebral cortex architecture. Brain Research, 95, 475–496.
Szentágothai, J. (1978) The neuron network of the cerebral cortex: A functional interpretation. Proceedings of
the Royal Society of London, series B, 201, 219–248.
Szentágothai, J. (1979) Local neuron circuits of the neocortex. In: F.O. Schmitt and F.G. Worden (eds), The
Neurosciences. Fourth Study Program, Cambridge, Massachusetts: MIT Press, pp. 399–415.
Szentágothai, J. and Arbib, M.A. (1974) Conceptual models of neural organization. Neuroscience Research
Program Bulletin, 12, 307–510.
Tamamaki, N., Fujimori, E. and Takauji, R. (1997) Origin and route of tangentially migrating neurons in the
developing neocortical intermediate zone. Journal of Neuroscience, 17, 8313–8323.
Valverde, F. (1968) Structural changes in the area striata of the mouse after enucleation. Experimental Brain
Research, 5, 274–292.
Valverde, F. (1971) Short axon neuronal subsystems in the visual cortex of the monkey. International Journal of
Valverde, F. (1976) Aspects of cortical organization related to the geometry of neurons with intra-cortical axons.
Journal of Neurocytology, 5, 509–529.
Valverde, F. (1978) The organization of area 18 in the monkey. A Golgi study. Anatomy and Embryology, 154,
305–334.
Valverde, F. (1983) A comparative approach to neocortical organization based on the study of the brain of the
hedgehog (Erinaceus europaeus). In: S. Grisolía, C. Guerri, F. Samson, S. Norton and F. Reinoso-Suárez
(eds), Ramón y Cajal’s Contribution to the Neurosciences, Amsterdam: Elsevier, pp. 149–170.
Valverde, F. (1985) The organizing principles of the primary visual cortex in the monkey. In: A. Peters and
E.G. Jones (eds), Cerebral Cortex, Vol. 3. Visual Cortex, New York and London: Plenum Press, pp. 207–257.
Valverde, F. (1986) Intrinsic neocortical organization: some comparative aspects. Neuroscience, 18, 1–23. Val-
verde, F. and Facal-Valverde, M.V. (1986) Neocortical layers I and II of the hedgehog (Erinaceus
europaeus). I. Intrinsic organization. Anatomy and Embryology, 173, 413–430.
Valverde, F., De Carlos, J.A., López-Mascaraque, L. and Doñate-Oliver, F. (1986) Neocortical layers I and II of
the hedgehog (Erinaceus europaeus). II. Thalamo-cortical connections. Anatomy and Embryology, 175,
167–179.

Valverde, F., Facal-Valverde, M.V., Santacana, M. and Heredia, M. (1989) Development and differentiation of
early generated cells of sublayer VIb in the somatosensory cortex of the rat. A combined Golgi and auto-
radiographic study. Journal of Comparative Neurology, 290, 118–140.
Valverde, F. (1998) Golgi Atlas of the Postnatal Mouse Brain. Wien-New York: Springer.
Welker, C. (1976) Receptive fields of barrels in the somatosensory neocortex of the rat. Journal of Comparative
Neurology, 166, 173–190.
White, E.L. (1979) Thalamocortical synaptic relations: A review with emphasis on the projections of specific
thalamic nuclei to the primary sensory areas of the neocortex. Brain Research Reviews, 1, 275–311.
Winfield, D.A. and Powell, T.P.S. (1976) The termination of thalamo-cortical fibers in the visual cortex of the
cat. Journal of Neurocytology, 5, 269–281.
Wise, S.P. and Jones, E.G. (1976) The organization and postnatal development of the commissural projections in
the rat somatic sensory cortex. Journal of Comparative Neurology, 168, 313–344.
Woolsey, C.N. (1960) Organization of cortical auditory system: a review and a synthesis. In: G.L. Rasmussen
and W.F. Windle (eds), Neural Mechanisms of the Auditory and Vestibular Systems, Springfield, Illinois:
Ch. Thomas, pp. 165–180.
Woolsey, T.A. and Van der Loos, H. (1970) The structural organization of layer IV in the somatosensory region
(SI) of mouse cerebral cortex. Brain Research, 17, 205–242.
Woolsey, T.A., Dierker, M.L. and Wann, D.F. (1975) Mouse SmI cortex: qualitative and quantitative classific-
ation of Golgi-impregnated barrel neurons. Proceedings of the National Academy of Sciences, U.S.A., 72,
2165–2169.
Záborsky, L. and Wolff, J.R. (1982) Distribution patterns and individual variations of callosal connections in the
albino rat. Anatomy and Embryology, 165, 213–232.
Zhu, Y., Li, H.-S., Zhou, L., Wu, J.Y. and Rao, Y. (1999) Cellular and molecular guidance of GABAergic
neuronal migration from an extracortical origin to the neocortex. Neuron, 23, 473–485.

11 Laminar Continuity between Neo- and Meso-Cortex:
The Hypothesis of the Added Laminae in the Neocortex
Robert Miller1 and Rupa Maitra2
1
Otago Centre for Theoretical Studies in Psychiatry and Neuroscience,
c/o Department of Anatomy and Structural Biology, School of Medical Science,
University of Otago, P.O. Box 913, Dunedin, New Zealand
Tel: 0064-3-4797357; FAX: 0064-3-479-7254; e-mail: robert.miller@stonebow.otago.ac.nz
2
Department of Anatomic Pathology, Wellington Hospital, Wellington, New Zealand
A major subdivision of the mammalian cerebral cortex is between those regions lying dorsal to the
rhinal fissure (dorsal and lateral in primate) and those parts lying ventral to it (ventromedial in pri-
mates). With the exception of the hippocampal formation, both parts are described conventionally as
six-layered structures. However, in this chapter, the question is raised whether the laminar pattern
above and below the rhinal fissure is essentially the same. Evidence relating to this question includes
that from comparative anatomy of the cortex, its development, and especially that showing the
extent to which different chemical markers are continuous across the rhinal fissure. The conclusion
is reached that the middle laminae of the neocortex (laminae III and/or IV) are missing in the region
below the rhinal fissure, where the remaining laminae “fill out” the thickness of the cortex. The slice
of cortical tissue “added” in the neocortex may have biophysical and connectional properties which
enable cell assemblies to form in the neocortex. This appears to be a unique endowment of the mam-
malian cortex, permitting cortical tissue to acquire more detailed spatiotemporal representations than
in the cortical rudiment of submammalian species.
KEYWORDS: brain development, comparative anatomy, cerebral cortex, laminae, neurochemical

markers
1. INTRODUCTION
The cerebral neocortex is a laminated structure usually described in terms of six cellular
layers, which, in the main, are continuous from one cytoarchitectural area to another.
There may be differences in thickness and cell density of laminae between areas, and
laminae are missing in some areas (e.g. lamina IV is missing in the primary motor area).
Moreover, for descriptive purposes, laminae may be subdivided in some areas (e.g. lamina
IV in the primate visual cortex). Nevertheless the basic six-layered arrangment is a scheme
which has general application across areas and across all mammalian species.
The neocortex (sometimes referred to as “isocortex”) forming most of the convexity of
the hemispheres is distinguished from the more medial parts of cortex, referred to in
Broca’s terminology as the “limbic lobe”. However, the word “limbic” has a variety of
definitions. As a morphological entity the limbic lobe includes the hippocampal formation
(Ammon’s horn, the dentate gyrus and subiculum) which is of far simpler cellular arrange-
ment than the neocortex. The limbic lobe also includes a variety of other areas of cortex,
219
220 Robert Miller and Rupa Maitra
such as the cingulate, entorhinal and piriform cortex, whose immediate appearance is
more like that of the neocortex.
The neocortex varies between species, from a small unfolded structure in the smaller
mammals to a large and much-folded surface layer with complex patterns of sulci and
gyri, the latter occurring especially in the animals with large forebrains, such as primates
and cetaceae. Despite this great variation across species, there is one sulcus which is con-
stant across all mammalian species, the rhinal fissure. This is visible early in development,
though its position in the adult varies across species. It separates the neocortex, which lies
dorsal and medial to the rhinal fissure (dorsal and lateral in primates), from the “limbic”
regions just referred to, which lie ventrally and/or medially to this fissure. In small-
brained mammals, the rhinal sulcus is located on the ventro-lateral aspects of the cortex,
running roughly antero-posteriorly. In larger-brained mammals, such as primates, the
neocortex is relatively expanded compared with the limbic cortex, and, as a result, the
rhinal fissure lies on the ventral aspect of the hemispheres, but, as in animals with smaller
brains, runs roughly antero-posteriorly.
The regions of the “limbic” cortex which have grossly similar appearance to the neo-
cortex are sometimes referred to as the “mesocortex” (Sarnat and Netsky, 1974), to indicate
that they are transitional between typical neocortex and the cortex of the hippocampal
formation. Other terms are sometimes used to indicate these subdivisions, the term “allocor-
tex” referring to the hippocampal formation, and the term “periallocortex” refering to the
transitional regions, entorhinal, cingulate and piriform cortex (Chronister and White, 1977)2.
In terms of laminar architecture the transitional mesocortex (or periallocortex), lying
ventral or medial to the rhinal fissure has a striking feature which distinguishes it from the
neocortex. This is very easily discernible, and applies universally across mammalian
species. This distinction is found in lamina II, the most superficial of the cell-rich layers:
in the neocortex this contains many small pyramidal cells, somewhat more densely-
packed than in the underlying lamina III, but nevertheless not outstandingly prominent.
In contrast, ventral to the rhinal fissure, lamina II is outstanding, as a very dense layer of
neurones (Braitenberg and Schüz, 1991). This can be seen easily, with the naked eye, in
Nissl stained sections, as shown in the atlases from mouse, rat (see Figure 11.1), rabbit,
cat, dog, monkey, human and other species.
In atlases of the rat brain, Nissl-stained sections reveal an interesting pattern of trans-
formation across the rhinal fissure from the pattern typical of the neocortex to that typical
of the mesocortex. This is depicted in Figure 11.1. Here it can be seen that the cell-dense
lamina II of the mesocortex is in continuity not only with a corresponding lamina II of the
neocortex, but also with a prominent layer of small cells, densely packed in the neocortex
much deeper than lamina II, approximately in the position of lamina IV. This layer grad-
ually “slopes” superficially as it approaches the rhinal fissure from the neocortical side,
becoming continuous with lamina II in the mesocortex. It can also be seen in Figure 11.1
that the deeper layer of the neocortex slopes to the surface in the dorso-medial extent of
the neocortex, in the region of transition with the cingulate cortex. In the rabbit (Urban
and Richard, 1972; Figure 11.2A) the transition between neocortex and mesocortex
ventral to the rhinal fissure is more abrupt, but the convergence of the small-celled layers
2
The term “allocortex” is sometimes used collectively to refer to the hippocampus and cortical regions such as
entorhinal, cingulate and piriform cortices (the latter otherwise known as “periallocortex” (See Stephan and
Andy, 1970)).

Laminar Continuity Across the Rhinal Fissure 221
Figure 11.1. Section through rat brain, showing the continuity of layer IV of neocortex with layer II of the
cingulate cortex dorsomedially and of the entorhinal cortex ventrolaterally. (The border between the small-
celled, densely-packed layer IV and the underlying layer V in the neocortex is indicated by a black arrow.)
Layer III of neocortex thus appears as an addition to the cortical architecture of cingulate and entorhinal cortex
(reproduced with permission from Paxinos and Watson [1986] The rat brain in stereotaxic coordinates, 2nd ed,
Figure 26R, courtesy of Academic Press).
upon layer II of the mesocortex is still evident. Exactly the same “sloping” cell layer can
be seen in the cortex of other species, such as the cat (Figure 11.2B), the marmoset
(Stephan et al., 1980; Figure 11.2C) and the human (Braak and Braak, 1992; Figure
11.2D), in the region of the rhinal fissure.

A B
C D
Figure 11.2. Details of laminar arrangements in the region of the rhinal fissure in rabbit (A: reproduced with
permission from Urban, I. and Richard, P. (1972) A stereotaxic atlas of the New Zealand rabbit’s brain., p. 64;
courtesy of C.C. Thomas, Springfield, Illinois); cat (B: kindly provided by C.J. Heath), marmoset (C: repro-
duced with permission from Stephan, H., Baron, G. and Schwertfeger, W.K. (1980), The brain of the common
marmoset. A stereotaxic atlas. Springer Verlag, Figure A8.5) and from human (D: reprinted from Neuroscience
Research, Vol. 15, H. Braak and E. Braak, The human entorhinal cortex: normal morphology and lamina-
specific pathology in various diseases, pp. 6–31, Figure 6A. Copyright [1992], with permission of Elsevier
Science). In A and B the mesocortex is located in the lower part of the cortical tissue; in C and D it is found in
the right hand side of the Figure. (In D, the transition to the entorhinal cortex in humans is shown. The layers
cannot be fully equated with those in the other three Nissl-stained sections, since D uses a different stain
[lipofuscin]. The convergence of neocortical layers—to the left—upon the superficial cell layer of the entorhinal
cortex—to the right—is clearly seen).
This simple observation gives rise to a broad hypothesis which is investigated in this
paper, namely that there is a fairly regular pattern of transformation of the cell layers in the
neocortex to give those in the mesocortex. In particular, it is argued that some of the
middle laminae in the neocortex are not represented in the mesocortex (either ventrally in
the entorhinal cortex, or dorsomedially in the cingulate cortex). They are, as it were,
“additions” to convert a more basic cortical laminar architecture found in the mesocortex
into that typical of the neocortex. The continuity of mesocortical lamina II to neocortical
(approximately) lamina IV is then one out of a number of consequences of the laminar
reorganization between the two types of cortex.

This broad hypothesis can be evaluated using a variety of types of evidence: compara-
tive, developmental, cytological, connectional, but most of all in terms of the layering of
numerous chemical markers, which have been described in many publications in the last
20 years. As will be seen such evidence does give striking evidence of a regular trans-
formation of laminae between the neocortex and mesocortex, such that the deeper and
most superficial laminae of the neocortex are continuous into the mesocortex, and the
neocortical layers between these two appear to be missing in the mesocortex.
Amongst the laminar differences which could be studied are a variety of specializations
which are local to particular cortical areas. For instance, in visual cortex of primate there
are many very specialized laminar features, and a characteristic pattern of cytochrome
oxidase staining. In lamina III, this forms an arrangement of cytochrome oxidase-rich
“blobs”. Such an arrangement of cytochrome oxidase-rich “blobs” is also found in the
entorhinal cortex, but in lamina II rather than lamina III (Hevner and Wong-Riley, 1992).
Such features are not general characteristics of the cortex, and are taken here to represent
local specializations. Details such as these, interesting though they are, are not the main
focus of the present paper. The real focus is data about laminae which can be generalized
across the cortical mantle, and thus allow comparison between neocortical and meso-
cortical regions in their entireties.
The main part of this paper explores the above hypothesis in a strictly morphological
context. Differences between the neo- and mesocortical regions can thus be considered as
probable consequences of differences in the program of brain development between
cortical regions. However, the largest body of data assimilated to test the hypothesis is
concerned with the laminar distribution of chemical markers, especially those associated
with various neurotransmitter systems. Some of these chemical markers may be related to
developmental processes (Berger-Sweeney and Hohmann, 1997), and have little func-
tional role in the adult. However, the markers of major neurotransmitter systems are likely
to have important functional roles in the adult. Thus, if a major distinction between
neocortex and mesocortex can be drawn, based on different laminar distribution of such
chemical markers, it is also likely to have important large-scale implications for the dif-
ferences in cybernetic operation of neocortex vs mesocortex. This topic is not explored in
depth here, although a few suggestions will be offered at the end of the paper.
2. COMPARATIVE ASPECTS OF LAMINAR ARRANGEMENT
Comparison of the laminar arrangement of the mammalian cortex with that of reptiles
(presumably related to the evolutionary antecedents of mammals) is a useful point of
departure. Reiner (1993) has discussed this topic in relation to the primitive cortex of the
turtle. This cortex has a trilaminar arrangement not unlike that of Ammon’s horn (see Figure
11.3). Thus it has a single cell layer containing pyramidal cells (whose projections
descend to the diencephalon and brainstem), a very thick molecular layer superficial to it,
containing the apical dendrites of the pyramidal cells, and, deep to the pyramidal cell
layer, a basal dendritic zone. In the turtle (according to Smith et al., 1980), thalamic inputs
terminate exclusively in the outer third of the molecular layer, that is, on the distal parts of
the pyramidal cell apical dendrites, where they show their distal arborization. This is
a notable contrast with the mammalian neocortex, in which the cell-rich lamina IV is the
main destination of projections from sensory thalamo-cortical nuclei, while projections

Figure 11.3. Laminar transformation from ancestral reptilian cortex to typical mammalian neocortex (from
Reiner, 1993; Figure 8).
from other thalamic nuclei can be distributed in a variety of complex patterns amongst
various laminae, mainly those rich in cells (Steriade et al., 1990), but not focussed on the
most superficial layers of the cortex. The arrangement of thalamic projections to the cor-
tex in turtles is however, very similar to their distribution to the mammalian hippocampus
(Herkenham, 1978; Wouterlood et al., 1990) where thalamic afferents also terminate in
a cell-sparse layer, amongst the distal dendritic arborizations of pyramidal cells. In the
mammalian mesocortex, the pattern of termination of thalamic afferents is complex, but in
some ways intermediate between those of the hippocampal formation and the neocortex:
thalamic afferents terminate in a variety of laminae, amongst which lamina I (the most
superficial) is more prominent than for thalamic afferents to the neocortex (entorhinal
cortex: Robertson and Kaitz, 1981; Room and Groenewegen, 1986; Yanigahara et al., 1987;
retrosplenial cortex: Van Groen and Wyss, 1992; cingulate cortex: Shibata, 1993). Reiner
(1991) also points out that the pyramidal cell layer in the turtle cortex largely lacks the
interregional and interhemispheric projection neurones so abundant in the mammalian
neocortex (Voneida and Ebbesson, 1969; Hall, 1971; Lohman and Mentink, 1972; Ebner,
1976). In mammals, laminae II and III of the neocortex are those from whose neurones
such long-distance connections arise in greatest profusion.
Taking together these differences between mammalian and turtle cortex, in both
thalamocortical and cortico-cortical connections, the suggestion arises that the pyramidal
cell layer of the turtle cortex corresponds to lamina V of the mammalian neocortex (both
of which have projections descending below the forebrain); and cortical layers II, III and
perhaps IV in the mammalian neocortex are in some sense an addition to the simpler
cortex found in reptiles, providing the cellular substrate for a massive associative network
of cortico-cortical connections, typical of the mammalian neocortex.
Viewed from this perspective, the status of the mammalian mesocortex becomes
somewhat uncertain. Some areas of the mesocortex (e.g. cingulate) have lamina V pyram-
idal cells whose projections descend to the brainstem. For other mesocortical regions
(e.g. the entorhinal cortex), there is negligible evidence for such descending connections,
and probably they do not exist. Mesocortical regions do have prolific long cortico-
cortical connections, but their pattern is quite complex, and it cannot be maintained

that the superficial laminae are of special importance as origin or termination of such
connections.
Reiner (1991) adds a third argument in support of his conclusions: he describes the
distribution of chemically specified neurones in the reptile cortex and compares them with
the mammalian neocortex. He suggests that chemically-specified cell types found only in
laminae II-IV of mammalian neocortex (Cholecystokinin-8-, Vasoactive Intestinal Poly-
peptide-, and Choline Acetyl Transferase-positive neurones) are absent in the equivalent
cortical region of turtles, whereas those prominent in the mammalian neocortical laminae
I, V and VI (e.g. Substance P-, Neuropeptide Y- and Somatostatin-positive neurones) are
found in the reptile cortex. Such evidence is used to suggest that laminae II-IV are an
added component in the mammal. However, this argument is less convincing than the
strictly morphological ones: for instance other evidence suggests that CCK or its frag-
ments are found in all cortical laminae, though admittedly more abundantly in laminae II
and III than in the deeper laminae (Peters et al., 1983; Cho et al., 1993). Neurones which
can be labelled for somatostatin mRNA are common in both the supra- and infragranular
layers in all regions of rat neocortex (Garrett et al., 1994). Substance P immunoreactive
neurones can be found in all layers of rat neocortex, except lamina I, with preference for
II/III, according to Kaneko et al. (1994). These distributions disagree with those given by
Reiner (1991). Nevertheless, the first two of Reiner’s arguments give an interesting
indication of the evolutionary processes leading to the distinctively mammalian neocortex.
3. DEVELOPMENTAL RELATION BETWEEN LAMINAE

OF THE MESO- AND NEOCORTEX
In neurodevelopmental terms, the deeper layers of the neocortex mature earlier than the
superficial layers, and the mesocortex matures before the neocortex. Thus deep neocorti-
cal laminae can be expected to develop at the same time as more superficial mesocortical
laminae. This could constitute a developmental basis for the apparent continuity between
lamina IV of the neocortex and lamina II of the mesocortex: once the latter superficial
cell-rich mesocortical lamina has formed no further layers of cells are formed superficial
to it, while the neocortex accumulates further cell layers superficial to lamina IV. A few
papers give detailed support for such a developmental pattern: Smart and Smart (1982)
exposed mouse embryos to tritiated thymidine at various stages of pregnancy. Labelling
on the eleventh day of pregnancy labelled lamina IV of the neocortex, and lamina II of the
mesocortex. Similar experiments were reported by Sanderson and Weller (1990) and
Sanderson and Aitken (1990) in marsupial possums, which are born at a much earlier
developmental stage, making the experiment easier to perform. The results were similar:
injection at one stage of development labels a continuous band of neurones extending
from lamina IV in the medial neocortex, sloping up to lamina III as the rhinal fissure is
approached and constituting lamina II in the mesocortex.
A cytological feature of the adult cortex allows one to offer a more specific hypothesis
about correspondance of laminae between neo- and mesocortex. In the neocortex, lamina
IV contains a variety of neurone called the “spiny stellate” cell, similar to pyramidal cells,
but lacking prominent apical dendrites. Such cells are very densely packed in primary
sensory areas, but are also found in lower density in lamina IV of most other neocortical
areas. Similar cells are also characteristic of lamina II of the neocortex, although not

generally distinguished as “spiny stellate” because, in a superficial position such as lamina

II, it is not remarkable that long apical dendrites are absent. In the mesocortex, lamina II
also has dense congregations of spiny stellate cells, but there is no corresponding con-
gregation of such cells in the middle layers of the mesocortex. Thus, one may suggest that
the laminae “added on” in the neocortex (especially lamina III) are “inserted” within the
original lamina II, so that, in the adult, neocortical layers II and IV are both derived from
the original lamina II, and have cytological features in common (abundance of spiny stel-
late cells). This does not happen in the mesocortex, which may account for the special
prominence of lamina II in the mesocortex, rich in densely-packed spiny stellate cells.
4. NEUROCHEMICAL LAMINAR MARKERS
4.1. Methodological Issues

In the last fifteen years, histo- and immunochemical methods have become available for
labelling a wide variety of chemical markers in brain tissue. Some of these are transmitters,
others are their receptors, or the messenger RNA (or fragments of mRNA) involved in
synthesis of receptor protein, while yet others are a miscellaneous collection of other
neurochemical molecules. As a result of the development of these methods, a large number
of papers has been published, giving textual descriptions, or photographic illustrations of
the distribution of markers across the laminae of the cerebral cortex.
Most of such data comes from rodents, especially rats. This serves the purposes of the
present paper well: in a single coronal section of the forebrain it is possible to illustrate the
laminar distribution in the neocortex, in the parts of the mesocortex ventral to the rhinal
fissure, and in the dorso-medial parts of the mesocortex such as cingulate cortex. In add-
ition, because of the simplicity of the cortex in such animals, it is easier to grasp the overall
laminar patterns than in animals with highly folded cortices. Some data are available from
carnivores and primates, but in such animals it is only occasionally possible to piece
together a coherent picture from which to compare mesocortex and neocortex. In the Tables
presented below, most data are from rats or other rodents, but a few are included from
carnivores and primates.
The studies included in the tables below represent only a selection of those which have
been published. A primary basis for the selection of chemical markers to be included is
that data are available on cortical laminar distribution in both the neocortex, and at least
one mesocortical region. The latter is usually the entorhinal cortex, seen ventral to the
rhinal fissure in coronal sections. For many chemical markers, clear evidence is also avail-
able for the cingulate cortex, and in a number of cases, where serial sections through the
cortex are illustrated, it is clear that the laminar arrangement of the marker is different in
the anterior cingulate cortex from the posterior cingulate regions. The latter details are
sometimes included in the tables, where the conclusions are relatively clear.
Data presented in the tables below are derived from both the illustrations shown in the
specific reports, and from the corresponding statements made in the text of these papers.
The tabulated data correspond (with one exception) to bands of heavy labelling relative to
the surrounding deeper or more superficial laminae. When such a prominent band is tabu-
lated, it is not meant to imply that there is no labelling in adjacent or unmentioned lam-
inae, but rather that they are more weakly labelled than the laminae mentioned in the

tables. The broad hypothesis under examination is thus evaluated by the continuity (or
lack of it) of these oustanding bands between neo- and mesocortex, rather than by presence
or absence (in an absolute sense) of label in particular bands. The exception is the banding
of the Muscarinic-2 receptor, which is found in all laminae of the neocortex, with the
exception of laminae IV. This “gap” is not continuous into the entorhinal cortex, so the
tabulated data refers to the “gap” of lower density labelling rather than to the presence of a
band of higher density.
Many technical details determine the contrast and definition of the laminar distributions
revealed in the papers cited. Most important is the particular ligand (tritiated or immuno-
logical) used to label a particular receptor or other marker: different ligands may differ in
the specificity for the receptor of interest, even when measures are taken to occlude label-
ling to irrelevant receptors, with an excess of non-tritiated ligands. Another technical
matter which affects the clarity of the banding is the concentration of ligand used. A few
papers (e.g. Xia and Haddad, 1992) present illustrations of adjacent brain sections,
exposed to a series of different concentrations of the ligand of a receptor. This shows that
the contrast in the banding across the cortical laminae varies substantially according to
this variable. The definition of the laminar binding in published illustrations undoubtedly
also reflects such things as exposure time when autoradiography is used, as well as other
details of photographic processing.
The conclusion presented in the table below reflects the consensus results between
different papers which sometimes use a variety of ligands. It is assumed that suboptimal
techniques (with respect to details such as those just mentioned) can reduce the contrast of
labelled bands in the cortex, but are not likely to produce discrete bands when none exists.
Therefore, the findings represented in the Tables give greater emphasis to those illustra-
tions in which laminar patterns are revealed with greatest clarity. Occasionally there is
clear disagreement between different studies of the same marker, and this is indicated in
footnotes in the tables.
Only rarely do the studies cited identify laminae of distribution of a neurochemical
marker by rigorous comparison with Nissl-stained sections. There may thus be minor
inaccuracies in defining laminae, especially for the middle laminae of the cortex, and
when contrast between labeling of adjacent bands is poor. In addition, in the mesocortex,
the definition of laminae is less well established than in the neocortex. Thus, in the meso-
cortex laminar detail is sometimes referred to in broad terms as “superficial”, “middle” or
“deep”.
4.2. Classification of Patterns of Banding

In tabulating the data, banding patterns are classified into five different groups, shown in
separate tables. Four of these make up patterns which are clearly those predictable from
the broad hypothesis described above. The fifth constitutes miscellaneous exceptions
which do not fit this hypothesis. When, as sometimes happens, a particular ligand is dis-
tributed in more than one band, these being separated in different laminae, the bands for
that ligand in different laminae may have appropriate places in more than one of the
tables. In such cases, the labelling for such a ligand appears more than once in the tables,
with each of its bands of labelling identified in the corresponding tables as “deep” “middle”
or “superficial”. It is recognized that there may be significant differences between species
in laminar distribution of a ligand, and separate entries are therefore made for different

species, including humans. For a particular combination of ligand and species, entries are
made in the tables only when there is data on laminar distribution from at least one paper
for both the neocortex and one of the areas of mesocortex.
4.2.1. Banding Pattern I: Bands in deep laminae of neocortex continuous into

deep laminae of the mesocortex
The broad hypothesis particularly concerns lamina IV and laminae superficial to it. There-
fore, the laminar transformation between neocortex and mesocortex, if it exists, might
have little effect on the deepest laminae. Hence one might expect that ligands which label
a band in lamina V and/or VI of the neocortex would label a band in a similar position of
the mesocortex. Table 11.1 and Figure 11.4 shows, this to be the case: many ligands
(cholinergic, monoaminergic, peptidergic, as well as an adenosine receptor) which label
lamina V or VI of the neocortex also label deep layers of the entorhinal cortex. The data
on the cingulate cortex is less complete, and there may be differences between anterior
and posterior cingulate cortex, and inconsistencies in the relative distribution between the
two when different ligands are compared. There are a few exceptions relating to neocortical
laminae V and VI, which are included in Table V: Dopamine D1 receptors, and dopamine
fibres constitute deep bands in the neocortex, which extend to fill all laminae in the entorhi-
nal cortex. The dopamine- and adenosine 3′5′-monophosphate-regulated phosphoprotein
Table 11.1. PATTERN I: Ligands which label only lamina VI or VI/V of neocortex, and only
deep layers of mesocortex
Chemical marker Sp. Neoctx Entorhinal Cingulate Refs Comments (ref)
N (deep) R V deep ant: deep 15, 62, 72

post: ?A
N-mRNA, R V, VI deep NR 80
alpha-2 submit
Bungarotoxin R VI deep ant: deep 16, 26 presynaptic location on
receptors (deep) post: A ACh fibres (71)
Musc-2 (deep) R VI deep NR 70
Musc-3 mRNA R VI deep NR 12, 77
(deep)
D-1 (deep) R V-VI VI ant: VI 14, 60, 81 located extra-synaptically
and in dendritic spines (69)
D-1 mRNA (deep) R VI V-VI NR 22
D-2 R V(VI) deep ant: deep 9
post: sup
5HT-2 (deep) R V-VI deep A 28, 51
CCK (deep) GP VI deep ant: deep 87
post:
deep
CCK (deep) R VI NR NR 67 located in cortico-thalamic
neurones
CCK mRNA R VI, V deep ant: deep 63, 65 neuronal labelling
post: A
Ad-1 (deep) R V-VI deep deep 19, 24, 35, 83
Notes, and key to authors: see Table 11.5.

1 2
3 4
Figure 11.4. Ligands whose binding to the cortex shows continuity between deep layers of neocortex and
deep layers of mesocortex. 1. Bungarotoxin (reproduced with permission from Clarke, P.B.S. (1985), Nicotinic
acid binding in rat brain: autoradiographic comparison of [3H]acetylcholine, [3H]nicotine and [3H]-alpha-
bungarotoxin. Journal of Neuroscience, 5, 1307–1315, Figure 3b, right). 2. Tachikinin (reproduced with per-
mission of Blackwells Science, Ltd., from Bergstrom, L., Torrens, Y., Saffroy, M., Beaujouan, J.C., Lavielle, S.,
Chassaing, G., Morgat, J.L., Glowinski, J. and Marquet, A. (1987) (3H) Neurokinin B and I251-boulton hunter
eledoisin label identical tachykinin binding sites in the rat brain. Journal of Neurochemistry, 48(1), 125–133.
Figure 11.7 upper right). 3. Muscarinic-3 mRNA (reproduced with permission from Buckley, N.J., Bonner, T.I.
and Brann, M.R. 1988, Localization of a family of muscarinic receptor mRNAs in rat brain. Journal of Neuro-
science 8, 4646–4652; Figure 2 col 3, B3). 4. CCK mRNA (reproduced with permission, from Schiffman, S.N.
and Vanderhaeghen, J.J. (1991), Distribution of cells containing mRNA encoding cholecystokinin in the rat
central nervous system. Journal of Comparative Neurology, 304, 219–233, Figure 1f).
(known as “DARPP-32”) labels a band in lamina VI of the neocortex, which does not
continue into the entorhinal cortex. Nevertheless, there is substantial support for the broad
hypothesis in so far as it has implications for laminae V and VI.
4.2.2. Banding Pattern 2: Bands in the deeper half of the neocortex which come
to fill all cell-rich laminae of the mesocortex
According to the hypothesis of this paper, the neocortex is envisaged to have laminae
added which are not represented in the mesocortex. However, the mesocortex (particularly

Table 11.2. PATTERN II: Ligands which label laminae V & VI or IV, V & VI of neocortex,
and come to fill all cell-rich laminae of mesocortex
AChE (deep, Ms IV-VI or IV II-VI II-VI 29

intermediate)
AChE R IV(V-VI) all ant: all 13 colocalized in GABA
post: II, neurones (27A)
V-VI
5HT-1a R V, VI all III-VI 66, 76
Kainate (deep) R V-VI all III-VI 41, 73
the entorhinal cortex) is not significantly reduced in overall thickness compared with the
neocortex. Therefore the neocortical laminae on either side of this added component
would be expected to “fill out” the thickness of the cortex, where they extend into the
mesocortex. It is therefore predicted that bands of labelling in the deeper half of the neo-
cortex would come to fill out layers of mesocortex which are thicker than in the neocortex.
A few markers (Table 11.2 and Figure 11.5, parts 1 and 2) clearly show this pattern,
including acetylcholinesterase, serotonin 1a receptors, and the deep band of kainate recep-
tors. It should however be noted that in monkeys, acetylcholinesterase has a heterogen-
eous distribution in both entorhinal and cingulate cortex, which does not clearly fit the
predicted pattern (Table 11.5). Another exception (Table 11.5) is the somatostatin receptor
in rats, which is found in laminae IV to VI of neocortex, but only in the deep laminae of
the entorhinal cortex.
4.2.3. Banding Pattern 3: Bands in lamina IV and/or III of neocortex which terminate
abruptly on approach to the rhinal fissure or the cingulate cortex
The most important prediction from the broad hypothesis investigated here is that
chemical markers which label bands in the middle laminae of the neocortex (IV and/or
III), should show an abrupt termination on approach to either the rhinal fissure or the
cingulate cortex. A wide variety of markers show this pattern, and the published pictures
are sometimes of striking clarity. Markers whose bands in laminae III or IV terminate at
the rhinal fissure include cholinergic receptors (both nicotinic and muscarinic), aminer-
gic receptors (noradrenergic, serotoninergic, histaminergic), GABA and benzodiazepine
receptors, neurotensin and adenosine receptors. The pattern at the border between
neocortex and cingulate cortex is more complex, with a difference between anterior
and posterior cingulate cortex which is repeated for several markers: the neocortical
middle-depth band continues at a superficial level in the anterior cingulate cortex, but is
missing in the posterior cingulate cortex. Table 11.3 shows more detail of the ligands
whose banding fits this pattern, and Figure 11.6 gives illustrative examples. Apparent
exceptions to the prediction are found in primates (Table 11.5)—the DARPP-32 band in
lamina III of monkeys, and the 5HT-2 band in lamina III of humans. Both of these bands
continue, in the same or different laminar positions, into the entorhinal or cingulate
cortex.

1 2 3
DLG
ZI
Laminar Continuity Across the Rhinal Fissure

LH
4 5
Figure 11.5. Parts 1 and 2. Ligands binding in the deep half of the neocortex which come to fill out a greater thickness of the mesocortex. 1. Acetylcholine-esterase (repro-
duced with permission from Bogdanovic, N., Islam, A., Nilson, L., Bergstrom, L., Winblad, B. and Adem, A. (1993), Effects of nucleus basalis lesion on muscarinic receptor
subtypes. Experimental Brain Research, 97, 225–232, Figure 3, left side) 2. 5HT1a receptors, (reproduced with permission from Pompeiano, M., Palacios, J.M. and Mengod, G.
(1992), Distribution and cellular localization of mRNA coding for 5HT1a receptor in the rat brain: correlation with receptor binding. Journal of Neuroscience, 12, 440–453;
Figure 2D’). Parts 3, 4 and 5. Ligands binding in laminae I and II (or I-III) of neocortex, and continuing into superficial laminae of the mesocortex. 3. NMDA receptors
(reproduced with permission from Maragos, W.F., Penney, J.B. and Young, A.B. (1988) Anatomic correlation of NMDA and 3H-TCP-labeled receptors in rat brain. Journal of
Neuroscience, 8, 493–501; Figure 1F; note also detail of narrow unlabelled band in entorhinal cortex). 4. DARPP-32 (reprinted from Molecular Brain Research, M. Schalling
et al., Distribution and cellular localization of DARPP-32 mRNA in rat brain, pp. 139–149, Figure 1i, Copyright [1990] with permission of Elsevier Science). 5. Muscarinic-3
receptors (reprinted from Neuroscience, Vol. 40, M.T. Vilaro et al., Muscarinic cholinergic receptors in the rat caudate-putamen and olfactory tubercle belong predominantly to
231
the M4 class: in situ hybridization and receptor autoradiography evidence, pp. 159–167, Figure 2D. Copyright [1991], with permission from Elsevier Science).

Table 11.3. PATTERN III: Ligands in lamina IV and/or III of neocortex, which terminate
abruptly on approach to the rhinal fissure or cingulate cortex
N (middle) R III-IV A ant: sup 15, 16, 62, 72 located in both TC

post: A axons (57) and
cell bodies (61)
N mRNA R IV A A 80
(alpha-3 submit)
Musc-2 R IV(gap) no gap no gap 3, 37, 61, 70* located on thalamic
axons and cortical
cell dendrites (78)
NE-alpha1b R IV-V A ant: sup 39 localized in cell
mRNA post: A bodies
NE-beta2 R IV A ant: NR 32, 46
post: A
5HT-2 (middle) R IV A ant: sup 2, 7, 21, 28 in 5HT axons in S1,
post: A 43, 52, 68 M1 lam V (7)
Hist-1 R IV A A 50
GABAa R III-IV A ant: NR 49, 74+
post: A
BZ R IV A ant: sup 23, 74, 75
post: A
NT (middle) R IV A NR 42
Ad-1 (middle) R IV A NR 19, 20, 24,
35, 66, 83

* But see (82) for conflicting results
+
But see (49) for conflicting results
4.2.4. Banding Pattern 4: Bands in lamina I and/or II of neocortex (whether or not

lamina III is also labelled), narrowing to superficial bands in the mesocortex
It was noted above that, if the broad hypothesis about laminar reorganization between neo-
and meso-cortex is correct, neocortical laminae on either side of this added component
would be expected to “fill out” the thickness of the cortex, where they extend into the
mesocortex. This appeared to be true be for deep laminae. It might also then be expected
that there be some form of continuity between the most superficial laminae of the neocortex,
and superficial laminae of the mescortex. The details of this cannot be predicted from the
broad hypothesis.
However, available data show a tendency for several markers, which form a band in
laminae I and II (or I-III) of the neocortex to continue as a much narrower band within the
most superficial layer of the mesocortex. Ligands which show such a pattern in the
entorhinal cortex include a subunit of the nicotinic receptor mRNA, mRNA for muscarinic
receptors, DARPP-32 in rats, two types of glutamate receptor, and the mRNA for the CCK
receptor. Data for the cingulate cortex is less complete, and, where available, fits the
pattern in the entorhinal cortex in only some cases. Table 11.4 gives details of the ligands
whose banding fits this pattern, and Figure 11.5 (3–5) illustrates some examples (see also
Figure 11.4(4)). There are also some apparent exceptions: Muscarinic M-5 receptors, which
form a band in superfical layers of the neocortex appear to form no band in the entorhinal

1 2
3 4
Figure 11.6. Ligands binding in lamina III and or IV of neocortex, in a band which terminates on approaching
cingulate or entorhinal cortices. 1. Nicotinic receptors (reproduced with permission from Clarke, P.B.S. (1985),
Nicotinic acid binding in rat brain: autoradiographic comparison of [3H]acetylcholine, [3H]nicotine and [3H]-
alpha-bungarotoxin. Journal of Neuroscience, 5, 1307–1315, Figure 3b). 2. 5HT-2 receptors (reprinted from
Brain Research, Vol. 453, M.E. Blue et al., Correspondance between 5HT2 receptors and serotonergic axons in
rat cortex. pp. 315–328, Figure 1A. Copyright [1988], with permission of Elsevier Science). 3. Noradrenergic
beta-2 receptors (from Ordway et al., 1988, Journal of Pharmacology and Experimental Therapeutics, Vol.
247, pp. 379–389, Figure 1; reproduced with permission). 4. Benzodiazepine receptors (reproduced with
permission of Blackwell Science, Ltd., from Niddam, R., Dubois, A., Scatton, B., Arbilla, S. and Langer, S.Z.
(1987), Autoradiographic localization of [3H]zolpidem binding sites in the rat CNS: comparison with the distri-
bution of [3H]nitrazepam binding sites. Journal of Neurochemistry, 49(3), 890–899, Figure 1b).
Table 11.4. PATTERN IV: Ligands with bands in neocortex lamina I and/or II (whether or not
lamina III is also labelled), narrowing to quite superficial band in the mesocortex
N mRNA alpha-2 R I-III I NR 80

subunit (sup)
Musc-3 mRNA R I-II I I 12, 77
Musc-4 mRNA R III I NR 12, 77
DARPP-32 R I-III I-II A 47, 64
NMDA R I-III I, III* ant: I-II 31, 38, 41
post: A
Kainate R I-II I I-II 41
CCK mRNA (sup) R I-II I ant: I 65
post: A
Notes, and key to authors: see Table 11.5

* In entorhinal cortex there is a sharply defined lamina between the NMDA-labelled laminae I and III which is
not labelled (see Figure 11.5(3)).

Table 11.5. EXCEPTIONS
Musc-2 (sup) R I-II/III A sup 37, 70

AChE fibre Mk IV or IV-VI H/C H/C 27, 40
D-1 (deep) C VI all all 14, 59, 60
Dopamine fibres R VI all all 17, 18, 54, 86
Dopamine fibres Mk H/C H/C H/C 1, 4, 36
DARPP-32 (deep) R VI A ant: deep 64
post: deep
DARPP-32 Mk II-III + V-VI II-III V-VI 5
neurones
NE-alpha-1 R IV, V(2) middle sup 33, 48 develops with
NE fibres (34)
NE-beta-1 R I-III A A 32, 46, 57
5HT-2 H III, V I, II-III, V III or III, V 30, 53*
QUIS R I-II I-II(III) NR 45
SS receptor R IV-V, VI deep ant: deep 58
post: deep
* But see (86) for differing results

Notes
Tables 11.1–11.5
Where labelling in adjacent laminae consists of two separate bands they are separated by a comma (e.g. “V, VI”)
When labelling within a single lamina is split into two separate bands it is indicated as “[2]”. When labelling is
continuous across laminae this is indicated with a “-” (e.g “V-VI”). References cited are those in which the
tabulated labelling is most clearly seen.
Abbreviations (Tables 11.1–11.5)
C = cat; GP = guinea pig; H = human; Mk = monkey; Ms = mouse; R = rat; NR = Not reported or pictures not
clear; Ad = adenosine; BZ = benzodiazepine receptors; CCK = cholecystokinin; D = dopamine; Musc = muscarinic;
Hist = histamine; N = nicotinic; NT = neurotensin; QUIS = quisqualate receptors; S1,M1= primary somato-
sensory and motor cortical areas; SS = somatostatin; A = ligand absent, or band of labelling not present;
H/C = complex and heterogeneous distribution.
Key to authors (Tables 11.1–11.5)
1. Akil and Lewis (1993)
2. Appel et al. (1990)
3. Aubert et al. (1992)
4. Berger et al. (1988)
5. Berger et al. (1990)
6. Bergstrom et al. (1987)
7. Blue et al. (1988)
8. Bogdanovic et al. (1993)
9. Bouthenet et al. (1987)
10. Bravo and Karten (1992)
11. Broide et al. (1995)
12. Buckley et al. (1988)
13. Butcher and Woolf (1984)
14. Camps et al. (1990)
15. Clarke et al. (1984)
16. Clarke et al. (1985)
17. Descarries et al. (1987)
18. Doucet et al. (1988)
19. Fastbom and Fredholm (1987)
20. Fastbom et al. (1987)
21. Fischette et al. (1987)
22. Fremeau et al. (1991)
23. Giradino et al. (1993)
24. Goodman and Synder (1982)

25. Hallanger et al. (1986)

26. Harfstrand et al. (1988)
27. Hedreen et al. (1984)
28. Hoffman et al. (1987)
29. Hohmann and Ebner (1985)
30. Hoyer et al. (1986)
31. Jacobson and Cottrell (1993)
32. Johnson et al. (1989)
33. Jones et al. (1985a)
34. Jones (1985b)
35. Lee and Reddington (1986)
36. Lewis et al. (1988)
37. Mash and Potter (1986)
38. Maragos et al. (1988)
39. McCune et al. (1993)
40. Mesulam et al. (1984)
41. Monoghan and Cotman (1982)
42. Moyse et al. (1987)
43. Nakada et al. (1984)
44. Niddam et al. (1987)
45. Olsen et al. (1987)
46. Ordway et al. (1988)
47. Ouimet et al. (1989)
48. Palacios et al. (1987)
49. Palacios et al. (1981a)
50. Palacios et al. (1981b)
51. Pazos and Palacios (1985)
52. Pazos et al. (1985)
53. Pazos et al. (1987)
54. Phillipson et al. (1987)
55. Pieribone et al. (1994)
56. Pompeiano et al. (1992)
57. Rainbow et al. (1984)
58. Reubi and Maurer (1985)
59. Richfield et al. (1987)
60. Richfield et al. (1989)
61. Rossner et al. (1995)
62. Sahin et al. (1992)
63. Savasta et al. (1988)
64. Schalling et al. (1990)
65. Schiffman and Vanderhaegen (1991)
66. Schroeder et al. (1989)
67. Senatorov et al. (1995)
68. Slater and Patel (1983)
69. Smiley et al. (1994)
70. Spencer et al. (1986)
71. Sugaya et al. (1991)
72. Swanson et al. (1987)
73. Unnerstall and Wamsley (1983)
74. Unnerstall et al. (1981)
75. Unnerstall et al. (1982)
76. Verge et al. (1986)
77. Vilaro et al. (1991)
78. Vogt and Burns (1988)
79. Vogt et al. (1990)
80. Wada et al. (1989)
81. Wamsley et al. (1991)
82. Wang et al. (1989)
83. Weber et al. (1988)
84. Weber et al. (1990)
85. Xia and Haddad (1992)
86. Yoshida et al. (1988)
87. Zarbin et al. (1983)

cortex. Noradrenergic beta-1 receptors are exceptional in a similar way. Quisqualate

receptors show continuity between neo- and entorhinal cortex, but the band of labelling
appears not to narrow in the entorhinal cortex as do other ligands included in Table 11.4.
5. SUMMARY AND COMMENT
Despite a number of exceptional cases, the broad hypothesis tested in this article receives
striking support. The detailed version of the hypothesis supported by the data in Table
11.1–11.4 is shown in Figure 11.7. According to this summary, lamina III and/or IV of the
neocortex are missing in the entorhinal cortex and other areas of mesocortex, while deeper
laminae of the neocortex appear to fill out the missing layers of the mesocortex. Layers
superficial to lamina III (and sometimes including lamina III) of the neocortex tend to be
reduced to a narrow band (laminae I and/or II) of the entorhinal cortex, and possibly other
regions of mesocortex. The slab of cortical tissue which is added in the neocortex appears
to be inserted within the original lamina II as it is found in the mesocortex3.
It is interesting also to note that the laminar transformation between neo- and meso-
cortex documented here in mammals is not the same as that advocated by Reiner et al.
(1993) between the rudimentary cortex of reptiles and the more advanced cortex of mammals.
Specifically, there is no superficial layer of cells (and associated chemical markers) in
common between reptiles and mammals (see Figure 11.3), whereas there is good evidence
of continuity of superficial cell layers and markers between neo- and mesocortex (Table
11.4, and Figure 11.5 [parts 3–5]).
Most of the data tabulated in Tables 11.1–11.5 refer to the rat. It may be asked how far
the laminar transformation between neo- and mesocortex summarised in Figure 11.7
applies in other species. Figure 11.2 shows that, as far as adult cellular arrangement goes,
there is considerable similarity between widely different mammalian species in the lam-
inar transformation in the neighbourhood of the rhinal fissure. There is also similarity in
the developmental sequences between different species (mouse, marsupial opossum).
However, it is not clear whether this similarity across species extends to the chemical
markers of different laminae compiled in Tables 11.1–11.5. In the few instances where
sufficent data are available for species with larger brains, the laminar pattern appears to be
more complex and more heterogeneous than in rats. In animals, with larger brains it may
not be possible to summarise the banding pattern of chemical markers in terms of simple
rules for laminar transformation between neocortex and mesocortex. Nevertheless, the
data compiled here referring mainly to animals with smooth brains, together with the other
data mentioned, suggests a systematic reorganization of laminae on either side of the
rhinal fissure.
Since many of the ligands whose banding is summarised in Figure 11.7 are major
neurotransmitters, with potent biophysical effects on cortical neurones, the laminar
3
Sanides (1970) comes close to this formulation of the laminar transformation between neo- and meso-cortex
(periallocortex in his terminology). He notes that the periallocortex receives thalamic and olfactory input to
superficial cell layers compared with lamina IV in the neocortex, and emphasises another description of
cellular arrangement of the periallocortex—as two cell layers, separated by a cell-sparse layer, the “lamina
dissecans”. He regards neocortical growth in mammals as the accumulation of concentric rings of growth
superimposed on the original limbic ring; and in this scheme the periallocortex is the first growth ring.

Neocortexx
Neocorte mesocortex
I
II
III
(IV)
IV
V/VI
rhinal
fissure
Figure 11.7. Schematic summary of laminar transformation between neo- and mesocortex.
transformation shown is likely to have important implications for large scale cybernetic
differences between neo- and mesocortex. It is premature to explore these differences in
detail. However, there is great scope for further investigation of the cybernetic significance
of the different markers. The connectional data referred to in the early part of this paper
imply that the laminae added in the neocortex, as part of the distinctively mammalian
cortex, have a much richer endowment of cortico-cortical connections than the meso-
cortex. This being so, it is likely that the mammalian cortex has developed specifically to
allow representation by complex combinations of linked co-active neurones—that is by
cell assemblies (to use Hebb’s terminology). Two sorts of information are needed before
the cybernetic differences between neo- and mesocortex can be unravelled in more detail:
(i) The cellular elements upon which each ligand is located—be they postsynaptic neur-
onal cell bodies or dendrites, presynaptic terminals of various kinds, or glial cells—need to
be identified. (ii) The biophysical role which each receptor, neurotransmitter or neuro-
modulator plays for the structure on which it is located needs to be made clear. For most of
the chemical markers listed in the tables above, the detail on either of these questions is
scanty or non-existant. In addition, the assimilation of such data into a coherent scheme of
the cybernetic differences between different regions of cortex depends on the develop-
ment of hypotheses about the dynamics of laminar interaction and their role in information
processing. A start was made to the latter enterprise in two recent papers (Miller, 1996a,b;
see also Miller, this volume). This included one example of how laminar differences in
distribution of a neurochemical marker may have cybernetic significance: the NMDA
receptors are believed to be of special importance in strengthening of excitatory synapses,
according to the Hebbian paradigm. They are particuarly abundant in laminae II and III of
the neocortex. This is one of the arguments for suggesting that these laminae are central in
the formation of neural assemblies, whose formation is envisaged to rely on strengthening
of excitatory synapses.
One further point of cybernetic significance is offered before finishing this paper: in
origin, the mesocortex has links with the olfactory system, while the mammalian neocor-
tex has links mainly with visual, auditory and somatosensory systems. The former of these

has, in principle, little ability to represent exact temporal patterns, since olfactory stimuli
do not give information which is fine-grained in the time dimension. Moreover, olfactory
information accesses the cortex directly without the relay through the thalamus. However,
the latter three sensory systems do have the ability to represent precise temporal patterns,
and are relayed to the cortex via the thalamus. It has recently been suggested (Miller,
1996b; Miller, this volume) that the interplay between thalamus and neocortex serves to
enhance the capacity of the neocortex to represent temporal information within cell
assemblies. Thus, the emergence, during evolution, of the mammalian neocortex can be
regarded as a development which allows representation of exact temporal structure, as
well as the “spatial” information content of cell assemblies.
REFERENCES
Akil, M. and Lewis, D.A. (1993) The dopaminergic innervation of monkey entorhinal cortex. Cerebral Cortex,
3, 533–550.
Appel, N.M., Mitchell, W.M., Garlick, R.K., Glennon, R.A., Teitler, M. and De Souza, E.B. (1990) Autoradio-
graphic characterization of (+/–)-1-(2,5-dimethoxy-4-[125I]iodophenyl)-2-aminopropane ([125I]DOI) bind-
ing to 5HT2 and 5HT1c receptors in rat brain. Journal of Pharmacology and Experimental Therapeutics,
255, 843–857.
Aubert, I., Cecyre, D., Gauthier, S. and Quirion, R. (1992) Characterization and autoradiographic distribution of
[3H]AF-DX384 binding to putative muscarinic M2 receptors in the rat brain. European Journal of Pharma-
cology, 217, 173–184.
Berger, B., Trottier, S., Verney, C., Gaspar, P. and Alvarez, C. (1988) Regional and laminar distribution of the
dopamine and serotonin innervation in the macaque cerebral cortex: a radioautographic study. Journal of
Berger, B., Febvret, A., Greengard, P. and Goldman-Rakic, P.S. (1990) DARPP-32, a phosphoprotein enriched
in dopaminoceptive neurons bearing dopamine D1 receptors: distribution in the cerebral cortex of the
newborn and adult rhesus monkey. Journal of Comparative Neurology, 299, 327–348.
Berger-Sweeney, J. and Hohmann, C.F. (1997) Behavioural consequences of abnormal cortical development:
insights into developmental disabilities. Behavioral Brain Research, 86, 121–142.
Bergstrom, L., Torrens, Y., Saffroy, M., Beaujouan, J.C., Lavielle, S., Chassaing, G., Morgat, J.L., Glowinski, J.
and Marquet, A. (1987) (3H)neurokinin B and I251-boulton hunter eledoisin label identical tachykinin
binding sites in the rat brain. Journal of Neurochemistry, 48, 125–133.
Blue, M.E., Yagaloff, K.A., Mamounas, L.A., Hartig, P.R. and Molliver, M.E. (1988) Correspondance between
5HT2 receptors and serotonergic axons in rat cortex. Brain Research, 453, 315–328.
Bogdanovic, N., Islam, A., Nilson, L., Bergstrom, L., Winblad, B. and Adem, A. (1993) Effects of nucleus
basalis lesion on muscarinic receptor subtypes. Experimental Brain Research, 97, 225–232.
Bouthenet, M.-L., Martres, M.-P., Sales, N. and Schwartz, J.-C. (1987) A detailed mapping of dopamine D-2
receptors in rat central nervous system by autoradiography with [125I]iodosulpride. Neuroscience, 20,
117–155.
Braak, H. and Braak, E. (1992) The human entorhinal cortex: normal morphology, and lamina-specific pathology
in various diseases. Neuroscience Research, 15, 6–31.
Braitenberg, V. and Schüz, A. (1991) Anatomy of the cortex statistics and geometry. Springer-Verlag: Berlin
Heidelberg.
Bravo, H. and Karten, H.J. (1992) Pyramidal neurons of the rat cerebral cortex, immunoreactive to nicotinic
acetylcholine receptors, project mainly to subcortical targets. Journal of Comparative Neurology, 320, 62–68.
Broide, R.S., O’Connor, L.T., Smityh, M.A., Smith, J.A.M. and Leslie, F.M. (1995) Developmental expression
of alpha-7 neuronal nicotinic receptor messenger mRNA in rat sensory cortex and thalamus. Neuroscience,
67, 83–94.
Buckley, N.J., Bonner, T.I. and Brann, M.R. (1988) Localization of a family of muscarinic receptor mRNAs in
rat brain. Journal of Neuroscience, 8, 4646–4652.
Butcher, L.L. and Woolf, N.J. (1984) Histochemical distribution of acetylcholinesterase in the central nervous
system: clues to the localization of cholinergic neurons. In: A. Bjorklund, T. Hokfelt and M.J. Kuhar (eds),
Handbook of chemical neuroanatomy, Vol. 3, Classical transmitters and transmitter receptors in the CNS.
Amsterdam: Elsevier, pp. 1–50.
Camps, M., Kelly, P.H. and Palacios, J.M. (1990) Autoradiographic localization of dopamine D1 and D2 recep-
tors in the brain of several mammalian species. Journal of Neural Transmission, 80, 105–127.

Cho, H.J. and Takagi, H. (1993) Cholecystokinin (CCK)-8-immunoreactivity in the piriform cortex of the rat
with special reference to its fine structures. Osaka City Medical Journal, 39, 75–92.
Chronister, R.B. and White, Jr., L.E. (1977) Fiber architecture of the hippocampal formation: Anatomy, projec-
tions and structural significance. In: R.L. Isaacson and K.H. Pribram (eds), The Hippocamus, Vol. 1,
Stucture and development. New York and London: Plenum Press, pp. 9–39.
Clarke, P.B.S., Pert, C.B. and Pert, A. (1984) Autoradiographic distribution of nicotine receptors in rat brain.
Clarke, P.B.S. (1985) Nicotinic acid binding in rat brain: autoradiographic comparison of [3H]acetylcholine,
[3H]nicotine and [3H]-alpha-bungarotoxin. Journal of Neuroscience, 5, 1307–1315.
Descarries, L., Lemay, B., Doucet, G. and Berger, B. (1987) Regional and laminar density of the dopamine
innervation in adult rat cerebral cortex. Neuroscience, 21, 807–824.
Doucet, G., Descarries, L., Audet, M.A., Garcia, S. and Berger, B. (1988) Radioautographic method for quanti-
fying regional monoamine innervations in the rat brain. Application to the cerebral cortex. Brain Research,
441, 233–259.
Ebner, F.F. (1976) The forebrain of reptiles and mammals. In: R.B. Masterton, M.E. Bitterman, C.B.G. Campbell
and N. Hutton (eds), Evolution of brain and behaviour in vertebrates. Hillsdale, N.J.: Erlbaum, L. Associates,
pp. 147–167.
Fastbom, J. and Fredholm, B.B. (1987) Regional differences in the GTP-dependence of adenosine receptor
binding in rat brain. Acta Physiologica Scandinavica, 131, 467–469.
Fastbom, J., Pazos, A. and Palacios, J.M. (1987) The distribution of adenosine A1 receptors and 5′-nucleotidase
in the brain of some commonly used experimental animals. Neuroscience, 22, 813–826.
Fischette, C.T., Nock, B. and Renner, K. (1987) Effects of 5,7-dihydroxytryptamine on serotonin1 and serotonin
2 receptors throughout the rat central nervous system using quantitative autoradiography. Brain Research,
421, 263–279.
Fremeau, R.T., Duncan, G.E., Fornaretto, M.-G., Dearry, A., Gingrich, J.A., Breese, G.R. and Caron, M.G.
(1991) Localization of D1 dopamine receptor mRNA in brain supports a role in cognitive, affective and
neuroendocrine aspects of dopaminergic neurotransmission. Proceedings of the National Academy of
Sciences, U.S.A., 88, 3772–3776.
Garrett, B., Finsen, B. and Wree, A. (1994) Parcellation of cortical areas by in situ hybridization for somatostatin
mRNA in the frontal, parietal, occipital and temporal regions. Anatomy and Embryology, 190, 389–398.
Giradino, L., Zanni, M., Velardo, A., Amato, G. and Calza, L. (1993) Effect of sertraline treatment on benzo-
diazepine receptors in the rat brain. Journal of Neural Transmission, General Section, 94, 31–41.
Goodman, R.R. and Snyder, S.H. (1982) Autoradiographic localization of adenosine receptors in rat brain using
[3H]cyclohexyladenosine. Journal of Neuroscience, 2, 1230–1241.
Hall, J.A. (1971) Efferent projections from the general cortex in the turtle (Pseudemys scripta). Masters Thesis,
Brown University, U.S.A.
Hallanger, A.E., Wainer, B.H. and Rye, D.B. (1986) Colocalization of gamma-aminobutyric acid and acetyl-
cholinesterose in rodent cortical neurons. Neuroscience, 19, 763–769.
Harfstrand, A., Adem, A., Fuxe, K., Agnati, L., Andersson, K. and Nordberg, A. (1988) Distribution of nicotinic
cholinergic receptors in the rat tel- and diencephalon: a quantitative receptor autoradiographical study using
[3H]-acetylcholine, [alpha-125]bungarotoxin and [3H]nicotine. Acta Physiologica Scandinavica, 132, 1–14.
Hedreen, J.C., Uhl, G.R., Bacon, S.J., Famborough, D.M. and Price, D.L. (1984) Acetycholinesterase-immuno-
reactive axonal network in monkey visual cortex. Journal of Comparative Neurology, 226, 246–254.
Herkenham, M. (1978) The connections of the nucleus reunions thalami: Evidence for a direct thalamo-
hippocampal pathway in the rat. Journal of Comparative Neurology, 177, 589–610.
Hevner, R.F. and Wong-Riley, M.T.T. (1992) Entorhinal cortex of the human, monkey and rat: metabolic map as
revealed by cytochrome oxidase. Journal of Comparative Neurology, 326, 451–469.
Hoffman, B.J., Scheffel, U., Lver, J.R., Karpa, M.D. and Hartig, P.R. (1987) N1-methyl-2–125I-Lysergic acid
diethylamide, a preferred ligand for in vitro and in vivo characterization of serotonin receptors. Journal of
Neurochemistry, 48, 115–124.
Hohmann, C.F. and Ebner, F.F. (1985) Development of cholinergic markers in mouse forebrain. I. Choline
acetyltransferase enzyme activity and actylcholinesterase histochemistry. Developmental Brain Research,
23, 225–241.
Hoyer, D., Pazos, A., Probst, A. and Palacios, J.M. (1986) Serotonin receptors in the human brain. II. Characteriza-
tion and autoradiographic localization of 5HT1c and 5HT2 recognition sites. Brain Research, 376, 97–107.
Jacobsen, W. and Cottrell, G.A. (1993) Rapid visualization of NMDA receptors in the brain: characterization of
(+)-3-[125I]-iodo-MK-801 binding to thin sections of rat brain. Journal of Neuroscience Methods, 46,
17–27.
Johnson, E.W., Wolfe, B.B. and Molinoff, P.B. (1989) Regulation of subtypes of beta-adrenergic receptors in rat
brain following treatment with 6-hydroxydopamine. Journal of Neuroscience, 9, 2297–2305.
Jones, L.S., Gauger, L.L. and Davis, J.N. (1985a) Anatomy of brain alpha1-adrenergic receptors: In vitro auto-
radiography with [125I]-HEAT. Journal of Comparative Neurology, 231, 190–208.

Jones, L.S., Gauger, L.L., Davis, J.N., Slotkin, T.A. and Bartolome, J.V. (1985b) Postnatal development of brain
alpha-1-adrenergic receptors: in vitro autoradiography with [125]HEAT in normal rats and rats treated with
alpha-difluoromethylornithine, a specific irreversible inhibitor of ornithine decarboxylase. Neuroscience,
15, 1195–1202.
Kaneko, T., Shigemoto, R., Nakanishi, S. and Mizuno, N. (1994) Morphological and chemical characteristics of
substance P receptor-immunoreactive neurons in the rat neocortex. Neuroscience, 60, 199–211.
Lee, K.S. and Reddington, M. (1986) Autoradiographic evidence for multiple CNS binding sites for adenosine
derivatives. Neuroscience, 19, 535–549.
Lewis, D.A., Foote, S.L., Goldstein, M. and Morrison, J.H. (1988) The dopaminergic innervation of monkey
prefrontal cortex: a tyrosine hydroxylase immunohistochemical study. Brain Research, 449, 225–243.
Lohman, A.H. and Mentink, G.M. (1972) Some cortical connections of the tegu lizard (tupinambis teguixin).
Maragos, W.F., Penney, J.B. and Young, A.B. (1988) Anatomic correlation of NMDA and 3H-TCP-labeled
receptors in rat brain. Journal of Neuroscience, 8, 493–501.
Mash, D.C. and Potter, L.T. (1986) Autoradiographic localization of M1 and M2 muscarine receptors in the rat
brain. Neuroscience, 19, 551–564.
McCune, S.K., Voigt, M.M. and Hill, J.M. (1993) Expression of multiple alpha adrenergic receptor subtype
messenger mRNAs in the adult rat brain. Neuroscience, 57, 143–151.
Mesulam, M.-M., Rosen, A.D. and Mufson, E.J. (1984) Regional variations in cortical cholinergic innervation:
chemoarchitectonics of acetylcholinestarase-containing fibers in the macaque brain. Brain Research, 311,
245–258.
Miller, R. (1996a) Neural assemblies and laminar interactions in the cerebral cortex. Biological Cybernetics, 75,
253–261.
Miller, R. (1996b) Cortico-thalamic interplay and the security of operation of neural assemblies and temporal
chains in the cerebral cortex. Biological Cybernetics, 75, 263–275.
Monoghan, D.T. and Cotman, C.W. (1982) The distribution of [3H]kainic acid binding sites in rat CNS as
determined by autoradiography. Brain Research, 252, 91–100.
Moyse, E., Rostene, W., Vial, M., Leonard, K., Mazella, J., Kitabgi, P., Vincent, J.-P. and Beaudet, A. (1987)
Distribution of neurotensin binding sites in rat brain: a light microscopic radioautographic study using
monoiodo[125I]tyr3-neurotensin. Neuroscience, 22, 525–536.
Nakada, M.T., Wieczorek, C.M. and Rainbow, T.C. (1984) Localization and characterization by quantitative
autoradiography of [125I]LSD binding sites in rat brain. Neuroscience Letters, 49, 13–18.
Niddam, R., Dubois, A., Scatton, B., Arbilla, S. and Langer, S.Z. (1987) Autoradiographic localization of
[3H]zolpidem binding sites in the rat CNS: comparison with the distribution of [3H]nitrazepam binding
sites. Journal of Neurochemistry, 49, 890–899.
Olsen, R.W., Szamraj, O. and Houser, C.R. (1987) [3H]AMPA binding to glutamate receptor subpopulations in
rat brain. Brain Research, 402, 243–254.
Ordway, G.A., Gambarana, C. and Frazer, A. (1988) Quantitative autoradiography of central beta adrenoceptor
subtypes: comparison of the effects of chronic treatment with desipramine or central administered
i-isoproterenol. Journal of Pharmacology and Experimental Therapapeutics, 247, 379–389.
Ouimet, C.C., Hemmings, H.C. and Greengard, P. (1989) DARPP-32, a cyclic AMP-regulated phosphoprotein
enriched in dopamine-innervated brain regions. II. Immunocytochemical localization in rat brain. Journal of
Palacios, J.M., Wamsley, J.K. and Kuhar, M.J. (1981a) High affinity GABA receptors-autoradiographic localiza-
tion. Brain Research, 222, 285–307.
Palacios, J.M., Wamsley, J.K. and Kuhar, M.J. (1981b) The distribution of histamine H1 receptors in the rat
brain: an autoradiographic study. Neuroscience, 6, 15–37.
Palacios, J.M., Hoyer, D. and Cortes, R. (1987) Alpha-1 adrenoceptors in the mammailian brain: similar pharma-
cology but different distribution in rodents and primates. Brain Research, 419, 65–75.
Pazos, A. and Palacios, J.M. (1985) Quantitative autoradiographic mapping of serotonin receptors in the rat
brain. I. Serotonin-1 receptors. Brain Research, 346, 205–230.
Pazos, A., Cortes, R. and Palacios, J.M. (1985) Quantitative autoradiographic mapping of serotonin receptors in
the rat brain. II. Serotonin-2 receptors. Brain Research, 346, 231–249.
Pazos, A., Probst, A. and Palacios, J.M. (1987) Serotonin receptors in the human brain-IV. Autoradiographic
mapping of serotonin-2 receptors. Neuroscience, 21, 123–139.
Peters, A., Miller, M. and Kimerer, L.M. (1983) Cholecystokinin-like immunoreactive neurons in rat cerebral
cortex. Neuroscience 8, 431–448.
Phillipson, O.T., Kilpatrick, I.C. and Jones, M.W. (1987) Dopaminergic innervation of the primary visual cortex
in the rat, and some correlations with human cortex. Brain Research Bulletin, 18, 621–633.
Pieribone, V.A., Nicholas, A.P., Dagerlind, A. and Hokfelt, T. (1994) Distribution of alpha-1 adrenoceptors
in rat brain revealed by in situ hybridization experiments utilizing subtype-specific probes. Journal of

Pompeiano, M., Palacios, J.M. and Mengod, G. (1992) Distribution and cellular localization of mRNA coding
for 5HT1a receptor in the rat brain: correlation with receptor binding. Journal of Neuroscience, 12, 440–453.
Rainbow, T.C., Parsons, B. and Wolfe, B.B.(1984) Quantitative autoradiography of beta-1 and beta-2 adrenergic
receptors in rat brain. Proceedings of the National Academy of Science, U.S.A., 81, 1585–1589.
Reiner, A. (1991) A comparison of neurotransmitter-specific and neuropeptide-specific neuronal cell types
present in the dorsal cortex in turtles with those present in the isocortex in mammals: Implications for the
evolution of Isocortex. Brain Behavior and Evolution, 38, 53–91.
Reiner, A. (1993) Neurotransmitter organisation and connections of turtle cortex: Implications for the evolution
of mammalian isocortex. Comparative Biochemistry and Physiology, 104A, 735–748.
Reubi, J.C. and Maurer, R. (1985) Autoradiographic mapping of somatostatin receptors in the rat central nervous
system and pituitary. Neuroscience, 15, 1183–1193.
Richfield, E.K., Debowey, D.L., Penney, J.B. and Young, A.B. (1987) Basal ganglia and cerebral cortex distribu-
tion of dopamine D1 and D2 receptors in neonatal and adult cat brain. Neuroscience Letters, 73, 203–208.
Richfield, E.K., Young, A.B. and Penney, J.B. (1989) Comparative Distributions of dopamine D-1 and D-2
receptors in the cerebral cortex of rats, cats and monkeys. Journal of Comparative Neurology, 286, 409–426.
Robertson, R.T. and Kaitz, S. (1981) Thalamic connections with limbic cortex.I. Thalamocortical projections.
Room, P. and Groenewegen, H.J. (1986) Connections of the parahippocampal cortex in the cat. II. Subcortical
Afferents. Journal of Comparative Neurology, 251, 451–473.
Rossner, S., Schliebs, R., Perez-Polo, J.R., Wiley, R.G. and Bigl, V. (1995) Differential changes in cholinergic
markers from selected brain regions after specific immunolesion of rat cholinergic basal forebrain system.
Journal of Neuroscience Research, 40, 31–43.
Sahin, M., Bowen, W.D. and Donghue, J.P. (1992) Location of nicotinic and muscarinic cholinergic and u-opiate
receptors in rat cerebral neocortex: evidence from thalamic and cortical lesions. Brain Research, 579,
135–147.
Sanderson, K.J. and Aitken, L.M. (1990) Neurogenesis in a marsupial: The brush-tailed possum (trichosurus
vulpecula) I. Visual and auditory pathways. Brain Behavior and Evolution, 35, 325–338.
Sanderson, K.J. and Weller, W.L. (1990) Gradients of neurogenesis in possum neocortex. Developmental Brain
Research, 55, 269–274.
of neocortex evolution. In: C.R. Noback and W. Montagna (eds), The Primate Brain. New York: Appleton-
Century-Crofts, pp. 137–208.
Sarnat, H.B. and Netsky, M.G. (1974) Evolution of the Nervous System. Oxford University Press, Oxford.
Savasta, M., Palacios, J.M. and Mengod, G. (1988) Regional localization of the mRNA coding for the neuro-
peptide cholecystokinin in the rat brain studied by in situ hybridization. Neuroscience Letters, 93, 132–138.
Schalling, M., Djurfeldt, M., Hokfelt, T., Ehrlich, M., Kurihara, T. and Greengard, P. (1990) Distribution and
cellular localization of DARPP-32 mRNA in rat brain. Molecular Brain Research, 7, 139–149.
Schiffman, S.N. and Vanderhaegen, J.J. (1991) Distribution of cells containing mRNA encoding cholecystokinin
in the rat central nervous system. Journal of Comparative Neurology, 304, 219–233.
Schroeder, H., Zilles, K., Maelicke, A. and Hajos, F. (1989) Immunohisto- and cytochemical localization of
cortical nicotinic cholinoceptors in rat and man. Brain Research, 502, 287–295.
Senatorov, V.V., Trudeau, V.L. and Bin, H. (1995) Expression of cholecystokinin mRNA in combined fluores-
cence in situ hybridization and retrograde tracing study in the ventrolateral thalamus of the rat. Brain
Research, Molecular Brain Research, 30, 87–95.
Shibata, H. (1993) Efferent projections from the anterior thalamic nuclei to the cingulate cortex in the rat.
Slater, P. and Patel, S. (1983) Autoradiographic distribution of serotonin receptors in rat brain. European Journal
of Pharmacology, 92, 297–298.
Smart, I.H.M. and Smart, M. (1982) Growth Patterns in the lateral wall of the mouse telencephalon. I. Autoradio-
graphic studies of the histogenesis of the isocortex and adjacent areas. Journal of Anatomy, 134, 273–298.
Smiley, J.F., Levey, A.I., Ciliax, B.J. and Goldman-Rakic, P.S. (1994) D1 dopamine receptor immunoreactivity
in human and monkey cerebral cortex: Predominant and extrasynaptic localisation in dendrite spines.
Proceedings of the National Academy of Sciences, U.S.A., 91, 5720–5724.
Smith, L.M., Ebner, F.F. and Colonnier, M. (1980) The thalamocortical projection in pseudemys turtles: A qual-
itative electron microscopic study. Journal of Comparative Neurology, 190, 445–461.
Spencer, D.G., Horvath, E. and Traber, J. (1986) Direct autoradiographic determination of M1 and M2 muscar-
inic acetylcholine receptor distribution in the rat brain: relation to cholinergic nuclei and projections. Brain
Research, 380, 59–68.
Stephan, H. and Andy, O.J. (1970) The allocortex in primates. In: C.R. Noback and W. Montagna (eds), The
Primate Brain. New York: Appleton-Century-Crofts, pp. 109–136.
Stephan, H., Baron, G. and Schwertfeger, W.K. (1980) The brain of the common marmoset. A stereotaxic atlas.
Springer Verlag, Berlin, Heidelberg.

Steriade, M., Jones, E.G. and Llinas, R.R. (1990) Thalamic oscillations and signalling. New York: Wiley
Interscience.
Sugaya, K., Downen, M. and Giacobini, E. (1991) Nucleus basalis lesions decraease alpha- and kappa-bungaro-
toxins binding in rat cortex. NeuroReport, 2, 177–180.
Swanson, L.W., Simmons, D.M., Whiting, P.J. and Lindstrom, J. (1987) Immunohistochemical localization of
neuronal nicotinic receptors in the rodent central nervous system. Journal of Neuroscience, 7, 3334–3342.
Unnerstall, J.R., Kuhar, M.J., Niehoff, D.L. and Palacios, J.M. (1981) Benzodiazepine receptors are coupled to
a subpopulation of gamma-aminobutyric acid (GABA) receptors. evidence from a quantitative autoradio-
graphic study. Journal of Pharmacology and Experimental Therapapeutics, 218, 797–804.
Unnerstall, J.R., Niehoff, D.L., Kuhar, M.J. and Palacios, J.M. (1982) Quantiative autoradiography using
[3H]ultrofilm: application to multiple benzodiazepine receptors. Journal of Neuroscience Methods, 6, 59–73.
Unnerstall, J.R. and Wamsley, J.K. (1983) Autoradiographic localization of high-affinity [3H]kainic acid
binding sites in the rat forebrain. European Journal of Pharmacology, 86, 361–371.
Urban, I. and Richard, P. (1972) A stereotaxic atlas of the New Zealand rabbit’s brain. C.C. Thomas, Spring-
field, Illinois.
Van Groen, T. and Wyss, J.M. (1992) Projections from the laterodorsal nucleus of the thalamus to the limbic and
visual cortices in the rat. Journal of Comparative Neurology, 324, 427–448.
Verge, D., Daval, G., Marcinkiewicz, M., Patey, A., El Mestimawy, S., Gozlan, H. and Hamon, M. (1986)
Quantitative autoradiography of multiple 5-HT1 receptor subtypers in the brain of control or 5,7-dihydroxy-
tryptamine-treated rats. Journal of Neuroscience, 8, 3474–3482.
Vilaro, M.T., Wiederhold, K.-H., Palacios, J.M. and Mengod, G. (1991) Muscarinic cholinergic receptors in the
rat caudate-putamen and olfactory tubercle belong predominantly to the M4 class: in situ hybridization and
receptor autoradiography evidence. Neuroscience, 40, 159–167.
Vogt, B.A. and Burns, D.L. (1988) Experimental localization of muscarinic receptor subtypes to cingulate
cortical afferents and neurons. Journal of Neuroscience, 8, 643–652.
Vogt, B.A., Plager, M.D., Crino, P.B. and Bird, E.D. (1990) Laminar distribution of muscarinic acetylcholine,
serotonin, GABA and opioid receptors in human posterior cingulate cortex. Neuroscience, 36, 165–174.
Voneida, T.J. and Ebbesson, S.O.E. (1969) On the origin and distribution of axons in the pallial commissures in
the Tegu Lizard (Tupinambis nigropunctatus). Brain Behavior and Evolution, 2, 467–481.
Wada, E., Wada, K., Boulter, J., Deneris, E., Heinemann, S., Patrick, J. and Swanson, L.W. (1989) Distribution
of alpha2, alpha3, alpha4 and beta2 neuronal nicotinic receptor subunit mRNAs in the central nervous
system: A hybridization histochemical study in the rat. Journal of Comparative Neurology, 284, 314–335.
Wamsley, J.K., Hunt, M.E., McQuade, R.D. and Alburges, M.E. (1991) [3H]SCH39166, a D1 dopamine recep-
tor antagonist: binding characteristics and localization. Experimental Neurology, 111, 145–151.
Wang, J.-X., Roeske, W.R., Hawkins, K.N., Gehlert, D.R. and Yamamura, H.I. (1989) Quantitative autoradio-
graphy of M2 muscarinic receptors in the rat brain identified by using a selective radioligand [3H]AF-DX
116. Brain Research, 477, 322–326.
Weber, R.G., Jones, C.R., Palacios, J.M. and Lohse, M.J. (1988) Autoradiographic visualization of A1-adenosine
receptors in brain and peripheral tissues of rat and guinea pig using 125I-HPIA. Neuroscience Letters, 87,
215–220.
Weber, R.G., Jones, C.R., Lohse, M.J. and Palacios, J.M. (1990) Autoradiographic visualization of A1 adenosine
receptors in rat brain with [3H]8-cyclopentyl-1,3-dipropylxanthine. Journal of Neurochemistry, 54, 1344–1353.
Wouterlood, F.G., Saldana, E. and Witter, M.P. (1990) Projection from the nucleus reuniens thalami to the
hippocampal region: Light and electron microscopic tracing study in the rat with the anterograde tracer
phaseolus vulgaris-leucoagglutinin. Journal of Comparative Neurology, 296, 179–203.
Xia, Y. and Haddad, G.G. (1992) Ontogeny and distribution of GABAa receptors in rat brainstem and rostral
brain regions. Neuroscience, 49, 973–989.
Yanigahara, M., Niimi, K. and Ono, K. (1987) Thalamic projection to the hippocampal and entorhinal areas in
the cat. Journal of Comparative Neurology, 266, 122–141.
Yoshida, M., Sakai, M., Kani, K., Nagatsu, I. and Tanaka, M. (1988) The dopaminergic innervation as observed
by immunohistochemistry using anti-dopamine serum in the rat cerebral cortex. Experientia, 44, 700–702.
Zarbin, M.A., Innis, R.B., Wamsley, J.K., Snyder, S.H. and Kuhar, M.J. (1983) Autoradiographic localization of
cholecystokinin receptors in rodent brain. Journal of Neuroscience, 3, 877–906.

Part IV
FUNCTIONAL EQUIVALENCE
BETWEEN AREAS

12 Cross-Modal Plasticity as a Tool for Understanding
the Ontogeny and Phylogeny of Cerebral Cortex
Sarah L. Pallas
Department of Biology, Georgia State University, P.O. Box 4010, Atlanta, GA 30302
Correspondence: S.L. Pallas, Tel: 404-651-1551; FAX: 404-651-2509;
e-mail: spallas@GSU.edu
There has been a continuing debate on the relative roles of nature and nurture in specifying cortical
areal identity, and it shows no signs of abatement. The cross-modal plasticity paradigm provides
a unique and direct way to address the controversy, and has provided important insights. This
approach has shown that diverting visual afferents to auditory thalamus from birth results in an audi-
tory cortex with reorganized circuitry, with visual responses typical of those seen in visual cortex,
and with the ability to mediate visual behaviour. This work provides strong evidence that sensory
afferents have the ability to organize cortical circuits for the purpose of their own optimum process-
ing. The ability of sensory inputs to drive the organization of cortical circuitry may also facilitate
recovery from perinatal brain damage, and compensation by sensory substitution in deaf or blind
individuals. Furthermore, the cross-modal approach provides insights into how new cortical areas
may have arisen, or how changes in peripheral sensory systems could have been accommodated by
existing cortical areas during mammalian evolution.
KEYWORDS: areal specification, auditory cortex, cortical development, cross-modal plasticity,

ferret, sensory cortex, visual cortex
1. INTRODUCTION
Philosopher-scientists as far back as Galen (see Changeux, 1986, p. 8) understood that one
of the most fundamental organizing principles of the mammalian neocortex is its parcella-
tion into anatomically and functionally distinct areas. Neocortical areas are similar in many
ways, including types of cellular constituents, the presence of six cell layers, and the laminar
organization of inputs and outputs. Neocortical areas differ from one another as well,
along several structural and functional dimensions, including connectivity, topography,
and response properties (Seitz, this volume). During evolution, neocortical parcellation
produces new areas where they did not previously exist. Although major advances in
understanding the brain have certainly been made in the last two centuries, it remains
unclear what factors drive the appearance of distinct neocortical areas during either develop-
ment or evolution (Pallas, 2001). Why has this issue remained so refractory to resolution?
With regard to evolutionary mechanisms, descriptive studies of fossil brain-cases
(Jerison, 1990) and comparative studies of the brains of extant species have provided a
great deal of insight into the evolution of new cortical areas. One reasonable future goal
for work in this field would be to develop an experimental system in which a novel
245
246 Sarah L. Pallas
cortical area can be induced in the laboratory on a more manageable time scale than that
allowed by evolutionary studies. Developmental approaches may allow this feat to be
accomplished.
The areal specification process is evident during development, in that areas are progres-
sively carved out of a fairly uniform sheet of embryonic precursor cells. New cortical
areas appear in temporal sequence during development (Luskin and Shatz, 1985; Bayer
and Altman, 1991) as they do spatially during evolution. However, unlike evolving
systems, developing systems have the advantage that they can be manipulated experimen-
tally. Evolution can operate through developmental mechanisms, with development
providing substrates for change and setting constraints on the degree of change over time.
The “evo-devo” approach of using developmental studies to inform evolutionary studies,
and vice versa, has been very fruitful in this regard (Goodman and Coughlin, 2000). Thus,
understanding how areal specification occurs during development may provide consider-
able insight into how new areas arose from simpler ancestral brains. In this review I will
argue that the cross-modal plasticity paradigm (Schneider, 1973; Frost, 1981; Sur et al.,
1988) is a uniquely valuable approach to this area of research.
2. NATURE OR NURTURE?
Investigators studying cortical development have more or less divided themselves into
two camps which fall along the classic nature-nurture fault line. That is, regionalization of
the cortical sheet could be due to intrinsic, preprogrammed differences (nature), or to
extrinsic factors such as thalamic inputs or the sensory information which they carry from
the periphery (nurture). Such dichotomous views have definite heuristic value, but they
tend to outlive their usefulness. We are now in a position to evaluate and synthesize the
contributions and shortcomings of both views. The idea that intrinsic and extrinsic pattern-
ing information act synergistically to specify an area is now gaining sway. Clearly, early
events prior to the formation of connections between the brain and the sensory organs are
necessarily controlled by intrinsic factors, but later events may be directed by extrinsic
factors, intrinsic factors, or both. Intrinsic and extrinsic factors are likely to interact in as
yet unknown ways in late development. Although activity-dependent processes have clas-
sically been thought of as being driven extrinsically, it is becoming clear that most brain
regions exhibit spontaneous patterned activity (Wong et al., 1993; Gu and Spitzer, 1997;
O’Donovan, 1999; Weliky and Katz, 1999; see Katz and Shatz, 1996, for review), and thus
both extrinsic and intrinsic control can be exerted electrically through neuronal activity.
Furthermore, specification events that are initially under intrinsic control may be reversible
at later stages by extrinsic information such as patterned sensory input (see below).
2.1. Source of the Debate

The swings in the number of adherents to the extrinsic- vs the intrinsic-control hypo-
theses of cortical regionalization have seemed almost comparable in amplitude to nature-
driven versus nurture-driven trends in child rearing practices (Bruer, 1999; Gopnik et al.,
1999). The most recent debate perhaps began with a review article by Pasko Rakic (1988)
in which he proposed that a fate map, or “protomap” of the cortical sheet exists in the
nascent ventricular zone. Subsequent reviews by Dennis O’Leary (1989) and my colleagues

Cross-Modal Plasticity Studies of Cortical Development 247
and I (Pallas, 1990; Sur et al., 1990) argued for what O’Leary called the “protocortex”
hypothesis—that the type of thalamic input or sensory activity that any given cortical
region receives is a primary determinant of its identity. Subsequent reviews have oscil-
lated primarily around these two views (Shatz, 1992; Walsh, 1993; Levitt et al., 1997;
Butler and Molnar, 1998), although a refreshingly synergistic review has appeared recently
(Kingsbury and Finlay, 2001).
2.2. What do we mean by Cortical Identity?

Areal specification, or parcellation, is a process that may be distinct from what some have
referred to as regionalization—the process whereby large regions of cortex come to have
different features from other regions, usually with respect to gene expression patterns (e.g.
Pimenta et al., 1996; Bishop et al., 2000). When we refer to specification of cortical
identity, do we mean regionalization or parcellation, and at what point do we consider that
an identity has been attained irreversibly?
One confound has been in viewing cortical identity as a unitary concept (Levitt et al.,
1997). There are undoubtedly numerous factors contributing to the establishment of the
multitudinous characteristics that make up each unique cortical area. Another issue is that
what is defined as an extrinsic or intrinsic factor has changed over time. Extrinsic can be
(and has been) defined relative to the cortex, or relative to the animal, and it is important
for investigators to be clear about their frame of reference. Events that occur prior to
ingrowth of thalamocortical afferents are necessarily independent of sensory experience,
but the importance of early, patterned, spontaneous activity is becoming apparent (Wong
et al., 1993; Ruthazer and Stryker, 1996; Kandler and Katz, 1998; Cook et al., 1999; Crair,
1999), and, depending on its source, could be considered an intrinsic or extrinsic factor.
These early events are thus under intrinsic control, whereas those that occur later could be
under extrinsic control, intrinsic control, or both. There could be, and certainly are, cortical
specification events that are intrinsically specified initially but that are later altered by
experience.
If agreement can be reached on the definition of cortical identity, the remaining
problem would be to determine the specific and differential roles of intrinsic and extrinsic
factors, and their reversibility, for each specification event. To accomplish this, the
research must be conducted in a species in which the thalamocortical pathway is formed
postnatally, or in which the contribution of prenatal, patterned spontaneous activity in
thalamocortical afferents can be defined.
3. EVIDENCE FAVOURING INTRINSIC OR EXTRINSIC SPECIFICATION
3.1. Evidence for Specification by Intrinsic Factors

To provide the most convincing evidence that cortical parcellation is controlled by
intrinsic factors, it would be desirable to identify different molecular markers or labels in
each cortical area, prior to the ingrowth of axons carrying information from extrinsic
sources. Several molecular markers are found only in certain cortical regions (e.g. Arimatsu
et al., 1992; Cohen-Tannoudji et al., 1994; Paysan et al., 1994, 1997; Pimenta et al., 1996;
Reinoso et al., 1996; Nakagawa et al., 1999; Rubenstein et al., 1999), and in some cases

248 Sarah L. Pallas
these markers are expressed at a time when they would be in a position to direct regional
fate independently of extrinsic information. However, few of these markers have been
found to be restricted to the confines of a single functionally defined cortical area (but see
Cohen-Tannoudji et al., 1994; Nothias et al., 1998).
Areal restrictions on markers are not a necessary condition for specification however,
because sharp boundaries between adjacent cortical areas could be achieved through dif-
ferential, threshold-type responses of the tissue to a gradient of a morphogen, or to several
nested morphogens, as occurs elsewhere along the neuraxis (Lumsden and Krumlauf,
1996; Tanabe and Jessell, 1996). Recently, several factors have been found in the forebrain
which are expressed in gradients, boundary zones, or compartments (Donoghue and Rakic,
1999; Rubenstein et al., 1999), including Emx genes ( Nakagawa et al., 1999; Bishop et al.,
2000), Otx2 (Nothias et al., 1998), ephrins (Mackarehtschian et al., 1999; Miyashita-Lin
et al., 1999), Pax6 (Bishop et al., 2000), cadherins (Nakagawa et al., 1999), and several
classes of transcription factors (Miyashita-Lin et al., 1999; Nakagawa et al., 1999). In
addition to the descriptive studies mentioned here, two recent papers using an experi-
mental approach (Bishop et al., 2000; Mallamaci et al., 2000) strongly support the idea
that morphogenic gradients can specify the cortex. They argue that knockout of the regula-
tory genes Emx2 or Pax6 result, respectively, in a respecification and/or prenatal loss of
the caudal or rostral portions of the cortical epithelium where the respective genes are
normally highly expressed, and in a disproportionate expansion of opposite areas of
cortex, where the genes exhibit low expression. The results suggest that the genes are con-
ferring regional identity on portions of cortex. Whether such a regional identity translates
into an identity for discrete cortical areas remains an open question.
Knockout experiments suggest that at least some of these marker patterns can be esta-
bished independent of thalamocortical input (Miyashita-Lin et al., 1999; Nakagawa et al.,
1999; Bishop et al., 2000), but some apparently require thalamic input for upregulation
and maintenance of the marker (Nothias et al., 1998; Gitton et al., 1999). Unfortunately,
many knockout mice die perinatally (Guillemot et al., 1993; Miyashita-Lin et al., 1999;
Bishop et al., 2000), preventing determination of the role of intrinsic factors in specifica-
tion of cortical features which appear postnatally, such as cytoarchitecture, local connec-
tivity, and circuit formation—the very features most typically thought of as defining
cortical identity. Hopefully the use of conditional knockouts will provide more conclusive
information in the near future.
3.2. Evidence for Specification by Extrinsic Factors

For the purposes of this section, “extrinsic” refers to information carried by thalamocor-
tical afferents, whether or not that information is based on patterns of electrical activity,
and whether or not activity is driven by sensory experience. Given that intrinsic molecular
markers are capable of establishing at least some aspects of regional and possibly areal
identity of the neocortex on their own (Miyashita-Lin et al., 1999), what is the role of
extrinsic information in areal specification? There is substantial evidence that certain areal
features can be altered by manipulating extrinsic information, such as sensory experience.
It is not clear whether this means that an initial intrinsically-specified areal identity is
altered by manipulations of thalamic input, whether the thalamic input establishes the
original identity, or whether only certain aspects of cortical identity are affected, leaving
earlier-established aspects intact. In at least one case, expression of a molecular marker

seems to be triggered by thalamic input (Levitt et al., 1997). However a causal relation-
ship between thalamocortical innervation, marker expression, and areal specification is
difficult to establish.
3.2.1. Specificity of thalamocortical targeting

It is clear that there are detailed recognition mechanisms guiding thalamic axons to the
proper cortical region (Ghosh et al., 1990; Suzuki et al., 1997; Castellani et al., 1998;
Inoue et al., 1998; Mann et al., 1998), because they, unlike most other central sensory
pathways, exhibit considerable projection specificity and tend not to make ectopic projec-
tions even if other cortical areas are deafferented and thus available as targets (Crandall
and Caviness, 1984; Miller et al., 1991). This projection specificity could allow the
thalamocortical pathway to act as a framework for neocortical parcellation during devel-
opment or evolution. Whether thalamocortical axons are instructed by or are instructing
areal identity is unknown, however.
3.2.2. Thalamocortical activity plays an important role

Although it is possible that the primary role of thalamus in pathfinding is to carry patterning
information to its cortical targets independent of activity, it appears that neural activity is
required to obtain correct thalamocortical targeting (Catalano and Shatz, 1998), pointing
out the necessity for both activity-independent and activity-dependent processes. How-
ever, thalamocortical activity-dependent processes are essential for normal development.
Coordinated, spontaneous activity in the input pathway prior to sensory organ function
(Wong et al., 1993; Weliky and Katz, 1999) can drive the formation of crude circuits, but
these are later refined by external sensory cues (see Katz and Shatz, 1996, for review;
Ruthazer and Stryker, 1996; Sengpiel et al., 1998). The potential ability of thalamic axons
to specify cortex could thus depend on patterning information at their source, on matching
markers in the axons and their corresponding target, on their activity pattern, or a com-
bination of factors.
In practice, it has been difficult to demonstrate a definitive role for patterned sensory
activity in neocortical parcellation. Previous approaches have employed primarily sensory
deprivation or heterotopic transplantation. Sensory deprivation can markedly affect the
cytoarchitecture of visual cortex (Dehay et al., 1991; Rakic et al., 1991), and heterotopic
transplantation of one embryonic cortical region into the place of another can cause the
donor tissue to develop host-specific cytoarchitectonic and subcortical connectivity
patterns (O’Leary and Stanfield, 1989; Schlaggar and O’Leary, 1991). Although thalamic
afferents may be the causal factor in the cortical changes observed in the deprivation and
transplantation paradigms, there are alternative interpretations: the bilateral enucleation
paradigm causes massively increased cell death in striate cortex (Rakic, 1988), and this
alone would cause marked changes in cytoarchitecture, especially that of layer 4 (Finlay
and Pallas, 1989). The cortical transplantation results are also difficult to interpret:
although the switch from donor to host characteristics could well result from the change in
the source of thalamic input, there are many other potential organizing factors that change
with a change in location, such as corticocortical inputs, developmental timing, and distri-
bution of morphogens, to name a few. Any or all of these could contribute to the altered
appearance of the donor tissue, and thus all that can be rigorously concluded is that the

250 Sarah L. Pallas
presence of different thalamocortical afferents is correlated with the appearance of host-

specific features in the donor tissue. Whether there remain some donor-specific
characteristics in the transplant that has not been thoroughly investigated. Experience may
alter only later-developing characteristics of cortex, but then it should be asked at what
stage one can consider that cortical “identity” has been respecified by extrinsic factors?
At some point the argument becomes a semantic one, although one could argue sensibly
that early choices made by molecular markers bias later ones made by extrinsic factors.
The importance of extrinsic factors in cortical areal specification would be strongly
supported by a demonstration that they could specify cortex independently, or could
respecify cortex that had previously been specified by intrinsic factors. An insightful
experiment in the marsupial Monodelphis domestica, in which large portions of caudal
neocortex were ablated prior to thalamocortical invasion, showed that the remaining
cortical sheet contained the same cortical areas present in normal animals (Huffman et al.,
1999). The cortical areas were arranged in the same spatial pattern with respect to each
other on this smaller cortical surface, and were innervated by the same thalamic nuclei as
usual, although the thalamocortical axons had to shift their projection paths significantly
to reach the shifted location of their target. These results may mean that the cortical tissue
can compress a predetermined “fate map”, and/or that areal fate can be specified or
respecified by the thalamocortical innervation. This result is in some respects reminiscent
of the results from Emx2 and Pax6 knockout mice (Bishop et al., 2000; Mallamacci et al.,
2000). A similar result has been obtained in the retinocollicular system of hamsters, where
it has been observed that a complete but compressed visual map forms in superior colliculi
partially-ablated postnatally (Finlay et al., 1979; Pallas and Finlay, 1989). Our recent
evidence shows that NMDA receptor blockade does not interfere with compression of
retinocollicular maps, although it does block conservation of single-cell receptive field
properties (Huang and Pallas, 1999, 2000, 2001). These results raise the possibility that a
prenatally-determined molecular “fate map” can be altered postnatally, and/or that fate can
be specified or respecified by afferent innervation. In these cases, as in those mentioned
above, it is difficult to distinguish the relative roles of intrinsic and extrinsic factors in the
compression, although they are consistent with a critical role for thalamocortical afferent
inputs in areal specification.
3.2.3. The cross-modal plasticity paradigm

Our work on cross-modal plasticity in ferrets, and related work in hamsters, provides
another approach with important advantages for investigating the role of thalamocortical
afferents in areal specification. The cross-modal “rewiring” of retinal axons into the audi-
tory thalamus (medial geniculate nucleus or MGN) provides patterned spontaneous and
visually-driven activity from the retina to primary auditory cortex during its development.
The manipulation is performed prior to thalamocortical ingrowth, without manipulating
the thalamocortical projection. With this approach, one can address whether patterned
activity can specify neocortex independent of the source and identity of the thalamocortical
axons, and so can gain insight into whether any existing morphogens can be overridden.
Evolution has provided a compelling illustration of the thalamic influence on areal
specification. Mole rats (Spalax ehrenberghi) are burrowing animals whose fossorial habit
is associated with a reduction in the visual pathway, and a dependence on communication
using seismic signals (Rado et al., 1989). This species is blind to form and motion due to

the vestigial form of their eye (which are subcutaneous), although they can detect light for
the entrainment of circadian rhythms (Cooper et al., 1993). Visual responses cannot be
detected in their geniculocortical pathway, but instead the lateral geniculate nucleus
(LGN) and visual cortex respond to auditory stimulation (Bronchti et al., 1989, 1991; Heil
et al., 1991). Anatomical studies of these animals have revealed that the “unused” LGN is
invaded by the auditory modality via the inferior colliculus (Doron and Wollberg, 1994),
although the visual cortex retains normal cytoarchitectural characteristics (Wollberg,
personal communication). The behavioural role of the auditory-to-visual pathway is not yet
known.
The visual-to-auditory cross-modal rewiring procedure is essentially a developmental
situation which is the inverse of the mole rat system. The advantage of a developmental
approach is that one can view not just the end point but the entire process of parcellation
on a manageable time scale, and manipulate it experimentally. In the discussion below
I will summarize the results of the cross-modal plasticity work and its clinical implications
as well as the insights it provides into the development and evolution of neocortex.
4. HOW DIFFERENT ARE DIFFERENT SENSORY CORTICAL AREAS?
Primary sensory cortical areas, such as primary visual (V1), somatosensory (S1), and audi-
tory (A1) cortices, have several characteristics in common, as addressed by others in this
volume. The presence of 6 layers and a similar complement of cell types and laminar pattern
of connectivity are examples. Visual and somatosensory cortex are similar in many ways
due to their role in representing a two-dimensional spatial surface (viz, the retina and the
skin surface). However, the auditory cortex is arranged in a quite different way as a result of
representing the cochleae, which in itself is ignorant of the spatial location of stimuli.
4.1. The Topographic Map in A1 is Different from the Map in V1 or S1

The visual cortex maps the retina in two dimensions, azimuth and elevation (Sherman and
Spear, 1982), and somatosensory cortex maps the two-dimensional skin surface (Mount-
castle, 1957). In contrast, the auditory cortex maps the cochlea in one dimension, that of
frequency; the orthogonal axis contains an isofrequency representation (Merzenich and
Knight, 1975; Aitkin et al., 1984). Although other stimulus features are mapped in AI
(Schreiner and Mendelson, 1990; Schreiner et al., 1992; Mendelson et al., 1993; Langner
et al., 1997; Recanzone et al., 1999), many are the result of a computational mapping and
are thus different in character from the tonotopic map, which is a direct reflection of the
sensory epithelium.
4.2. Connectivity Patterns in A1 and V1 are Clustered along Different Dimensions

Beyond the fact that different cortices make connections with different brain structures,
there are basic differences in the organization of connections. Connections throughout the
auditory pathway reflect its one-dimensional tonotopic organization, so that thalamocor-
tical (Andersen et al., 1980; Morel and Imig, 1987; McMullen and de Venecia, 1993),
callosal (Imig and Brugge, 1978) and horizontal intracortical connections (Matsubara and
Phillips, 1988; Wallace and Bajwa, 1991; Gao and Pallas, 1999) form slabs or strips which

252 Sarah L. Pallas
interconnect neurons with similar frequency tuning or binaural organization across the
tonotopic map in A1. In contrast, connectivity patterns in V1 are point-to-point (Hubel
and Wiesel, 1962; Wiesel and Hubel, 1963), reflecting the two-dimensional organization
of the visual pathway (Innocenti, 1986; Callaway and Katz, 1990; Ruthazer and Stryker,
1994). Somatosensory cortex has widespread, radially-arranged clusters of intrinsic pro-
jections (Juliano et al., 1996; Sonty and Juliano, 1997) similar to that seen in visual cortex
and distinct from the auditory pattern. These specific connections are created from an
initially diffuse pattern under the influence of spontaneous cortical activity and sensory
cues (Feng and Brugge, 1983; Callaway and Katz, 1990, 1991; Ruthazer and Stryker,
1996).
4.3 There are Similarities and Differences in Receptive Field Properties

between V1, S1, and A1
Primary visual cortex (Hubel and Wiesel, 1962), primary somatosensory cortex
(Hyvärinen and Poränen, 1978; Essick and Whitsel, 1985), and primary auditory cortex
(Mendelson and Cynader, 1985; de Charms et al., 1998; Nelken and Versnel, 2000)
contain cells which are tuned to the direction and velocity of movement of a stimulus
across the sensory epithelium (movement of touch or light across the receptive field,
frequency modulated sweeps across the cochlea), and all contain cells which can be
inhibited if the stimulus activates more than a restricted portion of the sensory organ
(end-stopped cells in visual cortex, two-tone inhibition or sharpened frequency tuning in
auditory cortex). However, although both visual cortex (Hubel and Wiesel, 1962) and
somatosensory cortex (Phillips and Johnson, 1981) contain cells tuned to the orientation of
an elongated stimulus, and visual cortex contains simple and complex cells, whose
responses depend on the two-dimensional substructure of the receptive field, the auditory
cortex is necessarily lacking in such two-dimensional receptive field properties because its
sensory epithelium is not two-dimensional, nor is the thalamocortical projection between
MGN and primary auditory cortex. The presence of such response properties in a cross-
modal A1 would suggest that a visually-driven change in circuitry had occurred.
5. CROSS-MODAL PLASTICITY RESULTS IN STRUCTURAL

AND FUNCTIONAL ALTERATION OF SENSORY CORTEX
In cross-modal plasticity experiments, afferents of one sensory modality are induced to

innervate the thalamic nucleus of a different modality, which in turn carries the cross-
modal information to the sensory cortex (Figure 12.1). The affected cortical area then
responds to the new modality of sensory stimulation. These experiments were first per-
formed by Schneider (1973) and then Frost (1981, 1999) in hamsters. Combining removal
of visual targets with deafferentation of the somatosensory pathway in hamsters induces
retinal axons to project to nucleus VB, the somatosensory division of thalamus, whereas
combined visual target ablation and auditory system deafferentation produces retina-
to-MGN projections in hamsters (Frost, 1981). We have utilized the procedure in ferrets,
which are carnivores with sensory physiology similar to that of the cat, except that they
are born at an earlier stage of brain development, prior to ingrowth of thalamic afferents
and migration of layer 4 (Luskin and Shatz, 1985; Jackson et al., 1989). The lesions are

made before sensory information has gained access to cortex. Patterned visual input is not
available to ferrets until their eyes open and the optics clear, at one month of age. Their
auditory pathway becomes functional at the same age (Moore and Hine, 1992). As a con-
trol, we also included in each experiment a group of ferrets deafened by cochlear ablation
at two weeks of age. With normal ferrets as an additional control, we compared the
structure and function of A1 with auditory inputs, A1 with visual inputs, and A1 with no
auditory input.
The cross-modal approach represents a unique and powerful way to investigate cortical
parcellation. Unlike experiments which manipulate the number of inputs or the overall
amount of activity reaching the cortex, the cross-modal rewiring procedure addresses the
role of the spatiotemporal activity pattern of the inputs in parcellation. We can thus inves-
tigate whether thalamic inputs have a permissive or instructive role in cortical differenti-
ation, in other words, whether thalamic activity simply allows differentiation to proceed
by triggering other factors, or whether the pattern of activity guides cortical differentiation
in a particular direction (Crair, 1999). Unlike transplant experiments, we can change the
activity pattern without changing the local environment of the cortical cells. Our results
suggest that afferent input can organize cortical circuitry in a directed fashion that allows
recovery of function.
5.1. Anatomy
5.1.1. Retinofugal projections
The cross-modal “rewiring” in ferrets is done on postnatal day (P)1. Unilateral superior
colliculus (SC) ablation removes a main target of the retina, whereas inferior colliculus
(IC) ablation and severing of the brachium of the inferior colliculus deafferents MGN
(Figure 12.1). As a result, W-type retinal ganglion cells from both eyes invade MGN
(Pallas et al., 1994b; Pallas and Sur, 1994; Angelucci et al., 1996) and form eye-specific
clusters there (Roe et al., 1993; Angelucci et al., 1997). In ferrets, W-cells are found in SC
and the C-laminae of the LGN, and these project in turn to extrastriate cortex but not V1.
Thus one might not expect to see V1-type responses in the cross-modal A1. However, the
differences in response properties of X, Y and W cells are minor with respect to the
Retina Retina
LGN LGN
A1
A1
MGN SC MGN
SC
V1 IC IC V1
NORMAL X-MODAL
Figure 12.1. Neonatal cross-modal surgery. Neonatal lesion of midbrain SC and IC converts the normal
retinal projection (Left) to the cross-modal projection (Right).

254 Sarah L. Pallas
parameters we investigated (Sherman and Spear, 1982; Mulliken et al., 1984; Ferster,
1992).
In hamsters, all three ganglion cell types invade the MGN (Métin et al., 1995; Bhide
and Frost, 1999). Also, although there is an early exuberant projection from the retina to
the MGN in hamsters, the retino-MGN axons present in the adult come from fibres which
sprout in reaction to the lesions, and not from the exuberant projection.
In young normal animals, we found a few retino-MGN axons at P20, 30 and 40. This
was also true of early-deafened control animals. At P60 and in adults, however, retinal
axons were only found in the MGN of the early-deafened and the cross-modal animals
(Pallas and Moore, 1997; and in preparation). No retinal axons were found in the MGN of
57 normal adult control animals. We suggest that the anomalous projections are lost
during development in normal adult ferrets but are stabilized and expanded in cross-modal
and early-deafened adult ferrets.
Our observations in the deafened animals show a remarkable potential for plasticity in
retinal axon growth. Although the cochlea is several synapses away from the MGN, and
normal retinal targets were intact, retinal growth cones can detect and react to a lack of
auditory input to the MGN. However, the number of retino-MGN axons in deaf ferrets is
at least two orders of magnitude less than that seen in cross-modal ferrets, providing strong
support for the idea that the cross-modal surgery causes massive sprouting of retinal axons
into MGN. The data further support the theory that sensory inputs can take advantage of
available target space in the thalamus, and by inference in the cortex as well (see below),
providing a developmental substrate for accomodation of neocortical expansion during
evolution (Finlay and Darlington, 1995; Finlay et al., 1998; Darlington et al., 1999).
5.1.2. Thalamocortical projections

In normal animals, the thalamocortical projection from the visual thalamic nucleus LGN
projects in a point-to-point fashion to V1, maintaining the two-dimensional topography of
the retina, whereas MGN projects in a slab-to-slab fashion to A1, maintaining the one-
dimensional topography of the cochlea (Figure 12.2A, 12.2B). Although the rewiring sur-
gery reroutes retinofugal projections into MGN, the thalamocortical projection from MGN
projects as it normally does to A1 even though it is carrying visual information. That is, it
remains a one-dimensional projection (Pallas et al., 1990). It is thus hard to imagine how
the two-dimensional retinal topography could be transferred to A1 from the cross-modal
projection via MGN. Thus one would not expect to find a retinotopic map or receptive
field properties such as orientation tuning or simple and complex cells that depend on a
two-dimensional input in A1.
5.2. Physiology
5.2.1. Topography
Surprisingly, the cross-modal A1 does contain a two-dimensional map of visual space,
with visual azimuth represented on the tonotopic axis and visual elevation represented in
place of the isofrequency axis (Roe et al., 1990). This is true despite the lack of change in
the form of the thalamocortical pathway mentioned above (Pallas et al., 1990), suggesting
that the visual map present in the retina and the MGN must be recreated along the

Retina LGN Area 17
Cochlea MGN A1
C
Retina MGN A1
Figure 12.2. Schematic of thalamocortical projections in the (A) normal visual system, (B) normal auditory
system, and (C) cross-modal auditory system. Note the roughly point-to-point thalamocortical projection in the
visual system, in contrast to the highly overlapped slab-to-slab projection in the auditory system. In the cross-
modal A1, there is a map of visual space in MGN but the thalamocortical projection remains overlapped. We
hypothesize that the visual map is recreated in A1 by lateral inhibition within isofrequency laminae as shown.
isofrequency axis. This result is made even more surprising by the observation that a 2-D
retinotopic map exists in the MGN in the cross-modal ferrets (Roe et al., 1991). Thus the
visual map is imposed on the MGN by the manipulation, then one axis of the map
(the elevational axis) is discarded as the information travels to A1 in a thalamocortical
projection that is highly overlapped along the isofrequency axis (Pallas et al., 1990). In
primary auditory cortex, the second dimension of the visual map is recreated, presumably
by circuitry intrinsic to cortex (Figure 12.2C). Interestingly, the elevational axis is less
orderly than the azimuthal axis, and is inconsistent in its polarity, as one might expect for
a dynamically-created array as opposed to one that arises directly from thalamocortical
connectivity. We propose that the second dimension of the map is based on correlated
patterns of retinal activity which stabilize neighbouring neurones through Hebbian mecha-
nisms (Hebb, 1949; Katz and Shatz, 1996).
Cross-modal rewiring of retinal axons to VB allows interaction of two sensory struc-
tures whose topography is based on two-dimensional spatial information. Nonetheless, the
visual map that is made in S1 is one-dimensional from a surface view, mapping from
lower to upper visual field, excluding the nasotemporal axis (Metin and Frost, 1989). The
second dimension was not observed from surface mapping, but contributes in some way to
stimulus processing.

256 Sarah L. Pallas
5.2.2. Receptive field properties

In cross-modal hamsters, it was found that the somatosensory cortex could process visual
input. Direction- and orientation-tuned neurones were also found in proportions and tuning
strength similar to the normal S1 (Metin and Frost, 1989). Interestingly, some multiunit
recording sites showed both visual and somatosensory responsiveness.
Single-unit extracellular recordings from cross-modal ferrets (Roe et al., 1992) revealed
that neurones in AI responded sluggishly to visual stimulation and had large and diffuse
receptive fields, as might be expected for cells with W-type input. Lateral geniculate and
visual cortical neurones receiving input from W-cells are described as sluggish, with large
receptive fields, lower spatial frequency tuning and tuning to slower velocity of movement
than X or Y cells (Sherman and Spear, 1982; Mulliken et al., 1984; Vitek et al., 1985;
Price and Morgan, 1987; Roe et al., 1990; Ferster, 1992; Baker et al., 1998). However,
with respect to stimulus selectivity, cross-modal A1 neurones responded in much the same
way as normal visual cortical neurones (Roe et al., 1992; Law et al., 1988), suggesting
that these properties are organized primarily intracortically and not via input type. As
might have been expected based on pre-existing properties in normal A1 (see above),
many neurones were tuned to the direction of movement of a light stimulus, and were
inhibited by long edges. However, we also saw response properties that were never found
in normal A1. Specifically, neurones could be classified as simple or complex, based on
the two-dimensional substructure of their receptive fields. Many neurones were found to
be tuned to the orientation of a bar of light. These cells were not rare but rather were found
in approximately the same proportions as in normal ferret V1, and in both cortices all
orientation vectors were represented in equivalent distributions (this was also true for the
distribution of direction tuning vectors). Although this result in itself is quite surprising,
we also found that the strength of tuning to orientation in A1 was comparable to that in V1.
One possible explanation for the presence of visual responses in cross-modal AI would
be potential rerouting of visual inputs to AI as a byproduct of the rewiring surgery, which
could then confer their response properties on the cross-modal auditory structures. To
address this, we examined thalamocortical and corticocortical connections of AI in normal
and cross-modal ferrets (Pallas et al., 1990; Pallas and Sur, 1993). With the exception of
the visual inputs arising from the induced retino-MGN projection, AI did not have mono-
synaptic connections with visual thalamus, except for a few cells in the multimodal region
of the lateral posterior thalamus. Furthermore, AI did not send to or receive connections
from visual cortical areas, but rather was interconnected only with other auditory cortical
regions as in normal animals. Thus the only pathway likely to account for the visual
responses in cross-modal AI is the pathway induced from retina to MGN to AI. This inter-
pretation was further supported by our behavioural experiments (see below).
5.3. Two Alternative Hypotheses

These results demonstrated that some of the transformations in stimulus representation
that normally occur in V1 can also occur in AI of rewired ferrets, and they show that it is
not the case that AI and V1 are different by their inherent nature. Further experiments
were needed in order to determine whether the early visual inputs drive changes in A1’s
circuitry which allow it to process visual inputs de novo, or whether visual and auditory
cortex are fundamentally similar enough that they can process each other’s inputs without

alterations in circuitry. For example, perhaps the job of sensory cortex or cortex in general
is simply to enhance contrast in whatever input is received, and the lateral inhibitory
circuits, ubiquitous to sensory pathways, could accomplish this goal regardless of the type
of input, and without any change in cortical circuitry. In this case the output of the normal
auditory cortex would be different from the output of normal visual cortex only because
the input is different and not due to any difference in input processing. Providing the same
input would thus result in the same output.
It should be noted that the two hypotheses are not mutually exclusive; that is, some
aspects of cortical processing are likely to be held in common between the two cortical
areas, and other aspects are likely to show differences. The presence of a two-dimensional
visual map and orientation tuning certainly suggested that cortical circuitry was altered by
the visual inputs. Our more recent data provides more definitive evidence in this regard,
and support the hypothesis that sensory inputs can direct the construction of cortical
circuits for their own purposes.
5.4. Functional Connectivity

Decades of study of the functional anatomy of sensory cortex have revealed an underlying
modular organization. In the normal visual cortex, a number of different modules have
been described, such as ocular dominance columns, blobs of colour-sensitive cells, stripes
representing form, colour and motion pathways, and pinwheels representing orientation
tuning (LeVay et al., 1978; Livingstone and Hubel, 1988; Bartfeld and Grinvald, 1992;
Daw, 1998). In auditory cortex, well-described modules include isofrequency slabs (Reale
and Imig, 1980; Kelly et al., 1986) and binaural bands, which are arranged orthogonal to
the isofrequency axis (Middlebrooks et al., 1980; Kelly and Judge, 1994). The intercor-
tical and intracortical connectivity patterns of sensory cortices reflect their modular organ-
ization. For example, neurones which are matched in orientation tuning are interconnected
and form clusters of similarly-tuned neurones across the retinotopic map. In the auditory
cortex, neurones which are matched in frequency tuning and/or in binaural characteristics
are interconnected (Imig et al., 1986). We reasoned that in order that AI can support visual
function, the modular connectivity patterns in A1 would have to be reorganized to the
“visual pattern”, reflecting the retinotopic organization of the visual system and not the
tonotopic organization of the auditory system. We thus undertook a study of the callosal
and horizontal connectivity patterns in normal animals (auditory input to auditory cortex),
cross-modal animals (visual input to auditory cortex), and deafened animals (deafferentation
of auditory cortex).
5.4.1. Callosal connectivity

Callosal connections unite functionally and/or topographically-related cells in the two
cortical hemispheres, and are specific and different in each sensory cortical area. In rat and
cat visual cortex, callosal neurones connect areas representing the vertical meridian of the
visual field (Innocenti, 1986; Lewis and Olavarria, 1995; Olavarria and Van Sluyters,
1995; Bourdet et al., 1996). In the auditory cortex, neurones in the E-E binaural bands
(strips of neurones facilitated by binaural input) project across the callosum, but the neu-
rones inhibited by binaural stimulation (EI suppression cells) project primarily ipsilaterally
(Feng and Brugge, 1983). Callosal projections are exuberant early in postnatal development

258 Sarah L. Pallas
A B C
Figure 12.3. Tangential pattern of retrograde and anterograde callosal label resulting from HRP injections
throughout the contralateral A1. (A) Normal animal. (B) Cross-modal animal. (C) Deaf animal. Scale
bar = 2.5 mm.
(Innocenti et al., 1977; Feng and Brugge, 1983), and in visual cortex become progres-
sively more restricted under the influence of sensory experience (Innocenti and Frost,
1979). To examine the effect of the early visual inputs on connections between AI and its
contralateral homolog, we injected a tracer (wheat germ-agglutinated HRP, which travels
in both retrograde and anterograde directions) throughout A1 on one side of the brain (the
left side in normal and deafened ferrets, the unlesioned side in cross-modal ferrets) (Pallas
et al., 1999). We found that callosal connectivity in normal ferret AI is much like that in
cats, with 2–3 bands of neurones with callosal projections oriented along the tonotopic
axis (Figure 12.3A). In the cross-modal AI, a different pattern was observed (Figure
12.3B). Rather than bands of callosal terminals as in normal AI, significantly smaller,
rounder patches were found in a widespread arrangement, although they were absent from
medial AI. In contrast, deafened animals had callosal projections which were uniformly
and diffusely arranged (Figure 12.3C) These results show that sensory input is required for
refinement of the callosal pathway, and represent the first evidence that changing only the
modality and pattern of activity of sensory inputs, without changing the identity of the
thalamocortical fibres carrying the information, can direct the pattern of cortical inter-
connectivity.
The callosal patches are likely to connect neurones with similar visual or auditory
response properties, as in normal animals. Our pilot data (S. Pallas, D. Depireaux,
J. Simon and S. Shamma, unpublished) suggest that there is residual auditory input reach-
ing the cross-modal AI from the callosal pathway, the thalamocortical pathway, or both.
We propose that the pattern of callosal connections reflects a compromise in representing
visual and auditory information on one cortical surface. The absence of callosal projec-
tions from the medial AI suggests that AI contains segregated visual and auditory repre-
sentations, and thus a “new” cortical area, in order to prevent perceptual confusion. Such
a scenario would provide a valuable model for the early stages of cortical parcellation
driven by new inputs or by a modified pattern of activity during evolution.
5.4.2. Horizontal connectivity

In visual cortex, layer 2/3 pyramidal neurones interconnect with neurones that have
similar orientation tuning (Callaway and Katz, 1990). The regular spacing of orientation
columns produces a somewhat radial pattern of clustered intrinsic horizontal projections to
and from any one site (Gilbert and Wiesel, 1983; Matsubara et al., 1985, 1987; Rockland
and Lund, 1982; Callaway and Katz, 1990; Weliky and Katz, 1994; Ruthazer and Stryker,
1996) that represent mainly excitatory connections (Matsubara and Boyd, 1992).

Figure 12.4. Distribution of labelled boutons following a small injection of BDA in AI show its horizontal
connectivity patterns in (A) normal (B) cross-modal and (C) deaf animals.
In contrast, horizontal connections in the auditory cortex are arranged in elongated strips
along the isofrequency lines (Reale et al., 1983; Imig et al., 1986; Matsubara and Phillips,
1988; Ojima et al., 1991; Wallace and Bajwa, 1991). In visual cortex (Callaway and Katz,
1991), the pattern is sculpted from an initially diffuse pattern under the influence of visual
activity. We hypothesized that the visual input would prevent the formation of elongated
horizontal connections in cross-modal A1. Because of the role of intrinsic connections in
generating response properties, we proposed that alterations in these connections induced
by visual activity might be a substrate for the complex visual response properties which
we observed in cross-modal AI.
Small pressure injections of anterograde tracer (biotinylated dextran amine, BDA) in AI
(Gao and Pallas, 1999) revealed that the horizontal connections in normal ferrets were
organized as clusters of boutons along the isofrequency axis (Figure 12.4A), as in cats
(e.g. Matsubara and Phillips, 1988). In the cross-modal ferrets, bouton clusters in AI were
significantly increased in number, in scatter (both with respect to distance and angulation
away from the anteroposterior axis), and in the zone of influence (bouton coverage area)
after exposure of AI to the early anomalous visual inputs (Figure 12.4B). The bouton
clusters were arranged not as anteroposteriorly-elongated strips, but more radially around
the injection site, and into medial AI, an area devoid of boutons in normal A1. These
results suggest that the spatiotemporal activity pattern of sensory inputs, and not their
molecular identity or the identity of the cortical target, controls the organization of intra-
cortical circuitry. It is unlikely that the pattern results from a blockade of the develop-
mental refinement of horizontal connections, because of the increase in number of clusters
and their degree of refinement. Our studies of horizontal connections in deafened animals
support this interpretation (Figure 12.4C) (Gao et al., 1999a, and in preparation). In animals
deafened at P14 by bilateral cochlear ablation, horizontal connectivity is diffuse and not
well-organized into clusters. This diffuse state represents sprouting beyond an earlier
developmental state (Moershel and Pallas, 2001).
Our results are supported and confirmed by data from Sur and colleagues obtained by
using tracers in concert with optical recording (Sharma et al., 2000). Their results using
retrograde tracers were similar in some respects to ours using anterograde tracers; clusters
of cells retrogradely labeled by a tracer injection in A1 were very different in organization
to those seen in normal A1, and extended into medial AI rather than along the isofrequency
axis. They reported further that clusters of labeled cells were aligned with regions exhibit-
ing similar orientation tuning, much as they are in normal V1.

260 Sarah L. Pallas
5.4.3. A new cortical area?

Comparison of the pattern of callosal connectivity compared to the pattern of horizontal
connectivity in cross-modal A1 suggests a mirror-image relationship. Callosal connec-
tions are excluded from medial A1, whereas horizontal connections extend preferentially
into medial A1. Given that the lesions in these animals were largely unilateral, it seems
likely that A1 receives not only visual input from ipsilateral MGN, but also auditory input
from its contralateral homolog via callosal connections. As mentioned above, we have some
preliminary evidence that the cross-modal A1 is responsive to auditory stimuli (S. Pallas,
D. Depireaux, J. Simon and S. Shamma, unpublished). If this is the case, then what may
have occurred in these animals is the creation of a new cortical area in the medial portion
of the ectosylvian gyrus which represents visual information in two dimensions, and
a more lateral portion which contains a compressed version of the one-dimensional tono-
topic map. If this is indeed occurring, our paradigm mimics in developmental time the
addition of a new cortical pathway driven by the periphery. The ability to generate a new
cortical area during development would provide a powerful model system for the cortical
parcellation that occurred during mammalian brain evolution.
5.5. Activity Instructs Cortical Circuitry

The anatomical results showing visually-induced alterations of callosal and horizontal
connectivity, taken together with the physiological data showing visual response proper-
ties in AI, strongly suggest that patterned input activity plays an instructive and not merely
a permissive role in the development of modular organization in sensory cortex (Crair,
1999; Gao and Pallas, 1999). Does this explain how A1 is able to process visual informa-
tion? We suggest that changes in the modular organization of A1 are more likely to be a
reflection of and not the driving force behind the visual response properties in cross-modal
A1. The process of mapping stimulus attributes is likely to be independent of the process
of constructing the circuits underlying specific response properties, in that stimulus tuning
can be disrupted without affecting map topography (e.g. Weliky and Katz, 1997; Huang
and Pallas, 1999, 2000, 2001). How then do visual response properties get constructed?
5.6. Chemoarchitecture
To investigate possible alterations in cross-modal A1’s circuitry that might be the source
of the 2-D visual response properties, we returned to our model (Figure 12.2), which
suggested that there must be a means of selectively enhancing or suppressing some of
the overlapped inputs from MGN along the isofrequency axis. With the goal of testing
the hypothesis that visually-driven alteration of inhibitory circuits provided the substrate
for visual response properties and topography, we examined the number, distribution,
and morphology of inhibitory neurones in A1 of normal, cross-modal, and deafened
ferrets.
GABAergic neurones are involved in the sculpting of receptive field properties in
visual cortex (Sillito, 1975a,b; Berman et al., 1992; Crook and Eysel, 1992; Sato et al.,
1995, 1996; Allison et al., 1996; Crook et al., 1996, 1997, 1998; Das and Gilbert, 1999),
and in the auditory pathway (Le Beau et al., 1996; Wang et al., 2000), although the
auditory cortex has not been well-studied in this respect. GABA may also subserve

critical functions during early (Lauder, 1993; Berninger et al., 1995; Behar et al., 1996;
Antonopoulos et al., 1997) and late (Reiter and Stryker, 1988; Hendry and Carder,
1992; Rutherford et al., 1997; Hensch et al., 1998; Zheng and Knudsen, 1999) cortical
development. Other studies suggest GABA regulation is of major importance during
activity-dependent postnatal development (Rutherford et al., 1997; Hensch et al., 1998;
Zheng and Knudsen, 1999), suggesting a role for GABA in cortical circuit construction
and experience-dependent plasticity during late stages of development. It therefore
seemed a likely candidate as a driving force for changes in cortical circuits in the cross-
modal animals.
To determine whether changes in inhibitory circuitry were involved in the creation of
visual response properties in cross-modal A1, we examined the distribution of GABAergic
neurones and GABAA receptors in A1 and V1 of normal, cross-modal and deaf ferrets
during postnatal development. In addition, because identified subpopulations of GABAergic
neurones contain different calcium-binding proteins, we also examined the distribution of
parvalbumin (PV), which is found in basket and chandelier cells in layers 2–5 of cortex,
and calbindin (CB), which is found in double-bouquet and bipolar cells in supra- and
infragranular layers but not the granular layer of primary sensory cortex (Peters and Jones,
1984; Hendry et al., 1989; Hendry and Jones, 1991).
5.6.1. Normal animals

We described the development of GABA-immunoreactive (-ir) neurones in visual and
auditory cortex of normal ferrets to determine the time course of GABA expression, par-
ticularly whether it coincided with important events in cortical development. Morphology
of these neurones was similar to that in cats, and not demonstrably different in normal V1
vs. AI. Quantitative analysis revealed that the percentage of GABA-ir, PV-ir and CB-ir
neurones peaked late in postnatal development (Gao et al., 1999b, 2000), just prior to the
end of the critical period (Issa et al., 1999). Using quantitative receptor-binding autoradio-
graphy, we found a parallel increase in GABAA receptors at P60 in both visual and audi-
tory cortex (Pallas et al., 1994; and in preparation). These data suggest that inhibitory
circuitry is stabilized relatively late in cortical development, and thus that GABAergic
neurones could be malleable later in postnatal development and serve as a substrate for
plasticity.
5.6.2. Deaf animals

We examined the distribution of GABAA receptors and GABA, PV, and CB-ir neurones in
AI of animals that had been deafened at P14 by bilateral cochlear ablation (two weeks
prior to the onset of hearing: see Moore and Hine, 1992). Deafferentation of the auditory
cortex during development caused a marked decline in the proportion of GABA, PV and
CB-ir neurones (Gao et al., 1999a, and in preparation), and a decrease in GABAA receptor
binding (Pallas et al., 1994a, and in preparation). This is similar to what occurs in visual
cortex with deafferentation (Hendry and Jones, 1986, 1988; Blümcke et al., 1994; Hendry
et al., 1994; Huntsman et al., 1994; Rosier et al., 1995; Arckens et al., 1998) and may
relate to the tendency of neurones to regulate their overall level of excitablility (Ruther-
ford et al., 1997).

262 Sarah L. Pallas
Normal Cross-Modal
Figure 12.5. The morphology of calbindin-ir neurones is altered in cross-modal AI. Rather than arborizing
within a column as is typical of calbindin-containing double-bouquet neurones, the dendrites extend laterally
across columns.
5.6.3. Cross-modal animals

Contrary to the results in deaf ferrets, there was no measurable decline in GABAA receptor
binding in AI in cross-modal ferrets. Also, the proportion of GABAergic neurones was not
significantly different from normal. However, CB neurones increased in density and were
more widely distributed across all layers of cross-modal cortex, including layer 4, and
many CB neurones had an atypical morphology (Figure 12.5): Rather than arborizing
vertically within a cortical column, their dendrites extended horizontally across columns,
where they would be in a position to suppress information gathered from across the isofre-
quency axis. These results are intriguing in relation to our hypothesis that visually-driven
activity from early in development changes inhibitory circuitry in AI. To suppress the
overlapped thalamocortical inputs along the isofrequency axis, increased lateral inhibition
along that axis and within layer 4 would be expected. We hypothesize that such visually-
driven changes in inhibitory circuitry could be responsible for the creation of visual
response properties in AI. If this is true, it would further support the idea that visual inputs
can cause active changes in target circuitry to arrange for their own effective processing.
5.7. Behaviour
Our psychophysical studies showed that the novel pathway from the retina to the auditory
system is capable of subserving visual function (Merzenich, 2000; von Melchner et al.,
2000). Unilaterally-lesioned cross-modal animals were trained through their unlesioned
hemisphere (the monocular segment of the left eye) to distinguish auditory from visual
stimuli and report accordingly in a forced-choice paradigm. When tested through the
cross-modal pathway (monocular segment of the right eye), they immediately and con-
sistently defined visual stimuli as visual, not auditory. Lesion of any visual structures
remaining from the early rewiring surgery (superior colliculus, lateral posterior nucleus,
LGN, visual cortex) did not affect the behavioural choice, showing that it was not
dependent upon visual pathways other than the one induced through MGN. In contrast,
post-testing lesion of auditory cortex eliminated the perceptual ability on the lesion side

only, and the animals reported visual or auditory percepts randomly in their right mon-
ocular field.
These results contradict centuries of philosophical and scientific thought on the local-
ization of function in the brain and the development of perception. It was expected that
any neuronal activity occurring in the auditory pathway would be interpreted by the ferret
as auditory, not visual, according to Müller’s labelled line theory. In our experiments,
because the percept depended on the input source, and not the labelled line, the animals
could “see” with their auditory pathway. Although the acuity of this pathway is much
lower than that of the normal visual pathway, the animals can still distinguish different
grating frequencies, indicating that not only can they detect light, but that the new path-
way subserves some degree of pattern vision.
6. AREAL SPECIFICATION: PREPROGRAMMED VERSUS

POTENTIAL FATE
Our results and those of others employing the cross-modal plasticity paradigm strongly
support the notion that sensory inputs and patterned activity play an important role in the
areal specification of the neocortex. But how are we to reconcile this with the demon-
strated role of molecular factors intrinsic to the brain, or the several studies that have
revealed the presence of area-specific modules prior to or in the absence of activity?
The answer may lie in the timing and the degree of specification at each developmental
stage. Early events, prior to connections between the brain and the sensory organs, are
necessarily controlled by intrinsic factors. Differential expression of molecular markers
may specify the dorsoventral and rostrocaudal axes as in spinal cord (Tanabe and Jessell,
1996), and construct a preliminary scaffold for the cortex, setting up regional (but perhaps
not areal) markers that guide thalamic afferents to their approximate termination site
(Chenn et al., 1997; Mann et al., 1998; Bishop et al., 2000). These factors may bias the
brain toward its usual final state and restrict its potential somewhat. Later events may be
directed by extrinsic information, intrinsic factors, or both. Furthermore, specification
events that are initially under intrinsic control may be modifiable or reversible at later
stages by extrinsic information. The dependence of late stages of cortical development on
correlated activity patterns suggests that morphogenic markers are no longer present at
that time, that the brain is no longer sensitive to them, or that they can be overridden.
A study of the temporal and spatial overlap of markers and their receptors could begin to
address this question. In any case, this malleability allows the brain to be exquisitely sens-
itive to environmental influences, a critical factor in cortical development and plasticity as
well as in neocortical evolution.
7. SENSORY SUBSTITUTION STUDIES: CLINICAL RELEVANCE
What is the relevance of cross-modal plasticity beyond our unique experimental situation?
The paradigm is also a model for perinatal brain damage and can reveal how the brain
compensates for a loss of (or a change in) sensory input. It is well-known that sensory
systems even in adults can reorganize in response to altered sensory stimulation

264 Sarah L. Pallas
(e.g. Merzenich et al., 1996), but it is also possible for compensatory reorganization to
occur across different cortical areas into other sensory representations.
Clinical evidence relevant to this issue includes the observation that in blind and deaf
patients, the “unused” cortical area can be taken over and used by a different sensory
modality (Kujala et al., 1997; see Kujala et al., 2000, for review), leading to increased
ability in the other modality compared to normal (Muchnik, 1991; Lessard et al., 1998;
Roder et al., 1999). Studies in humans with early-onset deafness showed that event-related
potentials could be recorded in putative auditory areas in response to visual stimuli
(Neville et al., 1983; Neville and Lawson, 1987; Neville, 1990). Additionally, Braille
reading in the early-blind activates visual cortex (Sadato et al., 1996; Cohen et al., 1997,
1999) and auditory localization tasks activate the occipital cortex in the blind (Weeks
et al., 2000). Sensory substitution depends on attentional mechanisms, and can occur to
some extent in individuals with late-onset deficits as well (Kujala et al., 1997).
Cross-modal plasticity has also been demonstrated in other animal models. Rauschecker,
Wollberg and their colleagues have shown that sensory substitution can occur in visually-
compromised animals (Rauschecker, 1997, 1999; Yaka et al., 1999), and in some cases
performance in the intact modality is enhanced (King and Parsons, 1999). Loss of vision
through deprivation or enucleation resulted in auditory activation of visual cortex and
improved auditory localization ability. The converse manipulation has also been
performed. When the cochlea was damaged early in life, visual responses were seen in
auditory cortex (Rebillard et al., 1977; Rauschecker and Korte, 1993). It has not been
clear in each case, however, whether the novel inputs reorganize cortical circuitry, or
whether they simply take advantage of existing circuits. Furthermore, it is not known
whether the invasion and reorganization process is dependent on the pattern of activity or
the identity of the thalamocortical afferents.
Our results suggest that activity-dependent control over cortical circuit formation by
thalamocortical afferents, based on correlated activity within neighbouring regions of the
sensory epithelium, may explain a number of these observations. Furthermore, it may pro-
vide a general mechanism for cortical regionalization at the circuit level of organization.
Such an afferent-driven organization rule would have broad implications for both the
development and the evolution of mammalian sensory cortex.
8. CONCLUSION
Cortical areas may indeed be interchangeable to a significant extent, but only if appropri-
ate sensory experience is provided before the end of the critical period. Results from
experiments using the cross-modal plasticity paradigm reveal that the modality and pat-
tern of activity in the inputs to cortex can play an instructive as well as a permissive role in
the organization of cortical circuitry during development. Cross-modal plasticity may also
provide a substrate for neocortical evolution, allowing the adaptive incorporation of
changes in peripheral or central sensory structures. If the basis of the critical period in
brain tissue and the limitations on axon pathfinding and synapse formation are understood,
then we may eventually be in a position to offer clinical therapy for brain injuries that
occur later in development, teaching the remaining cortical tissue how to substitute for the
damaged regions.

ACKNOWLEDGEMENTS
We thank Barbara Finlay and the editors for critical reading of the manuscript. This work
was supported by grants to S.L.P. from the National Institutes of Health, the National
Science Foundation, the Whitehall Foundation, the Fight for Sight Inc. Research Division
of Prevent Blindness America, and the Georgia Research Alliance.
REFERENCES
Aitkin, L., Irvine, D.R.F. and Webster, W.R. (1984) Central neural mechanisms of hearing, Handbook of Physi-
ology: The Nervous System III. Bethesda: American Physiological Society, pp. 675–737.
Allison, J.D., Kabara, J.F., Snider, R.K., Casagrande, V.A. and Bonds, A.B. (1996) GABAB-receptor-mediated
inhibition reduces the orientation selectivity of the sustained response of striate cortical neurons in cats.
Visual Neuroscience, 13, 559–566.
Andersen, R.A., Knight, P.L. and Merzenich, M.M. (1980) The thalamocortical and corticothalamic connections
of AI, AII and the anterior auditory field (aaf) in the cat: Evidence for two largely segregated systems of
connections. Journal of Comparative Neurology, 194, 663–701.
Angelucci, A., Clascá, F. and Sur, M. (1996) Anterograde axonal tracing with the subunit B of cholera toxin:
A highly sensitive immunohistochemical protocol for revealing fine axonal morphology in adult and
neonatal brains. Journal of Neuroscience Methods, 65, 101–112.
Angelucci, A., Clascá, F., Bricolo, E., Cramer, K.S. and Sur, M. (1997) Experimentally induced retinal projec-
tions to the ferret auditory thalamus: Development of clustered eye-specific patterns in a novel target.
Antonopoulos, J., Pappas, I.S. and Parnavelas, J.G. (1997) Activation of the GABAA receptor inhibits the
proliferative effects of BFGF in cortical progenitor cells. European Journal of Neuroscience, 9, 291–298.
Arckens, L., Eysel, U.T., Vanderhaegen, J.J., Orban, G.A. and Vandesande, F. (1998) Effect of sensory deaffer-
entation on the GABAergic circuitry of the adult cat visual system. Neuroscience, 83, 381–391.
Arimatsu, Y., Miyamoto, M., Nihonmatsu, I., Hirata, K., Uratani, Y., Hatanaka, Y. and Takiguchi-Hayashi, K.
(1992) Early regional specification for a molecular neuronal phenotype in the rat neocortex. Proceedings of
the National Academy of Science, U.S.A., 89, 8879–8883.
Baker, G.E., Thompson, I.D., Krug, K., Smyth, D. and Tolhurst, D.J. (1998) Spatial-frequency tuning and genic-
ulocortical projections in the visual cortex (areas 17 and 18) of the pigmented ferret. European Journal of
Bayer, S.A. and Altman, J. (1991) Neocortical development. New York: Raven Press.
Bartfeld, E. and Grinvald, A. (1992) Relationships between orientation-preference pinwheels, cytochrome
oxidase blobs and ocular dominance columns in primate striate cortex. Proceedings of the National Acad-
emy of Science, U.S.A., 89, 11905–11909.
Behar, T.N., Li, Y.X., Tran, H.T., Ma, W., Dunlap, V., Scott, C. et al. (1996) GABA stimulates chemotaxis and
chemokinesis of embryonic cortical neurons via calcium-dependent mechanisms. Journal of Neuroscience,
16, 1808–1818.
Berman, N.J., Douglas, R.J. and Martin, K.A.C. (1992) GABA-mediated inhibition in the neural networks of
visual cortex. In: R.R. Mize, R.E. Marc and A.M. Sillito (eds), GABA in the Retina and Central Visual
System (Progress in Brain Research, Vol. 90), New York: Elsevier, pp. 443–476.
Berninger, B., Marty, S., Zafra, F., Da Penha Berzaghi, M., Thoenen, H. and Lindholm, D. (1995) GABAergic
stimulation switches from enhancing to repressing bdnf expression in rat hippocampal neurons during
maturation in vitro. Development, 121, 2327–2335.
Bhide, P.G. and Frost, D.O. (1999) Intrinsic determinants of retinal axon collateralization and arborization
patterns. Journal of Comparative Neurology, 411, 119–129.
Bishop, K.M., Goudreau, G. and O’Leary, D.D.M. (2000) Regulation of area identity on the mammalian
neocortex by Emx2 and Pax6. Science, Washington, 288, 344–349.
Blümcke, I., Weruaga, E., Kasas, S., Hendrickson, A.E. and Celio, M.R. (1994) Discrete reduction patterns of
parvalbumin and calbindin D-28k immunoreactivity in the dorsal lateral geniculate nucleus and the striate
cortex of adult macaque monkeys after monocular enucleation. Visual Neuroscience, 11, 1–11.
Bourdet, C., Olavarria, J.F. and Van Sluyters, R.C. (1996) Distribution of visual callosal neurons in normal and
strabismic cats. Journal of Comparative Neurology, 366, 259–269.
Bronchti, G., Heil, P., Scheich, H. and Wollberg, Z. (1989) Auditory pathway and auditory activation of primary
visual targets in the blind mole rat (Spalax ehrenbergi): I. 2-deoxyglucose study of subcortical centers.

266 Sarah L. Pallas
Bronchti, G., Rado, R., Terkel, J. and Wollberg, Z. (1991) Retinal projections in the blind mole rat: A WGA-HRP
tracing study of a natural degeneration. Developmental Brain Research, 58, 159–170.
Bruer, J.T. (1999) The Myth of the First Three Years: A New Understanding of Early Brain Development and
Lifelong Learning. San Francisco: Free Press.
Butler, A.B. and Molnar, Z. (1998) Development and evolution in nervous systems: Development and evolution
of ideas. Trends in Neuroscience, 21, 177.
Callaway, E.M. and Katz, L.C. (1990) Emergence and refinement of clustered horizontal connections in cat
Callaway, E.M. and Katz, L.C. (1991) Effects of binocular deprivation on the development of clustered
horizontal connections in cat striate cortex. Proceedings of the National Academy of Science, U.S.A., 88,
745–749.
Castellani, V., Yue, Y. and Bolz, J. (1998) Dual action of a ligand for eph receptor tyrosine kinases on specific
populations of axons during the development of cortical circuits. Journal of Neuroscience, 18, 4663–4672.
Catalano, S.M. and Shatz, C.J. (1998) Activity-dependent cortical target selection by thalamic axons. Science,
Washington, 281, 599–562.
Changeaux, J.-P. (1986) Neuronal Man. New York: Oxford University Press.
Chenn, A., Braisted, J., McConnell, S.K. and O’Leary, D.D.M. (1997) Development of the cerebral cortex:
Mechanisms controlling cell fate, laminar and areal patterning and axonal connectivity. In: W.M. Cowan,
T.M. Jessell and S.L. Sipursky (eds), Molecular and Cellular Approaches to Neural Development. New
York: Oxford University Press, pp. 440–473.
Cohen, L.G., Celnik, P., Pascual-Leone, A., Cornwell, B., Faiz, L., Dambrosias, J. et al. (1997) Functional
relevance of cross-modal plasticity in blind humans. Nature, London, 389, 180–186.
Cohen, L.G., Weeks, R.A. and Hallett, M. (1999) Period of susceptibility for cross-modal plasticity in the blind.
Annals of Neurology, 45, 451–460.
Cohen-Tannoudji, M., Babinet, C. and Wassef, M. (1994) Early determination of a mouse somatosensory cortex
marker. Nature, London, 368, 460–463.
Cook, P.M., Prusky, G. and Ramoa, A.S. (1999) The role of spontaneous retinal activity before eye opening in
the maturation of form and function in the retinogeniculate pathway of the ferret. Visual Neuroscience, 16,
491–501.
Cooper, H.M., Herbin, M. and Nevo, E. (1993) Ocular regression conceals adaptive progression of the visual
system in a blind subterranean mammal. Nature, London, 361, 156–159.
Crair, M.C. (1999) Neuronal activity during development: Permissive or instructive? Current. Opinion in
Crandall, J.E. and Caviness, V.S. (1984) Thalamocortical connections in newborn mice. Journal of Comparative
Neurology, 228, 542–556.
Crook, J.M. and Eysel, U.T. (1992) GABA-induced inactivation of functionally characterized sites in cat visual
cortex (area 18): Effects on orientation tuning. Journal of Neuroscience, 12, 1816–1825.
Crook, J.M., Kisvárday, Z.F. and Eysel, U.T. (1996) GABA-induced inactivation of functionally character-
ized sites in cat visual cortex (area 18): Effects on direction selectivity. Journal of Neurophysiology, 75,
2071–2088.
Crook, J.M., Kisvárday, Z.F. and Eysel, U.T. (1997) GABA-induced inactivation of functionally characterized
sites in cat striate cortex: Effects on orientation tuning and direction selectivity. Visual Neuroscience, 14,
141–158.
Crook, J.M., Kisvárday, Z.F. and Eysel, U.T. (1998) Evidence for a contribution of lateral inhibition to orienta-
tion tuning and direction selectivity in cat visual cortex: Reversible inactivation of functionally character-
ized sites combined with neuroanatomical tracing techniques. European Journal of Neuroscience, 10,
2056–2075.
Darlington, R.B., Dunlop, S.A. and Finlay, B.L. (1999) Neural development in metatherian and eutherian
mammals: Variation and constraint. Journal of Comparative Neurology, 411, 359–368.
Das, A. and Gilbert, C.D. (1999) Topography of contextual modulations mediated by short-range interactions in
primary visual cortex. Nature, London, 399, 655–661.
Daw, N.W. (1998) Neurobiology: Columns, slabs and pinwheels. Nature, London, 395, 20.
deCharms, R.C., Blake, D.T. and Merzenich, M.M. (1998) Optimizing sound features for cortical neurons.
Science, Washington, 280, 1439.
Dehay, C., Horsburgh, G., Berland, M., Killackey, H. and Kennedy, H. (1991) The effects of bilateral enucle-
ation in the primate fetus on the parcellation of visual cortex. Developmental Brain Research, 62, 137–141.
Donoghue, M.J. and Rakic, P. (1999) Molecular evidence for the early specification of presumptive functional
domains in the embryonic primate cerebral cortex. Journal of Neuroscience, 19, 5967–5979.
Doron, N. and Wollberg, Z. (1994) Cross-modal neuroplasticity in the blind mole rat Spalax ehrenbergi:
A WGA-HRP tracing study. NeuroReport, 5, 2697–2701.
Essick, G.K. and Whitsel, B.L. (1985) Factors influencing cutaneous direction sensitivity: A correlative psycho-
physical and neurophysiological investigation. Brain Research Reviews, 10, 213–230.

Feng, J.Z. and Brugge, J.F. (1983) Postnatal development of auditory callosal connections in the kitten. Journal
Ferster, D. (1992) The synaptic inputs to simple cells of the cat visual cortex. In: R.R. Mize, R. Marc and
A. Sillito (eds), GABA in the Retina and Central Visual System (Progress in Brain Research, Vol. 90).
Amsterdam: Elsevier, pp. 423–441.
Finlay, B.L., Schneps, S.E. and Schneider, G.E. (1979) Orderly compression of the retinotectal projection
following partial tectal ablation in the newborn hamster. Nature, London, 280, 153–154.
Finlay, B.L. and Pallas, S.L. (1989) Control of cell number in the developing mammalian visual system.
Progress in Neurobiology, 32, 207–234.
Finlay, B.L. and Darlington, R.B. (1995) Linked regularities in the development and evolution of mammalian
brains. Science, Washington, 268, 1578–1584.
Finlay, B.L., Hersman, M.N. and Darlington, R.B. (1998) Patterns of vertebrate neurogenesis and the paths of
vertebrate evolution. Brain Behavior and Evolution, 52, 232–242.
Frost, D.O. (1981) Ordered anomalous retinal projections to the medial geniculate, ventrobasal and lateral
posterior nuclei. Journal of Comparative Neurology, 203, 227–256.
Frost, D.O. (1999) Functional organization of surgically created visual circuits. Restorative Neurology and
Gao, W.-J., Moore, D.R. and Pallas, S.L. (1999a) Bilateral cochlear ablation in neonatal ferrets prevents refine-
ment of horizontal connectivity in primary auditory cortex. Society for Neuroscience Abstracts, 25, 1771.
Gao, W.-J., Newman, D.E., Wormington, A.B. and Pallas, S.L. (1999b) Development of inhibitory circuitry in
visual and auditory cortex of postnatal ferrets: Immunocytochemical localization of GABAergic neurons.
Gao, W.-J. and Pallas, S.L. (1999) Cross-modal reorganization of horizontal connectivity in auditory cortex
without altering thalamocortical projections. Journal of Neuroscience, 19, 7940–7950.
Gao, W.-J., Wormington, A.B., Newman, D.E. and Pallas, S.L. (2000) Development of inhibitory circuitry
in visual and auditory cortex of postnatal ferrets: Immunocytochemical localization of calbindin and
parvalbumin-containing neurons. Journal of Comparative Neurology, 422, 140–157.
Ghosh, A., Antonini, A., McConnell, S.K. and Shatz, C.J. (1990) Requirement for subplate neurons in the
formation of thalamocortical connections. Nature, London, 347, 179–181.
science, 3, 1116–1133.
Gitton, Y., Cohen-Tannoudji, M. and Wassef, M. (1999) Role of thalamic axons in the expression of H-2Z1,
a mouse somatosensory cortex specific marker. Cerebral Cortex, 9, 611–620.
Goodman, C.S. and Coughlin, B.C. (2000) Special feature: The evolution of evo-devo biology. Proceedings of
the National Academy of Science, U.S.A., 97, 4424–4425.
Gopnik, A., Meltzoff, A.N. and Kuhl, P.K. (1999) The Scientist in the Crib: Minds, Brains and how Children
Learn. New York: William Morrow and Co.
Gu, X.N. and Spitzer, N.C. (1997) Breaking the code: Regulation of neuronal differentiation by spontaneous
calcium transients. Developmental Neuroscience, 19, 33–41.
Guillemot, F., Lo, L.C., Johnson, J.E., Auerbach, A., Anderson, D.J. and Joyner, A.L. (1993) Mammalian
achaete scute homolog 1 is required for the early development of olfactory and autonomic neurons. Cell,
75, 463–476.
Hebb, D.O. (1949) The Organization of Behavior. New York: John Wiley.
Heil, P., Bronchti, G., Wollberg, Z. and Scheich, H. (1991) Invasion of visual cortex by the auditory system in
the naturally blind mole rat. NeuroReport, 2, 735–738.
Hendry, S.H.C. and Jones, E.G. (1986) Reduction in number of immunostained GABAergic neurones in
deprived-eye dominance columns of monkey area 17. Nature, London, 320, 750–753.
Hendry, S.H.C. and Jones, E.G. (1988) Activity-dependent regulation of GABA expression in the visual cortex
of adult monkeys. Neuron, 1, 701–712.
Hendry, S.H.C., Jones, E.G., Emson, O.C., Lawson, D.E.M., Heizmann, C.W. and Streit, P. (1989) Two classes
of cortical GABA neurons defined by differential calcium binding protein immunoreactivities. Experi-
mental Brain Research, 76, 467–472.
Hendry, S.H.C. and Jones, E.G. (1991) GABA neuronal subpopulations in cat primary auditory cortex:
Co-localization with calcium binding proteins. Brain Research, 543, 45–55.
Hendry, S.H.C. and Carder, R. (1992) Organization and plasticity of GABA neurons and receptors in monkey
visual cortex. In: R.R. Mize, R. Marc and A. Sillito (eds), GABA in the Retina and Central Visual System
(Progress in Brain Research, Vol. 90). Amsterdam: Elsevier, pp. 477–502.
Hendry, S.H.C., Huntsman, M.-M., Viñuela, A., Möhler, H., De Blas, A.L. and Jones, E.G. (1994) GABAA
receptor subunit immunoreactivity in primate visual cortex: Distribution in macaques and humans and
regulation by visual input in adulthood. Journal of Neuroscience, 14, 2383–2401.
Hensch, T., Fagiolini, M., Mataga, N., Stryker, M., Baekkeskov, S. and Kash, S. (1998) Local GABA circuit
control of experience-dependent plasticity in developing visual cortex. Science, Washington, 282, 1504–1508.

268 Sarah L. Pallas
Huang, L. and Pallas, S.L. (1999) NMDA receptor blockade in developing superior colliculus prevents map
plasticity without blocking retinotectal transmission. Society for Neuroscience Abstracts, 25, 1005.
Huang, L. and Pallas, S.L. (2000) Visual response properties in the superior colliculus after NMDA receptor
blockade during early development. Society for Neuroscience Abstracts, 26, 119.11
Huang, L. and Pallas, S.L. (2001) NMDA antagonists in the superior colliculus prevent developmental plasticity
but not visual transmission or map compression. Journal of Neurophysiology, 86, 1179–1194.
Huffman, K.J., Molnar, Z. and Krubitzer, L. (1999) Formation of cortical fields on a reduced cortical sheet.
Huntsman, M.M., Isackson, P.J. and Jones, E.G. (1994) Lamina-specific expression and activity-dependent
regulation of seven GABAA receptor subunit mRNAs in monkey visual cortex. Journal of Neuroscience,
14, 2236–2259.
Hyvärinen, J.L. and Poränen, A. (1978) Movement-sensitive and direction and orientation-selective cutaneous
receptive fields in the hand area of the post-central gyrus in monkeys. Journal of Physiology (London), 283,
523–537.
Imig, T.J. and Brugge, J.F. (1978) Sources and terminations of callosal axons related to binaural and frequency
maps in primary auditory cortex of the cat. Journal of Comparative Neurology, 182, 637–660.
Imig, T.J., Reale, R.A., Brugge, J.F., Morel, A. and Adrian, H.O. (1986) Topography of cortico-cortical connec-
tions related to tonotopic and binaural maps of cat auditory cortex. In: F. Lepore, M. Ptito and H.H. Jasper
(eds), Two Hemispheres-One Brain: Functions of the Corpus Callosum. New York: Alan R. Liss, Inc.,
pp. 103–115.
Innocenti, G.M., Fiore, L. and Caminiti, R. (1977) Exuberant projection into the corpus callosum from the visual
cortex of newborn cats. Neuroscience Letters, 4, 237–242.
Innocenti, G.M. and Frost, D.O. (1979) Effects of visual experience on the maturation of the efferent system to
the corpus callosum. Nature, London, 280, 231–234.
Innocenti, G.M. (1986) General organization of callosal connections in the cerebral cortex, In: A. Peters and
E.G. Jones (eds), Sensory-Motor Areas and Aspects of Cortical Connectivity. Cerebral Cortex, Volume 5.
New York: Plenum Press, pp. 291–353.
Inoue, T., Tanaka, T., Suzuki, S.C. and Takeichi, M. (1998) Cadherin-6 in the developing mouse brain: Expres-
sion along restricted connection systems and synaptic localization suggest a potential role in neuronal
circuitry. Developmental Dynamics, 211, 338–351.
Issa, N.P., Trachtenberg, J.T., Chapman, B., Zahs, K.R. and Stryker, M.P. (1999) The critical period for ocular
dominance plasticity in the ferret’s visual cortex. Journal of Neuroscience, 19, 6965–6978.
Jackson, C.A., Peduzzi, J.D. and Hickey, T.L. (1989) Visual cortex development in the ferret. I. Genesis and
migration of visual cortical neurons. Journal of Neuroscience, 9, 1242–1253.
Jerison, H.J. (1990) Fossil brains and the evolution of the neocortex. In: B.L. Finlay, G. Innocenti and H. Scheich
(eds), The Neocortex: Ontogeny and Phylogeny. NATO Advanced Research Workshop. New York: Plenum.
Juliano, S.L., Palmer, S.L., Sonty, R.V., Noctor, S. and Hill II, G.F. (1996) Development of local connections in
ferret somatosensory cortex. Journal of Comparative Neurology, 374, 259–277.
Kandler, K. and Katz, L.C. (1998) Relationship between dye coupling and spontaneous activity in developing
ferret visual cortex. Developmental Neuroscience, 20, 59–64.
Katz, L.C. and Shatz, C.J. (1996) Synaptic activity and the construction of cortical circuits. Science, Washington,
274, 1133–1138.
Kelly, J.B., Judge, P.W. and Phillips, D.P. (1986) Representation of the cochlea in primary auditory cortex of the
ferret (Mustela putorius). Hearing Research, 24, 111–115.
Kelly, J.B. and Judge, P.W. (1994) Binaural organization of primary auditory cortex in the ferret (Mustela
putorius). Journal of Neurophysiology, 71, 904–913.
King, A.J. and Parsons, C.H. (1999) Improved auditory spatial acuity in visually deprived ferrets. European
Kingsbury, M.A. and Finlay, B.L. (2001) The cortex in multidimensional space: Where do cortical areas come
from? Developmental Science, 4, 125–142.
Kujala, T., Alho, K., Huotilainen, M., Ilmoniemi, R.J., Lehtokoski, A., Leinonen, A., Rinne, T., Salonen, O.,
Sinkkonen, J., Standertskjold Nordenstam, C.G. and Naatanen, R. (1997) Electrophysiological evidence
for cross-modal plasticity in humans with early- and late-onset blindness. Psychophysiology, 34, 213–216.
Kujala, T., Alho, K. and Näätänen, R. (2000) Cross-modal reorganization of human cortical functions. Trends in
Langner, G., Sams, M. and Schulze, H. (1997) Frequency and periodicity are represented in orthogonal maps in
the human auditory cortex: Evidence from magnetoencephalography. Journal of Comparative Physiology,
A, 181, 665.
Lauder, J.M. (1993) Neurotransmitters as growth regulatory signals: Role of receptors and second messengers.

Law, M.I., Zahs, K.R. and Stryker, M.P. (1988) Organization of primary visual cortex (area 17) in the ferret.
Le Beau, F.E.N., Rees, A. and Malmierca, M.S. (1996) Contribution of GABA- and glycine-mediated inhibition
to the monaural temporal response properties of neurons in the inferior colliculus. Journal of Neurophysi-
ology, 75, 902–919.
Lessard, N., Pare, M. and Lassonde, M. (1998) Early-blind human subjects localize sound sources better than
sighted subjects. Nature, London, 395, 278–280.
LeVay, S., Stryker, M.P. and Shatz, C.J. (1978) Ocular dominance columns and their development in layer IV
of the cat’s visual cortex: A quantitative study. Journal of Comparative Neurology, 179, 223–244.
Levitt, P., Barbe, M.F. and Eagleson, K.L. (1997) Patterning and specification of the cerebral cortex. Annual
Review of Neuroscience, 20, 1–24.
Lewis, J.W. and Olavarria, J.F. (1995) Two rules for callosal connectivity in striate cortex of the rat. Journal of
Livingstone, M. and Hubel, D.H. (1988) Segregation of form, color, movement and depth: Anatomy, physiology
and perception. Science, Washington, 240, 740–749.
Lumsden, A. and Krumlauf, R. (1996) Patterning the vertebrate neuraxis. Science, Washington, 274, 1109–1115.
Luskin, M.B. and Shatz, C.J. (1985) Neurogenesis of the cat’s primary visual cortex. Journal of Comparative
Neurology, 242, 611–631.
Mackarehtschian, K., Lau, C.K., Caras, I. and McConnell, S.K. (1999) Regional differences in the developing
cerebral cortex revealed by ephrin-a5 expression. Cerebral Cortex, 9, 601–610.
Mallamaci, A., Muzio, L., Chan, C.-H., Parnavelas, J. and Boncinelli, E. (2000) Area identity shifts in the early
cerebral cortex of Emx–/– mutant mice. Nature Neuroscience, 3, 679–686.
Mann, F., Zhukareva, V., Pimenta, A., Levitt, P. and Bolz, J. (1998) Membrane-associated molecules guide
limbic and nonlimbic thalamocortical projections. Journal of Neuroscience, 18, 9409–9419.
Matsubara, J.A., Cynader, M., Swindale, N.V. and Stryker, M. P. (1985) Intrinsic projections within visual
cortex: Evidence for orientation-specific local connections. Proceedings of the National Academy of
Science, U.S.A., 82, 935–939.
Matsubara, J.A., Cynader, M.S. and Swindale, N.V. (1987) Anatomical properties and physiological correlates
of the intrinsic connections in cat area 18. Journal of Neuroscience, 7, 1428–1446.
Matsubara, J.A. and Phillips, D.P. (1988) Intracortical connections and their physiological correlates in the
primary auditory cortex (AI) of the cat. Journal of Comparative Neurology, 268, 38–48.
Matsubara, J.A. and Boyd, J.D. (1992) Presence of GABA-immunoreactive neurons within intracortical patches
in area 18 of the cat. Brain Research, 583, 161–170.
McMullen, N.T. and de Venecia, R.K. (1993) Thalamocortical patches in auditory neocortex. Brain Research,
620, 317–322.
Mendelson, J.R. and Cynader, M.S. (1985) Sensitivity of cat primary auditory cortex (al) neurons to the direction
and rate of frequency modulation. Brain Research, 327, 331–335.
Mendelson, J.R., Schreiner, C.E., Sutter, M.L. and Grasse, K.L. (1993) Functional topography of cat primary
auditory cortex: Responses to frequency-modulated sweeps. Experimental Brain Research, 94, 65–87.
Merzenich, M.M. and Knight, P.L. (1975) Representation of cochlea within primary auditory cortex in the cat.
Merzenich, M., Wright, B., Jenkins, W., Xerri, C., Byl, N., Miller, S. and Tallal, P. (1996) Cortical plasticity
underlying perceptual, motor and cognitive skill development: Implications for neurorehabilitation. Cold
Spring Harbor Symposium on Quantitative Biology, 61, 1–8.
Merzenich, M.M. (2000) Seeing in the sound zone. Nature, London, 404, 820–821.
Métin, C. and Frost, D.O. (1989) Visual responses of neurons in somatosensory cortex of hamsters with experi-
mentally induced retinal projections to somatosensory thalamus. Proceedings of the National Academy of
Science, U.S.A., 86, 357–361.
Métin, C., Irons, W.A. and Frost, D.O. (1995) Retinal ganglion cells in normal hamsters and hamsters with
novel retinal projections. I. Number, distribution and size. Journal of Comparative Neurology, 353,
179–199.
Middlebrooks, J.C., Dykes, R.W. and Merzenich, M.M. (1980) Binaural response-specific bands in primary
auditory cortex (AI) of the cat: Topographical organization orthogonal to isofrequency contours. Brain
Research, 181, 31–48.
Miller, B.M., Windrem, M.S. and Finlay, B.L. (1991) Thalamic ablations and neocortical development: Alter-
ations in thalamic and callosal connectivity. Cerebral Cortex, 1, 241–261.
Miyashita-Lin, E.M., Hevner, R., Montzka Wasserman, K., Martinez, S. and Rubenstein, J.L.R. (1999) Early
neocortical regionalization in the absence of thalamic innervation. Science, Washington, 285, 906–909.
Moerschel, D.J. and Pallas, S.L. (2001) Early postnatal formation of clustered horizontal connections in ferret
auditory cortex. Society for Neuroscience Abstracts, 27, 903–908.
Moore, D.R. and Hine, J.E. (1992) Rapid development of the auditory brainstem response threshold in individual
ferrets. Developmental Brain Research, 66, 229–235.

270 Sarah L. Pallas
Morel, A. and Imig, T.J. (1987) Thalamic projections to fields A, AI, P and VP in the cat auditory cortex.
Mountcastle, V.B. (1957) Modality and topographic properties of single neurons of cat’s somatic sensory cortex.
Muchnik, C. (1991) Central auditory skills in blind and sighted subjects. Scandinavian Journal of Audiology, 20,
19–23.
Mulliken, W.H., Jones, J. P. and Palmer, L.A. (1984) Receptive field properties and laminar distribution of x-like
and y-like simple cells in cat area 17. Journal of Neurophysiology, 52, 350–371.
Nakagawa, Y., Johnson, J.E. and O’Leary, D.D.M. (1999) Graded and areal expression patterns of regulatory
genes and cadherins in embryonic neocortex independent of thalamocortical input. Journal of Neuro-
science, 19, 10877.
Nelken, I. and Versnel, H. (2000) Responses to linear and logarithmic frequency-modulated sweeps in ferret
primary auditory cortex. European Journal of Neuroscience, 12, 549–562.
Neville, H.J., Schmidt, A. and Kutas, M. (1983) Altered visual-evoked potentials in congenitally deaf adults.
Neville, H. and Lawson, D. (1987) Attention to central and peripheral visual space in a movement detection task:
An event-related potential and behavioral study. I. Congenitally deaf adults. Brain Research, 405, 268–283.
Neville, H.J. (1990) Intermodal competition and compensation in development. Evidence from studies of the
visual system in congenitally deaf adults. Annals of the New York Academy of Science, 608, 71–87.
Nothias, F., Fishell, G. and Ruiz i Altaba, A. (1998) Cooperation of intrinsic and extrinsic signals in the elabora-
tion of regional identity in the posterior cerebral cortex. Current Biology, 8, 459–462.
O’Donovan, M.J. (1999) The origin of spontaneous activity in developing networks of the vertebrate nervous
system. Current Opinion in Neurobiology, 9, 94–104.
Ojima, H., Honda, C. and Jones, E.G. (1991) Patterns of axon collateralization of identified supragranular
pyramidal neurons in the cat auditory cortex. Cerebral Cortex, 1, 80–94.
Olavarria, J.F. and Van Sluyters, R.C. (1995) Overall pattern of callosal connections in visual cortex of normal
and enucleated cats. Journal of Comparative Neurology, 363, 161–176.
O’Leary, D.D.M. (1989) Do cortical areas emerge from a protocortex? Trends in Neuroscience, 12, 400–406.
O’Leary, D.D.M. and Stanfield, B.B. (1989) Selective elimination of axons extended by developing cortical
neurons is dependent on regional locale: Experiments utilizing fetal cortical transplants. Journal of Neuro-
science, 9, 2230–2246.
Pallas, S.L. and Finlay, B.L. (1989) Conservation of receptive-field properties of superior colliculus cells after
developmental rearrangements of retinal input. Visual Neuroscience, 2, 121–135.
Pallas, S.L. (1990) Cross-modal plasticity in sensory cortex: Visual responses in primary auditory cortex in
ferrets with induced retinal projections to the medial geniculate nucleus. In: B.L. Finlay, G. Innocenti and
H. Scheich (eds), The Neocortex: Ontogeny and Phylogeny. NATO Advanced Research Workshop. New
York: Plenum, pp. 205–218.
Pallas, S.L., Roe, A.W. and Sur, M. (1990) Visual projections induced into the auditory pathway of ferrets.
I. Novel inputs to primary auditory cortex (AI) from the LP/pulvinar complex and the topography of the
MGN-AI projection. Journal of Comparative Neurology, 298, 50–68.
Pallas, S.L. and Sur, M. (1993) Visual projections induced into the auditory pathway of ferrets. II. Cortico-
cortical connections of primary auditory cortex with visual input. Journal of Comparative Neurology, 337,
317–333.
Pallas, S.L. and Sur, M. (1994) Morphology of retinal axon arbors induced to arborize in a novel target, the
medial geniculate nucleus. II. Comparison with axons from the inferior colliculus. Journal of Comparative
Neurology, 349, 363–376.
Pallas, S.L., Booth, V. and Cynader, M. (1994a) Development and plasticity of GABA circuitry in primary
visual and primary auditory cortex in ferrets. Society for Neuroscience Abstracts, 20, 875.
Pallas, S.L., Hahm, J. and Sur, M. (1994b) Morphology of retinal axons induced to arborize in a novel target, the
medial geniculate nucleus. I. Comparison with arbors in normal targets. Journal of Comparative Neurology,
349, 343–362.
Pallas, S.L. and Moore, D.R. (1997) Retinal axons arborize in the medial geniculate nucleus of neonatally-
deafened ferrets. Society for Neuroscience Abstracts, 23, 1994.
Pallas, S.L., Littman, T. and Moore, D.R. (1999) Cross-modal reorganization of callosal connectivity in auditory
cortex without altering thalamocortical projections. Proceedings of the National Academy of Science,
U.S.A., 96, 8751–8756.
Pallas, S.L. (2001) Intrinsic and extrinsic factors that shape neocortical specification. Trends in Neuroscience,
24, 417–423.
Paysan, J., Bolz, J., Mohler, H. and Fritschy, J.-M. (1994) GABAA receptor a1 subunit, an early marker for area
specification in developing rat cerebral cortex. Journal of Comparative Neurology, 350, 133–149.
Paysan, J., Kossel, A. and Fritschy, J.-M. (1997) Area-specific regulation of gamma-aminobutyric acid type
a receptor subtypes. Proceedings of the National Academy of Science, U.S.A., 94, 6995–7000.

Peinado, A. (2000) Traveling slow waves of neural activity: a novel form of network activity in developing
neocortex. Journal of Neuroscience, 200, RC54.
Peters, A. and Jones, E.G. (1984) Cerebral Cortex Volume 1: Cellular Components of the Cerebral Cortex. New
York: Plenum Press.
Phillips, J.R. and Johnson, K.O. (1981) Tactile spatial resolution. II. Neural representation of bars, edges and
gratings in monkey afferents. Journal of Neurophysiology, 46, 1192–1203.
Pimenta, A.F., Reinoso, B.S. and Levitt, P. (1996) Expression of the mRNAs encoding the limbic system-
associated membrane protein (LAMP) 2. Fetal rat brain. Journal of Comparative Neurology, 375, 289–302.
Price, D.J. and Morgan, J.E. (1987) Spatial properties of neurones in the lateral geniculate nucleus of the
pigmented ferret. Experimental Brain Research, 68, 28–36.
Rado, R., Himelfarb, M., Arensburg, B., Terkel, J. and Wollberg, Z. (1989) Are seismic communication signals
transmitted by bone conduction in the blind mole rat? Hearing Research, 41, 23–29.
Rakic, P. (1988) Specification of cerebral cortical areas. Science, Washington, 241, 170–176.
Rakic, P., Suner, I. and Williams, R.W. (1991) A novel cytoarchitectonic area induced experimentally within the
primate visual cortex. Proceedings of the National Academy of Science, U.S.A., 88, 2083–2087.
Rauschecker, J.P. and Korte, M. (1993) Auditory compensation for early blindness in cat cerebral cortex.
Rauschecker, J.P. (1997) Mechanisms of compensatory plasticity in the cerebral cortex. In: H.-J. Freund,
B.A. Sabel and O.W. Witte (eds), Brain Plasticity. Philadelphia: Lippincott-Raven Publishers, pp. 137–146.
Rauschecker, J.P. (1999) Auditory cortical plasticity and sensory substitution. In: J. Grafman and Y. Christen
(eds), Neuronal plasticity: Building a Bridge from the Laboratory to the Clinic. Berlin: Springer-Verlag,
pp. 53–63.
Reale, R.A., Brugge, J.F. and Feng, J.Z. (1983) Geometry and orientation of neuronal processes in cat primary
auditory cortex (AI) related to characteristic-frequency maps. Proceedings of the National Academy of
Science, U.S.A., 80, 5449–5453.
Reale, R.A. and Imig, T.J. (1980) Tonotopic organization in auditory cortex of the cat. Journal of Comparative
Neurology, 192, 265–291.
Rebillard, G., Carlier, E., Rebillard, M. and Pujol, R. (1977) Enchancement of visual responses in the primary
auditory cortex of the cat after an early destruction of cochlear receptors. Brain Research, 129, 162–164.
Recanzone, G.H., Schreiner, C.E., Sutter, M.L., Beitel, R.E. and Merzenich, M.M. (1999) Functional organiza-
tion of spectral receptive fields in the primary auditory cortex of the owl monkey. Journal of Comparative
Neurology, 415, 460–481.
Reinoso, B.A., Pimenta, A.F. and Levitt, P. (1996) Expression of the mRNAs encoding the limbic system-
associated membrane protein (LAMP) 1. Adult rat brain. Journal of Comparative Neurology, 375, 274–288.
Reiter, H.O. and Stryker, M.P. (1988) Neural plasticity without postsynaptic action potentials: Less-active inputs
become dominant when kitten visual cortical cells are pharmacologically inhibited. Proceedings of the
National Academy of Science, U.S.A., 85, 3623–3627.
Rockland, K.S. and Lund, J.S. (1982) Widespread intrinsic connections in the tree shrew visual cortex. Science,
Washington, 215, 1532–1534.
Roder, B., Teder-Salejarvi, W. and Neville, H.J. (1999) Improved auditory spatial tuning in blind humans.
Nature, London, 400, 162.
Roe, A.W., Garraghty, P.E. and Sur, M. (1990) Terminal arbors of single on-center and off-center retinal ganglion
cell axons within the ferret’s lateral geniculate nucleus. Journal of Comparative Neurology, 288, 208–242.
Roe, A.W., Pallas, S.L., Hahm, J. and Sur, M. (1990) A map of visual space induced in primary auditory cortex.
Roe, A.W., Hahm, J. and Sur, M. (1991) Experimentally induced establishment of visual topography in auditory
thalamus. Society for Neuroscience Abstracts, 17, 898.
Roe, A.W., Pallas, S.L., Kwon, Y. and Sur, M. (1992) Visual projections routed to the auditory pathway in ferrets:
Receptive fields of visual neurons in primary auditory cortex. Journal of Neuroscience, 12, 3651–3664.
Roe, A.W., Garraghty, P.E., Esguerra, M. and Sur, M. (1993) Experimentally induced visual projections to the audi-
tory thalamus in ferrets: Evidence for a W cell pathway. Journal of Comparative Neurology, 334, 263–280.
Rosier, A.M., Arckens, L., Demeulemeester, H., Orban, G.A., Eysel, U.T., Wu, Y.J. et al. (1995) Effect of
sensory deafferentation on immunoreactivity of GABAergic cells and on GABA receptors in the adult cat
visual cortex. Journal of Comparative Neurology, 359, 476–489.
Rubenstein, J.L.R., Anderson, S., Shi, L., Miyashita-Lin, E., Bulfone, A. and Hevner, R. (1999) Genetic control
of cortical regionalization and connectivity. Cerebral Cortex, 9, 524–532.
Ruthazer, E.S. and Stryker, M.P. (1996) The role of activity in the development of long-range horizontal
connections in area 17 of the ferret. Journal of Neuroscience, 16, 7253–7269.
Rutherford, L.C., DeWan, A., Lauer, H.M. and Turrigiano, G.G. (1997) Brain-derived neurotrophic factor
mediates the activity-dependent regulation of inhibition in neocortical cultures. Journal of Neuroscience,
17, 4527–4535.
Sadato, N., Pascual-Leone, A., Grafman, J., Ibañez, V., Deiber, M.P., Dold, G., Hallett, M. (1996) Activation of
the primary visual cortex by Braille reading in blind subjects. Nature, London, 380, 526–528.

272 Sarah L. Pallas
Sato, H., Katsuyama, N., Tamura, H., Hata, Y. and Tsumoto, T. (1995) Mechanisms underlying direction select-
ivity of neurons in the primary visual cortex of the macaque. Journal of Neurophysiology, 74, 1382–1394.
Sato, H., Katsuyama, N., Tamura, H., Hata, Y. and Tsumoto, T. (1996) Mechanisms underlying orientation
selectivity of neurons in the primary visual cortex of the macaque. Journal of Physiology (London), 494,
757–771.
Schlaggar, B.L. and O’Leary, D.D.M. (1991) Potential of visual cortex to develop an array of functional units
unique to somatosensory cortex. Science, Washington, 252, 1556–1560.
Schneider, G.E. (1973) Early lesions of the superior colliculus: Factors affecting the formation of abnormal ret-
inal projections. Brain, Behavior and Evolution, 8, 73–109.
Schreiner, C.E. and Mendelson, J.R. (1990) Functional topography of cat primary auditory cortex: Distribution
of integrated bandwidth. Journal of Neurophysiology, 64, 1442–1459.
Schreiner, C.E., Mendelson, J.R. and Sutter, M.L. (1992) Functional topography of cat primary auditory cortex:
Representation of tone intensity. Experimental Brain Research, 92, 105–122.
Sengpiel, F., Gödecke, I., Stawinski, P., Hübener, M., Löwel, S. and Bonhoeffer, T. (1998) Intrinsic and environ-
mental factors in the development of functional maps in cat visual cortex. Neuropharmacology, 37, 607–621.
Sharma, J., Angelucci, A. and Sur, M. (2000) Induction of visual orientation modules in auditory cortex. Nature,
London, 404, 841–847.
Shatz, C.J. (1992) Dividing up the neocortex. Science, Washington, 258, 237–238.
Sherman, S.M. and Spear, P.D. (1982) Organization of visual pathways in normal and visually-deprived cats.
Physiology Reviews, 62, 738–855.
Sillito, A.M. (1975a) The contribution of inhibitory mechanisms to the receptive field properties of neurones in
the striate cortex of the cat. Journal of Physiology (London), 250, 305–329.
Sillito, A.M. (1975b) The effectiveness of bicuculline as an antagonist of GABA and visually evoked inhibition
in the cat’s striate cortex. Journal of Physiology (London), 250, 287–304.
Sonty, R.V. and Juliano, S.L. (1997) Development of intrinsic connections in cat somatosensory cortex. Journal
Sur, M., Garraghty, P.E. and Roe, A.W. (1988) Experimentally induced visual projections into auditory thalamus
and cortex. Science, Washington, 242, 1437–1441.
Sur, M., Pallas, S.L. and Roe, A.W. (1990) Cross-modal plasticity in cortical development: Differentiation and
specification of sensory neocortex. Trends in Neuroscience, 13, 227–233.
Suzuki, S.C., Inoue, T., Kimura, Y., Tanaka, T. and Takeichi, M. (1997) Neuronal circuits are subdivided by
differential expression of type-II classic cadherins in postnatal mouse brains. Molecular and Cellular
Tanabe, Y. and Jessell, T.M. (1996) Diversity and pattern in the developing spinal cord. Science, Washington,
274, 1115–1123.
Vitek, D.J., Schall, J.D. and Leventhal, A.G. (1985) Morphology, central projections and dendritic field
orientation of retinal ganglion cells in the ferret. Journal of Comparative Neurology, 241, 1–11.
von Melchner, L.S., Pallas, S.L. and Sur, M. (2000) Visual behavior induced by retinal projections directed to
the auditory pathway. Nature, London, 404, 871–875.
Wallace, M.N. and Bajwa, S. (1991) Patchy intrinsic connections of the ferret primary auditory cortex.
Walsh, C. (1993) Cell lineage and regional specification in the mammalian neocortex. Perspectives in Develop-
mental Neurobiology, 1, 75–80.
Wang, J., Caspary, P. and Salvi, R.J. (2000) GABA-A antagonist causes dramatic expansion of tuning in primary
auditory cortex. NeuroReport, 11, 1137–1140.
Weeks, R., Horwitz, B., Aziz-Sultan, A., Tian, B., Wessinger, C.M., Cohen, L.G., Hallett, M. and Rauschecker,
J.P. (2000) A positron emission tomographic study of auditory localization in the congenitally blind.
Weliky, M. and Katz, L.C. (1994) Functional mapping of horizontal connections in developing ferret visual
cortex: Experiments and modeling. Journal of Neuroscience, 14, 7291–7305.
Weliky, M. and Katz, L.C. (1997) Disruption of orientation tuning in visual cortex by artificially correlated
neuronal activity. Nature, London, 386, 680–685.
Weliky, M. and Katz, L.C. (1999) Correlational structure of spontaneous neuronal activity in the developing
lateral geniculate nucleus in vivo. Science, Washington, 285, 599–604.
Wiesel, T.N. and Hubel, D.H. (1963) Single cell responses in striate cortex of kittens deprived of vision in one
eye. Journal of Neurophysiology, 26, 1003–1017.
Wong, R.O.L., Meister, M. and Shatz, C.J. (1993) Transient period of correlated bursting activity during devel-
opment of the mammalian retina. Neuron, 11, 923–938.
Yaka, R., Yinon, U. and Wollberg, Z. (1999) Auditory activation of cortical visual areas in cats after early visual
deprivation. European Journal of Neuroscience, 11, 1301.
Zheng, W. and Knudsen, E. (1999) Functional selection of adaptive auditory space map by GABAA-mediated
inhibition. Science, Washington, 284, 962–965.

13 Do Primary Sensory Areas Play Analogous Roles
in Different Sensory Modalities?
Hubert R. Dinse1 and Christoph E. Schreiner2
1
Institute for Neuroinformatics, Dept. of Theoretical Biology, Group for Experimental
Neurobiology, Ruhr-University Bochum, Germany
2
Coleman Laboratory, W.M. Keck Center for Integrative Neuroscience, Sloan Center for
Theoretical Neurobiology, University of California San Francisco, San Francisco, USA
Correspondance: Hubert R. Dinse, PhD, Institute for Neuroinformatics,
Dept. of Theoretical Biology, Group Experimental Neurobiology, Ruhr-University Bochum,
ND 04, D-44780 Bochum, Germany
Tel: +49-234-32 25565; FAX: +49-234-32-14209
e-mail: Hubert.dinse@neuroinformatik.ruhr-uni-bochum.de
We compare the properties of primary auditory, visual and somatosensory cortex in order to discuss
whether the extraction and mapping of stimulus features shows equivalencies across sensory modal-
ities. We propose that modality-specific differences in “early” cortical properties are largely a
consequence of differences in receptor properties, and that early cortical processes and their imple-
mentation are similar across sensory modalities. This view is based largely upon the striking similar-
ities of receptive field organization found in visual, auditory and somatosensory areas with respect
to the spatio-temporal distribution of excitation and inhibition. By the same token, inspection of the
shape and properties of cortical point-spread functions suggest substantial equivalencies across
areas. Analysis of temporal aspects of sensory processing indicates that differences can be attributed
to differences in subcortical and peripheral processing, resulting mainly from anatomical constraints
and particularities. On the other hand, as exemplified by spontaneous activity and intrinsic oscillatory
activity, differences found between cortical layers or those found for state-dependencies, such as
wakefulness, attention or sleep, outweigh possible modality and area-specific differences. The common
feature of multiple parametric maps observed in early sensory areas might reflect a rather general
principle of cortical organization that allows the combination of local operations with a continuous
representation of elemental parameters of the environmental scene, maintaining local neighbourhood
relationships.
KEYWORDS: auditory, cortical maps, distributed processing, latencies, oscillations, parametric

maps, primary sensory areas, receptive fields, receptors, somatosensory, spontaneous activity,
temporal aspects of processing, temporal integration, thalamus, vision
1. INTRODUCTION
1.1. Role of Primary Sensory Cortices

What is the role of the representation and processing of the sensory environment in prim-
ary cortical fields? How do the principles underlying sensory processing differ among
cortical areas? These questions are significant, since the current dominant model of cortex
often uses the visual system as a frame of reference, thus, potentially biasing the general view.
273
274 Hubert R. Dinse and Christoph E. Schreiner
To address these questions, we made a comparison between different sensory modalities.

The structural framework in different sensory cortices is similar with respect to many
aspects (Rockel et al., 1980) such as the global arrangement of synapses, axons, excitatory and
inhibitory cell types, lamination, as well as vertical and horizontal connections (as discussed
in some detail in other chapters). This could lead to the conclusion that the common features
of cortical machinery implement algorithms, or a set of universal operations, rather than
unique, modality-specific operations. On the other hand, cortical subregions have specific
anatomical and physiological distinctions, and critical differences in their functional
attributes. The clearest distinctions between subregions of sensory cortex relate to the
modality-specificity of receptive field properties and the resulting functional organizations.
In fact, visual, auditory, and somatosensory cortex have functional characteristics and
parameter representations so different that it may seem unlikely that common cortical
processes and mechanisms can subserve such diversity (e.g. see Luria, 1973). However,
analysis of the functional properties of cortical neurones, in conjunction with the modality-
specific subcortical processes reveals that many sensory cortical attributes reflect
analogous, if not identical, local and distributed processing mechanisms. Each modality
imposes some specific constraints on these processes and on their functional cortical
representation. From a global perspective, we propose that primary sensory cortex in differ-
ent modalities performs many of the same tasks. The underlying processes are based
on analogous operations in different modalities, namely, activity- and correlation-based
integration of convergent information by means of well-coordinated excitatory and inhibi-
tory interactions.
1.2. Hypothesis
Different sensory cortices perform similar functions based on common rules in processing
elements and mechanisms. Substantive modality-specific differences in peripheral signal
properties are transformed subcortically to provide functional equivalency at the cortex.
In the cortex, homogeneity of processing is evident in the spatio-temporal arrangement of
excitatory and inhibitory interactions. Receptor-specific characteristics that remain at the
cortical level are implemented by local arrangements, and not by the basic principles of
the processing which is applied.
1.3. Synopsis
We discuss first the parallels and differences between vision, touch, and audition for
signal properties of the environmental scene, receptors and receptor organs, and principles
of subcortical processing. Here we will emphasize some properties of the auditory system,
in order to highlight modality-specific differences in subcortical processes. We, then,
compare properties of cortical processing with respect to modality-specific aspects of
peripheral processing. We next assess receptive field properties, cortical maps and tem-
poral processing in three primary areas—AI, VI, and SI—to test the notion of a unifying
principle of early cortical processing. Next, we summarize experimental findings of cross-
modal plasticity that support the idea of general, rather than modality-specific, cortical
operational principles. We conclude by discussing briefly some potential consequences
for future studies of sensory processing.

Functional Equivalence of Sensory Areas 275
2. MODALITY CHARACTERISTICS OF SIGNALS AND

SUBCORTICAL SYSTEMS
2.1. The Environmental Scene

Sensory organs scan the surrounding environmental scene in different physical realms,
such as electromagnetic waves, air-pressure waves, and mechanical forces. Our inter-
action with the environment depends critically on the identification of objects within it.
Delineation of objects, and formation of frame-independent representations of objects
enables us to learn their effects on us, to attach behavioural significance to these objects,
and to interact with them (Koffka, 1935; Gibson, 1966). The non-chemical senses—
vision, touch, and audition—all have to solve the same problem: analysis of the environ-
mental scene, or identification of objects in complex backgrounds. Characteristic and
contingent features in the scene need to be integrated so that object recognition and back-
ground perception can occur. For example, features defining a picture of a cat, a sculpture
of a cat, or the word “cat”, for vision, touch and hearing, respectively, each need to be
deconvoluted, coded, analyzed, identified and unified. “High level” processes that deal
with the actual analysis of the scene, rely on higher-order correlations in some feature-
space and on the comparison with previously stored information about the properties
of objects, backgrounds, and events. “Low level” processes, in contrast, extract the
information from the receptors, to construct the appropriate feature-space in which the
separation of object and background can occur. We argue that the former processes
are largely accomplished subcortically, and in the primary sensory cortices. They rely on
lower-order correlations among spatial and temporal domains within and across different
receptor types and sensory surfaces. The rest of this section refers only to these low-level
processes.
2.2. Signal Statistics

Visual, tactile, and auditory objects have a number of properties in common such as edges
(definable in space or time), and attributes of their internal features such as colour, texture,
and coherent or incoherent frequency and/or amplitude modulations. Object motion
relative to the environment, and across the receptor surface, provides data optimal for the
extraction of relationships between different object properties. Different physical stimulus
dimensions are represented by the nature of receptor organs. Each modality represents the
lower-order statistical nature of the signals. Acoustic, somatosensory and visual ecologies
share statistical similarities, and embody important differences. Thus, objects that reflect
light may also generate and reflect sound, so that the spatial statistics of acoustic and
visual environments may have similar scaling behaviour. On the other hand, differences
between the visual and acoustic worlds exist: Objects that are visually opaque can be
acoustically transparent. In combination with the different generation mechanisms for
light and sound, acoustic and visual ecologies may have some different statistical pro-
perties that, consequently, entail distinct differences in sensory processing requirements.
However, the resulting representational differences of an object by several modalities have
to be brought into alignment at some level of processing, in order to allow multi-sensory
integration which ensures perceptual unity and consistency in decision-making and
behaviour (Stein and Meredith, 1990).

The statistical properties of the modality-specific environments can influence two

major aspects of neuronal coding: what is represented and how is it encoded (e.g.
deCharms and Zador, 2000). The what-aspect is reflected in the nature of the representational
dimensions in feature space, as discussed above. Receptive field analysis (see section 3,
below) reveals which features a neurone responds to. Each neurone represents specific
values along several feature dimensions, depending on the probability of the occurrence of
that stimulus aspect, as well as on the behavioural relevance of that range of values. An
important influence of the stimulus statistics on receptive fields is demonstrated in the
effect of temporal correlation among independent stimuli. An illuminating example is the
effect of the artifical fusion of neighbouring digits of a hand (Clark et al., 1988) on its
cortical representation. Normally, each finger is mapped in primary somatosensory cortex
by non-overlapping receptive fields, that is, the neurones respond to stimulation of one
finger alone. After surgical fusion of the skin of two adjacent fingers and extended use of
the fused fingers by the monkey, the cortical receptive field boundaries cover both fingers,
i.e. the normal discontinuity in the representation of the fingers is altered. This change in
receptive field property relates to the high degree of spatio-temporal, correlated input to
the fused fingers compared to the uncorrelated input to the fingers when they are used
independently. In general, the extraction of spatio-temporal correlations and coherences
across the multi-dimensional stimulus space is a critical principle, and underlies many
aspects of the generation and representation of feature dimensions related directly to
object properties.
Stimulus-response analysis can determine how a cell responds to a stimulus and helps
one to understand how stimulus information is coded by the spike train. Approaches from
information theory allow one to estimate the precision of the neuronal code, and the quality
of the decoding scheme used by neurones and their networks (Nicolelis, 1996; Zhang and
Sejnowski, 1999; Doetsch, 2000). This defines how environmental information is coded.
Several studies (reviewed in Bialek et al., 1991; Atick, 1992) have shown that early visual
stations, such as the retina and the lateral geniculate nucleus, are optimized—in an
information-theoretic sense—for transmitting information about scenes that have natural
spatio-temporal statistics. Psychophysical image-discrimination experiments in humans
find that performance is best when the pictures have natural second-order spatial statistics
(Parraga et al., 2000). Similar conclusions have been drawn for the auditory system using
a mutual information metric for stimulus and spike trains. Attias and Schreiner (1998)
found that neurones in the inferior colliculus of the cat code naturalistic stimuli more
efficiently than stimuli with non-naturalistic distributions. Similar results have been found
in the auditory systems of frog (Rieke et al., 1995) and song-bird (Theunissen and Doupe,
1989). In the somatosensory system problems in evaluating the statistical properties of
tactile stimuli have slowed progress.
In combination, these considerations illuminate the idea that many aspects of the
natural statistics of the environmental scene are imposed upon, and constrain sensory
information processing. Some of these statistical constraints can be quite similar for
optical, acoustical, and mechanical aspects of objects, and it is not unreasonable to
assume that the neuronal processes underlying vision, audition, and somatic-sensation
have developed similar mechanisms to exploit these properties for central nervous
representation and coding. Conversely, a number of stimulus characteristics remain very
much modality-specific and require special processing strategies to utilize and integrate
that information.

Table 13.1.
Parameter Audition (A) Touch (S) Vision (V)
Dimensionality 1 2 2
Laterality + – +
Receptor Types 1 >4 2
Latency short short long
2.3. The Receptors

All receptors have in common the fact that they encode intensity, duration and location of
external stimuli. The most obvious distinction between the different sensory modalities
are the different designs and spatial arrangements of the receptors that underlie respon-
siveness to a specific form of physical energy and its transformation into neuronal activity.
The auditory system uses essentially a single receptor type, the inner hair cell, to convey
information to the central nervous system about air pressure changes. The hair cells are
arranged in a single line along the cochlea, that is, the receptor sheet is one-dimensional.
Since the ear is a paired organ it provides laterality-specific information to assess
correlated activity between the two sides. The eyes are also paired organs. The retina is
a two-dimensional receptor sheet with two receptor types (rods and cones). The somato-
sensory system has a two-dimensional receptor sheet, and uses many more receptor types
than vision and audition. More than ten receptor-types serve its function, with four recep-
tors for mechano-sensation alone.
While each modality is highly specific, each modality also shares some properties with
the other receptors. The main correspondences are in the dimensionality of the receptor
sheet (S = V); in the laterality (A = V); and in the latency (A = S) domain. These parallels
imply that each modality has a unique feature that distinguishes it from the others. The
receptor-level based differences between the modalities create distinct processing regimes
that exploit these properties to optimize feature dimensions for object-oriented scene
analysis. The differences in these optimized processing schemes are reflected in the
organization and function of the subsequent processing stages. However, the resulting
feature maps, despite their many transformations, conserve the plan of the primary receptor
sheet, at least up to the primary cortical level.
2.4. Sub-Thalamocortical Processing

Structural design, cell types, local circuits and function of the sub-thalamic processing
stages, as well as the number of synapses between receptor and cortex differ, between the
three modalities:
Vision: Retina –> Thalamus –> Cortex

Touch: Spinal cord –> Brain stem –> Thalamus –> Cortex
Audition: Spiral Ganglion –> Medulla/Pons –> Midbrain –> Thalamus –> Cortex
In the visual system, several transformations take place before the information arrives
in the primary visual cortex. Within the retina, the two receptor types interact with several

local-circuit cells, and give rise to at least two (and likely more) anatomically and func-
tionally distinct pathways. The local circuit cells are a common feature in all modalities,
where they modify sensory processing and recombine information internally. The preferred
stimulus information conveyed by the two main visual streams differs in various ways,
including colour information, receptive field size, stimulus contrast, and temporal aspects
(Ungerleider and Mishkin, 1982; Livingstone and Hubel, 1988; Maunsell, 1992; Merigan
and Maunsell, 1993; Ungerleider and Haxby, 1994; Hendry and Reid, 2000). In higher
mammals, the stream from the retinal ganglion cells reaches the thalamus directly, which
then projects to the primary visual cortex. Thus, the main sub-thalamic station which
introduces functional varieties and distinctions is at the retinal level, without further
brainstem or midbrain processing.
In the somatosensory system, many receptor types encode tactile information. Their
pathways remain parallel (that is, they do not converge), as they relay discriminable sensations,
including fine touch, vibration, pain and temperature. The slowly adapting system, for
example, is essential for tactile form recognition. This division of labour establishes, from
the outset, a multidimensional representation of the tactile environment. Overall, the
pathways from the receptors to the primary sensory cortical areas—via one brainstem
station and the thalamus—provides, through considerable computational effort, a stepwise
transformation from an isomorphic representation to a distributed one, in primary
somatosensory cortex.
In the auditory system, an even larger amount of task-oriented processing is accom-
plished below the thalamo-cortical level. Many brainstem stations and one obligatory
midbrain nucleus are involved before the information reaches the thalamus and cortex. In
addition to frequency coding and information about timing, which originate directly from
the activity patterns of the receptor sheet (the Organ of Corti in the cochlea), at least three
other basic features have been computed by the time the information reaches the thalamus:
Sound localization: This dimensions uses differences in timing, and in spatial distri-
bution of response strength between the receptor surfaces of the two ears to code
spatial position. Unlike the visual and somatosensory system, a direct correspondence
between activation of the receptor surface and object position in the external world is
not possible. However, the position of an auditory object relative to the position of the
two ears, in combination with the physics of sound propagation around the head,
evokes unique activity patterns in the two cochleae. Differences in response timing
and intensity between the two sides are used to compute the location of a sound
source. This computation is capable of locating external sound sources in coordinates
that are compatible with those of the other sensory modalities. Thus, in the superior
colliculus, auditory space and visual space maps are superimposed (Middlebrooks
and Knudsen, 1984).
Spectral integration: In hearing, the extent of spectral integration, or, equivalently,
spatial integration in relation to the receptor surface, varies from narrow to broad.
However, subsets of midbrain and cortical neurones display properties that are similar
to the psychophysical properties of “critical bands” including an integration bandwidth
that is insensitive to intensity changes (Ehret and Merzenich, 1988; Ehret and Schreiner,
1997). This property underlies phenomena, such as loudness perception and stimulus
discrimination. While it is not understood how this property is extracted, it is not only
a consequence of receptor and receptor-surface interactions, but must be generated in

subsequent stations. By the midbrain, this process is essentially completed (Ehret and
Merzenich, 1988).
Periodicity: The analysis of coherent temporal modulations across the receptor
surface underlies the identification of sound sources (e.g. a particular speaker), the
distinction between simultaneously active sound sources (“cocktail party effect”),
and it contributes to the perceptual grouping of acoustic objects (ranging, for example,
from isolated vowels to words). While the neuronal basis of periodicity analysis is not
understood fully, it involves several steps and is largely accomplished subcortically
(Langner, 1997).
The many stations of the auditory brainstem probably reflect the number and computa-
tional complexity of feature dimension required before reaching the cortex. In contrast, the
many receptor types in the somatosensory system, and the complex local retinal processing
schemes in vision seem to accomplish the task of generating sufficient feature dimensions
without equally extensive brainstem processing. Overall then, the sub-thalamocortical
stations compute several aspects from the receptor surface activity, that are optimized for
representing and analyzing the environment in a framework that allows further refinement
and alignment of the various representations of the external world.
2.5. Thalamus
The thalamus is an obligatory nucleus for all three modalities. It has subdivisions in each
modality that segregate sensory information prior to cortical processing, and it is a hub for
descending (cortico-thalamic) control. The thalamic nuclei have relatively few cell types,
mainly one type of output neurone and one or two types of local interneurone. It often is
referred to as a relay nucleus, though this is an oversimplification since each nucleus
modulates or transforms the signal passing through it. Indeed, the projections from sub-
thalamic stations can remain anatomically and physiologically distinct. Examples of main-
tained segregation and pathway-specificity include magnocellular (fast-conducting) and
parvocellular (slow-conducting) divisions in vision and audition (Hendry and Reid, 2000);
tonotopic vs non-tonotopic pathways in audition (Winer, 1992); information for eye and
ear laterality; cutaneous representation of face and body; slowly and rapidly adapting
responses to cutaneous stimulation; and processing of discriminative sensation (high
resolution) versus epicritic sensation (low resolution but with urgent behavioural qualities),
such as pain and temperature sense. A possible reason for such segregation is that the
different “channels” need to be optimized spatially and temporally at the cortical input
level, for instance for binocular fusion, or higher spectral integration. From this point of
view, the sensory thalamic nuclei may serve as staging areas for cortical processing.
An important functional property of thalamic neurones is their high synaptic security,
which enables reliable transfer at high temporal precision. Thus, thalamic neurones follow
stimulus repetition rates up to >100 Hz with time-locked responses, in somatic sensation
and audition. Thalamic function may be less related to local receptive field modification
and feature extraction then to state-dependent modulation and gating of activity forwarded
to the cortex (for reviews see: Casagrande and Norton, 1991; Winer et al., 1999). Strong
modulatory influences from cortical feedback, and many extra-cortical sources have been
described (e.g. Steriade and Llinas, 1988). Functionally, such complex modulatory systems
may provide state-dependent synchronization or de-synchronization among different

cortical inputs (Traub et al., 1999), attentional amplification of receptive field properties
(Zhang et al., 1997; Yan and Suga, 1999), and, overall, an affective colouring of the sensory
experience. These modulatory effects are similar across modalities, suggesting a common
operating principle at this gateway to cortex. This also suggests that the content and
configuration of the modality-specific information in the thalamus is at an equivalent level
of abstraction and integration. Thus, consequences of the modulatory processes imposed
on the sensory stream at the thalamus operate on the same level of processing for the three
modalities. For an excellent survey of thalamic function, see the review by Sherman and
Guillery (2001).
3. MODALITY CHARACTERISTICS OF PRIMARY CORTICAL AREAS
3.1. Receptive Fields

Since its introduction more than 60 years ago, the concept of receptive fields (RFs) has
constituted a powerful tool for the analysis of sensory systems. Hartline (1938) wrote:
“The region of the retina which must be illuminated to produce a response in a particular
nerve cell is termed the receptive field of that cell”. Although we perceive the world as
unitary, the neuronal elements can analyze only small portions of the environment. Knowing
structure, organization and properties of RFs is indispensable for understanding their
specific contribution to signal processing and their role in brain function. Consequently,
receptive fields are widely used to define and map cortical neural activity into a parametric
feature-space derived from either stimulus or computed variables. This analysis allows the
characterization of the processing capacities of single cells. Historically, RF properties
have been described in terms of “feature-detectors” and filters operating hierarchically at
increasing levels of complexity and specificity. Below, some extensions and revisions of
this original scheme are discussed. Figure 13.1, depicts characteristics of single cell
receptive fields in visual, auditory, and somatosensory cortex. These differ greatly in
shape, dimension, and dimensionality. In all of these examples, activity is represented
systematically in the parametric space of the respective modality, such as the visual field,
the “frequency space”, and the skin location. More distinct and refined RF forms emerge
in so-called “functional” spaces embedded in the respective modality. Here, neural activity
is plotted as a function of a graded variation of a selected stimulus parameter, and the
ensuing RFs are often referred to as “tuning curves”. This approach allows both quantitative
and qualitative analyses of an example of neuronal sensitivity to a given stimulus. Well-
known examples include orientation tuning in vision, intensity tuning in audition, and
grating resolution in somatic sensation.
Is there a common principle of RF organization despite these dissimilarities? By defin-
ition, RFs exist in the parametric space of the respective modality. Thus, the “coordinates”
must be different and depend on the parameters under investigation. Inevitably there are
substantial and substantive differences, since parameter spaces differ across modalities.
Closer inspection of RFs and tuning curves across modalities reveals at least one com-
mon aspect of RF organization: integration of excitatory and inhibitory inputs. Accordingly,
the question of a possible common type of RF organization can be resolved by analysis of
the spatial distribution and interaction of excitation and inhibition. Figures 13.2 and 13.3
show examples from visual, somatosensory, and auditory cortex. Although the RF structures

cortical representation cortical representation

of visual hemifield of skin surface
VI SI
receptive fields
receptive fields
visual hemifield skin surface
cortical representation
of cochlea
AI
receptive fields
cochlea
Figure 13.1. Examples of receptive fields (RFs) depicted in parametric space recorded in visual (top left),
somatosensory (top right) and auditory cortex (bottom). As a rule, RFs in all modalities cover a certain
circumscribed area in visual field, skin surface or frequencies, respectively, thereby providing a “window” to
the outside world. Receptive fields are mapped by inserting microelectrodes, into the cortex (usually the
middle layers), to record action potentials from single cells or multiple unit activity from small clusters of
neurones. The receptive field is defined as that region on the sensory surface where stimulation evokes
action potentials. This procedure maps activity recorded in the cortex into the stimulus space, which allows
an easy and systematic way of parametric analysis. When moving the electrode to an adjacent location in the
cortex, a systematic shift in the corresponding receptive field location will be encountered. A complete
topographic map can be obtained when a large number of electrode penetrations is combined in such a way
that the penetration coordinates are related to the corresponding receptive field coordinates. The inverse
approach is taken when cortical activity distributions are measured. In contrast to the above, a fixed stimulus,
ideally a small, “point-like” stimulus, is applied, and the entire activity in the cortex evoked by that stimulus
is measured. This type of activity distribution is often referred to as “point spread function—PSF”. Techno-
logies often employed for this kind of analysis are optical imaging and fMRI. However, it should be noted
that the PSF can be obtained using microelectrodes. In this case, single or multiple neurone activity evoked
by the “point-like” stimulus is recorded, and its spatial distribution is derived from a systematic mapping
at different locations. In theory, RFs and point-spread functions are the corresponding counterparts of
a mapping rule that describes how input is represented in a topographic map. In practice, however, due to
differences in threshold and due to particularities in methodological constraints, the two descriptors of
cortical representations may yield different results.

A B C D
E F G H
Figure 13.2. Two-dimensional RF profiles recorded in somatosensory cortex (top) and visual cortex (bottom).
Top: RF structures observed in the hand representation of monkey area 3b. Black indicates excitation, white
suppression. The sample RFs are meant to illustrate the wide range of combinations of excitatory and inhibitory
areas. A–D and G, H: single inhibitory regions located on the trailing side (i.e. towards the distal aspects of the
hand) of the excitatory RF region. E and F: two regions of inhibition on opposite sides of the excitatory region.
The line through each RF passes through the excitatory and inhibitory peaks (modified according to diCarlo
et al., 1998). Bottom: Two-dimensional response profiles of 2 typical simple RFs to illustrate variability in sizes,
shapes, and placements of individual subregions (modified according to Jones and Palmer, 1987). The letter
above each RF corresponds to the one-dimensional profiles shown in Figure 13.3. Reproduced with permission.
vary widely across modalities, they typically have a central region of excitation flanked by
surrounding, or offset inhibitory regions (Figure 13.2). The two-dimensional nature of
both retina and skin surface makes it challenging to distinguish visual from somatosensory
RFs, because the structure of excitatory and inhibitory subregions in RFs of VI and SI

A B C D
E F G H
A B A
BF
1 2 4 16 32
KHz
C D B
0.5 1 BF 8
KHz
Figure 13.3. One-dimensional RF profiles recorded in somatosensory (top), visual (bottom left) and auditory
(bottom right) cortices. The plots display the relative intensities of the excitatory and inhibitory components.
Scale bars in bottom left panel indicate 1 degree. BF: best frequency. The letter above each RF in the top and
bottom left panels corresponds to the two-dimensional profiles shown in Figure 13.2. Modified according
DiCarlo et al. (1998) (top), Jones and Palmer (1987) (bottom left), and Shamma and Versnel (1995) (bottom right).
Reproduced with permission.
neurones is quite similar (Jones and Palmer, 1987; DiCarlo et al., 1998). An analogous
sub-field structure has been described for auditory cortical neurones (Shamma and Versnel,
1987). However, due to the one-dimensional nature of the receptor surface of the cochlea,
the RF structure found in AI is also one-dimensional, with inhibitory ‘sidebands’. A striking
resemblance between RF structures of all three modalities emerges in one-dimensional
RF-profiles (Figure 13.3). This demonstrates a common principle of sensory processing,
with a complex spatial arrangement of excitatory and inhibitory processes that, together,
appear crucial for processing performed in each area.
A next logical step in discerning similarities in RF properties is to study RF organization
directly in cortical coordinates, i.e. in the spatial distribution of dendritic activation.
Unfortunately, because of technical difficulties, there is very little information available. It
has been hypothesized that the often-dramatic forms of functional selectivity might arise
from asymmetries and anisotropies at a cellular level. However, in spite of the improvement
of anatomical staining techniques, this issue has not been sufficiently resolved. Early
papers reported a substantial correlation between cell shape and functional cell types (i.e.
“simple” and “complex”) in the visual cortex (van Essen and Kelly, 1973; Lin et al., 1979).

In a subsequent study, the dendritic arborizations of horseradish peroxidase-filled cells

were reconstructed in three dimensions. There was no consistent relationship between
orientation selectivity and the tangential bias of the dendritic tree. The width of the receptive
fields was compared to the equivalent “width” of the tangential extent of the dendrites,
and there was no significant relationship between the two widths. Accordingly, the tangen-
tial arrangement of the dendritic field does not appear to be important in determining the
orientation selectivity, or the size of the receptive fields of neurones in the cat visual cortex
(Martin and Whitteridge, 1984). Because of its peculiar morphology, the Meynert cell
type has been suspected to mediate “direction” selectivity (Livingstone, 1998). In her
model, excitatory synapses are activated sequentially along the asymmetric basal dendrites
of the large pyramidal cells of Meynert. However, recent modeling studies indicated that
even when the electrotonic asymmetries in the dendrites were extreme, as in cortical
Meynert cells, the biophysical properties of single neurones could contribute only partially
to the directionality of cortical neurones (Anderson et al., 1999). Accordingly, the authors
suggested that most of the computation of direction of motion, over the range of velocities
observed, must rely on network mechanisms, most probably using the local recurrent
circuits of cortex.
In a series of studies, extracellular and intracellular recordings were made from neurones
in the cat visual cortex, in order to compare the subthreshold membrane potentials, reflect-
ing the input to the neurone, with the output from the neurone seen as action potentials
(Douglas et al., 1988, 1991; Martin, 1988). The authors failed to find direct experimental
evidence for the hypothesis that the selectivity of visual cortical neurones depends on
shunting inhibition. More generally, they concluded that the intracellular recordings do
not support models of directionality and orientation that rely solely on strong inhibitory
mechanisms to produce stimulus selectivity. Taken together, most of the selectivity
observed in cortical sensory areas may be functional rather than structural.
Other aspects of cortical processing may be shared by all modalities. These aspects
relate to the crucial role of context and nonlinearities. Neurones in primary visual cortex
have been characterized with respect to key physical features such as visual field location,
orientation, motion direction, ocular dominance, and spatial frequency. These approaches
allowed the analysis of neural representations within parameter spaces that are explicitly
defined by physical stimulus attributes. However, many visual illusions, such as the
perception of illusory contours (Kanizsa, 1976; von der Heydt et al., 1984; Ramachandran
et al., 1994; Sheth et al., 1996; Mendola et al., 1999), indicate that the visual system must
contain representations within parameter spaces without a physical counterpart. This
agrees with the observation that single neurones exhibit complex, non-predictable behav-
iour, dependent on stimulus context (for review see Gilbert et al., 2000; Kapadia et al.,
2000). The complex spatio-temporal response properties are plastic, and can be altered by
stimuli outside the receptive field centre, or even outside the classical receptive field
(Allman et al., 1985; Dinse, 1986; Gilbert and Wiesel, 1990; Sillito et al., 1995). Horizontal
circuits within the cortex contribute significantly to local cortical processing (Bolz and
Gilbert, 1989). Thus, cortical representations of peripheral activity deviate significantly
from a simple feedforward re-mapping of sensory space (Ferster and Miller, 2000).
Particularly for the visual system, complex subsystems dedicated either to form or motion
processing have been identified (Ungerleider and Mishkin, 1982; Livingstone and Hubel,
1988; Maunsell, 1992; Merigan and Maunsell, 1993; Ungerleider and Haxby, 1994).
These pathways originate in the retina and can be traced serially in the visual pathway

up to functionally specific high-level areas. Similar anatomical substrates may exist for
the other modalities. However, the actual degree of functional segregation or cross-talk
between such subsystems is unknown (Romanski et al., 1999; Recanzone et al., 1999).
In any event, the concept of two basic elements of sensory processing—form and
motion—holds across modalities. Form is derived from two-dimensional representation
of objects, and is best understood in the visual and somatosensory systems. In the audi-
tory system, form contains the one-dimensional frequency space, as well as the time axis,
in order to accommodate the spectral-temporal extent and structure of events such as
vocalizations.
Objects move relative to the receiver and the background. Movement is represented as
a systematic spatio-temporal displacement. Many lines of evidence suggest that the
analysis of such displacements is crucial for the processing of objects (Metzger, 1932;
MacKay, 1958; Reichardt, 1961; Burr, 1980). In both the visual and somatosensory modal-
ities, perceptual illusions of apparent motion exist (Ramachandran and Anstis, 1983, 1986;
Geldard and Sherrick, 1972; Kirman, 1974; Evans and Craig, 1991). In the auditory system,
two types of movement can be distinguished: movement across the receptor surface and
movement in external space. Frequency sweeps are directly analogous to visual and somato-
sensory motion with reference to the receptor surface, and result in similar perceptual illu-
sions (e.g. Warren, 1970; Bregman, 1990). However, frequency sweeps that are part of an
auditory object (e.g. in formant transitions of stop-consonants, or in diphthongs) are not
necessarily equivalent to motion of external objects. Analysis of the motion of an external
object in the auditory sense involves a complex interplay of spectral and temporal changes
across the receptor surfaces of both ears. Perhaps, the neural machinery for the specialized
processing of motion shares similar algorithms across modalities such as directional
selectivity and temporal correlation/sequencing of activity across long distances of the
receptor surfaces.
From these considerations of response profiles, it follows that RF characteristics, as
viewed in a modality-specific framework can obscure the underlying mechanisms, which
can be expressed in more general terms, spanning across modalities, such as spatial and
temporal interactions of the distributions of excitatory and inhibitory responses.
3.2. Spatial Processing

3.2.1. Cortical maps
Early investigators of sensory cortical areas agreed that these areas re-map certain
aspects of the external world, thereby preserving the local neighbourhood relationships in
the environment. These representations are known as retinotopic (visual), somatotopic
(touch) and cochleotopic (auditory) maps. All constitute systematic parametric represen-
tations across cortical space. Given the differences in the respective receptor arrays of the
retina, skin and cochlea, these maps differ accordingly in design, and are not a direct,
one-to-one representation of the world. This is particularly obvious on a fine spatial scale,
where the considerable scatter of RF position is larger than, or in the same range as the
required systematic shifts due to a topographic gradient in the map (Hubel and Wiesel,
1962; Albus, 1975; Sutter and Schreiner, 1991). On a larger scale, a clear topographic grad-
ient is present, though distorted. The retinotopic gradient of the cortical map of the visual
field is overlaid by other “functional maps”. Such maps contain an orderly arrangement

of stimulus attributes for portions of the respective retinal locations (Hübener et al.,
1997; Kim et al., 1999) but also exhibit discontinuities that result from the discrete
organization of several of such attributes (Das and Gilbert, 1997). Visual functional maps
for orientation of moving gratings (Blasdel and Salama, 1986; Swindale et al., 1987;
Bonhoeffer and Grinvald, 1991), direction of motion (Weliky et al., 1996), and the
spatial frequency of a moving grating (Shoham et al., 1997; Kim et al., 1999; Issa et al.,
2000) are now available and undoubtedly yet more such examples will be discovered as
our knowledge of cortical representation deepens. In the auditory cortex, there is also
evidence for multiple functional maps, that transcend simple frequency representations
(Schreiner, 1998; Schreiner et al., 2000). Thus, maps of sharpness of tuning, preferred
intensity, direction of FM sweeps, and onset latencies have been suggested.
A comparable system for the somatosensory cortex is the spatial segregation of slowly
and rapidly adapting mechanoreceptors which form an “overlay” to the general somato-
topic cortical map. However, evidence is still scant for other parameter maps analogous
to those in audition and vision.
Besides these feature maps, in modalities with a paired arrangement of receptor organs
(vision and hearing), there are maps for the inputs from the two organs: ocular dom-
inance maps (Wiesel et al., 1974; LeVay et al., 1978) and binaural bands (Imig and
Adrian, 1977; Middlebrooks et al., 1980). No such lateralization maps are known for the
somatosensory system. Interestingly, “disparity maps” have been described as an example of
higher-order functional maps, that combine the information from the lateralization maps
to form an additional parameter space (Burkitt et al., 1998). A similar situation may exist
for the auditory cortex where laterality information is readily expressed in binaural bands,
whereas explicit maps of spatial information are less obvious (Furukawa et al., 2000).
As noted above, parametric representations must reflect differences of the stimulus
spaces. Thus, all maps known in the different modalities differ decisively from one
another. A conventional way to obtain insight into the cortical coordinates of these maps is
to study the spatial distribution of activation patterns in cortex. The activity pattern can be
used to derive the so-called “cortical point spread function” (Fischer, 1972; Capuano and
McIlwain, 1981; McIlwain, 1986). In contrast to functional maps, cortical activity is not
represented in a parametric space, either by topography or in the stimulus space, but is
directly recorded in cortical dimensions (cf. Figure 13.1). Due to technical restrictions,
single electrode penetration maps can only reveal an approximate or incomplete picture of
the point spread function (PSF). As new imaging techniques—such as optical imaging or
the recent development of non-invasive imaging technology for humans such as PET or
fMRI—emerge we can derive spatially continuous forms of the PSF, although the meas-
urement itself is necessarily indirect because of the still-uncertain correlation between
neural activation and cortical metabolism.
PSFs obtained from activity patterns recorded in VI, SI and AI under comparable
imaging conditions with simple stimuli (small squares of light, small indentations of the
digit skin surface, and tone bursts, each stimulus applied at moderate intensity) share some
properties (see Figure 13.4). As a rule, the simple stimuli used are meant to be an experi-
mentally-feasible equivalent of a “point” on the receptor surface. Although the different
stimuli used are not qualitatively scalable, they activate only a small portion of the recep-
tor surfaces. While a direct comparison of absolute size of the PSFs is not possible due to
the incomensurate nature of the different stimuli, several basic characteristics of PSFs are
shared across modalities:

AI SI VI
Figure 13.4. Point-spread functions, recorded by means of optical imaging in auditory (left), somatosensory
(middle), and visual (right) cortices. AI: tone burst stimulation at 6 kHz delivered at 40 dB SPL, cat primary audi-
tory cortex, scale bar 1 mm. SI: cutaneous stimulation of digit 3 of rat hindpaw, probe diameter, 1 mm, indenta-
tion, 250 to 500 µm, rat primary somatosensory cortex, scale bar 1 mm. VI: square of light (0.4 deg visual
angle), flashed for 25 ms, luminance 0.9 cd/m2 against background of 0.002 cd/m2, cat primary visual cortex,
scale bar 0.5 mm. AI modified according to Dinse et al. (2000), SI and VI unpublished data of B. Godde, T. Hilger
and H.R. Dinse. (see Color Plate 5)
1. size: PSFs recorded in primary cortices involve a great portion of the total primary
area, irrespective of simple, spatially or spectrally restricted stimuli (Grinvald et al.,
1994; Das and Gilbert, 1995 (VI); Godde et al., 1995; Chen-Bee and Frostig, 1996
(SI); Bakin et al., 1996; Dinse et al., 1997, 2000 (AI)).
2. symmetry: PSFs usually are not simply concentric distributions of activity around the
core of the stimulation. Rather, they are spatially highly asymmetric.
3. compactness: PSFs are patchy, indicating a non-monotonic decay of activity with dis-
tance from the center of stimulation.
While the PSFs recorded in VI and SI are rather similar, the PSFs recorded in AI differ
from the others in terms of their shape (cf. Figure 13.4). PSFs for VI and SI are circular, in
line with the two-dimensional receptor array arrangement. In contrast, PSFs recorded in
AI have an elliptic shape. This might reflect the constraints of the one-dimensional cochlear
receptor array, the specific central projection pattern of auditory nerve fibres (Brown and
Ledwith, 1990), and specific central circuits. As in the RF comparisons, PSF analyses reveal
many similarities once the differences in receptor sheet dimensionality are considered.
3.2.2. Topography and distributed activity

Functional and structural segregation into subregions (patchiness) represents another
common property of sensory cortical fields. Many thalamo-cortical and cortico-cortical
projections target several, non-contiguous regions. Correspondingly, neurones representing
similar functional parameters also tend to cluster in small subregions rather then exhibiting
smooth spatial gradients. Such patchiness has been used as evidence against the existence of
strict parameter gradients in the various sensory cortical fields. However, considerations of
self-organizing models (e.g. Kohonen and Hari, 1999; Swindale, 2000) suggest that both
topographical gradients and local patches are necessary consequences of self-organizing
algorithms optimized for representing several behaviourally relevant dimensions of environ-
mental scenes. These functional maps are overlaid so as to ensure that many, if not all,

combinations of the different parameters are represented in cortex. Recent theoretical studies
show that geometrical factors do not constrain the ability of the cortex to represent combina-
tions of parameters in spatially superimposed maps of similar periodicity. Considerations of
uniform coverage suggest an upper limit of six or seven maps. A higher limit, of about nine
or ten, may be imposed by the numbers of neurones or minicolumns available to represent
each feature within a given cortical microdomain (Swindale, 2000; Swindale et al., 2000).
Thus, several feature dimensions can be expected to be represented across VI, AI, and SI.
The set of all values in a given dimension is not equally represented; ranges that are
especially important for behaviour are expanded (Suga, 1984; Recanzone et al., 1993). To
provide the optimal range of combinations between different information-bearing para-
meters, complex spatial relationships between the various parameter maps are necessary
(Obermayer et al., 1990; Swindale, 1991). This general principle is evident in the gross
similarity of maps in the different sensory cortical fields. From this point of view, each
neurone, and each location in VI, AI, and SI can be understood as representing a specific set
of many independent variables in the sensory environment. Topographically, each location
on the cortical surface corresponds to a specific intersection of several systematic maps.
Mathematically this forms a response vector, with specific direction and length in a multi-
dimensional parameter space (e.g. Lennie, 1998). Based on anatomical studies (Lund et al.,
1993), similar conclusions have been drawn: the size-match between axonal patches and
dendritic arbors should result in a maximal diversity of dendritic sampling (Malach, 1994).
There is agreement that physical attributes of sensory stimuli are encoded as activity
levels in populations of neurones. Reconstruction or decoding describes the inverse problem,
in which the physical attributes are estimated from neural activity. Reconstruction
methods have been regarded useful, first in quantifying how much information about the
physical attributes is present in a neural population, and second, in providing insight into
how the brain might use activity arising from many neurones (Nicolelis, 1996; Zhang and
Sejnowski, 1999; Doetsch, 2000).
Most of what we know in the visual cortex about functional maps, beyond orientation
preference, comes from optical imaging techniques. This method allows the simultaneous
assessment of entire maps, covering many square millimeters of cortex. Functional maps
are calculated by finding, for each cortical location, the preferred parameter value, viz.,
that causing the largest change in light absorption. This approach employs a particular
reconstruction scheme to decode information from the averaged activity distribution. It
implicitly assumes that thresholded activity reveals pertinent aspects in brain functioning.
However, this “winner-takes-all” approach is only one possible functional interpretation
of distributed neural activity.
Are there alternatives? Recently, it became evident that a critical step for the investigation
of how distributed cell assemblies process behaviourally-relevant information is the
introduction of data analysis methods that could identify functional neuronal interactions
within high-dimensional data sets (cf. Nicolelis, 1999). In fact, the introduction of the multi-
neurone/multi-site recording technique made it possible to record from a large number of
neurons simultaneously, even in awake and behaving animals (Abeles, 1991; Nicolelis et al.,
1997). This approach allows the exploration of dynamically-maintained distributions of
activity, including aspects of cooperativity between neurones, on a single trial basis. For
example, in a study devoted to exploring the representation of tactile information in three
areas of the primate somatosensory cortex (areas 2, 3b and SII), small neural ensembles (30–40
neurones) of broadly-tuned somatosensory neurones were sufficient to identify correctly the

location of a single tactile stimulus on a single trial (Nicolelis et al., 1998). Interestingly,
each of these cortical areas could use different combinations of encoding strategies, such as
mean firing rate, or temporal patterns of ensemble firing, to represent the location of a tactile
stimulus. Thus, several distinct ensembles of broadly tuned neurones, in different regions of
the somatosensory cortex contain information about the location of a tactile stimulus. In
contrast to functional maps, distributed representations are characterized by poor selectivity
of their constituents, and accordingly, little topographic specificity. Instead, it is assumed
that each neurone contributes, in a weighted form, to each possible stimulus configuration.
Another (albeit related) approach was recently introduced in order to account for popula-
tion activity at the level of spikes, recorded in early areas of sensory cortex. The goal was
to visualize and to analyze cortical activity distributions, in the coordinates of the respective
stimulus space, in order to explore cooperative processes (Dinse et al., 1996; Jancke et al.,
1996; Kalt et al., 1996; Erlhagen et al., 1999; Jancke et al., 1999; Jancke, 2000). Instead of
asking how accurately the parameter of, for example, stimulus location can be reconstructed
or decoded, the main interest was on analyzing interaction-based deviations of population
representations, dependent on defined variations of stimulus configurations. Despite the
fact that very different types of stimuli were employed, (squares of light versus indentations
on the digit of the hand), comparable distance-dependent interactions could be demonstrated
for visual and somatosensory cortex. This, again, suggests modality-independent modes of
processing and representation of distance-dependencies (Dinse and Jancke, 2001a).
Taken together, there is convincing experimental evidence that sensory cortices contain
large pools of neurones that act in concert to represent aspects of the outside world.
Depending on the methodology used, the outcome emphasizes either the aspect of “para-
metric maps” or the notion of “distributed representations”. In this scenario, parametric
maps are often regarded as a mass activity-based feature detector, i.e. a rather robust rep-
resentation of certain parameter regimes, invariant against further contextual influences.
Accordingly, this view provides some problems concerning how plastic adaptive capacities
can be implemented. The notion of “distributed representations” allows a higher degree of
flexibility, at the cost of a less rigid representation of feature spaces. On the other hand, it
has even been suggested that distributed representations might reflect a lack of orderly
representational maps. Only further experiments and more refined techniques will solve
these conceptual discrepancies in functional mapping. However, there are no clear modality
specific differences in these general questions regarding principles of cortical organization.
3.3. Temporal Processing

Many aspects of temporal processing across modalities can be compared rather directly,
because the time-axis reflects an absolute measure, rather than a parametric distribution of
activity which would allow differences between modalities in input energy. We will now
consider response latencies, repetition rate coding, RF-dynamics (i.e. the spatio-temporal,
or spectro-temporal organization of RFs), and oscillatory behavior.
3.3.1. Response latencies

Cortical response latencies are basically determined pre-cortically, by the kinetics of the
receptors, by subcortical processing and by properties of the propagating axons. There are

considerable differences in the temporal properties of the different receptors, and in the number
of subcortical stages, which in turn results in significant timing differences across modalities.
Axon properties such as myelination and size are comparable across modalities (Brown, 1987;
Hunt and McIntyre, 1960). For the visual system, fast and slow conducting fibres constitute
two segregated information channels that may play differing roles in form and motion
processing (Ungerleider and Mishkin, 1982; Dreher et al., 1976; Bullier and Henry, 1980;
Petersen et al., 1988). By contrast, segregated pathways in the somatosensory system that also
may contribute to form and motion processing, that is, neurones with slowly and rapidly adapting
responses, are not distinguished by different conduction times (Tremblay et al., 1993).
Comparing cortical response latencies across modalities reveals significant differences,
even given the uncertainty of comparing and scaling the stimuli (see above discussion of
Point spread function). In any case, auditory and somatosensory latencies have the shortest
latencies (ranging 10 to 20 msec), and visual latencies are the longest (40 to 60 ms). The
somatosensory system has an unusual property: cortical latencies vary with skin position
(in contrast to the ears and eyes, which have a fixed distance with respect to the cortex).
Response latencies to hindleg stimulation can be twofold longer than latencies following
vibrissa or foreleg stimulation. Thus, latencies in S1 reflect the distance between skin site
and cortex. Other cortical factors such as synaptic integration and threshold behaviour
may contribute to response latencies. Such factors probably do not contribute significantly
to timing differences of peripheral and subcortical origin.
The significant differences across modalities in the time at which sensory information
arrives in the cortex provide a clear modality-specificity. However, it also poses a problem
when information from different modalities has to be combined to yield an integrative,
behaviourally-meaningful output. This raises the question of how information from different
modalities is combined. In one possible scenario, the slowest modality could set the pace
(for an account on insula contributions to auditory-visual interactions via short-latency
connections with the tectal system see Bushara et al., 2001). Compensatory effects on
sensory processing in blind subjects provide some insights. Late components of auditory
event-related potentials (ERPs) have shorter latencies in blind than in sighted humans
(Naveen et al., 1998; Niemeyer and Starlinger, 1981; Röder et al., 1996). In auditory dis-
crimination tasks ERPs show an enhanced amplitude recovery in blind subjects as compared
to normal sighted subjects, indicating an improved ability to process fast sequences (Röder
et al., 1999). Studies on the speed of language processing revealed a similar superiority for
blind people, as compared to normal subjects (Röder et al., 2000). On the other hand, the
texture segmentation and visual search capacities in deaf subjects did not exceed that of
age-and gender-matched hearing subjects. Rather, deaf school children showed deficits in
visual processing, which were partially compensated in adult deaf subjects (Rettenbach et al.,
1999). These findings from blind and deaf subjects indicate that sensory deprivation has no
general effect on processing times, resulting either in acceleration or deceleration. The data
are compatible with the hypothesis that the multi-sensory processing speed in normal subjects
arises from integrative processes, which are dominated by the slowest information stream.
Once the modality with the longest latencies is removed, sensory processing is accelerated.
3.3.2. Repetition rate coding

Temporal integration, as expressed in repetition rate coding or sequence representation, is
the capacity of cortical cells to respond to consecutive stimuli. In the periphery and the

thalamus many neurones in each modality can follow high repetition rates, of more than
100 events/s. In contrast, cortical cells usually have a low-pass characteristic which limits
the following of more than 10 to 20 events/s, in all modalities (Movshon et al., 1978;
Creutzfeldt et al., 1980; Foster et al., 1985; Simons, 1985; Philipps et al., 1989; Nelson,
1991,a,b,c; Gardner et al., 1992; Reid et al., 1992; Merzenich et al., 1993; Brosch and
Schreiner, 2000; Krukowski, 2000). It has been proposed that specific properties of the
NMDA-receptor channel kinetics mediate this reduction in stimulus-following capacity
(Thomson and West, 1993; Crair and Malenka, 1995; Denham, 2000; Krukowski, 2000).
In addition, GABAergic mechanisms might contribute to this property (Dykes et al.,
1984) as well as fatigue and depletion of synaptic transmission (Chance et al., 1998,
Galarreta and Hestrin, 1998; Wang and Kaczmarek, 1998; Buonomano, 1999, 2000). Pre-
sumably, these all act in concert to constrain the representation of fast event sequences in
the cortex. Further work shows that active behaviour such as exploration can modulate the
temporal processing capacities, compared to passive stimulation (Fanselow and Nicolelis,
1999; Moore et al., 1999), and that plastic reorganization alters and modifies temporal
processing (Kilgrad and Merzenich, 1998; Buonomano, 1999; Dinse and Merzenich,
2002). Thus, in the cortex there is a significant reduction in repetition frequency response
compared to subcortical processing. Such uniformity is another argument for homogeneity
of cortical sensory processing across modalities.
3.3.3. RF-dynamics
It has been known for decades that RFs have complex spatio-temporal behaviour (in the
visual cortex) or spectro-temporal behaviour (in the auditory cortex) involving the time
domain. This is seen, when the complete temporal profile of neurone responses is assessed
(de Boer and Kuyper, 1968; Podvigin et al., 1974; van Gisbergen et al., 1975; Aertsen and
Johannesma, 1981; Eggermont et al., 1981; Krause and Eckhorn, 1983; Jones and Palmer,
1987; Shevelev, 1987; Best et al., 1989; Dinse et al., 1990; for an update of more recent
results see: Eckhorn et al., 1993; DeAngelis et al., 1993; Dinse, 1994; Wörgötter and Eysel,
2000; Dinse and Jancke, 2001b; Dinse, 2001). The substantial changes in RFs over time
indicate a significant interaction of space and time—or of spectrum and time—as well as
temporal dependence of tuning characteristics (Dinse et al., 1990, 1991; Dinse and Schreiner,
1996; Ringach et al., 1997; Ghazanfar and Nicolelis, 2001). Dynamic RFs are a common
denominator in all sensory modalities. Their basic dynamic properties are influenced by
aspects of latencies that reflect subcortical processes and by aspects of response duration.
Response duration is governed by feedforward, feedback, and lateral interactions (von Seelen
et al., 1986; Krone et al., 1986; Dinse, 1994; Dinse and Schreiner, 1996; Omurtag et al., 2000).
These interactions constitute an essential and unifying feature of cortical functions.
Another interpretation of RF dynamics attempts to link cortical processing dynamics and
dynamics of the environment (Dinse et al., 1993; Dinse, 1994; Wiemer et al., 2000).
Sensory signals typically have complex time-variant properties, which are manifested on
a variety of time scales, that may be reflected in the dynamics of RFs. In the auditory sys-
tem, one can differentiate timing based on the period of syllable sequences, in the range of
hundreds of ms, from the duration of consonants and formant transitions, in the range of
only 10 to 20 ms, from onset information, whose precision is in the millisecond range (for
single events). In the visual system, the timing schedule provided by eye movements also
imposes ecologically relevant dynamics. For the somatosensory system the movement of

objects across the receptor surface reflects largely the velocity of limb movements relative
to the objects (in the range of hundreds of milliseconds) and, by vibrational correlates,
also reflects the texture properties (in the range of tens of milliseconds). RF dynamics may
represent specific adaptations for processing inherently time-variant signals specific for
each modality. A common feature of cortical signal processing would undergo some specific
adaptations to match the requirements of the signal space.
3.3.4. Oscillations
Perhaps the most widely recognized but least understood electrophysiological activity of
the cerebral cortex is its characteristic electrical oscillations, which comprise a broad spec-
trum of periodic events from high-frequency oscillations (30–90 Hz, the so-called gamma
range) to frequencies as low as seconds or minutes (besides ultradian and circadian
rhythms). Stimulus-evoked oscillatory responses in the 10 Hz range have been studied for
several decades at cortical and subcortical levels (Bishop and O’Leary, 1936; Chang,
1950; Andersen and Andersson, 1968). However, at present the analysis of these low
frequency oscillations is purely phenomenological, and it is uncertain whether the different
low-frequency patterns seen in different brain regions have common origins and functions,
despite their temporal similarities (Dinse et al., 1990; Kopecz et al., 1993; Ahmed, 2000;
Cotillon et al., 2000; Miller and Schreiner, 2000). Accordingly, their functional role
remains obscure. Low-frequency oscillatory events are phase-locked to sensory stimulation,
which explains their appearance as distinct oscillatory peaks in post-stimulus-time histo-
grams (PSTHs). They cover frequencies between 5 to 20 Hz, encompassing several EEG
bands. A detailed analysis of area- and modality-specific properties of low frequency
oscillatory patterns recorded in 4 visual cortical areas, as well as in AI and SI (Dinse et al.,
1997) found low-frequency oscillations in all areas. These differed in probability of
occurrence, and each area had a characteristic frequency range of its own. Thus, in VI,
frequencies ranged from 8 to 22 Hz, in AI from 6 to 10 Hz, and in SI from 10 to 20 Hz.
Accordingly, the phenomenon of low-frequency oscillations is very general and found in
all areas. There is a striking uniformity of the overall pattern, but there are substantial
differences in the sequence and timing of the individual oscillatory peaks, differences that
reflect an area-specific signature of the oscillations. Interestingly, human auditory and
visual EPs also differ in their amplitude-frequency characteristics: Frequency maxima for
visual stimuli were significantly higher than for auditory stimuli (Schurmann and Basar,
1999). Recent inactivation experiments in AI revealed that both the auditory thalamus,
and the auditory sector of the thalamic reticular nucleus, but not the auditory cortex, have
a role in the genesis of a specific type of low frequency stimulus-evoked oscillations
(Cotillon and Edeline, 2000). The tendency to oscillate can be modified through external
stimuli with appropriately chosen spatio-temporal properties (Miller and Schreiner, 2000).
These findings suggest a common drive or origin of this oscillatory behaviour, that is
modified by localized parameters.
High-frequency gamma oscillations were described first for the olfactory bulb by Free-
man and co-workers (Freeman, 1968; Freeman and Skarda, 1985; Eeckman and Freeman,
1990). They have since been found in visual (Eckhorn et al., 1988; Gray et al., 1989),
auditory (Franowicz and Barth, 1995; MacDonald and Barth, 1995; Sukov and Barth,
1998) and somatosensory (MacDonald and Barth, 1995; Jones and Barth, 1997; Hashimoto,
2000) cortex, and they have gained much attention due to their possible involvement in

“feature binding” and higher cognitive processes (Eckhorn, 1994; Singer, 1998; Sauve,
1999). In contrast to low-frequency oscillations, high-frequency oscillations are not stimulus-
locked, so they are only detectable in single trial analyses, or by computation of the power
spectra of spike trains, and not in PSTHs that are based on a temporal averaging. High-
frequency oscillations cover a wide frequency range, from the higher beta range (~20 Hz)
to >100 Hz. There is debate about the stimulus-specificity of these oscillations and
uncertainty about the features and states that drives them optimally. While moving bars of
light or moving gratings drive high-frequency oscillations strongly, flash stimuli are
unable to induce oscillations, although they have the same perceptual relevance (Tovee
and Rolls, 1992).
Several studies have emphasized the importance of high-frequency oscillations in tasks
relevant to behaviour, and there is ample evidence for a task-specific emergence of such
oscillations (Hamada et al., 1999). Comparing oscillations across several neocortical areas
shows that the spatial patterns formed by synchronous activity change when the contin-
gency of reinforcement is reversed, vary with respect to stimuli, and had a dependence on
context and learning, as seen in the olfactory bulb and prepyriform cortex (Barrie et al.,
1996). In monkeys performing a motion-discrimination task, significant temporal correla-
tions exist between simultaneously recorded pairs of neurones, in areas MT and MST and
other extrastriate cortical areas. Interestingly, temporal decorrelation of MT and MST
neurones could be used to detect the stimulus, but synchronization did not convey specific
information about its direction of motion and was, thus, unlikely to contribute to behav-
ioural performance (de Oliveira et al., 1997).
While the functional role of this type of oscillation remains unresolved (Eeckman and
Freeman, 1990; Ghose and Freeman, 1992; Kirschfeld, 1992; Tovee and Rolls, 1992;
Kirschfeld et al., 1996, Ghose and Freeman, 1997; Singer, 1998; Engel et al., 1999), it is
clear that even single cells display complex patterns of oscillatory behaviour across
the whole frequency spectrum (Nunez et al., 1992; von Krosigk et al., 1993; Gray and
McCormick, 1996; Jones et al., 2000). Accordingly, the origin of high-frequency oscilla-
tions is usually considered as a combination of cellular and network properties. For the
present purposes, there appear to be no modality-specific effects, since high frequency
oscillations are found in all areas discussed. The variability from cell to cell, or across
behavioural states certainly exceeds any modality-specific aspect. However, comparative
studies across modalities might reveal more specific insights into the functional role and
relevance of cortical oscillatory patterns.
3.4. Spontaneous Activity

The most conspicuous neural activity in sensory cortices occurs in the absence of sensory
stimulation. The interpretation of this “spontaneous” activity, as either random background
noise, or as functionally-relevant signal, has remained controversial. The high level of
spontaneous discharge emanating from the retina, the cochleae, and several mechano-
receptors may reflect true noise, since it is believed to result from the stochastic nature of
transmitter release from the receptors (Koerber et al., 1966; Rodieck, 1967). On the other hand,
it has been argued that “ongoing activity” in higher brain centres might contain codes for
global states and conditions, that reflect meaningful aspects of brain function, as yet
unknown to the experimenter, and usually uncontrolled (Perkel and Bullock, 1968; but
see: Miller and Schreiner, 2000). Spontaneous activity has also been shown to play an

important role in creating and maintaining connections in the developing nervous system
(Shatz and Stryker, 1988; Penn et al., 1998). Straightforward ways of analyzing spontaneous
activity are to utilize spike count and frequency, while more sophisticated tools include
analysis of interval distribution. In the case of a completely random series of action potentials,
the interval distribution follows a Poisson-process. As neurones are characterized by
refractory periods, such processes are more adequately described as renewal processes
(Cox and Miller, 1965; Moore et al., 1966; Perkel et al., 1967; Wilbur and Rinzel, 1983).
Even after decades of extensive study there is still controversy about the nature of the
underlying interval distribution, as well as the nature of variability and stationarity (Softky
and Koch, 1993; Shadlen and Newsome, 1998; Nawrot et al., 2000). However, perhaps more
importantly, spontaneous activity in central nervous stations can show pronounced devia-
tions from the predictions of a Poisson-process, as reflected in spontaneous oscillations.
A general scheme holds that the spontaneous activity is lowest at cortical levels, but
increase successively when one descends a sensory pathway. According to Herz and
coworkers (1964), retinal ganglion cells discharge at >30 spikes/s, geniculate (thalamic)
neurones at about 15 spikes/s, and visual cortical cells at about 5 spikes/s. Similar con-
ditions have been reported for auditory and somatosensory system (Kiang et al., 1965;
Goldstein et al., 1968; Willis et al., 1975; Peschanski et al., 1980), although skin receptors
can vary more widely in terms of spontaneous activity (Bergmans and Grillner, 1969;
Dykes, 1975; Johnson and Hsiao, 1992).
There is some agreement that the dynamic state of the brain is reflected in the level and
character of spontaneous activity (Evarts et al., 1962; Noda and Adey, 1970; Burns and
Webb, 1982; Miller and Schreiner, 2000; Steriade, 2000), which appears to hold across
modalities. For example, increasing depth of anaesthesia leads to decrease of spontaneous
firing rates from 2.5–11 Hz during light anaesthesia, to 0–2.5 Hz in deep anaesthesia
(Armstrong-James and George, 1988). This implies that the level and character of ongoing-
activity is subject to substantial modifications, making it difficult to assign a characteristic
pattern of spontaneous activity to a given area. In fact, there are very few studies devoted
to unravelling possible modality-specific aspects of spontaneous activity (cf. Eggermont,
1990). Swadlow (1990) published a series of papers on ongoing activity in awake rabbits,
recorded in different cortical areas. He classified cortical recording sites according to layer
and projection pattern. For the forelimb representation of somatosensory cortex he
reported discharge frequency differing substantially between laminae, by a factor of 5,
with highest rates found in layer V. Similar lamina-specific differences were found in
visual cortex recordings in the same preparation (Swadlow, 1988). Interestingly, intra-areal
comparison of the forelimb and vibrissae representations in the somatosensory cortex
confirmed laminar differences, but these differences were higher—up to a factor of 15—in
the vibrissae representation (Swadlow, 1989). According to these data, differences in spon-
taneous discharge rate appear comparable between areas for a given layer, but can be large
across layers, or within a single sensory representation. Average spontaneous activity
recorded in the motor cortex was about 11 Hz ranging from 0.4 to 26 Hz (Glass and Wollberg,
1973). These authors reported a slight correlation with cortical depth suggesting that this
might be due to layer-dependent variation in excitability or richness of dendritic arborization.
With regard to higher-level areas, prefrontal cortical areas are involved in the temporal
organization of behaviour, and are discussed in relation to working memory (cf. for
reviews Fuster, 2000; Goldman-Rakic, 1996). Spontaneous activity levels typically are in
the range of about 5 Hz (Harden et al., 1998; Gulledge and Jaffe, 1998). One key aspect of

working memory functions consists of the temporary holding of goal-relevant information

about sensory stimuli, even if that information is no longer present at the time of response
generation. In this sense, prefrontal neurones have the property of “memory cells”, in being
active during delay periods, suggested by changes in firing rates. Accordingly, the level of
ongoing activity is strongly modulated, by a factor of 10, in a highly task-specific way.
However, it should be noted that data about comparable firing rates do not imply that
variability and interval distributions are comparable as well.
Independent of the level of spontaneous activity, the instantaneous frequency undergoes
dramatic variation on a very short time scale (Werner and Mountcastle, 1963; Tolhurst
and Movshon, 1983; Holt et al., 1996; Gutkin and Ermentrout, 1998), which is also true
for stimulus-driven activity (Tomko and Crapper, 1974; Schiller et al., 1976; Manley and
Mueller-Preuss, 1978; Heggelund and Albus, 1978; Vogels et al., 1989). Recent development
of real-time imaging made it possible to monitor large-scale spatio-temporal changes of
cortical activity on a millimeter scale, and at a millisecond time resolution, by means of
voltage-sensitive dyes (Lieke et al., 1989). Recording both ongoing and light-evoked
spatio-temporal activity patterns in cat visual cortex revealed that the variability of evoked
activity appeared deterministic, resulting from the dynamics of ongoing activity. It has
been suggested that these dynamics might reflect the instantaneous state of cortical net-
works. In spite of the large variability, evoked responses in single trials could be predicted
by linear summation of the deterministic response and the preceding ongoing activity
(Arieli et al., 1996). Tsodyks et al. (1999) were able to demonstrate that the firing rate of
a spontaneously active single neurone depends strongly on the instantaneous spatial
pattern of ongoing population activity in a large cortical area. At the level of cortical
maps, very similar spatial patterns of population activity were observed both when the
neurone fired spontaneously, and when it was driven by its optimal stimulus. Accordingly,
these studies provide a direct link between spontaneous activity and cortical representations
arising from sensory stimulation. The processes underlying the transformation of random
and uncorrelated peripheral activity to central spontaneous activity (that can show spatial
and temporal correlations) remains to be resolved.
In conclusion, spontaneous activity is a typical characteristic of primary cortical areas.
It appears comparable in frequency, particularly when compared to sensory periphery or
subcortical stages which all discharge at a much higher frequency. Spontaneous activity at
the cortical level does reflect different neurological states and functional properties of
neuronal assemblies. There is little evidence for modality-specific characteristics. It remains
to be seen whether the striking correlation of the dynamics of spontaneous activity pattern
with the organization of sensory processing exists in non-visual modalities as well.
4. CROSS-MODAL PLASTICITY
We have noted some functional and potential conceptual similarities between primary
sensory cortical areas. It should then be possible, in principle, to substitute one cortical
area for another, if the machinery and the implemented processing algorithms in different
sensory cortices are indeed the same (or at least very similar). Building on work by Frost
and Metin (Frost and Métin, 1985; Métin and Frost, 1989), Sur and colleagues (Sur et al.,
1990; Sharma et al., 2000; von Melchner et al., 2000) tested this hypothesis directly (see
also the chapter by S. Pallas, this volume). In immature ferrets, they eliminated the

auditory input to the auditory thalamus, and redirected retinal fibres to the auditory portion
of the thalamus, which in turn projected to the auditory cortex. The auditory cortex,
completely deprived of auditory input, but supplied with information originating in the
retina, developed functional properties resembling VI receptive fields and topographies.
Single cells were tuned for direction- and orientation-selectivity, and showed organized
retinotopic and orientation maps (including “pinwheels”). Moreover, horizontal connec-
tions in the rewired and remodelled AI were more like those in normal VI than those in
normal AI. The rewired animals responded to visual stimulation as if they experienced
vision, and did not treat such stimuli as an auditory event (von Melchner et al., 2000).
If the experiment involved auditory input to the deafferented VI, there remains the
question of which auditory input is necessary to change VI into a replica of AI. It appears
unlikely that input from the auditory nerve alone would suffice to generate all AI proper-
ties in VI, since major feature dimensions are not yet computed at this peripheral level.
Rather, the output from the midbrain station, the inferior colliculus, would seem to be the
more appropriate input to the visual thalamus needed to induce the formation of functionally
correct AI properties in VI. It alone has an equivalent set of basic feature dimensions to
generate a sufficiently complete and computationally accessible cortical representation of
sound. This proposition awaits direct experimental investigation.
The elegant experiments described above provide direct evidence that different cortical
processing systems may be interchangeable, and are capable of implementing the processes
and algorithms used in other modalities. Since developmental and reorganizational
plasticity undoubtedly was involved in the process of “rewiring” the auditory cortex (e.g.
Buonomano and Merzenich, 1998), the experiments do not prove that the same algorithms
are implemented in AI and VI, but only that AI can implement the VI algorithms. Further-
more, the experiments show that sensory and perhaps perceptual qualities are modality-
specific, and not cortical field-specific. Such findings strongly support the hypothesis
proposed at the outset of this chapter, that the different primary sensory cortical fields can
be viewed as analogous structures anatomically, and perhaps more importantly, functionally.
5. GENERAL CONCLUSION AND PREDICTIONS
We have reviewed evidence germane to the hypothesis that the different sensory modalit-
ies perform similar tasks, under similar statistical constraints imposed by sensory input,
with the same structural elements, and, possibly, similar circuits: The common task is to
transform receptor images to cortical representations of external objects preparatory to
action. While the peripheral sensory systems have different structural adaptations at the
receptor level, the neuronal implementations of the required algorithms at the early cor-
tical level show much congruence and may be interchangeable between modalities.
The main lines of argument in support of this notion were as follows:
(a) Functional and structural heterogeneity in the subcortical systems: The role of the
subcortical system may be understood as a transformation of receptor surface
information into a multi-dimensional parameter space. This process adapts the different
sensory streams to a thalamo-cortical stage that applies common schemes for
information processing and modulation. The activity across a receptor surface codes
information about local and global correlations in the scene, shown by coactivation

and sequential activation of the receptors. This permits higher-order stimulus correl-
ations, including contextual events, to be integrated in this process. Multiple subcorti-
cal systems construct relevant, functionally-independent feature dimensions. Since
the nature of the stimuli and the corresponding receptors differ, the nature of these
subcortical processing stages also differs between the modalities.
(b) Functional and structural homogeneity in the thalamo-cortical system: The role
of the thalamus and primary sensory cortex is, according to this view, to establish
a compatible representation of different feature dimensions, that allows extraction of
object form, attribute, and location relative to the background, for an analysis of the
environmental scene. Subsequent sensory cortical areas use these base features to
analyse the global features of objects and to establish object representations that
become less dependent on the perceptual frame-of-reference. Furthermore, higher
processes establish the relationship of objects to the environment and the self, perform
multi-sensory integration, assign and evaluate the behavioural significance according
to context, and as a basis for action.
(c) Computational homogeneity in the thalamo-cortical system: At the thalamo-cortical
interface a similar set of operations is implemented that involves thalamo-cortical,
cortico-cortical and cortico-fugal network interactions. The main mechanisms of
these algorithms include the spatio-temporal arrangement of excitatory and inhibitory
processes in columnar and horizontal networks, that operate on spatially- and temporally-
related streams of input.
While only few of the actual processing schemes have been dissected and modeled
in detail, the processing that is accomplished at the thalamo-cortical level appears to
be largely a local, differentiating analysis, as opposed to a global, integrative one.
A case in point is the contrast-independent processing of orientation-selectivity of visual
cells (for review see Ferster and Miller, 2000; Anderson et al., 2000). The algorithms
used in this task are based on local correlations and anti-correlations of inputs, as
expressed in local excitatory and inhibitory circuits. Such computations use temporal
and spatial information arising from peripheral receptors. Correlations among the
activity in the receptor space are exploited to establish computationally-useful and
independent dimensions, that are representative of external objects and scenes, and
not just the receptor-surface activity. These computed entities may serve several
needs of central processing: (i) They precede a general representation of the external
world; (ii) They are a basis for the determination of object form and position; (iii)
They create reliable and stabilized feature properties, as reflected in the intensity and
contrast invariance of orientation tuning in the visual cortex, or in independent tuning
properties of some auditory neurones in the presence of background noise ; (iv) Such
computed features allow subsequent multi-sensory integration, by transcending a pure
receptor representation of the sensory events; (v) Finally, they allow the assignment
of significance to particular environmental conditions, and ultimately, the emergence
of different perceptual attributes and the initiation of behaviour. Certain of these tasks
may take place beyond the primary sensory field.
(d) Multi-sensory integration: The integration of information from several sensory
modalities has many advantages for the individual, including increase of salience,
removal of ambiguities, and unified object characterization and perception (e.g. Stein
and Meredith, 1990). To accomplish such integration, the information from the
different modalities has to be represented at an equivalent level of abstraction, and in

compatible frames-of-reference. Many aspects of multi-sensory integration take place

after the early cortical representation suggesting that major aspects of representational
equivalence have been established at that level. For example, it has been demon-
strated that auditory-visual stimulus-onset asynchrony activates a large-scale neural
network of insular, posterior parietal, prefrontal, and cerebellar areas (Bushara et al.,
2001) permitting auditory-visual interaction phenomena such as the ventriloquist, and
the McGurk illusions. Subsequent processing can then employ similar mechanisms
for similar tasks. Furthermore, it would be advantageous if attentional and emotional
modulation of these representations takes place at equivalent stages of processing, to
maintain compatibility between sensory systems. The common nature of the modulatory
influence at the thalamus suggests that the early cortical processing operates at equivalent
levels across the three modalities (cf. Sherman and Guillery, 2001).
(e) Cross-modal equivalence: The induction of visual receptive field properties and map
organization in AI after re-routing of sub-thalamic channels convincingly demonstrates
that the normally-observed functional differences between early sensory cortical areas
is not a consequence of immutable field-specificity but largely due to subthalamic
preprocessing. The processing capacities in the primary cortical fields therefore have
the same potential for expressing specific algorithms. However, they can function
somewhat differently according to the actual input.
From these general principles of cortical processing equivalencies, a number of specific

predictions can be made. Among them are the following:
• Given the joint 2-dimensionality for vision and touch, functional maps should be present
in SI that are comparable to the many aspects described for VI, including orientation
sensitivity, and direction selectivity.
• The emergence of spatial frequency maps (relative to the receptor sheet) should also
be a common factor. Spatial frequency in the visual system can be regarded as equiva-
lent to the spacing of frequency bands in the auditory system. Given that this aspect is
coded in functional maps in VI and AI, a corresponding spatial-frequency coding scheme
should be present in SI.
• The visual cortex shows a relationship between spatial frequency maps and orientation
maps. A similar relationship could be predicted for the somatosensory cortex. (One
should keep in mind that map expression and map interrelations can be quite weak on
a local and cellular scale.)
• Ocular dominance bands in the visual cortex have a specific relationship to orient-
ation maps and, perhaps, to spatial frequency maps. Cortical domains with low spatial
frequency tend to lie in the centre of the ocular dominance columns (Hübener et al.,
1997). Accordingly, the position of binaural bands and spatial frequency/bandwidth
distribution in auditory cortex may show a systematic relationship as well.
• The inhibitory push-pull mechanisms invoked for contrast-independent orientation
selectivity (Troyer et al., 1998) may also be found in somatosensory orientation
processing and, perhaps, in auditory processing of spectral-shape information.
• “Feature maps” are not invariant, resistant to variations of stimulus configuration,
context or timing parameters. For example, direction selectivity of visual neurones is
strongly dependent on the direction of a moving texture background (Orban et al., 1987;
Dinse and Krüger, 1990). Also, in SI and AI, there is evidence that spatial/spectral

and probably also temporal topography can be modified substantially by behavioural

discrimination training (e.g. Recanzone et al., 1992, 1993; Wang et al., 1995; Beitel
et al., 1999). The cortical mechanisms that govern these processes should hold for all
modalities.
The common feature of parametric maps observed in early sensory areas might reflect
a rather general principle of cortical organization and function, potentially holding also for
all other cortical areas. The fundamental processing step consists of a local operation,
modified (contextually) by long-range connections. As this type of processing is performed
within a two-dimensional sheet, it warrants the combination of local operations based on
excitatory and inhibitory interactions, with a continuous representation of parameters,
maintaining local neighbourhood relationships. While local processing obeys per se rules
of proximity, the two-dimensional sheet allows one to define proximities of various kinds.
For sensory areas, the “representation” of the outside world implicates both two-dimension-
ality, and proximities within the various types of physical worlds. In this sense this scheme
is highly intuitive for early cortical representation, where it is reasonably clear what is
represented. However, this scheme has been shown to hold also for intermediate states: An
example is the highly specialized and detailed maps as described by Suga and coworkers in
echolocating bats (e.g. Suga, 1984). Because of the highly specialized ultrasound
environment of bats it was possible to deduce and then to identify higher-order maps that
contain ordered representations of echo frequency and echo delay.
The main problem in higher cortical areas arises from the fact that the behaviourally or
computationally relevant parameter space is unknown and, in principle, is difficult to deduce.
The nature and properties of such parameter spaces need to be determined to understand
cortical processing outside the primary sensory or motor domains. For example, it is likely
that there are highly abstract parameter spaces representing a profile of a face in terms of
its emotional expression. In any case, the basic principle is to compute and assemble
behaviourally relevant aspects of proximity, or similarity and dissimilarity in the projected
parameter space.
In principle, “parametric mapping” can be regarded as equivalent to “distributed process-
ing”. Differences that have been pointed out between these two concepts are mainly
methodological, and arise largely from peculiarities of the reconstruction algorithms (e.g.
the optical imaging-based feature maps are just due to an “iceberg effect”, by ignoring
a large portion of neural activation). While “parametric-mapping” is more intuitive because
it relates directly the sensory representations to known physical features, we believe that the
concept of “distributed representation” encapsulates better the representational and com-
putational principles expressed in the brain, and this concept holds equally in lower (or early)
and higher cortical areas.
The appearance of modality-specific differences and modality-independent properties
of brain organization and function offers an opportunity to investigate systematically the
underlying functional and structural constraints and consequences. Among the questions
that can be addressed from a more general point of view are the following: What is the
basis and consequence of modality-specificities? Why and how are certain general aspects
of cortical processing modified in particular ways? What is the consequence of such
modifications for the evolution of higher brain functions in humans?
The emphasis of common structures, mechanisms, and operational goals in primary
sensory cortex, in support of the notion of functional equivalency, does not imply a perfect

correspondence across areas. An evolutionary advantage and consequence of the apparent

homogeneity of cortical organization is the richness in potential network characteristics
that can be realized. The substrate for cortical processing can apply in a flexible manner,
during development and learning, to a wide range of environmental and behavioural con-
ditions. However, the influence of these conditions on the different modalities may be
quite similar. In any case, it remains to be seen whether the local circuit interactions and
functions are identical in different modalities or merely constitute similar classes with dif-
ferent solutions in each modality. We are now ready to discover the local network properties
and the algorithms which are implemented for a number of primary cortical processes in
different modalities. A refinement of the modality-comparative approach to this difficult
but exciting and fundamental problem of brain function could decipher this mystery.
ACKNOWLEDGEMENTS
We thank Dr. Jeffery Winer for his insightful comments on a draft of this chapter. Supported
by grants DC02260 and NS35438 (to C.E.S.) and DFG 334 / 10 and 15 (to H.D.).
REFERENCES
Abeles, M. (1991) Corticonics. Neural circuits in the cerebral cortex. Cambridge, UK: Cambridge University Press.
Aertsen, A. and Johannesma, P. (1981) The spectro-temporal receptive field. A functional characterization of
auditory neurons. Biological Cybernetics, 42, 133–143.
Ahmed, B. (2000) Low frequency damped oscillations of cat visual cortical neurones. NeuroReport, 11,
1243–1247.
Albus, K. (1975) A quantitative study of the projection area of the central and the paracentral visual field in area 17
of the cat. I. The precision of the topography. Experimental Brain Research, 24, 159–179.
Allman, J., Miezin, F. and McGuiness, E.L. (1985) Stimulus specific responses from beyond the classical receptive
field. Annual Review of Neuroscience, 8, 407–430.
Andersen, P. and Andersson, S.A. (1968) Physiological basis of the alpha rhythm. Appleton-Century-Crofts,
New York.
Anderson, J.C., Binzegger, T., Kahana, O., Martin, K.A. and Segev, I. (1999) Dendritic asymmetry cannot
account for directional responses of neurons in visual cortex. Nature, Neuroscience, 2, 820–824.
Anderson, J.S., Lampl, I., Gillespie, D.C. and Ferster, D. (2000) The contribution of noise to contrast invariance
of orientation tuning in cat visual cortex. Science, Washington, 290, 1968–1972.
Arieli, A., Sterkin, A., Grinvald, A. and Aertsen, A. (1996) Dynamics of ongoing activity: explanation of the large
variability in evoked cortical responses. Science, Washington, 273, 1868–1871.
Armstrong-James, M. and George, M.J. (1988) Influence of anesthesia on spontaneous activity and receptive field
size of single units in rat Sm1 neocortex. Experimental Neurology, 99, 369–387.
Atick, J.J. (1992) Could information theory provide an ecological theory of sensory processing? Network Computa-
tion in Neural Systems, 3, 213–251.
Attias, H. and Schreiner, C.E. (1998) Coding of naturalistic stimuli by auditory midbrain neurons. In: M.I. Jordan
et al. (eds), Advances in Neural Information Processing Systems, Vol. 10. Cambridge, MA: MIT Press,
pp. 103–109.
Bakin, J.S., Kwon, M.C., Masino, S.A., Weinberger, N.M. and Frostig, R.D. (1996) Suprathreshold auditory
cortex activation visualized by intrinsic signal optical imaging. Cerebral Cortex, 6, 120–130.
Barrie, J.M., Freeman,W.J. and Lenhart, M.D. (1996) Spatiotemporal analysis of prepyriform, visual, auditory
and somesthetic surface EEGs in trained rabbits. Journal of Neurophysiology, 76, 520–539.
Beitel, R.E., Schreiner, C.E. and Merzenich, M.M. (1999) Spectral receptive field plasticity in auditory cortex of
owl monkeys trained with sinusoidally amplitude modulated (SAM) tones. Association for Research in
Otolaryngology, Abstracts (U.S.A.), Midwinter meeting, 21, 201.
Bergmans, J. and Grillner, S. (1969) Reciprocal control of spontaneous activity and reflex effects in static and
dynamic flexor gamma-motoneurones revealed by an injection of DOPA. Acta Physiologica Scandinavica,
77, 106–124.

Best, J., Mallot, H.A., Krüger, K. and Dinse, H.R. (1989) Dynamics of visual information processing in cortical
systems. In: L. Personnaz and G. Dreyfus (eds), Neural Networks. From Models to Application. IDSET,
Paris, pp. 107–116.
Bialek, W., Rieke, F., Van Steveninck, R.R. and Warland, D. (1991) Reading a neural code. Science, Washington,
252, 1854–1857.
Bishop, G.H. and O’Leary, J. (1936) Components of the electric responses to the optic cortex of the rabbit.
American Journal of Physiology, 117, 292–308.
Blasdel, G.G. and Salama, G. (1986) Voltage-sensitive dyes reveal a modular organization in monkey striate
cortex. Nature, London, 321, 579–585.
Bolz, J. and Gilbert, C.D. (1989) The role of horizontal connections in generating long receptive fields in the cat
visual cortex. European Journal of Neuroscience, 1, 263–268.
Bonhoeffer, T. and Grinvald, A. (1991) Iso-orientation domains in cat visual cortex are arranged in pinwheel-
like patterns. Nature, London, 353, 429–431.
Bregman, A.S. (1990) Auditory Scene Analysis: The Perceptual Organization of Sound. Cambridge, Mass.:
Bradford Books, MIT Press.
Brosch, M. and Schreiner, C.E. (2000) Sequence sensitivity of neurons in cat primary auditory cortex. Cerebral
Cortex, 10, 1155–1167.
Brown, M.C. and Ledwith, J.V. (1990) Projection of thin (type–II) and thick (type–I) auditory nerve fibers onto
the cochlear nucleus of the mouse. Hearing Research, 49, 105–118.
Brown, M.C. (1987) Morphology of labeled afferent fibers in the guinea pig cochlea. Journal of Comparative
Neurology, 260, 591–604.
Bullier, J. and Henry, G.H. (1980) Ordinal position and afferent input of neurons in monkey striate cortex.
Buonomano, D.V. and Merzenich, M.M. (1998) Cortical plasticity: From Synapses to Maps. Annual Review of
Buonomano, D.V. (1999) Distinct functional types of associative long-term potentiation in neocortical and
hippocampal pyramidal neurons. Journal of Neuroscience, 19, 6748–6754.
Buonomano, D.V. (2000) Decoding temporal information: A model based on short-term synaptic plasticity.
Burkitt, G.R., Lee, J. and Ts’o, D.Y. (1998) Functional organization of disparity in visual area V2 of the
macaque monkey. Society for Neuroscience Abstracts, 24, 1978.
Burns, B.D. and Webb, A.C. (1982) Mechanisms governing the relation between activity in the cerebral cortex
and level of arousal. Proceedings of the Royal Society, London, series B, 214, 325–334.
Burr, D. (1980) Motion smear. Nature, London, 284, 164–165.
Bushara, K.O., Grafman, J. and Hallett, M. (2001) Neural correlates of auditory-visual stimulus onset asyn-
chrony detection. Journal of Neuroscience, 21, 300–304.
Capuano, U. and McIlwain, J.T. (1981) Reciprocity of receptive field images and point images in the superior
colliculus of the cat. Journal of Comparative Neurology, 196, 13–23.
Casagrande, V.A. and Norton, T.T. (1991) The lateral geniculate nucleus: A review of its physiology and func-
tion. In: A.G. Leventhal, (ed.), The Neural Basis of Visual Function, Volume 4, of Vision and Visual
Disfunction series (ed. J.R. Cronley-Dillon), London: MacMillan Press, pp. 41–84.
Chance, F.S., Nelson, S.B. and Abbott, L.F. (1998) Synaptic depression and the temporal response characteristics
of V1 cells. Journal of Neuroscience, 18, 4785–4799.
Chang, H.T. (1950) The repetetive discharges of cortico-thalamic reverberating circuit. Journal of Neurophysi-
ology, 13, 235–257.
Chen-Bee, C.H. and Frostig, R.D. (1996) Variability and interhemispheric asymmetry of single-whisker func-
tional representations in rat barrel cortex. Journal of Neurophysiology, 76, 884–894.
Clark, S.A., Allard, T.T., Jenkins, W.M. and Merzenich, M.M. (1988) Receptive fields in the body-surface map
in adult cortex defined by temporally correlated input. Nature, London, 332, 444–445.
Cotillon, N. and Edeline, J.M. (2000) Tone-evoked oscillations in the rat auditory cortex result from interactions
between the thalamus and reticular nucleus. European Journal of Neuroscience, 12, 3637–3650.
Cotillon, N., Nafati, M. and Edeline, J.M. (2000) Characteristics of reliable tone-evoked oscillations in the rat
thalamo-cortical auditory system. Hearing Research, 142, 113–130.
Cox, D.R. and Miller, H.D. (1965) The Theory of Stochastic Processes. New York: John Wiley.
Crair, M.C. and Malenka, R.C. (1995) A critical period for long-term potentiation at thalamocortical synapses.
Nature, London, 375, 325–328.
Creutzfeldt, O., Hellweg, F.C. and Schreiner, C. (1980) Thalamocortical transformation of responses to complex
auditory stimuli. Experimental Brain Research, 39, 87–104.
Das, A. and Gilbert, C.D. (1995) Long-range horizontal connections and their role in cortical reorganization
revealed by optical recording of cat primary visual cortex. Nature, London, 375, 780–784.
Das, A. and Gilbert, C.D. (1997) Distortions of visuotopic map match orientation singularities in primary visual

de Boer, F. and Kuyper, P. (1968) Triggered correlation. I.E.E.E., 15, 169–179.

de Oliveira, S.C., Thiele, A. and Hoffmann, K.P. (1997) Synchronization of neuronal activity during stimulus
expectation in a direction discrimination task. Journal of Neuroscience, 17, 9248–9260.
DeAngelis, G.C., Ohzawa, I. and Freeman, R.D. (1993) Spatiotemporal organization of simple-cell receptive
fields in the cat’s striate cortex. I. General characteristics and postnatal development. Journal of Neuro-
physiology, 69, 1091–1117.
deCharms, C. and Zador, A. (2000) Neural representations and the cortical code. Annual Review of Neuro-
science, 23, 613–847.
Denham, S.L. (2000) Cortical Synaptic Depression and Auditory Perception. In: S. Greenberg and M. Slaney
(eds), Computational Models of Auditory Function, NATO ASI Series. Amsterdam: IOS Press.
DiCarlo, J.J., Johnson, K.O. and Hsiao, S.S. (1998) Structure of receptive fields in area 3b of primary somatosensory
cortex in the alert monkey. Journal of Neuroscience, 18, 2626–2645.
Dinse, H.R. (1986) Foreground-background-interaction—Stimulus dependent properties of the cat’s area 17, 18
and 19 neurons outside the classical receptive field. Perception, 15, A6.
Dinse, H.R. and Krüger, K. (1990) Contribution of area 19 to the foreground-background-interaction of the cat.
An analysis based on single cell recordings and behavioural experiments. Experimental Brain Research, 82,
107–122.
Dinse, H.R., Krüger, K. and Best, J. (1990) A temporal structure of cortical information processing. Concepts in
Dinse, H.R., Krüger, K., Mallot, H.P. and Best, J. (1991) Temporal structure of cortical information processing:
cortical architecture, oscillations, and non-separability of spatio-temporal receptive field organization.
In: J. Krüger (ed.), Neuronal Cooperativity. Berlin, Heidelberg: Springer Verlag, pp. 67–104.
Dinse, H.R., Spengler, F., Godde, B. and Hartfiel, B. (1993) Dynamic aspects of cortical functions: Processing
and plasticity in different sensory modalities. In: A. Aertsen (ed.), Brain Theory: Spatio-temporal aspects of
brain functions. Amsterdam: Elsevier, pp. 209–230.
Dinse, H.R. (1994) A time-based approach towards cortical functions: Neural mechanisms underlying
dynamic aspects of information processing before and after postontogenetic plastic processes. Physica D,
75, 129–150.
Dinse, H.R. and Schreiner, C.E. (1996) Dynamic frequency tuning of cat auditory cortical neurons: specific
adaptations to the processing of complex sounds. In: W. Ainsworth and S. Greenberg (eds), Proceedings of
the workshop on: The auditory basis of speech perception. (European Speech Communication Association
ETRW Series). Keele, U.K., Keele University, pp. 45–48.
Dinse, H.R., Jancke, D., Akhavan, A.C., Kalt, T. and Schöner, G. (1996) Dynamics of population representations
of visual and somatosensory cortex based on spatio-temporal stimulation. In: S.I. Amari, L. Xu, L.W. Chan,
I. King and K.S. Leung (eds), Progress in Neural Information Processing. International Conference on
Neural Information Processing (ICONIP’96). Singapore: Springer, pp. 1285–1290.
Dinse, H.R., Krüger, K., Akhavan, A.C., Spengler, F., Schöner, G. and Schreiner, C.E. (1997) Low-frequency
oscillations of visual, auditory and somatosensory cortical neurons evoked by sensory stimulation. Inter-
national Journal of Psychophysiology, 26, 205–227.
Dinse, H.R., Godde, B., Hilger, T., Reuter, G., Cords, S.M., Lenarz, T. and von Seelen, W. (1997) Optical
imaging of cat auditory cortex cochleotopic selectivity evoked by acute electrical stimulation of a multi-
channel cochlear implant. European Journal of Neuroscience, 9, 113–119.
Dinse, H.R., Godde, B., Hilger, T. and Schreiner, C.E. (2000) Optical imaging of relationships between func-
tional topographies in cat auditory cortex. Institut für Neuroinformatik, Ruhr-Univ. Bochum, Internal
Report IRINI 2000–01, pp. 1–11.
Dinse, H.R. and Jancke, D. (2001a) Comparative population analysis of cortical representations in parametric
spaces of visual field and skin: a unifying role for nonlinear interactions as a basis for active information
processing across modalities. In: M.A.L. Nicolelis (ed.), Advances in Population Coding, (Progress in
Brain Research, Vol. 130). Elsevier, pp. 155–173.
Dinse, H.R. and Jancke, D. (2001b) Time-variant processing in V1: Indication for qualitative transforma-
tional steps from microscopic (single cell) to mesoscopic (population) levels. Trends in Neuroscience,
24, 203–205.
Dinse, H.R. (2002) Multiple and variant timescales in dynamic information processing. Behavioural and Brain
Sciences, in press
Dinse, H.R. and Merzenich, M.M. (2002) Adaptation of Inputs in the Somatosensory System. In: M. Fahle and
T. Poggio (eds), Perceptual Learning, MIT Press, Boston, MA, pp. 19–42.
Doetsch, G.S. (2000) Patterns in the brain. Neuronal population coding in the somatosensory system. Physiology
and Behavior, 69, 187–201.
Douglas, R.J., Martin, K.A. and Whitteridge, D. (1988) Selective responses of visual cortical cells do not depend
on shunting inhibition. Nature, London, 332, 642–644.
Douglas, R.J., Martin, K.A. and Whitteridge, D. (1991) An intracellular analysis of the visual responses of
neurones in cat visual cortex. Journal of Physiology (London), 440, 659–696.

Dreher, B., Fukada, Y. and Rodieck, R.W. (1976) Identification, classification and anatomical segregation of
cells with X-like and Y-like properties in the lateral geniculate nucleus of old-world primates. Journal of
Physiology, (London), 258, 433–452.
Dykes, R.W. (1975) Afferent fibers from mystacial vibrissae of cats and seals. Journal of Neurophysiology, 38,
650–662.
Dykes, R.W., Landry, P., Metherate, R. and Hicks, T.P. (1984) Functional role of GABA in cat primary
somatosensory cortex: shaping receptive fields of cortical neurons. Journal of Neurophysiology, 52,
1066–1093.
Eckhorn, R., Bauer, R., Jordan, W., Brosch, M., Kruse, W., Munk, M. and Reitboeck, H.J. (1988) Coherent oscil-
lations: a mechanism of feature linking in the visual cortex? Multiple electrode and correlation analyses in the
cat. Biological Cybernetics, 60, 121–130.
Eckhorn, R., Krause, F. and Nelson, J.I. (1993) The RF cinematogram. A cross-correlation technique. Biological
Cybernetic, 69, 37–55.
Eckhorn, R. (1994) Oscillatory and non-oscillatory synchronizations in the visual cortex and their possible roles
in associations of visual features. In: J. van Pelt, M.A. Corner, H.B.M. Uylings and F.H. Lopes da Silva
(eds), The self-organizing brain: from growth cones to functional networks (Progress in Brain Research,
Vol. 102). Amsterdam: Elsevier, pp. 405–426.
Eeckman, F.H. and Freeman, W.J. (1990) Correlations between unit firing and EEG in the rat olfactory system.
Eggermont, J.J., Aertsen, A., Hermes, D.J. and Johannesma, P.I.M. (1981) Spectro-temporal characterization of
auditory neurons: redundant or necessary? Hearing Research, 5, 109–121.
Eggermont, J.J. (1990) The Correlative Brain: Theory and Experiment in Neural Interaction. Berlin, Heidelberg:
Springer Verlag.
Ehret, G. and Merzenich, M.M. (1988) Complex sound analysis (Frequency resolution, filtering and spectral
integration) by single units of the inferior colliculus. Brain Research Reviews, 13, 139–163.
Ehret, G. and Schreiner, C.E. (1997) Frequency resolution and spectral integation (critical band analysis) in single
units of the cat primary auditory cortex. Journal of Comparative Physiology, A 181, 635–650.
Engel, A.K., Fries, P., Konig, P., Brecht, M. and Singer, W (1999) Temporal binding, binocular rivalry, and
consciousness. Consciousness and Cognition, 8, 128–151.
Erlhagen, W., Bastian, A., Jancke, D., Riehle, A. and Schöner, G. (1999) The distribution of neuronal population
activation (DPA) as a tool to study interaction and integration in cortical representations. Journal of Neuro-
science Methods, 94, 53–66.
Evans, P.M. and Craig, J.C. (1991) Tactile attention and the perception of moving tactile stimuli. Perception and
Psychophysics, 49, 355–364.
Evarts, E.V., Bental, E., Bihari, B. and Huttenlocher, P.R. (1962) Spontaneous discharge of single neurons
during sleep and waking. Science, Washington, 135, 727–728.
Fanselow, E.E. and Nicolelis, M.A. (1999) Behavioral modulation of tactile responses in the rat somatosensory
system. Journal of Neuroscience, 19, 7603–7616.
Ferster, D. and Miller, K.D. (2000) Neural mechanisms of orientation selectivity in the visual cortex. Annual
Reviews of Neuroscience, 23, 441–471.
Fischer, B. (1972) Optical and neuron basis of visual image transfer: a uniform mathematical treatment of the
retinal image and the excitability of the retinal ganglion cells with the aid of the linear systems theory.
Vision Research, 12, 1125–1144.
Foster, K.H., Gaska, J.P., Nagler, M. and Pollen, D.A. (1985) Spatial and temporal frequency selectivity of
neurones in visual cortical areas V1 and V2 of the macaque monkey. Journal of Physiology (London), 365,
331–363.
Franowicz, M.N. and Barth, D.S. (1995) Comparison of evoked potentials and high-frequency (gamma-band)
oscillating potentials in rat auditory cortex. Journal of Neurophysiology, 74, 96–112.
Freeman, W.J. (1968) Patterns of variation in waveform of averaged evoked potentials from prepyriform cortex
of cats. Journal of Neurophysiology, 31, 1–13.
Freeman, W.J. and Skarda, C.A. (1985) Spatial EEG patterns, non-linear dynamics and perception: the Neo-
Sherringtonian view. Brain Research Reviews, 10, 147–175.
Frost, D.O. and Metin, C. (1985) Induction of functional retinal projections to the somatosensory system. Nature,
London, 317, 162–164.
Furukawa, S., Xu, L. and Middlebrooks, J.C. (2000) Coding of sound-source location by ensembles of cortical
neurons. Journal of Neuroscience, 20, 1216–1228.
Fuster, J.M. (2000) Executive frontal functions. Experimental Brain Research, 133, 66–70.
Galarreta, M. and Hestrin, S. (1998) Frequency-dependent synaptic depression and the balance of excitation and
inhibition in the neocortex. Nature, Neuroscience, 1, 587–594.
Gardner, E.P., Palmer, C.I., Hamalainen, H.A. and Warren, S. (1992) Simulation of motion on the skin. V. Effect
of stimulus temporal frequency on the representation of moving bar patterns in primary somatosensory
cortex of monkeys. Journal of Neurophysiology, 67, 37–63.

Geldard, F.A. and Sherrick, C.E. (1972) The cutaneous “rabbit”: a perceptual illusion. Science, Washington, 178,
178–179.
Ghazanfar, A.A. and Nicolelis, M.A. (2001) The structure and function of dynamic cortical and thalamic receptive
fields. Cerebral Cortex, 11, 183–193.
Ghose, G.M. and Freeman, R.D. (1992) Oscillatory discharge in the visual system: does it have a functional role?
Ghose, G.M. and Freeman, R.D. (1997) Intracortical connections are not required for oscillatory activity in the
visual cortex. Visual Neuroscience, 14, 963R–979R.
Gibson, J.J. (1966) The Senses Considered as Perceptual Systems. London: George Allen and Unwin.
Gilbert, C., Ito, M., Kapadia, M. and Westheimer, G. (2000) Interactions between attention, context and learning
in primary visual cortex. Vision Research, 40, 1217–1226.
Gilbert, C.D. and Wiesel, T.N. (1990) The influence of contextual stimuli on the orientation selectivity of cells in
primary visual cortex of the cat. Visual Research, 30, 1689–1701.
Glass, I. and Wollberg, Z. (1973) Spontaneous activity of single neurons in the precruciate area of the cat’s
sensory motor cortex. Brain Research, 64, 391–396.
Godde, B., Hilger, T., von Seelen, W., Berkefeld, T. and Dinse, H.R. (1995) Optical imaging of rat somato-
sensory cortex reveals representational overlap as topographic principle. NeuroReport, 7, 24–28.
Goldman-Rakic, P.S. (1996) Regional and cellular fractionation of working memory. Proceedings of the
National Academy of Sciences, U.S.A., 93, 13473–13480.
Goldstein, M.H., Hall, J.L.2d and Butterfield, B.O. (1968) Single-unit activity in the primary auditory cortex of
unanesthetized cats. Journal of the Acoustic Society of America, 43, 444–455.
Gray, C.M., König, P., Engel, A.K. and Singer, W. (1989) Oscillatory responses in cat visual cortex exhibit
inter-columnar synchronization which reflects global stimulus properties. Nature, London, 338, 334–337.
Gray, C.M. and McCormick, D.A. (1996) Chattering cells: superficial pyramidal neurons contributing to the
generation of synchronous oscillations in the visual cortex. Science, Washington, 274, 109–113.
Grinvald, A., Lieke, E.E., Frostig, R.D. and Hildesheim, R. (1994) Cortical point-spread function and long-range
lateral interactions revealed by real-time optical imaging of macaque monkey primary visual cortex.
Gulledge, A.T. and Jaffe, D.B. (1998) Dopamine decreases the excitability of layer V pyramidal cells in the rat
prefrontal cortex. Journal of Neuroscience, 18, 9139–9151.
Gutkin, B.S. and Ermentrout, G.B. (1998) Dynamics of membrane excitability determine interspike interval
variability: a link between spike generation mechanisms and cortical spike train statistics. Neural Computation,
10, 1047–1065.
Hamada, Y., Miyashita, E. and Tanaka, H. (1999) Gamma-band oscillations in the “barrel cortex” precede rat’s
exploratory whisking. Neuroscience, 88, 667–671.
Harden, D.G., King, D., Finlay, J.M. and Grace, A.A. (1998) Depletion of dopamine in the prefrontal cortex
decreases the basal electrophysiological activity of mesolimbic dopamine neurons. Brain Research, 794,
96–102.
Hartline, H.K. (1938) The response of single optic nerve fibres of the vertebrate eye to illumination of the retina.
American Journal of Physiology, 121, 400–415.
Hashimoto, I. (2000) High-frequency oscillations of somatosensory evoked potentials and fields. Clinical Neuro-
physiology, 17, 309–320.
Heggelund, P. and Albus, K. (1978) Response variability and orientation discrimination of single cells in striate
cortex of cat. Experimental Brain Research, 32, 197–211.
Hendry, S.H.C. and Reid, R.C. (2000) The Koniocellular Pathway in Primate Vision. Annual Review of Neuro-
science, 23, 127–153.
Herz, A., Creutzfeldt, O. and Fuster, J. (1964) Statistical properties of neuron activity in the ascending visual
system [in German]. Kybernetik, 2, 61–71.
Holt, G.R., Softky, W.R., Koch, C. and Douglas, R.J. (1996) Comparison of discharge variability in vitro and
in vivo in cat visual cortex neurons. Journal of Neurophysiology, 75, 1806–1814.
Hübener, M., Shoham, D., Grinvald, A. and Bonhoeffer, T. (1997) Spatial relationships among three columnar
systems in cat area 17. Journal of Neuroscience, 17, 9270–9284.
Hunt, C. and McIntyre, A.K. (1960) An analysis of fiber diameter and receptor characteristics of myelinated
cutaneous afferent fibers in cat. Journal of Physiology (London), 153, 99–112.
Imig, T.J. and Adrian, H.O. (1977) Binaural columns in the primary field (AI) of cat auditory cortex. Brain
Research, 138, 241–257.
Issa, N.P., Trepel, C. and Stryker, M.P. (2000) Spatial frequency maps in cat visual cortex. Journal of Neuro-
science, 20, 8504–8514.
Jancke, D., Akhavan, A.C., Erlhagen, W., Giese, M., Steinhage, A., Schöner, G. and Dinse, H.R. (1996)
Population coding in cat visual cortex reveals nonlinear interactions as predicted by a neural field model.

In: Ch. von der Malsburg, W. von Seelen, J.C. Vorbrüggen and B. Sendhoff (eds), Artificial Neural
Networks—ICANN 96. Berlin, Heidelberg: Springer Verlag, pp. 641–648.
Jancke, J., Erlhagen, W., Dinse, H.R., Akhavan, A.C., Giese, M., Steinhage, A. and Schöner, G. (1999) Para-
metric population representation of retinal location: Neuronal interaction dynamics in cat primary visual
Jancke, D. (2000) Orientation formed by a Spot’s Trajectory: a two-dimensional population approach in primary
visual cortex. Journal of Neuroscience, 20, RC86.
Johnson, K.O. and Hsiao, S.S. (1992) Neural mechanisms of tactual form and texture perception. Annual Review
Jones, J.P. and Palmer, L.A. (1987) The two-dimensional spatial structure of simple receptive fields in cat striate
cortex. Journal of Neurophysiology, 58, 1187–1211.
Jones, M.S. and Barth, D.S. (1997) Sensory-evoked high-frequency (gamma-band) oscillating potentials in
somatosensory cortex of the unanesthetized rat. Brain Research, 768, 167–176.
Jones, M.S., MacDonald, K.D., Choi, B., Dudek, F.E. and Barth, D.S. (2000) Intracellular correlates of
fast (>200 Hz) electrical oscillations in rat somatosensory cortex. Journal of Neurophysiology, 84,
1505–1518.
Kalt, T., Akhavan, A.C., Jancke, D. and Dinse, H.R. (1996) Dynamic population coding in rat somatosensory
cortex. Society for Neuroscience Abstracts, 22, 105.
Kanizsa, G. (1976) Subjective contours. Scientific American, 234, 48–52.
Kapadia, M.K., Westheimer, G. and Gilbert, C.D. (2000) Spatial distribution of contextual interactions in
primary visual cortex and in visual perception. Journal of Neurophysiology, 84, 2048–2062.
Kiang, N.Y., Watanabe, T., Thomas, E.C. and Clark, L.F. (1965) Discharge patterns of single fibers in the cat
auditory nerve. MIT, Cambridge.
Kilgard, M.P. and Merzenich, M.M. (1998) Plasticity of temporal information processing in the primary auditory
cortex. Nature, Neuroscience, 1, 727–731.
Kim, D.S., Matsuda, Y., Ohki, K., Ajima, A. and Tanaka, S. (1999) Geometrical and topological relationships
between multiple functional maps in cat primary visual cortex. NeuroReport, 10, 2515–2522.
Kirman, J.H. (1974) Tactile apparent movement: the effects of number of stimulators. Journal of Experimental
Psychology, 103, 1175–1180.
Kirschfeld, K. (1992) Oscillations in the insect brain: do they correspond to the cortical gamma-waves of
vertebrates? Proceedings National Academy of Science, U.S.A., 89, 4764–4768.
Kirschfeld, K., Feiler, R. and Wolf-Oberhollenzer, F. (1996) Cortical oscillations and the origin of express
saccades. Proceedings of the Royal Society of London, series B, 263, 459–468.
Koerber, K.C., Pfeiffer, R.R., Warr, W.B. and Kiang, N.Y. (1966) Spontaneous spike discharges from single
units in the cochlear nucleus after destruction of the cochlea. Experimental Neurology, 16, 119–130.
Koffka, K. (1935) Principles of Gestalt Psychology. London: Routledge and Kegan Paul.
Kohonen, T. and Hari, R. (1999) Where the abstract feature maps of the brain might come from. Trends in
Kopecz, K., Schöner, G., Spengler, F. and Dinse, H.R. (1993) Dynamic properties of cortical, evoked (10 Hz)
oscillations. Theory and experiment. Biological Cybernetics, 69, 463–473.
Krause, F. and Eckhorn, R. (1983) Receptive fields for motion determined for different types of cat visual
neurons. Neuroscience Letters, S14, 209.
Krone, G., Mallot, H.A., Palm, G. and Schüz, A. (1986) Spatio-temporal receptive fields: a dynamical
model derived from cortical architectonics. Proceedings of the Royal Society of London, series B, 226,
421–444.
Krukowski, A.E. (2000) A model of cat primary visual cortex and its thalamic input. Doctoral Dissertation,
University of California, San Francisco.
Langner, G. (1997) Temporal processing of pitch in the auditory system. Journal of New Music Research, 26,
116–132.
Lennie, P. (1998) Single units and visual cortical organization. Perception, 27, 889–935.
LeVay, S., Stryker, M.P. and Shatz, C.J. (1978) Ocular dominance columns and their development in layer IV of
the cat’s visual cortex: a quantitative study. Journal of Comparative Neurology, 179, 223–244.
Lieke, E.E., Frostig, R.D., Arieli, A., Ts’o, D.Y., Hildesheim, R. and Grinvald, A. (1989) Optical imaging of
cortical activity: real-time imaging using extrinsic dye-signals and high resolution imaging based on slow
intrinsic-signals. Annual Review of Physiology, 51, 543–559.
Lin, C.S., Friedlander, M.J. and Sherman, S.M. (1979) Morphology of physiologically identified neurons in the
visual cortex of the cat. Brain Research, 172, 344–348.
Livingstone, M. and Hubel, D. (1988) Segregation of form, color, movement, and depth: anatomy, physiology,
and perception. Science, Washington, 240, 740–749.
Livingstone, M.S. (1998) Mechanisms of direction selectivity in macaque V1. Neuron, 20, 509–526.

Luria, A.R. (1973) The working brain. London: Penguin.

MacDonald, K.D. and Barth, D.S. (1995) High frequency (gamma-band) oscillating potentials in rat somato-
sensory and auditory cortex. Brain Research, 694, 1–12.
MacKay, D.M. (1958) Perceptual stability of a stroboscopically lit visual field containing self-luminous objects.
101–104.
Manley, J.A. and Mueller-Preuss, P. (1978) Response variability in the mammalian auditory cortex: an objection
to feature detection? Federation Proceedings, 37, 2355–2359.
Martin, K.A. and Whitteridge, D. (1984) The relationship of receptive field properties to the dendritic shape of
neurones in the cat striate cortex. Journal of Physiology (London), 356, 291–302.
Martin, K.A. (1988) The Wellcome Prize lecture. From single cells to simple circuits in the cerebral cortex.
Quarterly Journal of Experimental Physiology, 73, 637–702.
Maunsell, J.H. (1992) Functional visual streams. Current Opinion in Neurobiology, 2, 506–510.
McIlwain, J.T. (1986) Point images in the visual system: new interest in an old idea. Trends in Neuroscience, 9,
354–358.
Mendola, J.D., Dale, A.M., Fischl, B., Liu, A.K. and Tootell, B.H. (1999) The representation of illusory and real
contours in human cortical visual areas revealed by functional magnetic resonance imaging. Journal of
Neuroscience, 19, 8560 –8572.
Merigan, W.H. and Maunsell, J.H. (1993) How parallel are the primate visual pathways? Annual Review of
Merzenich, M.M., Schreiner, C., Jenkins, W. and Wang, X. (1993) Neural mechanisms underlying temporal
integration, segmentation, and input sequence representation: some implications for the origin of learning
disabilities. Annals of the New York Academy of Science, 682, 1–22.
Métin, C. and Frost, D.O. (1989) Visual responses of neurons in somatosensory cortex of hamsters with experi-
mentally induced retinal projections to somatosensory thalamus. Proceedings of the National Academy of
Science, U.S.A., 86, 357–361.
Metzger, W. (1932) Versuch einer gemeinsamen Theorie der Phänomene Fröhlichs und Hazelhoffs und Kritik
ihrer Verfahren zur Messung der Empfindungszeit. Psychologische Forschung, 16, 176–200.
Middlebrooks, J.C., Dykes, R.W. and Merzenich, M.M. (1980) Binaural response-specific bands in primary
auditory cortex (AI) of the cat: Topographical organization orthogonal to isofrequency contours. Brain
Research, 181, 31–48.
Middlebrooks, J.C. and Knudsen, E.I. (1984) A neural code for auditory space in the cat’s superior colliculus.
Miller, L.M. and Schreiner, C.E. (2000) Stimulus-based state control in the thalamocortical system. Journal of
Moore, C,I., Nelson, S.B. and Sur, M. (1999) Dynamics of neuronal processing in rat somatosensory cortex.
Moore, G.P., Perkel, D.H. and Segundo, J.P. (1966) Statistical analysis and functional interpretation of neuronal
spike data. Annual Reviews of Physiology, 28, 493–522.
Movshon, J.A., Thompson, I.D. and Tolhurst, D.J. (1978) Spatial and temporal contrast sensitivity of neurones in
areas 17 and 18 of the cat’s visual cortex. Journal of Physiology (London), 283, 101–120.
Naveen, K.V., Srinivas, R. and Nirmala, K.S. (1998) Differences between congenitally blind and normally
sighted subjects in the P1 component of middle latency auditory evoked potentials. Perceptual and Motor
Skills, 86, 1192–1194.
Nawrot, M.P., Riehle, A., Aertsen, A. and Rotter, S. (2000) Spike Count Variability in Motor Cortical Neurons.
European Journal of Neuroscience, 12 (suppl.11), abstr. 225.27.
Nelson, S.B. (1991a) Temporal interactions in the cat visual system. I. Orientation-selective suppression in the
visual cortex. Journal of Neuroscience, 11, 344–356.
Nelson, S.B. (1991b) Temporal interactions in the cat visual system. II. Suppressive and facilitatory effects in the
lateral geniculate nucleus. Journal of Neuroscience, 11, 357–368.
Nelson, S.B. (1991c) Temporal interactions in the cat visual system. III. Pharmacological studies of cortical
suppression suggest a presynaptic mechanism. Journal of Neuroscience, 11, 369–380.
Nicolelis, M.A., Ghazanfar, A.A., Faggin, B.M., Votaw, S. and Oliveira, L.M. (1997) Reconstructing the
engram: simultaneous, multisite, many single neuron recordings. Neuron, 184, 529–537.
Nicolelis, M.A., Ghazanfar, A.A., Stambaugh, C.R., Oliveira, L.M., Laubach, M., Chapin, J.K., Nelson, R.J. and
Kaas, J.H. (1998) Simultaneous encoding of tactile information by three primate cortical areas. Nature,
Nicolelis, M.A.L. (1996) Beyond maps: a dynamic view of the somatosensory system. Brazilian Journal of
Medical and Biological Research, 29, 401–412.
Nicolelis, M.A.L. (1999) Methods in Neural Ensemble Recordings (M.A.L. Nicolelis, ed.). Baton Rouge: CRC
Press.

Niemeyer, W. and Starlinger, I. (1981) Do blind hear better? Investigations on auditory processing in congenital
early acquired blindness. II. Central functions. Audiology, 20, 510–515.
Noda, H. and Adey, W.R. (1970) Firing of neuron pairs in cat association cortex during sleep and wakefulness.
Nunez, A., Amzica, F. and Steriade, M. (1992) Voltage-dependent fast (20–40 Hz) oscillations in long-axoned
neocortical neurons. Neuroscience, 51, 7–10.
Obermayer, K., Ritter, H. and Schulten, K. (1990) A principle for the formation of the spatial structure of cortical
feature maps. Proceedings of the National Academy of Science, U.S.A., 87, 8345–8349.
Omurtag, A., Kaplan, E., Knight, B. and Sirovich, L. (2000) A population approach to cortical dynamics with an
application to orientation tuning. Network, 11, 247–260.
Orban, G.A., Gulyas, B. and Vogels, R. (1987) Influence of a moving textured background on direction selectiv-
ity of cat striate neurons. Journal of Neurophysiology, 57, 1792–1812.
Parraga, C.A., Troscianko, T. and Tolhurst, D.J. (2000) The human visual system is optimised for processing the
spatial information in natural visual images. Current Biology, 10, 35–38.
Penn, A.A., Riquelme, P.A., Feller, M.B. and Shatz, C.J. (1998) Competition in retinogeniculate patterning
driven by spontaneous activity. Science, Washington, 279, 2108–2112.
Perkel, D.H. and Bullock, T.H. (1968) Neural Coding. Neuroscience Research Program Bulletin, 6, 221–348.
Perkel, D.H., Gerstein, G.L. and Moore, G.P. (1967) Neuronal spike trains and stochastic point processes. I.
The single spike train. Biophysical Journal, 7, 391–418.
Peschanski, M., Guilbaud, G. and Gautron, M. (1980) Neuronal responses to cutaneous electrical and noxious
mechanical stimuli in the nucleus reticularis thalami of the rat. Neuroscience Letters, 20, 165–170.
Petersen, S.E., Miezin, F.M. and Allman, J.M. (1988) Transient and sustained responses in four extrastriate
visual areas of the owl monkey. Experimental Brain Research, 70, 55–60.
Phillips, D.P., Hall, S.E. and Hollett, J.L. (1989) Repetition rate and signal level effects on neuronal responses to
brief tone pulses in cat auditory cortex. Journal of the Acoustic Society of America, 85, 2537–2549.
Podvigin, N.F., Kuperman, A.M. and Chuyeva, I.V. (1974) Space-time properties of excitation and inhibition
and wave processes in the receptive fields of the external geniculate body of the cat. Biofizika, 19, 341–346.
Ramachandran, V.S. and Anstis, S.M. (1983) Perceptual organization in moving patterns. Nature, London, 304,
529–531.
Ramachandran, V.S. and Anstis, S.M. (1986) The perception of apparent motion. Scientific American, 254, 102–109.
Ramachandran, V.S., Ruskin, D., Cobb, S. and Rogers-Ramachandran, D. (1994) On the perception of illusory
contours. Vision Research, 34, 3145–3152.
Recanzone, G.H., Merzenich, M.M., Jenkins, W.M., Grajski, K. and Dinse, H.R. (1992) Topographic reorgan-
ization of the hand representation in cortical area 3b of owl monkeys trained in a frequency discrimination
task. Journal of Neurophysiology, 67, 1031–1056.
Recanzone, G.H., Schreiner, C.E. and Merzenich, M.M. (1993) Plasticity in the frequency representation of
primary auditory cortex following discrimination training in adult owl monkeys. Journal of Neuroscience,
13, 87–103.
Recanzone, G.L., Schreiner, C.E., Sutter, M.L., Beitel, R. and Merzenich, M.M. (1999) Functional organization
of spectral receptive fields in the primary auditory cortex of the owl monkey. Journal of Comparative
Neurology, 415, 460–481.
Reichardt, W. (1961) Autocorrelation, a principle for the evaluation of sensory information by the central nerv-
ous system. In: W.A. Rosenblith (ed.), Sensory Communication. New York: MIT Press, pp. 303–317.
Reid, R.C., Victor, J.D. and Shapley, R.M. (1992) Broadband temporal stimuli decrease the integration time of
neurons in cat striate cortex. Visual Neuroscience, 9, 39–45.
Rettenbach, R., Diller, G. and Sireteanu, R. (1999) Do deaf people see better? Texture segmentation and visual
search compensate in adult but not in juvenile subjects. Journal of Cognitive Neuroscience, 11, 560–583.
Rieke, F., Bodnar, D.A. and Bialek, W. (1995) Naturalistic stimuli increase the rate and efficiency of informa-
tion-transmission by primary auditory afferents. Proceedings of the Royal Society, London, series B, 262,
259–265.
Ringach, D.L., Hawken, M.J. and Shapley, R. (1997) Dynamics of orientation tuning in macaque primary visual
Rockel, A.J., Hiorns, R.W. and Powell, T.P.S. (1980) The basic uniformity in structure of the neocortex. Brain,
103, 221–244.
Röder, B., Rosler, F., Henninghausen, E. and Nacker, F. (1996) Event-related potentials during auditory and
somatosensory discrimination in sighted and blind human subjects. Cognitive Brain Research, 4, 77–93.
Röder, B., Rosler, F. and Neville, H.J. (1999) Effects of interstimulus interval on auditory event-related
potentials in congenitally blind and normally sighted humans. Neuroscience Letters, 264, 53–56.
Röder, B., Rosler, F. and Neville, H.J. (2000) Event-related potentials during auditory language processing in
congenitally blind and sighted people. Neuropsychologia, 38, 1482–1502.
Rodieck, R.W. (1967) Maintained activity of cat retinal ganglion cells. Journal of Neurophysiology, 30,
1043–1071.

Romanski, L.M., Tian, B., Fritz, J., Mishkin, M., Goldman-Rakic, P.S. and Rauschecker, J.P. (1999). Dual streams of
auditory afferents target multiple domains in the primate prefrontal cortex. Nature, Neuroscience, 2, 1131–1136.
Sauve, K. (1999) Gamma-band synchronous oscillations: recent evidence regarding their functional significance.
Consciousness and Cognition, 8, 213–224.
Schiller, P.H., Finlay, B.L. and Volman, S.F. (1976) Short-term response variability of monkey striate neurons.
Schreiner, C.E. (1998) Spatial distribution of responses to simple and complex sounds in primary auditory
cortex. Audiology and Neurotology, 3, 104–122.
Schreiner, C.E., Read, H.L. and Sutter, M.L. (2000) Modular organization of frequency integration in primary
auditory cortex. Annual Review of Neuroscience, 23, 501–529.
Schurmann, M. and Basar, E. (1999) Alpha oscillations shed new light on relation between EEG and single
neurons. Neuroscience Research, 33, 79–80.
Shadlen, M.N. and Newsome, W.T. (1998) The variable discharge of cortical neurons: implications for connec-
tivity, computation, and information coding. Journal of Neuroscience, 18, 3870–3896.
Shamma, S.A. and Versnel, H. (1995) Ripple analysis in ferret primary auditory cortex. II. Prediction of unit
responses to arbitrary spectral profiles. Auditory Neuroscience, 1, 207–320.
Sharma, J., Angelucci, A. and Sur, M. (2000) Induction of visual orientation modules in auditory cortex. Nature,
London, 404, 841–847.
Shatz, C.J. and Stryker, M.P. (1988) Prenatal tetrodotoxin infusion blocks segregation of retinogeniculate affer-
ents. Science, Washington, 242, 87–89.
Sherman, S.M. and Guillery, R.W. (2001) Exploring the thalamus. Academic Press: San Diego.
Sheth, B.R., Sharma, J., Rao, S.C. and Sur, M. (1996) Orientation maps of subjective contours in visual cortex.
Science, Washington, 274, 2110 –2115.
Shevelev, I.A. (1987) Temporal processing of orientation signals in visual cortex. Sensornje Sistemje, 1,
73–82.
Shoham, D., Hübener, M., Schulze, S., Grinvald, A. and Bonhoeffer, T. (1997) Spatio-temporal frequency domains
and their relation to cytochrome oxidase staining in cat visual cortex. Nature, London, 385, 529–533.
Sillito, A.M., Grieve, K.L., Jones, H.E., Cudeiro, J. and Davis, J. (1995) Visual cortical mechanisms detecting
focal orientation discontinuities. Nature, London, 378, 492–496.
Simons, D.J. (1985) Temporal and spatial integration in the rat S1 vibrissa cortex. Journal of Neurophysiology,
54, 615–635.
Singer, W. (1998) Consciousness and the structure of neuronal representations. Philosophical Transactions of
the Royal Society of London, series B, 353, 1829–1840.
Softky, W.R. and Koch, C. (1993) The highly irregular firing of cortical cells is inconsistent with temporal integ-
ration of random EPSPs. Journal of Neuroscience, 13, 334–350.
Stein, B.E. and Meredith, M.A. (1990) Multisensory integration. Neural and behavioral solutions for dealing
with stimuli from different sensory modalities. Annals of the New York Academy of Sciences, 608, 51–65.
Steriade, M. (2000) Corticothalamic resonance, states of vigilance and mentation. Neuroscience, 101,
243–276.
Steriade, M. and Llinas, R.R. (1988). The functional states of the thalamus and the associated neuronal interplay.
Physiology Reviews, 68, 649–742.
Suga, N. (1984) The extent to which biosonar information is represented in the bat auditory cortex. In:
G.M. Edelman, W.E. Gall and W.M. Cowan (eds), Dynamic Aspects of Neocortical Function. New York:
Wiley and Sons, pp. 315–373.
Sukov, W. and Barth, D.S. (1998) Three-dimensional analysis of spontaneous and thalamically evoked gamma
oscillations in auditory cortex. Journal of Neurophysiology, 79, 2875–2884.
Sur, M., Pallas, S.L. and Roe, A.W. (1990) Cross-modal plasticity in cortical development: Differentiation and
specification of sensory neocortex. Trends in Neuroscience, 13, 227–233.
Sutter, M.L. and Schreiner, C.E. (1991) Physiology and topography of neurons with multipeaked tuning curves
in cat primary auditory cortex. Journal of Neurophysiology, 65, 1207–1226.
Swadlow, H.A. (1988) Efferent neurons and suspected interneurons in binocular visual cortex of the awake
rabbit: receptive fields and binocular properties. Journal of Neurophysiology, 59, 1162–1187.
Swadlow, H.A. (1989) Efferent neurons and suspected interneurons in S-1 vibrissa cortex of the awake rabbit:
receptive fields and axonal properties. Journal of Neurophysiology, 62, 288–308.
Swadlow, H.A. (1990) Efferent neurons and suspected interneurons in S-1 forelimb representation of the awake
rabbit: receptive fields and axonal properties. Journal of Neurophysiology, 63, 1477–1498.
Swindale, N.V., Matsubara, J.A. and Cynader, M.S. (1987) Surface organization of orientation and direction
selectivity in cat area 18. Journal of Neuroscience, 7, 1414–1427.
Swindale, N.V. (1991) Coverage and the design of striate cortex. Biological Cybernetics, 65, 415–424.
Swindale, N.V. (2000) How many maps are there in visual cortex? Cerebral Cortex, 10, 633–643.
Swindale, N.V., Shoham, D., Grinvald, A., Bonhoeffer, T. and Hübener, M (2000) Visual cortex maps are
optimized for uniform coverage. Nature, Neuroscience, 3, 822–826.

Theunissen, F.E. and Doupe, A.J. (1998) Temporal and spectral sensitivity of complex auditory neurons in
nucleus HVc of male zebra finches. Journal of Neuroscience, 18, 3786–3802.
Thomson, A.M. and West, D.C. (1993) Fluctuations in pyramid-pyramid excitatory postsynaptic potentials
modified by presynaptic firing pattern and postsynaptic membrane potential using paired intracellular
recordings in rat neocortex. Neuroscience, 54, 329–346.
Tolhurst, D.J., Movshon, J.A. and Dean, A.F. (1983) The statistical reliability of signals in single neurons in cat
and monkey visual cortex. Vision Research, 23, 775–785.
Tomko, G.J. and Crapper, D.R. (1974) Neuronal variability: Non-stationary responses to identical visual stimuli.
Tovee, M.J. and Rolls, E.T. (1992) Oscillatory activity is not evident in the primate temporal visual cortex with
static stimuli. NeuroReport, 3, 369–372.
Traub, R.D., Jefferys, J.G.R. and Whittington, M.A. (1999) Fast Oscillations in Cortical Circuits. Cambridge,
Mass: MIT Press.
Tremblay, N., Bushnell, M.C. and Duncan, G.H. (1993) Thalamic VPM nucleus in the behaving monkey. II.
Response to air-puff stimulation during discrimination and attention tasks. Journal of Neurophysiology, 69,
753–763.
Troyer, T.W., Krukowski, A.E., Priebe, N.J. and Miller, K.D. (1998) Contrast-invariant orientation tuning in cat
visual cortex: thalamocortical input tuning and correlation-based intracortical connectivity. Journal of
Tsodyks, M., Kenet, T., Grinvald, A. and Arieli, A. (1999) Linking spontaneous activity of single cortical
neurons and the underlying functional architecture. Science, Washington, 286, 1943–1946.
Ungerleider, L.G. and Haxby, J.V. (1994) ‘What’ and ‘where’ in the human brain. Current Opinion in Neurobiol-
ogy, 4, 157–165.
Ungerleider, L.G. and Mishkin, M. (1982) Two cortical visual systems. In: D.J. Ingle, M.A. Goodale, R.J.W.
Mansfield (eds), Analysis of Visual Behavior. MIT Press, Cambridge, pp. 549–586.
Van Essen, D. and Kelly, J. (1973) Correlation of cell shape and function in the visual cortex of the cat. Nature,
London, 241, 403–405.
van Gisbergen, J.A., Grashuis, J.L., Johannesma, P. and Vendrik, A.J. (1975) Spectral and temporal characteris-
tics of activation and suppression of units in the cochlear nuclei of the anesthetized cat. Experimental Brain
Research, 23, 367–386.
Vogels, R., Spileers, W. and Orban, G.A. (1989) The response variability of striate cortical neurons in the behaving
monkey. Experimental Brain Research, 77, 432–436.
von der Heydt, R., Peterhans, E. and Baumgartner, G. (1984) Illusory contours and cortical neuron responses.
von Krosigk, M., Bal, T. and McCormick, D.A. (1993) Cellular mechanisms of a synchronized oscillation in the
thalamus. Science, Washington, 261, 361–364.
von Melchner, L., Pallas, S.L. and Sur, M. (2000) Visual behaviour mediated by retinal projections directed to
the auditory pathway. Nature, London, 404, 871–876.
von Seelen, W., Mallot, H.P., Krone, G. and Dinse, H.R. (1986) On information processing in the cat’s visual
cortex. In: G. Palm and A. Aertsen (eds), Brain Theory. Berlin, Heidelberg: Springer, pp. 49–79.
Wang, L.Y. and Kaczmarek, L.K. (1998) High-frequency firing helps replenish the readily releasable pool of
synaptic vesicles. Nature, London, 394, 384–388.
Wang, X., Merzenich, M.M., Sameshima, K. and Jenkins, W.M. (1995) Remodeling of hand representation in
adult cortex determined by timing of tactile stimulation. Nature, London, 378, 71–75.
Warren, R.M. (1970). Restoration of missing speech sounds. Science, Washington, 167, 392–393.
Weliky, M., Bosking, W.H. and Fitzpatrick, D. (1996) A systematic map of direction preference in primary
visual cortex. Nature, London, 379, 725–728.
Werner, G. and Mountcastle, V.B. (1963) The variability of central neural activity in a sensory system, and its
implications for the central reflection of sensory events. Journal of Neurophysiology, 26, 958–977.
Wiemer, J., Spengler, F., Joublin, F., Stagge, P. and Wacquant, S. (2000) Learning cortical topography from
spatiotemporal stimuli. Biological Cybernetics, 82, 173–187.
Wiesel, T.N., Hubel, D.H. and Lam, D.M. (1974) Autoradiographic demonstration of ocular-dominance columns
in the monkey striate cortex by means of transneuronal transport. Brain Research, 79, 273–279.
Wilbur, W.J. and Rinzel, J. (1983) A theoretical basis for large coefficient of variation and bimodality in
neuronal interspike interval distributions. Journal of Theoretical Biology, 105, 345–368.
Willis, W.D., Maunz, R.A., Foreman, R.D. and Coulter, J.D. (1975) Static and dynamic responses of spinotha-
lamic tract neurons to mechanical stimuli. Journal of Neurophysiology, 38, 587–600.
Winer, J. (1992) The functional architecture of the medial genicuate body and the primary auditory cortex.
In: D.B. Webster, A.N. Popper and R.R. Fay (eds), The Mammalian Auditory Pathway: Neuroanatomy. New
York: Springer, pp. 222–409.
Winer, J.A., Larue, D.T. and Huang, C.L. (1999) Two systems of giant axon terminals in the cat medial geniculate
body: Convergence of cortical and GABAergic inputs. Journal of Comparative Neurology, 413, 181–197.

Wörgötter, F. and Eysel, U. (2000) Context, state and the receptive fields of striate cortex cells. Trends in
Yan, J. and Suga, N. (1999) Corticofugal amplification of facilitative auditory responses of subcortical combination-
sensitive neurons in the mustached bat. Journal of Neurophysiology, 81, 817–824.
Zhang, K. and Sejnowski, T.J. (1999) A theory of geometric constraints on neural activity for natural three-
dimensional movement. Journal of Neuroscience, 19, 3122–3145.
Zhang, Y., Suga, N. and Yan, J. (1997) Corticofugal modulation of frequency processing in bat auditory system.

14 Plastic-Adaptive Properties of Cortical Areas
Hubert R. Dinse1 and Gerd Boehmer2
1
Institute for Neuroinformatics, Dept. of Theoretical Biology,
Group Experimental Neurobiology, Ruhr-University Bochum, Germany
2
Institute of Physiology and Pathophysiology, Gutenberg-University, Mainz, Germany
Correspondence to: Hubert R. Dinse, Institute for Neuroinformatics,
Dept. of Theoretical Biology, Group Experimental Neurobiology,
Ruhr-University Bochum, ND 04 D-44780 Bochum, Germany
Tel: +49-234-32-25565; FAX: +49-234-32-14209
e-mail: hubert.dinse@neuroinformatik.ruhr-uni-bochum.de
This chapter summarizes recent findings about plastic changes in adult early sensory and motor
cortices. We discuss mechanisms leading to enduring changes of synaptic efficacy and of neural
response behaviour in terms of receptive fields and cortical representational maps, with special
emphasis on behavioural and perceptual consequences of cortical reorganizations, after peripheral
lesion or injury, differential use and training. Given the assumption that the presence of plastic-
adaptive abilities are a prerequisite for coping successfully with an ever-changing environment,
we focus on comparative aspects, evaluating apparent similarities and dissimilarities emerging
across different modalities. Most of the material reviewed is from animal studies that allow the
study of adaptations and underlying mechanisms induced under a large variety of natural and
laboratory conditions, at all levels from channels and synapses, to groups of neurones and cortical
maps. Owing to the recent development of non-invasive imaging technologies, it has become
possible to explore the significance of cortical plasticity for humans, occurring in “every-day-
life”. Massive and enduring reorganizations are present for all areas and modalities discussed,
corroborating the view that cortical maps and response properties are in a permanent state of
use-dependent fluctuation. We discuss various mechanisms controlling synaptic plasticity, the
role of input statistics and attention, the top-down modulation of plastic changes, the “negative”,
(maladaptive) consequences of cortical reorganization, and the coding and decoding of adapta-
tional processes. Despite the convincing evidence for profound reorganizational changes in all
areas, specifically for injury-related plasticity, there exist also clear modality-specific differences,
an observation holding at both the cellular and the systemic level. Differences include magnitude
of changes, readiness of induceability and specificity of neural parameters that are affected. While
reorganization of somatosensory and auditory cortex appears to follow comparable rules and
constraints, adult visual cortex plasticity shows a number of particularities, indicating that visual
cortical maps might be more difficult to change. We discuss a number of possible explanations
based on different levels of abstraction. Among these are differences in control mechanisms of
synaptic plasticity, the limiting character of complex topological maps, and the possible limitations
of the metaphor of “use”, as a driving force of adult plasticity.
KEYWORDS: Bienenstock-Cooper model, cortical map, Hebbian rules, input probability, lesion,
LTP, LTD, maladaptivity, neural coding, perceptual learning, receptive field, sensory areas,
simultaneity, synaptic plasticity, synaptic control mechanism, training, top-down modulation
311
312 Hubert R. Dinse and Gerd Boehmer
1. GENERAL ASPECTS OF PLASTIC ADAPTIVE CAPACITIES
Heritable features evolving during evolutionary time spans are of ultimate advantage for
survival and are, without exception, structurally fixed. To cope successfully with the
ongoing changes of environmental conditions occurring during the lifespan of individuals,
additional mechanisms are required that allow rapid and effective adaptations that are not
specified by genetic constraints. In spite of the substantial amount of adaptational capacities,
systems must possess sufficient generic stability to allow secure processing. Conceivably,
there is a trade-off between modifiability and stability. In contrast to developmental plasti-
city, adaptations of adult brains do not, in the first place, rely on maturational and growth
processes. Specifically, for learning-induced alterations there is a consensus that there is
a crucial role for so-called functional plasticity, based on rapid and reversible modifica-
tions of synaptic efficacy. However, large-scale amputations have been shown to involve
sprouting and outgrowth of afferent connections into neighbouring regions at cortical and
subcortical levels (Florence et al., 1998; Jain et al., 2000). Technically, the term “adapta-
tion” is used in a rather neutral sense, i.e. there are no implicit assumptions whether “adap-
tational changes” might yield a positive or negative outcome. Given this general overview, it
appears conceivable that plastic-adaptive capacities of various forms represent a general
and ubiquitous cortical feature present in all sensory modalities as well as in higher
cortical areas. Before summarizing details and possible deviations from that scheme, some
basic properties of cortical plasticity, observable in early sensory areas are briefly
discussed.
1.1. Driving Forces that Lead to Adaptational Changes

What are crucial factors that are potentially effective in inducing changes of neural represen-
tations? We assume a dynamically-maintained steady state of representations that emerged
during development and adulthood from maturational and learning processes, reflecting the
history of adaptation to a “mean environment”. Mean environment is defined as the accumu-
lated and idiosyncratic experience of an individual. Adaptational processes are assumed to
operate on these representations, and long-lasting changes are likely to occur when sensory
input patterns are altered such that they deviate from the mean environment.
(a) A simple way to alter the average steady state is by changing the input statistics.
Features that are specifically effective in driving adaptational changes are simultaneity,
repetition, or more generally, spatio-temporal proximity. These changes occur without
involving attention or other cognitive processes. Accordingly, a class of non-cognitive
adaptations is largely based on bottom-up processing.
(b) Alternatively or additionally, attention can be drawn to a stimulus, thereby selecting it
in comparison to others. Furthermore, the relevance of a stimulus can change depend-
ent on context, history and behavioural task, thereby modifying the processing of the
physically-defined attributes. There is general agreement that modification of early
sensory processing by attention and stimulus relevance reflects top-down influences
arising from cognitive processes.
(c) Reinforcement of learning processes by reward or punishment usually accelerates
adaptational processes. Such influences are assumed to be mediated by specific brain
regions modifying early sensory processing.

Comparative Aspects of Cortical Plasticity 313
1.2. The Hebbian Metaphor

A central paradigm in the description and analysis of cortical plasticity is built around the
Hebbian concept (1949) stating that episodes of high temporal correlation between pre- and
postsynaptic activity are prerequisite for inducing changes of synaptic efficacy. Historic-
ally, the idea that cooperative processes are crucially involved in generating long-lasting
changes of excitability can be traced back to the 19th century (James, 1890). In fact, since
the time of Hebb, the aspect of simultaneity has become a metaphor in neural plasticity,
although the exact role of Hebbian mechanisms in use-dependent plasticity remains
controversial (Carew et al., 1984; Fox and Daw, 1993; Granger et al., 1994; Montague and
Sejnowski, 1994; Joublin et al., 1996; Buonomano and Merzenich, 1996; Edeline, 1996;
Cruikshank and Weinberger, 1996a,b; Ahissar et al., 1998). It has been suggested that the
definition of Hebbian mechanisms must be extended beyond “simultaneity”, in the sense
of strict coincidence, to cover all facets arising from learning processes. Such a definition
must include a large number of pre- and post synaptic patterns, as well as a broad time-
window for what neural systems regard as “simultaneous”.
1.3. Use-dependent Plasticity as a Basis for Perceptual and Motor Skills

One of the striking features of use-dependent plasticity is the correlation of cortical
changes with performance. The acquisition of skills has often been used as an index for
the build-up of implicit memories. Implicit memories are acquired automatically and with-
out consciousness. Many repetitions, over a long time and without higher-level cognitive
processes, are sufficient to improve perceptual and motor skills. This non-cognitive feature
in combination with many repetitions characterizes an important aspect of use-dependent
neural plasticity. It has been speculated that use-dependent plasticity might be strongly
related to, if not a substrate for, implicit memory function.
1.4. Perceptual Learning

Perceptual learning is the ability to improve perceptual performance by training and prac-
tice (cf. Gibson, 1953). In this sense, perceptual learning occurs largely independent of
conscious experience (cf. Fahle and Poggio, 2001). Perceptual learning is usually charac-
terized by a high specificity to stimulus parameters such as location or orientation of
a stimulus, with little generalization of what is learned to other locations or to other stimulus
configurations. Selectivity and locality of this type implies that the underlying neural
changes most probably occur within early cortical representations that contain well-ordered
topographic maps to allow for this selectivity, but where generalization with respect to
spatial location and orientation has not yet occurred. Transfer of newly acquired abilities
is considered an important marker of that processing level at which changes are most
likely to occur: limited generalization indicates high locality of effects in early represent-
ations. In contrast, transfer of learned abilities implies the involvement of higher process-
ing levels, as is often observed in task and strategy learning. There is increasing evidence
that changes in early cortical areas might be more directly linked to perceptual learning
than previously thought (Karni and Sagi, 1991; Recanzone et al., 1992a; Schoups et al., 1995;
Crist et al., 1997; Fahle, 1997; Fahle and Poggio, 2001). While there is a large literature on

perceptual learning in the visual system, much less data are available for comparable
experiments in the auditory and somatosensory system.
2. POST-ONTOGENETIC PLASTICITY OF CORTICAL MAPS

AND RECEPTIVE FIELDS
It is useful to distinguish between two different forms of adult plasticity:
• Lesion-induced plasticity subsumes the reorganization after injury and lesion,

induced either centrally or at the periphery. This type of plasticity refers to aspects of
compensation and repair of functions that have been acquired prior to injury or lesion.
• Training- and learning-induced reorganization is often denoted as “use-dependent
plasticity” and describes plastic changes parallel to the behavioural improvement of
performance, i.e. the acquisition of perceptual and motor skills.
In view of the fact that amputation changes the pattern of use entirely, a more accurate
distinction would be between “lesion-induced” vs “non-lesion-induced” plasticity. Just
how far the two forms are different, or possibly based on similar mechanisms, is a matter
of ongoing debate.
There are a number of detailed reviews providing an excellent overview covering all
facets of cortical plasticity in early sensory and motor areas (Merzenich et al., 1988; Kaas,
1991; Scheich et al., 1991; Eysel, 1992; Garraghty and Kaas, 1992; Sameshima and
Merzenich, 1993; Donoghue, 1995; Weinberger, 1995; Cruishank and Weinberger, 1996a;
Edeline, 1996; Dinse et al., 1997a; Kaas and Florence, 1997; Sanes and Donoghue, 1997;
Buonomano and Merzenich, 1998; Gilbert, 1998; Nicolelis et al., 1998a; Rauschecker,
1999; Recanzone, 2000; Dinse and Merzenich, 2002). In the following, the main focus is
on comparative aspects. It should be noted that this comparative approach is hindered by
the fact that there are, with rare exceptions, few studies explicitly exploring possible area- and
modality-specific properties of cortical plasticity.
2.1. Lesion-Induced Plasticity

Large-scale reorganizations were first described following digit amputation or deafferen-
tation in the primary somatosensory cortex of cats, monkeys and racoons (Kalaska and
Pomeranz, 1979; Kelahan et al., 1981; Rasmusson, 1982; Merzenich et al., 1983, 1984;
Wall and Cusick, 1984; Calford and Tweedale, 1988; Florence and Kaas, 1995; Kaas et al.,
1999). The main result was that the cortical territory representing the skin surface
removed by amputation or deafferentation did not remain silent, but was activated by
stimulation of bordering skin sites. Major topographic changes for cutaneous afferent
representations were limited to a cortical zone extending about 1 mm on either side of the
initial boundaries of the amputated digits. These early data indicated that the sensory
cortical representations in adults were not hard-wired, but retain a self-organizing capacity
operational throughout life (Merzenich et al., 1984).
Much more dramatic cortical reorganizations were reported after mapping the cortex of
monkeys that had undergone deafferentation of the dorsal roots (C2-T4) several years
before, thereby depriving a cortical area of over 1 cm2 of its normal input from the arm and

hand (Pons et al., 1991). These authors found that all of the deprived area had developed
novel responses to neighbouring skin areas, including the face and chin. It is now well
established that comparable large-scale remodeling occurs in human somatosensory and
motor cortical areas, weeks or months following limb amputation (Fuhr et al., 1992; Cohen
et al., 1993; Kew et al., 1994; Yang et al., 1994; Flor et al., 1995, 1998; Knecht et al., 1998)
implying that similar (if not identical) rules govern lesion-induced plastic reorganizations
in humans (see also section 2.5. “Therapeutic consequences of cortical plasticity”).
A series of lesion experiments performed a couple of years later in auditory and visual
system confirmed the tremendous capacities of the cortex for reorganization described for
SI. Several months after a restricted unilateral lesion of the cochlear of guinea-pigs, the
area of contralateral auditory cortex representing the lesioned frequency range was partly
occupied by an expanded representation of sound frequencies adjacent to the lesioned
frequency range. Thresholds at their new characteristic frequencies (CFs) were close to
normal (Robertson and Irvine, 1989). In extending these experiments, it was found that
unilateral restricted cochlear lesions in adult cats altered the topographic representations
of the lesioned cochleas along the tonotopic axis of primary auditory cortex, extending up
to 3 mm rostal to the area of normal representation, with no apparent topographic order
within this enlarged representation (Rajan et al., 1993). Interestingly, no comparable signs
of plastic changes of the frequency map were found in the dorsal cochlear nucleus of adult
cats following unilateral partial cochlear lesions (Rajan and Irvine, 1998). A striking over-
representation of the frequency corresponding to the border area of the cochlear lesion has
been observed after amikacin-induced cochlear lesions in primary auditory cortex of the
adult chinchilla. The amount of reorganization was similar in extent to that previously
seen during development (Kakigi et al., 2000).
Fairly large lesions (5 by 10 degrees of visual angle) of the retina markedly altered the
systematic representations of the contralateral eye in primary and secondary visual cortex,
when matched inputs from the ipsilateral eye were also removed. Cortical neurones that
normally have receptive fields in the lesioned region of the retina acquired new receptive
fields in portions of the retina surrounding the lesions (Kaas et al., 1990). In another study,
removal of visual inputs by focal binocular retinal lesions resulted in an immediate
increase in receptive field size for cortical cells with receptive fields near the edge of the
retinal scotoma. After a few months even the cortical areas that were initially silenced by
the lesion recovered visual activity, representing retinotopic loci surrounding the lesion
(Gilbert and Wiesel, 1992). Anatomical studies showed that the spread of geniculocortical
afferents is insufficient to account for the cortical recovery (Darian-Smith and Gilbert,
1995), indicating that the topographic reorganization within the cortex was largely due to
synaptic changes intrinsic to the cortex, most probably through the system of long-range
horizontal connections. In a series of studies, the reorganizational properties of adult
visual cortex following various forms of retinal injuries and lesions has been well established
(Schmid et al., 1996; Calford et al., 1999, 2000).
Quantitative studies of the response characteristics of visual neurones after retinal
lesions indicated that these neurones develop fairly normal processing features. After three
months of recovery, newly activated units exhibited strikingly normal orientation tuning,
direction selectivity, and spatial frequency tuning, when high-contrast stimuli were used.
However, contrast thresholds of most neurones were abnormally elevated, and the maximum
response amplitude under optimal stimulus conditions was significantly reduced. The results
suggest that the striate cortical neurones reactivated during topographic reorganization are

capable of sending functionally meaningful signals to more central structures, provided

that the visual scene contains relatively high contrast images (Chino et al., 1995).
In animals that were allowed to recover from a complete monocular deactivation for up
to several months, there was also rearrangement of the retinotopic maps. However, in this
case, most neurones in the deprived peripheral representation remained unresponsive to
visual stimuli even more than one year after treatment (Rosa et al., 1995). This is in
marked contrast with the extensive reorganization that is observed in the central repres-
entation of V1 after restricted retinal lesions. The low potential for reorganization of the
monocular sector of V1 demonstrates that the capacity for plasticity of mature sensory
representations varies with location in cortex: even small pieces of cortex, such as the
monocular crescent representations, may not reorganize completely if certain conditions are
not met. These results suggest the existence of natural boundaries that may limit the pro-
cess of reorganization of sensory representations.
Taken together, all sensory areas display a well-documented capacity for profound
reorganizations following peripheral lesions. However, there might exist differences in the
magnitude of changes. Whether this reflects some modality-specific limitation of the
visual cortex to reorganize after large retinal lesions requires further investigations. (For a
discussion of perceptual consequences of lesion-induced reorganizations, and general
aspects of maladaptive consequences see also section 2.5. below “Therapeutic conse-
quences of cortical plasticity”).
2.2. Training- and Learning-Induced Use-Dependent Reorganization

Perceptual skills improve with training (cf. Gibson, 1953). Accordingly, one of the key
questions in cortical plasticity is how cortical changes are linked to parallel changes of
perceptual and/or motor performance. This question requires the simultaneous assessment
of both neurophysiological and behavioural changes.
For example, Recanzone and coworkers showed that tactile frequency discrimination
training in the adult owl monkeys over several months led to a significant reduction of
frequency discrimination threshold (Recanzone et al., 1992a). When the cortical areas
representing the skin area of the trained fingers were mapped, large-scale cortical reorgan-
ization became apparent, which included changes of receptive fields and of topography of
cortical representational maps. Most notable, there was a significant correlation between
the enlargement of cortical territory representing the skin surface stimulated during training
and the improvement in performance, indicating a close relationship between cortical and
perceptual changes (Recanzone et al., 1992b). In addition, sinusoidal stimulation of the
trained skin elicited larger-amplitude responses, peak responses earlier in the stimulus
cycle, and temporally sharper responses, than did stimulation applied to control skin sites.
Analysis of cycle histograms for neuronal responses in area 3b revealed that the decreased
variance of each stimulus cycle could account for behaviourally-measured frequency dis-
crimination improvements (Recanzone et al., 1992c). With the somatosensory system as
an example, these data demonstrated for the first time a direct relation between cortical
plasticity and improvement of performance.
A largely identical approach was taken for an analysis of training-induced changes in
AI. Monkeys, trained for several weeks to discriminate small differences in the frequency
of sequentially presented tonal stimuli, revealed a progressive improvement in performance
with training. At the end of the training period, the tonotopic organization of Al was defined

electrophysiologically. The cortical representation, the sharpness of tuning, and the latency
of the response were greater for the behaviourally-relevant frequencies of trained monkeys
when compared to the same frequencies of control monkeys. Notably, the cortical area of
representation was the only studied parameter that was correlated with behavioural per-
formance. These results demonstrate that attended natural stimulation during a long-term
training protocol can modify the tonotopic organization of Al in the adult primate, and that
this alteration is correlated with changes in perceptual acuity (Recanzone et al., 1993).
While there is a large body of information about perceptual learning in the visual
domain in human subjects (see also section 2.7), little is known about parallel changes in
visual cortex. In a study on perceptual learning in humans (discrimination of orientations),
subjects showed a marked improvement over days, which was highly specific for position
and orientation (Vogels and Orban, 1985; Schoups et al., 1995). However, the precise
nature of the accompanying changes still remains unclear. As in the studies performed in
SI and AI, one could expect that there would be a recruitment of cells toward the trained
orientation. However, in contrast to the previous SI/AI studies, no comparable expansion in
representation was found. Instead, the proportion of cells recorded in primary visual cortex
(in monkeys trained to discriminate orientations) that preferred the orientation to which they
had been trained, was not larger than the proportion of cells preferring any other orientation
(Schoups, 2001). Parallel experiments using autoradiographic labelling of deoxyglucose
as an indirect marker of neural activity (based on the close relation between oxygen
consumption and firing activity) confirmed the electrophysiological data. No broadening
of the orientation columns was observed as a consequence of the perceptual learning, and
thus no recruitment occurred of cells responding to the trained orientation (Schoups, 2001).
Moreover, outside the primary visual cortex no major and systematic changes have been
found. Recordings in the inferotemporal cortex from rhesus monkeys trained to judge whether
or not two successively presented gratings differed in orientation revealed no consistent
effects either on the responsiveness or on the orientation tuning (Vogels and Orban, 1994).
More recent findings on the firing behaviour of neurones in striate cortex, recorded
from monkeys trained in an orientation discrimination task, provided evidence for rather
complex changes. In the population of trained neurones, those that preferred the trained
orientation exhibited a lower firing rate than the neurones preferring other orientations
(Schoups et al., 2000). At first sight this result seems counterintuitive. However, models
of perceptual learning involving orientation discrimination (Qian and Matthews, 1999)
predicted that lower firing rates by the neurones that prefer the trained orientation could
lead to selective changes in the tuning patterns of neurones that prefer the orientations bor-
dering the one trained, which then would lead to a better performance in the discrimination
task. These results raise some interesting possibilities related to the coding of plastic changes
(cf. also Dinse and Merzenich, 2002; Schoups, 2001), which will be discussed later (see also
section 2.6. “Coding of plastic changes”).
The paradigm of “modified use” as a determinant of cortical organization has been
applied in a large number of investigations, mostly performed in somatosensory cortex
(cf. Dinse and Merzenich, 2002), with few studies in other modalities. In this approach, plastic
changes are analyzed in a rather natural context, where the link between behaviour and
cortical reorganization is often less quantifiable, but still intuitively obvious. For example,
the implications of episodes of differential use, following nursing behaviour, occurring dur-
ing the normal life-span of an animal, have been shown in a study of lactating rats. The SI
representation of the ventral trunk skin was significantly larger than in matched postpartum

non-lactating or virgin controls (Xerri et al., 1994). After training squirrel monkeys on
a task involving retrieval of small objects, which required skilled use of the digits, their
motor digit representations expanded, whereas their evoked-movement wrist/forearm
representational zones contracted. In a second task, a monkey was trained in a key-turning
task. In this case, the representation of the forearm expanded, whereas the digit representa-
tional zones contracted. Movement combinations that were used more frequently after
training were selectively magnified (Nudo et al., 1996). Interestingly, repetitive motor
activity alone appeared not to produce functional reorganization of cortical maps indicating
that skill acquisition or motor learning is a prerequisite factor for induction (Plautz et al.,
2000; but see section 2.3.2. “Coactivation” on this issue).
In the auditory system, abnormal cochleotopic organization in the auditory cortex of
cats reared in a frequency-augmented environment has been observed (Stanton and Harrison,
1996). For the visual system, Sugita (1996) reported that V1 neurones in monkeys can
develop novel receptive fields to the ipsilateral hemifield after monkeys have worn reversing
spectacles for several months. These studies suggest that visual cortical neurones can in
fact acquire novel inputs, not only from neighbouring retinal areas, but also from distant
nonadjacent areas. This report contradicts earlier findings, according to which the visual
field representation in the striate cortex is rigidly prewired with reference to the anatomical
fovea (cf. Pöppel et al., 1987). It is well established that perceived orientation can be
influenced by previous adaptation to a tilted stimulus (tilt aftereffect), an illusion that
decays rapidly over time. Following short-term adaptation to one stimulus orientation,
systematic “rebound” shifts in orientation preference were observed, that included changes
in orientation tuning away from the adapting stimulus indicating the involvement of wide-
spread network interactions that mediate these effects (Dragoi et al., 2000).
The recent development of non-invasive imaging techniques has made it possible to
study the impact of modified use and practice in humans. Imaging studies performed over
the last few years provided overwhelming evidence that extensive use and practice result in
substantial changes of associated cortical representations. For example, in the somatosen-
sory cortex of blind Braille readers (Pascual-Leone and Torres, 1993; Sterr et al., 1998a,b)
and of string players (Elbert et al., 1995), selective enlargement was found for those cortical
territories representing the digits engaged in more extensive use, as exemplified by the
reading fingers (Braille readers) or the fingering digits (string player). In adults who were
studied before and after surgical separation of webbed fingers, a cortical reorganization of
the finger representation over several millimeters was observed (Mogilner et al., 1993),
a finding reminiscent to what had been reported some years ago for artificial induction of
syndactyly in monkeys (Clark et al., 1988). Subjects engaged in long-term perceptual
training in tactile discrimination revealed changes in responsivensss of the somatosensory
cortex (Spengler et al., 1997). When subjects received passive tactile stimulation of thumb
and little finger over a period of 4 weeks, the representations of the fingers in primary
somatosensory cortex were closer together after training. However, when subjects had to
discriminate stimuli, MEG imaging revealed that the digital representations were further
apart than before. Thus, the same prolonged repetitive stimulation produced two opposite
effects, suggesting that activation in the same region of cortex is specific to different tasks
(Braun et al., 2000).
In order to demonstrate the perceptual relevance of the neural changes induced by a tactile
coactivation protocol (see also section “2.3.2. Coactivation”), spatial discrimination per-
formance was investigated in human subjects who underwent a similar passive coactivation,

as described in the animal study by Godde et al. (1996). A small skin area on the index
finger was coactivated. Discrimination thresholds were used as a marker of reorganiza-
tional effects on human perception. It was found that two hours of coactivation were sufficient
to drive a significant improvement of the spatial discrimination performance (Godde et al.,
2000), demonstrating the potential role of pure input statistics for the induction of cortical
plasticity without involving cognitive factors such as attention or reinforcement. A com-
bined assessment of discrimination thresholds and recording of somatosensory evoked
potentials in human subjects revealed that the individual gain of discrimination perform-
ance was correlated with the amount of cortical reorganization, as inferred from the shifts
of the location of the “N20” dipole (Pleger et al., 2001).
For the human motor system, similar fast adaptational regulations have been reported:
using mapping of responses to transcranial magnetic stimulation (TMS), in human subjects
who had unilateral immobilization of the ankle joint (i.e. they had to wear a cast for
a couple of weeks), the area of motor cortex representing the tibialis anterior muscle were
significantly reduced compared to the representation of the unaffected leg. The amount of
areal reduction was correlated with the duration of immobilization, an effect rapidly
reversed by voluntary muscle contractions (Liepert et al., 1995). An hour of synchronous
movements of the thumb and foot resulted in a reduction of the distance of the centre of
gravity of their respective output maps in the primary motor cortex, whereas asynchronous
movements evoked no significant changes, indicating that similar principles of coactivation
hold for both the sensory and motor system (Liepert et al., 1999).
In highly skilled musicians, functional magnetic source imaging revealed an enlargement
of dipole moments for piano tones, but not for pure tones of similar fundamental frequency,
which was correlated with the age at which musicians began to practice (Pantev et al., 1998).
In addition, musicians with absolute pitch were characterized by distinct neural activities
in the auditory cortex (Hirata et al., 1999). Similarly, auditory cortical representations for
tones of different timbre (violin and trumpet) were enhanced compared to sine tones in
violinists and trumpeters, preferentially for timbres of the instrument of training (Pantev
et al., 2001). Reminiscent of the reorganizations after frequency-discrimination training in
monkeys (Recanzone et al., 1993), human subjects have been reported to show plastic
reorganization in the auditory cortex induced by frequency-discrimination training over
several weeks. Changes consisted of an increase of amplitude of the slow auditory evoked
(wave “N1m”) and mismatch field (Menning et al., 2000).
The human visual system is able to determine very precisely the relative positions of
objects in space. Using an artificial scotoma, by occluding part of the visual field, while
a pattern was shown over a surrounding region, resulted in severe mislocalization. This
was due to a strong bias toward the interior of the scotoma, indicating a significant short-term
cortical plasticity in adult human vision (Kapadia et al., 1994). In a psychophysical and
functional imaging study of adaptation to inverting spectacles, subjects showed rapid
adaptation of visuomotor functions within several days, but did not report return of upright
vision. This was corroborated by the functional magnetic resonance images (fMRI) that
failed to show alteration of the retinotopy of early visual cortical areas (Linden et al., 1999),
a finding in contrast to recent animal data demonstrating indeed a functional reversal in
area 17 after months of training (Sugita, 1996). Transcranial magnetic stimulation (TMS)
of the occipital cortex evokes the perception of phosphenes. In human subjects, a reduced
phosphene threshold was detected 45 min after a short period of light deprivation, and this
persisted for the whole deprivation period of 3 h. Similarly, fMRI showed increased visual

cortex activation after 60 min of light deprivation that persisted following 30 min of
re-exposure to light, demonstrating a substantial increase in visual cortex excitability
(Boroojerdi et al., 2000). It was speculated that such changes may underlie behavioural
gains such as a lowered visual recognition thresholds reported in humans associated with
light deprivation (Suedfeld, 1975).
The tilt aftereffect leads to wide-spread changes of visual cortex orientation maps
(Dragoi et al., 2000). When alphanumeric characters were presented to human subjects with
a clockwise tilt, they were perceived as less tilted than the same stimulus horizontally
inverted. In contrast, subjective perception of tilt magnitude for horizontally inverted non-
alphanumeric stimuli was similar to that for non-inverted stimuli reflecting a persistent
sensory recalibration of orientation perception as a result of previous long-term visual
experience (Whitaker and McGraw, 2000).
Taken together, these studies suggest that even small alterations in behaviour due to special
demands imposed in everyday life alter early cortical representations rapidly and revers-
ibly. The human studies summarized confirm the close relation between intensified or
altered use (on the one hand) and enlargement of associated cortical representational maps
(on the other hand), supporting the relevance of the concept of cortical plasticity for everyday
life. From a comparative point of view, it appears fair to state that while there is a large
body of information about use- and experience-related plastic changes in somatosensory
and auditory cortex, comparatively little is known about visual cortex.
2.3. The Role of Input Statistics

Human studies of the type summarized are very helpful in revealing signatures of cortical
plasticity under everyday life conditions. However, these studies are not designed to control
precisely for input pattern. Accordingly, it remains unclear what are the “driving factors”
leading to reorganization. In the case of the blind Braille readers, potential candidates are: the
frequency of finger usage, the spatial pattern of the Braille signs, the spatio-temporal pattern
arising when the finger is moved across the Braille signs, the level of attention, and the dur-
ation of practice. In addition, many lines of evidence have shown that cortical systems adapt
to input patterns characterized by different probabilities, implying that variations of input
statistics alone are sufficient to induce reorganization of cortical maps, i.e. without involving
cognitive processes such as those present in training protocols. Therefore, animal studies are
required that complement and extend human studies by a systematic variation of input pattern.
2.3.1. Intracortical microstimulation

Intracortical microstimulation (ICMS) is used to evoke selective motor responses by
applying current through microelectrodes inserted into defined regions of motor represen-
tations. More recently, this technique has been utilized to study short-term and reversible
plastic changes in various cortical regions, including motor (Nudo et al., 1990; Gu and
Fortier, 1996; Kimura et al., 1996), somatosensory (Dinse et al., 1990; Recanzone et al.,
1992d; Dinse et al., 1993; Spengler and Dinse, 1994; Joublin et al., 1996; Xing and
Gerstein, 1996; Dinse et al., 1997a; Heusler et al., 2000), auditory (Sil’kis and Rapoport,
1995; Maldonado and Gerstein, 1996a,b; Maldonado et al., 1998; Sakai and Suga, 2001) and
visual (Leonhardt et al., 1997; Godde et al., 2002) cortices, as well as thalamic relay nuclei
of the somatosensory system (Dinse et al., 1997a).

A specific advantage of ICMS is that it allows one to investigate locally the properties
of functional plasticity, independent of the peripheral and subcortical pathways and
independent of the constraints provided by particularities of a sensory pathway and its
preprocessing. In a typical ICMS experiment, repetitive electrical pulse trains of very low
currents (usually less than 10 µA) are delivered via a microelectrode. Based on theoretical
calculations, ICMS of that intensity activates a cortical volume of only 50 microns in diameter
(Stoney et al., 1968). The resulting synchronized discharges are assumed to be crucial for
mediating plastic changes. The short time scale and reversibility of ICMS-induced effects
support the hypothesis that modulations of synaptic efficiency in neuronal networks can
occur very rapidly without necessarily involving anatomical changes. Consequently, ICMS
is an ideal method for studying possible modality- and area-specific constraints of cortical
plasticity.
In the rat motor cortex, significant changes in representations of movement were observed
after a few hours of ICMS, these being fully reversible. Changes were characterized by
border shifts up to more than 500 microns (Nudo et al., 1990). Application of ICMS in the
hindpaw representation of the adult rat somatosensory cortex caused an overall but selective
expansion of receptive field size up to 1 mm around the ICMS site (Recanzone et al., 1992d;
Dinse et al., 1993; Spengler and Dinse, 1994). Receptive fields close to the stimulation
site were enlarged, and comprised large skin territories, always including the receptive
field at the ICMS-site, revealing a distance-dependent, directed enlargement towards the
ICMS-receptive field. Early ICMS-related reorganization could already be detected after
15 min of ICMS, and much greater effects emerged after 2 to 3 hours, which were reversible
within 6 to 8 hours after termination of ICMS (Dinse et al., 1993; Spengler and Dinse, 1994).
In the auditory cortex, ICMS induced fast changes in the tonotopic map, and in the
receptive field properties of cells at the electrically stimulated and adjacent electrodes.
There was an enlargement of the cortical domain tuned to the acoustic frequency that had
been represented at the stimulating electrode (Maldonado and Gerstein, 1996a,b; Maldonado
et al., 1998). Comparison of reorganization evoked by focal electric stimulation (ICMS) in
AI of an ecologically highly specialized animal (the mustached bat—Pteronotus parnellii)
and a non-specialized one (the Mongolian gerbil—Meriones unguiculatus) revealed differ-
ences in the ICMS-induced shifts of best frequency, implying differences between special-
ized and nonspecialized (ordinary) areas of the auditory cortex (Sakai and Suga, 2001).
ICMS-induced reorganization in somatosensory and visual cortex of pigmented rats
was compared in individual animals (Leonhardt et al., 1997, 1998). In visual cortex, ICMS
led to small (~20%), but significant expansions of receptive fields for a subpopulation of
neurones with small receptive fields pre-ICMS. Neurones characterized by initially large
RFs did not change. In contrast, RFs recorded in SI in the same animal, exhibited the
well-known several-fold enlargement. In addition, in visual cortex, the time-structure
of the neuronal responses was systematically altered, by suppressing late response com-
ponents, leading to profound changes in the temporal structure of the receptive field
dynamics. Comparable changes have not been observed in SI. As a further difference, in
visual cortex, the observed changes were not reversible within the observation period of
3–6 h after ICMS.
To further investigate the plasticity of functional maps in the visual cortex, orientation-
preference maps were recorded by means of optical imaging. A few hours of ICMS induced
major changes of orientation-preference maps in adult cats (Godde et al., 1999, 2002).
These results showed that orientation-preference maps undergo substantial expansions

reminiscent of non-visual cortical maps. However, changes were much more wide-spread
and enduring, indicating that the large-scale changes of the functional architecture resulted
from a restructuring of the entire underlying cortical network. Parallel electrophysio-
logical single cell recordings revealed distinct shifts of the individual orientation tuning
towards the preferred orientation present at the ICMS site. Again, no changes of receptive
field sizes were found.
Taken together, these results from ICMS experiments imply that sensory cortex, including
visual cortex, is modifiable in adults, both in terms of functional maps and in terms of single
cell properties. However, there appear to exist a number of differences, best documented
for somatosensory and visual cortex, concerning reversibility, spatial range of changes,
and neural response parameters most susceptible to modifications.
2.3.2. Coactivation
A number of protocols have been introduced in which neural activity, necessary to drive
plastic changes, was generated by an associative pairing protocol. In the pioneering studies
by Fregnac and coworkers (Fregnac et al., 1988, 1992), persistent functional changes in
response properties of single neurones of cat visual cortex were induced by a differential
pairing procedure, during which iontophoresis was used to increase artificially the visual
response for a given stimulus, and to decrease the response for a second stimulus. Neuronal
selectivity was nearly always displaced towards the stimulus paired with the reinforced
visual response, thereby leading to long-term modifications of orientation selectivity in about
one third of the neurones tested. The largest changes were obtained at the peak of the critical
period in normally reared and visually deprived kittens, but changes were also observed in
adults. From a conceptual point of view, these findings supported the role of temporal
correlation between pre- and postsynaptic activity in the induction of long-lasting modific-
ations of synaptic transmission in associative learning during development and in adults.
A similar modifiability of response properties of visual cortex neurones in adult cats has
been observed after a conditioning protocol, where the presentation of particular visual
stimuli was repeatedly paired with the iontophoretic application of either GABA or
glutamate to control postsynaptic firing rates (McLean and Palmer, 1998). The modification
in orientation tuning was not accompanied by a shift in preferred orientation, but rather,
responsiveness to stimuli at or near the positively-reinforced orientation was increased
relative to controls, and responsiveness to stimuli at or near the negatively-reinforced
orientation was decreased. These studies are in contrast to a previous study, where lasting
(maximal 1 h) modifications of the receptive fields of neurones in the visual cortex were
observed by pairing visual stimuli with iontophoretic application of the neuromodulators
acetylcholine and noradrenaline or the excitatory amino acids N-methyl-D-aspartate
(NMDA) and L-glutamate in kitten, but not in adult animals (Greuel et al., 1988).
While these experiments emphasize the potential capacities of adult visual cortex for
change of its response properties, it remains an interesting question, why comparable
experiments (drug-pairing) have been performed in the other modalities very rarely (Maalouf
et al., 1998; Shulz et al., 2000). Possibly, induction of plastic changes in adult visual
cortex might be more “difficult” to drive than in other areas using more natural types of
stimulation, without direct drug application (see also below). However, independent of
this speculation, the present findings from drug-pairing experiments point to a rather common
form of modifiability across cortical areas.

In contrast to the cellular pairing protocols, a number of studies utilized a pairing of

adequate (i.e. sensory) stimuli, for the somatosensory cortex (Diamond et al., 1993; Wang
et al., 1995; Godde et al., 1996, 2000) and visual cortex (Eysel et al., 1998). In the study
by Diamond et al. (1993) sensory experience was altered by a few days of “whisker
pairing”: whiskers D2 and either D1 or D3 were left intact, while all other whiskers were
trimmed. During whisker pairing, the receptive fields of cells in barrel D2 changed in
distinct ways: the response to the centre receptive field increased, the response to the
paired surround receptive field nearly doubled, and the response to all clipped, unpaired
surround receptive fields decreased. These findings indicate that a brief change in the
pattern of sensory activity induced by pairing of tactile stimuli can alter the configuration
of cortical receptive fields of adult animals.
To test the hypothesis that consistently non-coincident inputs may be actively segre-
gated from one another in their distributed cortical representations, monkeys were trained
to respond to specific stimulus sequence events (Wang et al., 1995). Animals received
temporally-coincident inputs across fingertips and fingerbases, but distal vs proximal
digit segments were non-coincidentally stimulated. Electrophysiological recordings in
somatosensory cortex (area 3b) showed that synchronously applied stimuli resulted in
integration of inputs in the cortical maps, whereas stimuli applied asynchronously were
segregated by two band-like zones, in which all neurones had multiple digit receptive
fields representing the stimulated skin surfaces. Interestingly, maps derived in the ventro-
posterior portion of the thalamus were not reorganized in an equivalent way, suggesting
that this particular type of representational plasticity appears to be cortical in origin.
In the study by Godde et al. (1996), receptive fields on the hindpaw of adult rats were
used for coactivation. These authors reported reversible reorganization consisting of
a selective enlargement of the cortical territory, and of the receptive fields representing the
co-stimulated skin fields. In addition, a large representation emerged that included a joint
representation of both skin sites. A control protocol applied to only a single skin site
evoked no changes indicating that coactivation was essential for induction (see also
“2.2. Training and learning induced reorganization”).
It has been stressed that passive stimulation, or repetitive motor activity alone appeared
not to produce comparable functional reorganization of cortical maps (Recanzone et al.,
1992b; Plautz et al., 2000). On the other hand, the coactivation studies reported here
showed a clear effect on cortical as well as on perceptual levels, in spite of the fact that
attention was not involved. One explanation is that during the coactivation protocol,
which was on average applied at a rate of 1 Hz for several hours, selected skin regions
were stimulated 10 000 times or more. This is a much stronger stimulation in terms of
stimulus number per time than the monkeys received during the passive discrimination
training. Conceivably, the intensity of the stimulation/movement protocol might be the
crucial factor responsible for its effectiveness.
For the visual system, early studies claimed that retinal stimulation alone does not induce
plastic changes (Buisseret et al., 1978). However, a recent coactivation study, performed
in mature visual cortex, revealed the capacity for significant changes of receptive field
organization: single cortical cells expanded their receptive fields, within minutes, into
previously-unresponsive regions, and changed their functional receptive field structure for
hours after associative co-stimulation of active and primarily-unresponsive regions (Eysel
et al., 1998). While this study corroborates the sensitivity of visual cortex to a coactivation
protocol consisting of natural sensory stimuli, the magnitude of changes are clearly below

those reported for SI. However, given the many differences in methodological details,
more experiments are required to provide a final answer about possible underlying modality-
specific differences.
2.3.3. Classical conditioning

Another type of associative learning, as exemplified by classical conditioning, has been
studied for decades, in several variations, in the auditory cortex (for review see Weinberger
et al., 1990; Weinberger, 1995). Using a classical conditioning protocol, a tone of a given
frequency (as the CS+) was paired with an aversive electrical shock. Tuning curves
recorded in the auditory cortex before and after conditioning revealed a shift in the best
frequencies in the direction of the frequency of the CS+; these shifts lasted up to a few
weeks and could be reversed by extinction training (Diamond and Weinberger, 1986:
Bakin and Weinberger, 1990; Edeline and Weinberger, 1993; Ohl and Scheich, 1996). Most
conspicuously, the approach of classical conditioning to studying aspects of cortical
plasticity appears to be entirely restricted to studies of the auditory system, ruling out any
comparative analysis, although the restriction to the auditory system might, in itself, hint
at some particular modality-specific constraints that might be worth exploring in more
detail. Anecdotally, classical conditioning was discovered by Durup and Fessard in the
visual cortex of humans many decades ago (cited by Weinberger et al., 1990). However, this
was conditioning of the alpha rhythm, and may have involved conditioning of subcortical
control mechanisms, rather than at the cortical level. Such conditioning, is likely to be
important, but is not the focus of the present chapter.
2.4. Pharmacological Modulation of Adult Plasticity

There are many sources modulating cortical responsiveness and plasticity. The major
source of cholinergic inputs that have long been implicated in learning and memory comes
from several groups of neurones within the basal forebrain, which receives inputs from
limbic and paralimbic structures. These inputs have been assumed to represent one example
of a top-down system providing modulatory information of higher-order—presumably
cognitive—processes. For example, in animal experiments, pairing of sensory stimulation
with electrical stimulation of the nucleus basalis has been shown to result in rapid and
selective reorganization (in the somatosensory cortex by Rasmusson and Dykes [1988], in
the auditory cortex by Edeline et al. [1994], Bakin and Weinberger [1996], Bjordahl
et al. [1998] and by Kilgard and Merzenich [1998a]). On the other hand, lesions of the
cholinergic system have been shown to prevent plastic reorganization in the somatosensory
cortex (Baskerville et al., 1997; Sachdev et al., 1998). However, using a whisker pairing
protocol, in which all but a few whiskers were trimmed, the animal’s active use of its
remaining intact whiskers can restore some aspects of plasticity in an acetylcholine-depleted
cortex (Sachdev et al., 2000). Direct administration of acetylcholine (ACh) to cortical
neurones facilitates or suppresses responses to sensory stimuli, and these effects can
endure well beyond the period of ACh application. In primary auditory cortex, analysis of
single neurone frequency receptive fields, before and after such pairing of acoustic stimula-
tion with ACh application revealed that in half of the cases, the receptive field alterations
were highly specific to the frequency of the tone previously paired with ACh (Metherate and
Weinberger, 1990). The involvement of neuromodulatory effects in cortical Hebbian-like

plasticity of acetylcholine (ACh) and noradrenaline (NE) was related to the timing of drug
applications, relative to the conditioning time, their local concentrations and/or the site of
application with respect to the relevant synapses (Ahissar et al., 1996).
Ocular dominance plasticity is strongly expressed in early postnatal life and is usually
assumed to be absent in the mature visual cortex. Local perfusion of kitten visual cortex
with 6-hydroxydopamine (6-OHDA) prevented the effects of monocular deprivation in
kittens, while locally perfused norepinephrine restored visual cortical ocular dominance
plasticity (Kasamatsu et al., 1979). The effect of norepinephrine perfusion was seen both
in kittens and, though to a lesser degree, in older animals which had outgrown the susceptible
period. More recently, it was demonstrated that activation of cAMP-dependent protein
kinase A could restore ocular dominance plasticity in visual cortex of adult cats (Imamura
et al., 1999). These findings indicate that various forms of visual cortex organization can
be affected, and persistently modified in adults. More generally, the former data are in line
with the well-documented catecholaminergic and cholinergic modulation of post-ontogenetic
cortical plasticity well established for auditory and somatosensory cortex.
2.5. Therapeutic Consequences of Cortical Plasticity

The final outcome of reorganizational processes must not necessarily be beneficial. There
is increasing evidence that abnormal perceptual experiences, such as the phantom limb
sensation, arise from reorganizational changes induced by the amputation of the limb (Flor
et al., 1995, 1998). In amputees, a number of perceptual correlates of cortical reorganiza-
tions have been described, such as a precise topographic mapping of the phantom onto the
face area, these being explained on the basis of the topography of the border of the face-hand
maps (Ramachandran et al., 1992; Halligan et al., 1993; Aglioti et al., 1997). In patients
with chronic pain, the power of the early evoked magnetic field, elicited by painful stimu-
lation, was elevated relative to that elicited by the same stimulation in healthy controls.
Furthermore, this enlargement was a function of the chronicity of pain (Flor et al., 1997).
Repetitive strain injuries, such as focal dystonia, have a high prevalence in workers who
perform heavy schedules of rapid alternating movements, or repetitive, sustained, coordin-
ated movements. It has been hypothesized that use-dependent plastic changes, as reviewed
in this chapter, may cause these injuries (Byl et al., 1996, 1997). Monkeys trained in repet-
itive hand closing and opening developed typical signs of movement control disorders.
Electrophysiological recordings within the primary somatosensory cortex revealed a
de-differentiation of cortical representations of the skin of the trained hand, manifested by
receptive fields that were 10 to 20 times larger than normal (Byl et al., 1996). A recent
study using MEG in musicians suffering from focal hand dystonia revealed a smaller
distance between the representations of the affected digits in somatosensory cortex,
compared to the same digits in non-musician controls (Elbert et al., 1998) indicating
similar neural changes in humans as a consequence of repetitive strain injuries.
The negative outcome of neuroplasticity may also play a major role in some forms
of age-related changes. It has been suggested that reorganizational processes lead to
maladaptive changes, as a result of walking impairments, developed in rats of high age as
a secondary response to muscle atrophy and other factors promoting limited agility (Spengler
et al., 1995; Jürgens and Dinse, 1997a; Dinse and Merzenich, 2002; Dinse, 2001).
In the auditory domain, some forms of dysfunctions in normal phonological processing,
which are critical to the development of oral and written language, have been speculated

to derive from initial difficulties in perceiving and producing basic sensory-motor informa-
tion in rapid succession, emphasizing the crucial role of temporal parameters. In fact,
when children with a particular type of language-based learning deficit were engaged in
adaptive training of their temporal processing skills, they showed a marked improvement
in their abilities to recognize brief and fast sequences of non-speech and speech stimuli.
This suggests that the reorganizational changes are specifically sensitive to temporal
parameters of the input (Tallal et al., 1993, 1996; Merzenich et al., 1996).
People with amputations often have the feeling that the amputated limb is still present
(phantom limb sensation). Subjective tinnitus, the hearing of a disturbing tone or noise in
the absence of a real sound source, shares many similarities with the sensation of phantom
limb, experienced by many amputees. Therefore, tinnitus has been thought of as an auditory
phantom phenomenon (Jastreboff, 1990; Lockwood et al., 1998; Rauschecker, 1999).
A marked shift of the cortical representation of the tinnitus frequency into an area adjacent
to the expected tonotopic location was observed in subjects suffering from tinnitus. Import-
antly, a strong positive correlation was found between the subjective strength of the tinnitus
and the amount of cortical reorganization (Mühlnickel et al., 1998), indicating that tinnitus
is related to plastic alterations in auditory cortex. Studies using 2-deoxyglucose autoradio-
graphy in gerbils treated with salicylate (known to generate tinnitus) demonstrated increased
activation in areas of the auditory cortex (Wallhäusser-Franke et al., 1996).
Cochlear implants (CI) are a frequent measure to provide sound perception in patients
with sensorineural hearing loss. Utilizing the critical period for speech acquisition, clinical
data suggest that children implanted before 2 years of age have an excellent chance of
acquiring speech understanding. For implanted children, maturational delays for cortically-
evoked potentials, that approximated the period of auditory deprivation prior to implant-
ation, were reported (Ponton et al., 1996). In cats implanted with multichannel intracochlear
electrodes, long-term electrical CI stimulation was found to induce substantial reorgan-
ization of cortical auditory maps, consisting of a selective enlargement of that territory
representing the frequency representations stimulated during chronic CI (Dinse et al.,
1997b). It was suggested that the outcome of these reorganizations was due to Hebbian
mechanisms, utilizing the simultaneity induced by the CI stimulation strategy (continuous
interleaved sampler) in which at high stimulation rates all frequency channels are stimu-
lated virtually simultaneously (viz within a single millisecond). Notably, the CIS strategy
has proven highly effective in human patients in providing a high level of open speech
understanding (Wilson et al., 1991). Similar reorganizational changes were observed in
cats deafened and chronically CI-stimulated as adults (Dinse et al., 1997b, 1998; Godde
et al., 1998). Major differences in cortical response distributions on the ectosylvian gyrus
of adult cats due to deafening were also observed in long-term deafened animals. The
authors speculated that these changes may reflect electrode-specific effects or reorganiza-
tional changes, as a consequence of the altered inputs (Raggio and Schreiner, 1999).
Recent imaging data obtained using positron emission tomography (PET), in pre-
lingually deaf patients before and after cochlear implantation support the relevance of
these animal findings for human patients (Lee et al., 2000). After cochlear implantation,
these authors found a positive correlation between the size of the hypometabolic area and
a hearing-capability score. Accordingly, several lines of evidence indicate that the underlying
plastic adaptational properties of cortical auditory neurones might provide the substrate
involved in mediating the highly variable improvement of open speech understanding
with practice, often observed in patients with hearing aids.

That adult visual cortex is indeed also capable of long-term changes, induced by simple
training procedures, comes from recent studies in which partially blind subjects obtained
some restitution of their visual field (Kasten et al., 1998). In visual restitution training,
visual stimuli were presented on a computer screen in such a manner that the majority of
stimuli appeared in the transition zone, which is usually located in the border region
between the intact and damaged visual-field, as well as near the border of the transition
zone and the defective field. When post-chiasma patients were trained for 6 months, (1 h per
day), subjects showed a 30% improvement in the ability to detect visual stimuli. In optic
nerve patients, the effects were even more pronounced. While in the past, partial blindness
after brain injury has been considered non-treatable, these data are in line with a profound
capacity of the visual cortex to reorganize, even in adults.
Of particular interest are findings on cross-modal plasticity in blind subjects, this
contributing to sensory compensation when vision is lost early in life (Cohen et al., 1997;
Weeks et al., 2000, for a general account of cross-modal plasticity; see also Pallas, this
volume). To identify differences in cross-modal reorganization, depending on the time of
onset of blindness, and thereby distinguishing effects due to ontogenetic or post-ontogenetic
plastic processes, blind subjects were studied by means of positron emission tomography,
to identify cerebral regions activated in association with Braille reading. In the congenitally
blind and early-onset blind groups, the occipital cortex was strongly activated, but this did
not occur in the late-onset blind group. These results indicate that the susceptible period
for this form of functionally relevant cross-modal plasticity does not extend beyond
14 years (Cohen et al., 1999). To determine whether the visual cortex receives input from the
somatosensory system during a Braille reading task, positron emission tomography (PET)
was used to measure activation in Braille readers blinded in early life. Blind subjects
showed activation of primary and secondary visual cortical areas during tactile tasks,
whereas normal controls showed deactivation. Importantly, a simple tactile stimulus that
did not require discrimination produced no activation of visual areas (Sadato et al., 1996).
Comparing behavioural and electrophysiological markers of spatial tuning within central
and peripheral auditory space in congenitally blind and normally sighted but blindfolded
adults, the hypothesis was tested that the effects of visual deprivation might be more
pronounced for processing peripheral sounds. In fact, blind participants displayed local-
ization abilities that were superior to those of sighted controls, but only when attending to
sounds in peripheral auditory space. Electrophysiological recordings obtained at the same
time revealed sharper tuning of early spatial attention mechanisms in the blind subjects.
Differences in the scalp distribution of brain electrical activity between the two groups
suggest compensatory reorganization in the blind, which may contribute to the improved
spatial resolution for peripheral sound sources (Röder et al., 1999).
Taken together, the maladaptive consequences of cortical plasticity are more and more
acknowledged as a major factor in various forms of dysfunctions, an assumption apparently
valid across modalities.
2.6. Coding of Plastic Changes

How are plastic changes coded? What neural response parameters are affected by the various
forms of manipulations leading to reorganizations, and the parallel changes of perception
and behavior? The studies discussed so far have in common that they almost exclusively
describe reorganizational changes in terms of receptive field size and in size of cortical

representational territory. Particularly, the new imaging techniques such as fMRI allow
one to study adaptational changes in humans, describing neural representations in terms of
activation size of cortical maps. At least one simple rule-of-thumb appears to hold:
extensive use leads to enlarged cortical territories, while limited use or no-use results in
a reduction of cortical representational size, indicating a form of proportionality between
representational area and use. Representational size correlates with the number of neurones
activated by a given task or stimulation. This view implies that enhanced performance is at
least partially achieved by recruitment of processing resources. However, a recent animal
study suggested that exceptions might exist (Polley et al., 1999b): allowing an animal to
use its deprived receptor organ in active exploration appeared to determine the direction of
plastic changes in the adult cortex. Further studies are needed to explore whether a similar
potential for a use-dependent direction of reorganizational changes holds true in normal,
nondeprived animals. From a more general point of view, this study suggests that the outcome
of plastic processes might depend on far more subtle constraints imposed by the individual
task than previously thought. In fact, as discussed for the visual cortex, a simple recruitment
after perceptual learning of orientation discrimination could not be demonstrated (see also
section 2.2. “Training and learning induced reorganization”).
More recently, temporal aspects of processing, i.e. aspects of coding in the time domain,
have been recognized as an additional and highly significant candidate code. As a con-
sequence, aspects of synchronicity and correlated activity have been intensively studied,
revealing that cooperativity among many neurones is indeed subject to profound modi-
fication during plastic reorganization. (This has been shown in the somatosensory cortex
by Dinse et al. [1990, 1993]; Faggin et al. [1997] and by Ghazanfar et al. [2000], in the
auditory cortex by Ahissar et al. [1992, 1998] and Maldonado et al. [1996a,b] and in the
motor cortex by Laubach et al. [2000]).
These findings imply that changes in temporal coding are crucial for our understanding
of use-dependent plasticity. Accordingly, a critical step for the investigation of how distrib-
uted cell assemblies process behaviourally-relevant information is therefore the introduction
of methods for data analysis that can identify functional neuronal interactions within high-
dimensional data sets (cf. Nicolelis, 1999). Laubach et al. (2000) applied such methods by
chronically recording from neuronal ensembles located in the rat motor cortex. Based on
such an elaborate approach they could demonstrate that motor learning was correlated with
an increase in the experimenter’s ability to predict a correct or incorrect single trial, based
on measures of neuronal ensemble activity such as firing rate, temporal patterns of firing, and
correlated firing.
On the other hand, temporal processing, i.e. the computation of sequential events, which
is particularly important in the auditory system, is still poorly understood. Under natural
conditions, stimuli never appear in isolation. Therefore, timing and sequencing impose
severe temporal constraints that modulate neurone responses (Zucker, 1989; Chance et al.,
1998). There is, in fact, clear experimental evidence that repetitively applied stimuli alter the
cortical response behaviour, as compared to a single stimulus (Gardner and Costanzo
1980; Lee and Whitsel, 1992; Dinse, 1994; Merzenich et al., 1993; Tommerdahl et al.,
1998; Polley et al., 1999a). So far, only a few studies have explored how far temporal
processing is affected and altered by plastic reorganizations. For example, as described
above, Recanzone et al. (1992c) demonstrated that behavioural training of a frequency-
discrimination task affected entrainment of repetitive stimuli in the somatosensory cortex.
To test whether experience can modify the maximum rate of following in adult rats, trains

of brief tones with random carrier frequency, but fixed repetition rate, were paired with
electrical stimulation of the nucleus basalis. This was continued 300–400 times per day
for 20–25 days. Pairing nucleus basalis stimulation with 5-p.p.s. stimuli markedly
decreased the cortical response to rapidly presented stimuli, whereas pairing with 15-p.p.s.
stimuli significantly increased the maximum cortical following rate, indicating an extensive
cortical remodeling of temporal response properties (Kilgard and Merzenich, 1998b).
In the studies on age-related changes of the hindpaw representation in old rats, the
neural response behaviour following repetitive stimulation was studied with trains of tact-
ile stimuli of variable interstimulus intervals (ISIs). Dramatic impairment of repetition
coding and input sequence representations were observed in old rats as compared to young
controls (Jürgens and Dinse, 1995), and comparable changes of the neural input sequence
representation were found in rats with artificially induced walking alterations (Jürgens and
Dinse, 1997b). As discussed in the next main section of this chapter, there is an extensive
literature about changes of temporal aspects in in-vitro studies.
Taken together, neural changes as a consequence of adaptational mechanisms include
a large number of both spatial and temporal parameters of sensory processing. However,
even under normal conditions, i.e. without involving adaptive processes, we have a poor
understanding of both sensory processing and how performance is coded. That is why it is
not clear what is meant by receptive field size. Is it “good” when a tuning curve gets
sharper? “Good” for what? As exemplified by the study of Schoups et al. (2000), perceptual
learning can be accompanied by rather unexpected changes: those neurones that preferred
the trained orientation exhibited a lower firing rate than the neurones preferring other
orientations. Conceivably, an apparent lack of plastic capacities might simply reflect hidden
changes in parameter regimens presently not recognized or understood.
2.7. Is There an Area-specificity of Particular Cortical Visual Cortex Plasticity?

Though not representative, inspection of the quantity of publications as provided by Medline
offers some insight into areas regarded important and accessible for scientific explorations.
Comparing papers published in the field of somatosensory and visual plasticity revealed
a number of interesting differences. When normalized to the absolute number of published
papers in both fields, the same percentage (about 5%) were devoted to the exploration of
“plasticity”. Searching among these papers for “developmental plasticity” revealed 18%
for visual, but only 5%” for somatosensory cortex, a discrepancy already noted by Wein-
berger (1995). “LTP” plus “NMDA” were found in 17% of visual cortex papers, but only in
6% dealing with somatosensory cortex. On the other hand, “adult reorganization” showed
up in 4% of visual, but in 24% of somatosensory cortex studies. Similarly, “adult plasticity
of cortical maps and receptive fields” was found in 8% of visual, but in 24% of somato-
sensory cortex papers. Finally, “perceptual learning” revealed 0.5% in the visual, but only
0.1% in the tactile modality. Combined, even if treated with ample caution, these data
imply the existence of some particular differences between these areas concerning plastic
changes.
These particularities are reflected in the discrepancy that, as shown above, training and
learning induce powerful cortical reorganizations, but most of what we know about
cortical plasticity in adults comes from experiments in somatosensory or auditory cortex.
However, as summarized in the next section, in-vitro studies using slice preparations
demonstrate very convincingly the existence of the whole spectrum of cellular mechanisms

in adult visual cortex crucial for mediating synaptic plasticity. Moreover, recent findings
on perceptual learning in the visual and other domains imply the modifiability of adult
visual cortex.
Although cortical maps are widely believed to emerge in the developing brain by activity-
dependent mechanisms, the apparent stability of the basic layout of orientation preference
maps of visual cortex has raised the suspicion that orientation-preference maps may be
governed not only by activity-dependent processes, but may even be pre-specified intrins-
ically (Blakemore, 1977; Sengpiel et al., 1998; Miller et al., 1999). In the visual cortex,
the precise match of orientation is a prerequisite for stereoscopic vision. Whether visual
experience is responsible for the match was tested in a reverse-suturing experiment, in
which kittens were raised so that both eyes were never able to see at the same time.
A comparison of the layout of the two maps formed under these conditions showed them
to be virtually identical. Considering that the two eyes never had common visual experience,
this indicates that correlated visual input is not required for the alignment of orientation
preference maps (Gödecke and Bonhoeffer, 1996). It was therefore suggested that the geometry
of functional maps in the visual cortex might be intrinsically determined, while the relative
strength of representation of different response properties can be modified through visual
experience (Sengpiel et al., 1998). On the other hand, kittens reared in a striped environment
responded to all orientations, but devoted up to twice as much cortical area to the experi-
enced orientation as to the orthogonal one. This effect has been attributed to the instructive
role of visual experience whereby some neurones shift their orientation preferences
toward the experienced orientation. Thus, although cortical orientation maps are remark-
ably rigid, in the sense that orientations that have never been seen by the animal are still
represented and occupy a large portion of the cortical territory, visual experience can
nevertheless alter neuronal responses to oriented contours (Sengpiel et al., 1999).
While these latter data refer to the critical sensitive period, there have been a number of
studies many years ago that reported a substantial capacity of the adult visual cortex to
reorganize. Creutzfeldt and coworkers reported that adult cats exposed to a visual environ-
ment consisting only of vertical stripes showed clear signs of plastic changes. The number
of neurones sensitive to the vertical orientations relative to those sensitive to horizontal
was markedly decreased (Creutzfeldt and Heggelund, 1975). In a previous study, Spinelli
and coworkers analyzed response properties of visual cortical neurones in kittens which
had viewed (from birth) horizontal lines with one eye and vertical lines with the other
eye. Neurones with horizontal preferred orientations could be activated only by the
eye exposed to horizontal lines, the analogous result was found for the vertical line. There
was a consistent lack of binocularity (Hirsch and Spinelli, 1970). In a continuing study,
these animals were re-exposed to a normal environment after about 2 months, for up to
19 months. After that period, the animals regained a high, but variable amount of binocularity
(Spinelli et al., 1972) indicating that plastic capacities of visual cortical units extend well
into adulthood (see their paper for an extensive discussion).
Taking the above results together, there appears at present to be a lack of striking readi-
ness for adult visual cortex to reorganize as reported for the other modalities. Just how far
these differences reflect a genuine area-specificity, or the outcome of the experimenters’
belief that the adult visual system displays a higher degree of rigidity, remains a matter of
debate. At any rate, evidence is accumulating that the adult visual cortex can be modified
as well, although the parameters affected, and the modes of induction of plastic changes
may be different.

3. CELLULAR MECHANISMS INVOLVED IN CORTICAL PLASTICITY
As far as neuronal mechanisms possibly involved in cortical plasticity are concerned,

Hebbian synapses are thought to play an important role not only in cortical development
in young animals, but also in cortical reorganization in adult animals (Buonomano and
Merzenich, 1998). In the Hebbian rule (Hebb, 1949) a synaptic input to a neurone is
strengthened when it repeatedly or persistently causes the postsynaptic neurone to dis-
charge. Thus, the examination of cellular mechanisms involved in neuronal plasticity is
mainly focused on the alteration of synaptic efficacy by adequate stimulation of axonal
inputs to small populations of neurones, or to single pyramidal cells.
3.1. Synaptic Plasticity: The Hippocampus

Several forms of activity-dependent alteration of synaptic efficacy have been established
since the first demonstration of long-lasting potentiation (LTP) of synaptic transmission
from axons of the perforant path to neurones in the dentate gyrus of the hippocampus (Bliss
and Lømo, 1973). In this study LTP was induced by tetanic stimulation in vivo. Subsequently,
most studies on mechanisms involved in the induction of hippocampal synaptic plasticity
were performed on brain slices. High-frequency stimulation (HFS; e.g. 1 sec at 100 Hz) of
the Schaffer collaterals of CA3 pyramidal cells results in LTP of EPSPs recorded from
pyramidal cells of the CA1 subfield, lasting at least 30 min and in most cases up to several
hours. This form of synaptic plasticity of the hippocampus has been intensively studied
(for reviews see Bliss and Collingridge, 1993; Malenka and Nicoll, 1993). Short-lasting
forms of enhancement of synaptic strength (for review see Zucker, 1989) include short-term
potentiation (STP) lasting up to 30 min, post-tetanic potentiation (PTP) lasting 30–40 s
(Malenka and Nicoll, 1993) and paired-pulse facilitation lasting less than a second (PPF;
Debanne et al., 1996). In contrast to HFS inducing LTP, low-frequency stimulation (LFS;
e.g. 900 pulses at 1 Hz) of Schaffer-collaterals results in long-term depression (LTD) of
EPSPs in CA1 pyramidal cells, i.e. in a sustained reduction of the synaptic efficacy (for
review see Bear and Abraham, 1996). In some experimental protocols, synaptic depression
of less than 30 min, or even less than a second, was observed. In analogy with STP and
PPF, these forms of transient depression were termed STD (Artola and Singer, 1993) and
PPD (Debanne et al., 1996), respectively. However, the term STD covers a variety of
physiological processes including postsynaptic mechanisms, e.g. desensitization of
neurotransmitter receptors, and presynaptic reduction of transmitter release from readily
releasable stores (for review see Zucker, 1989). The latter is also true for PPD.
With some remarkable exceptions (for review see Johnston et al., 1992), LTP as well as
LTD depend on activation of the NMDA type of glutamate receptor as indicated by the
blocking action of NMDA receptor antagonists applied during conditioning stimulation
(Collingridge et al., 1983; Mulkey and Malenka, 1992). In addition, the induction of either
LTP or LTD was demonstrated to depend on the intracellular concentration of calcium
(Mulkey and Malenka, 1992). According to the results of this study, LTD is induced by
a moderate increase of the intracellular calcium concentration, while a strong increase
of the intracellular calcium concentration results in the induction of LTP. Furthermore,
intracellular studies revealed that the induction of either LTP or LTD strongly depends on
the time-relation between presynaptic and postsynaptic activities (for review see Bliss and
Collingridge, 1993, and the following section).

3.2. Synaptic Plasticity: The Adult Neocortex

There is some evidence that the first and primary site of synaptic modification involved in
the reorganization of cortical maps is in the cortex (for review Buonomano and Merzenich,
1998). Thus, the focus of the present short review is on synaptic plasticity induced within
the cortex. In a study examining the susceptibility of the neocortex to the induction of
synaptic plasticity, compared to that in CA1 of the hippocampus, it was demonstrated
that—depending on the stimulation frequency—conditioning stimulation in cortical layer
IV induced either LTP (with HFS) or LTD (with LFS) in layer III of the adult rat visual
cortex (Kirkwood et al., 1993). Thus, common forms of synaptic plasticity can be induced
in the hippocampus and the neocortex. However, the amount of enhancement in neocortical
LTP is smaller and develops more slowly than in hippocampal LTP (Malenka, 1995). The
features of synaptic plasticity of neocortical neurones have been studied most frequently
in the primary visual cortex (Artola and Singer, 1987; Aroniadou and Teyler, 1991;
Kirkwood and Bear, 1994a,b; Bear, 1996), but also in the primary and secondary somatosen-
sory cortices (Castro-Alamancos et al., 1995; Kitagawa et al., 1997; Feldman, 2000; Heusler
et al., 2000; Kawakami et al., 2001), the primary auditory cortex (Kudoh and Shibuki,
1994, 1997) and the motor cortex (Baranyi and Szente, 1987; Castro-Alamancos et al.,
1995; Hess and Donoghue, 1994, 1996) as well as in the prefrontal cortex (Hirsch and Crepel,
1990). As with synaptic plasticity in the hippocampus, activation of NMDA glutamate
receptors is essential for the induction of most forms of neocortical LTP and LTD (e.g.
Kirkwood et al., 1993; Castro-Alamancos et al., 1995). Furthermore, by analogy with
hippocampal plasticity, the induction of either LTP or LTD in layer II/III of the neocortex
depends on the level of postsynaptic depolarization (Artola et al., 1990). Either LTP or
LTD could be induced by the same stimulation pattern, depending on the level of depolar-
ization during conditioning. Moreover, the same holds for the dependence of synaptic
plasticity on the concentration of intracellular free calcium (Tsumoto and Yasuda, 1996):
in neurones of the visual cortex, a stimulation pattern suitable to induce LTP was demon-
strated to induce LTD when the effective free calcium was reduced by calcium chelators,
i.e. substances binding free calcium (Hansel et al., 1997). In these neurones, the intracellular
calcium concentration was higher and decayed more slowly with stimulation protocols
inducing LTP than with stimulation protocols inducing LTD (Hansel et al., 1997). Quan-
titative results concerning the calcium concentration for the induction of either LTD or
LTP in the neocortex are not available. However, in the hippocampus, fura-2-based quan-
tification of calcium-dependence of the induction of synaptic plasticity revealed a calcium
threshold of about 180 nM Ca2+ for the induction of LTD and of about 540 nM for the
transition from LTD to LTP (Cormier et al., 2001). These results on the voltage- and calcium-
dependence of LTD and LTP are congruent with the “BCM” theory (Bienenstock, Cooper
and Munro, 1982). This model of synaptic plasticity proposes that active synapses are
potentiated when the total postsynaptic response exceeds a critical value, the modification
threshold θm, and that active synapses are depressed when the activity is less than θm (for
review see Bear, 1996).
Moreover, during conditioning the timing of evoked EPSPs and the postsynaptic action
potential is essential for the induction of either LTP or LTD (Feldman, 2000). LTP is
induced when both events coincide or when the EPSP leads the postsynaptic action potential,
whereas LTD is induced when the postsynaptic action potential leads the evoked EPSP.
The NMDA receptor seems to be ideally suited as a molecular coincidence-detector, since

it is activated by the presynaptic release of glutamate, only if the postsynaptic membrane

is sufficiently depolarized by other mechanisms. This associative signal may be provided
by strong synaptic activation or alternatively by postsynaptic action potentials backpropa-
gating along the apical dendrite (Paulsen and Sejnowski, 1999). Since the NMDA receptor
is part of a non-selective cation channel with an important calcium conductance (for
review see Collingridge and Lester, 1989; Kaczmarek et al., 1997), activation of the NMDA
receptor results in the activity-related increase of the intracellular calcium concentration
necessary for the induction of synaptic plasticity.
Recent studies indicate that the induction of synaptic plasticity is associated with
modulation of the number of functional synapses. During the induction of LTP in the
visual cortex, previously “silent” synapses can be activated, whereas during the induction of
LTD previously functional synapses can be inactivated (Voronin et al., 1996). In hippo-
campal neurones, “silent synapses” have NMDA receptors but lack AMPA receptors,
which can be acquired rapidly after induction of LTP (Isaac et al., 1995). Conversely,
induction of LTD results in an increase of the number of AMPA receptors but in an
unaltered number of NMDA receptors (Carroll et al., 1999; for review see Scannevin and
Huganir, 2000).
Recent studies revealed not only that long-term synaptic plasticity, i.e. LTP and LTD
could be induced in neocortical areas, but that also short-term alterations of synaptic strength
may occur. PPF of EPSPs as observed in the visual cortex (Volgushev et al., 1997) as well
as PPD of EPSPs as observed in the motor cortex (Thomson et al., 1993) are associated
with alterations of the release probability of neurotransmitter. However, conditions known
to increase the release of neurotransmitter were less effective in the neocortex compared
with the hippocampus, while conditions known to decrease the release probability were
similarly effective in the neocortex and hippocampus (Castro-Alamancos and Connors,
1997). These results were interpreted in terms of a relatively high probability of release in
the neocortex. Furthermore, STP and STD were also reported to occur in the neocortex
(Castro-Alamancos et al., 1995).
Thus, it is evident that the neocortex is susceptible to the induction of transient and
persistent modification of synaptic strength and that neocortical synaptic plasticity, in
spite of some important differences, shares many features with hippocampal synaptic
plasticity. However, there is accumulating evidence that—depending on the neocortical
area—axonal connections do not respond homogeneously to conditioning stimulation
under at least comparable conditions. Comparison of synaptic plasticity between functionally
and cytoarchitectonically different neocortical areas has revealed different susceptibility
to the induction of synaptic plasticity. Both, the granular somatosensory cortex and the
agranular motor cortex were equally capable of generating LTD as well as STD (Castro-
Alamancos et al., 1995). In contrast, the capability of the two areas to generate LTP was
unequal: in the somatosensory cortex, “theta burst” stimulation reliably induced LTP,
whereas it induced STP in the motor cortex. Induction of LTP in the motor cortex required
a reduction of the GABAA receptor-mediated intracortical inhibition, but the resulting
LTP still differed from that in the somatosensory cortex, e.g. by its slow onset. Similarly,
unequal capabilities to generate LTP were reported for the auditory and visual cortices
(Kudoh and Shibuki, 1997). In this study, whole cell recordings revealed that the post-
synaptic depolarization elicited by theta burst stimulation was significantly larger in the
auditory cortex than that in the visual cortex. These differences were diminished when
horizontal connections in supragranular layers were cut.

3.3. Reorganization of Cortical Maps: Associative Synaptic Plasticity and

Stabilization of Cortical Neuronal Networks
LTP and LTD have been implicated as the cellular mechanisms involved in the experience-
dependent reorganization of neocortical representational maps (e.g. Garraghty and Muja,
1996; Glazewski et al., 1996; Kirkwood et al., 1996). In this context, two basic properties
of LTP and LTD are of importance: (i) input-specificity, i.e. only synapses are modified
that were activated during stimulation of a given input (“homosynaptic plasticity”), and (ii)
associativity, i.e. a “weak” input can be modified if it is active at the same time as a separate
but convergent input is activated by tetanic stimulation (Bliss and Collingridge, 1993).
The latter feature may be of particular importance for the interaction of different inputs,
e.g. horizontal and vertical intracortical connections. Synaptic plasticity of horizontal con-
nections within cortical layer II/III is discussed as a mechanism possibly involved in
reorganization of cortical maps in the motor cortex (Hess and Donoghue, 1994, 1996), the
visual cortex (Hirsch and Gilbert, 1993), and the somatosensory cortex (Lee et al., 1991).
It was demonstrated in a behavioural study that motor skill learning is at least partly due to
LTP-like mechanisms in the motor cortex (Rioult-Pedotti et al., 1998). HFS applied to
intrinsic horizontal connections in layer II/III of different areas of the visual cortex, i.e. the
primary visual cortex and the inferotemporal cortex, resulted in the induction of LTP in
the inferotemporal cortex, whereas LTD was induced in the primary visual cortex
(Murayama et al., 1997). This difference in the susceptibility to the induction of LTP is
proposed to be due to differences in the distribution of neurochemicals, e.g. protein
kinase-C, implicated with the expression of LTP (for review see Elgersma and Silva, 1999).
Hebbian plasticity is indeed a powerful mechanism for the modification of synaptic
strength of vertical and horizontal cortical connections. However, Hebbian plasticity tends
to destabilize neuronal activity: strong inputs are further strengthened whereas weak
inputs are further weakened, resulting in either excessive discharge or inactivity of neu-
rones. Therefore, supplementary mechanisms are necessary to stabilize neuronal
responses within neuronal networks modified during sensory experience. The BCM pro-
posal (see above) of an activity-dependent shift of the LTD-LTP transition threshold θm,
depending on the level of the postsynaptic discharge rate provides such a stabilizing
mechanism: during high postsynaptic activity the threshold for LTP is high, making
depression easier and further potentiation more difficult; but after a period of low neuronal
activity the threshold for LTP should be low, resulting in a higher susceptibility to the
induction of LTP. Indeed, it was demonstrated in the visual cortex that LTP induced in
light-deprived animals was stronger than in control animals, whereas the magnitude of
LFS-induced LTD in light-deprived animals was significantly less compared to control
animals (Kirkwood et al., 1996). Weakening of LTD was reversible: the magnitude of
LTD returned nearly to control levels when light-deprived animals were exposed to light
for two days. These results were interpreted in terms of a light deprivation-dependent
promotion of LTP over LTD by a shift of the LTD-LTP transition threshold θm.
Additional candidate mechanisms important for stabilization of Hebbian plasticity have
been proposed: synaptic scaling, spike-timing dependent plasticity (STDP) and synaptic
redistribution (for review see Abbott and Nelson, 2000). Synaptic scaling, a mechanism
globally modifying synaptic strength, was demonstrated to occur in cultured neocortical
networks (Turrigiano et al., 1998). In these experiments, blocking spontaneous discharge
activity caused a multiplicative increase of synaptic strength in all afferents, whereas

enhancing spontaneous activity caused a multiplicative decrease of synaptic strength. An

important component of synaptic scaling is the postsynaptic modification of available
functional glutamate receptors. Activity manipulation scales both AMPA and NMDA
receptor-mediated transmission (Watt et al., 2000). Scaling of the NMDA receptor may
influence the influx of calcium and consequently the LTP-LTD induction threshold in
a manner comparable to that proposed by the BCM model.
The importance of STDP as a stabilizing mechanism is suggested by the observation
that the induction of either LTP or LTD depends on the temporal order of pre- and postsyn-
aptic activity (see above; Feldman, 2000). This mechanism contributes to the stabilization
of neuronal discharge (for review see Abbott and Nelson, 2000), since only inputs that
discharge in a narrow time window before the postsynaptic discharge are potentiated whereas
inputs that fire in a wider time window after the postsynaptic discharge are depressed.
When presynaptic action potentials arrive randomly in time with respect to the postsynaptic
discharge, LTD dominates over LTP. This was demonstrated in layer II/III pyramidal
neurones of the somatosensory (barrel) cortex (Feldman, 2000). In this study, random pairing
of pre- and postsynaptic activity resulted in an overall reduction of synaptic strength.
STDP leads to a non-uniform distribution of synaptic strengths and to irregular postsynaptic
firing on a reasonable average rate. Thus, STDP stabilizes Hebbian plasticity and leads to
a noisy but temporally-sensitive state (Abbott and Nelson, 2000).
The occurrence of synaptic redistribution as a stabilizing mechanism is suggested
by the observation that postsynaptic LTP acts presynaptically to modify the probability
of transmitter release (for review see Abbott and Nelson, 2000). In neocortical pyramidal
neurones, an increase of synaptic response was observed only when a synaptic input
occurred at low frequency, an effect that was interpreted in terms of a redistribution of
the available synaptic efficacy (Markram and Tsodyks, 1996). In the visual cortex,
postsynaptic intracellular tetanization resulted in LTP of inputs with strong PPF (low
release probability) but in LTD of inputs with small PPF (high release probability)
(Volgushev et al., 1997). This mechanism, probably involving a retrograde messenger,
allows Hebbian modification to act on transient activation without increasing the steady-
state response or the steady state excitability of postsynaptic neurones, since it increases
the probability of transmission early in a sequence of activity, but decreases the avail-
ability of releasable transmitter late in a sequence. Thus, this mechanism additionally
influences the short-term dynamics of synaptic transmission. Short-term synaptic
dynamics of vertical and horizontal intracortical connection as observed in the barrel
cortex may be involved in cortical reorganization by sensory experience (Finnerty, Roberts
and Connors, 1999).
Taken together, a variety of mechanisms may contribute to experience-dependent modi-
fication of cortical representational maps. Associative long-term synaptic modifications,
i.e. LTP and LTD of vertical and horizontal connections are suited to expand or shrink
receptive fields of cortical neurones. In this context the activation of “silent synapses” has
to be considered of high importance. Stabilization of the discharge level of cortical
neurones that is modified by Hebbian synaptic plasticity may be brought about by synaptic
scaling, spike timing-dependent synaptic plasticity and synaptic redistribution. The latter
and other short-term mechanisms of synaptic plasticity may also be important for the
selective sensitivity of cortical neurones to dynamic changes of afferent activity.
Furthermore, mechanisms involved in experience-dependent reorganization of cortical
representational maps may include short-term and long-term modification of the synaptic

efficacy of inhibitory input to cortical neurones, since it is generally accepted that

inhibitory surrounds and subthreshold contributions determine the receptive field size, i.e.
by a dynamically maintained balance between excitatory and inhibitory inputs. However,
studies on this subject are scarce: in the visual cortex long-term modification of inhibitory
synaptic transmission was demonstrated to occur in the developing visual cortex (Komatsu
and Iwakiri, 1993). This form of LTP is not voltage-dependent, but depends on the activa-
tion of GABAB and monoamine receptors (Komatsu, 1996). Short-term alterations of the
efficacy of inhibitory synapses have also been demonstrated to occur. PPF and PPD of
inhibitory afferents were induced in the somatosensory cortex depending on the pulse
interval (Fleidervish and Gutnick, 1995). PPF of IPSPs was induced when pulses were
delivered at a brief interval, while PPD was induced when the interval was long. Synaptic
depression induced by prolonged stimulation of afferents to cortical neurones was suggested
to depend on the depletion of synaptic vesicles (Galarreta and Hestrin, 1998). In these
experiments, sustained activation of neuronal afferents resulted in much weaker depression
of synaptic currents at inhibitory synapses than at excitatory ones. The differential depression
at excitatory and inhibitory synapses in the visual cortex indicates that the balance
between excitation and inhibition can change dynamically as a function of activity (Varela
et al., 1999). These alterations of the balance between excitation and inhibition may
contribute importantly to the reorganization of cortical maps.
4. SUMMARY AND OUTLOOK
We have reviewed recent findings about plastic changes in adult early sensory and motor
cortices, that were induced by different approaches including peripheral lesions, differential
use, and training. Generally, massive and enduring reorganizations have been described
for all areas discussed, confirming the contemporary view according to which all cortical
areas are modifiable, beyond the critical sensitive periods during development. The findings
demonstrate impressively that the sensorimotor cortical representations in adults are not
hard-wired, but retain a self-organizing capacity operational throughout life.
On the other hand, for all forms of plasticity described, there also exist distinct modality-
specific differences. These differences include the magnitude of changes, the readiness of
inducability, and the specificity of neural parameters that are affected. While plasticity in
somatosensory and auditory cortex share many features, many lines of evidence suggest
that visual cortex plasticity is characterized by a number of particularities. There exist also
a number of area and modality-specific properties of cellular mechanisms mediating plas-
ticity of synaptic transmission, indicating that dissimilarities observed at a systemic level
are also present at the cellular level. Assuming that cortical plasticity in adults represents
an ubiquitous feature required for survival of an individual, the emerging differences are
difficult to understand.
In the following, we offer a number of possible explanations touching on different
levels of abstraction of possible underlying mechanisms and functional constraints.
1. From a mechanistic point of view, different forms and magnitudes of plastic changes
might be due to differences in cellular, pharmacological and histochemical properties,
that reflect specific areal-specific constraints of the molecular equipment present in
an area (Huntley et al., 1994; Elgersma and Silva, 1999). While this can explain

existing differences in the outcome of plastic changes, the question remains, what are
the reasons for the emergence of such differences in cellular properties?
2. Besides cellular and pharmacological aspects, anatomical particularities can have
a decisive impact on the outcome of plastic changes. The network of horizontal long-
range connections is a particularly interesting candidate for mediating specific forms
of alterations of synaptic efficacy. Differences in the overall pattern of horizontal
connections in different areas would explain differences in reorganizations observed
both cellularly and at a systemic level.
3. From in-vitro experiments evidence has been accumulated that there exist a number
of different mechanisms controlling and stabilizing the outcome of plastic processes.
Among these, spike-timing effectively controls synaptic potentiation (Abbott and
Nelson, 2000). Given the profound area- and modality-specific differences in the
timing of the afferent inflow of information (see also Dinse and Schreiner, this volume),
it appears conceivable that such mechanisms are highly suited to govern the effectiveness
of input-dependent plastic reorganizations.
4. Similarly, the Bienenstock-Cooper-Munro model – BCM (Bienenstock, Cooper and
Munro, 1982) provides a mechanism that controls the transition from synaptic depres-
sion to synaptic potentiation, by means of the level of the postsynaptic discharge rate.
Active synapses are potentiated when the total postsynaptic response exceeds a critical
value, the modification threshold θm, and synapses are depressed when the activity is
less than θm. As the level of postsynaptic activity can be regulated in a complex way
by selecting and integrating inputs from many different sources, the BCM model can
potentially regulate the threshold for inducing diverse forms of plasticity, including
the failure to induce any changes.
5. Reorganization in early sensory areas is modified by so-called “top-down” routes
conveying information about cognitive and attentional aspects processed in high-level
areas, as exemplified by the modulatory action exerted by the cholinergic system. It is
possible that differences in the reorganizational outcome are due to a differential
sensitivity to this top-down modulation, thereby establishing a differential effectiveness
of input- vs. attentional-driven plasticity.
6. In somatosensory plasticity, the aspect of “use” and “no-use” provide the key features
that allows an easy and intuitive description and classification of plastic changes.
Given the obvious lack of typical “use-dependent” plasticity in the visual domain, it is
quite possible that the scheme of differential use is an inappropriate concept that does
not fit to the specific constraints of the visual system. By the same token, searching
for adequate driving forces that are particularly effective in the visual system might
allow one to reveal a specific form of visual cortex reorganization.
7. The visual cortex is characterized by a number of so-called functional maps, that are
overlaid across the retinotopic gradient, thereby generating a highly complicated
form of topological structure (Malach, 1994; Swindale et al., 2000). Up to now
comparable topological features have not been described for the somatosensory and
auditory cortex (see also Dinse and Schreiner, this volume). It is suggested that these
global topological constraints impose forces that stabilize the underlying cortical
networks (Wolf and Geisel, 1998), thereby limiting and restricting plastic changes,
particularly those of receptive fields. Consequently, only under severe circumstances
such as lesions or massive changes of input statistics, do distinct changes of receptive
fields or other parameters of neural response properties develop.

8. Neural changes as a consequence of adaptational mechanisms include a large variety

of both spatial and temporal parameters of neural response characteristics. However, even
under normal conditions, i.e. without involving adaptive processes, sensory processing
and how performance is coded is only poorly understood. As a consequence, perceptual
learning or training can be accompanied by rather unexpected changes. Conceivably,
an apparent lack of plastic capacities might simply reflect hidden changes in parameter
regimens presently not recognized or understood. In other words, there can be signifi-
cant changes, but we do not see them.
9. Finally, it is possible that part of the observed dissimilarities reflect genuine modality-
specific differences, building on important constraints associated to the processing of
sensory-specific information or constraints emerging from anatomical and morpho-
logical requirements, that in turn evolved in response to the processing requirements
of a sensory area. Consequently, comparative studies focusing on modality-specific
features of cortical plasticity will reveal insights into principles governing neocortical
organization.
REFERENCES
Abbott, L.F. and Nelson, S.B. (2000) Synaptic plasticity: taming the beast. Nature, Neuroscience, 3, 1178–1183.
Aglioti, S., Smania, N., Atzei, A. and Berlucchi, G. (1997) Spatio-temporal properties of the pattern of evoked
phantom sensations in a left index amputee patient. Behavioral Neuroscience, 111, 867–872.
Ahissar, E., Vaadia, E., Ahissar, M., Bergman, H., Arieli, A. and Abeles, M. (1992) Dependence of cortical
plasticity on correlated activity of single neurons and on behavioral context. Science (Washington), 257,
1412–1415.
Ahissar, E., Haidarliu, S. and Shulz, D.E. (1996) Possible involvement of neuromodulatory systems in cortical
Hebbian-like plasticity. Journal of Physiology (Paris), 90, 353–360.
Ahissar, E., Abeles, M., Ahissar, M., Haidarliu, S. and Vaadia, E. (1998) Hebbian-like functional plasticity in the
auditory cortex of the behaving monkey. Neuropharmacology, 37, 633–655.
Aroniadou, V.A. and Teyler, T.J. (1991) The role of NMDA receptors in long-term potentiation (LTP) and
depression (LTD) in rat visual cortex. Brain Research, 562, 136–143.
Artola, A., Bröcher, S. and Singer, W. (1990) Different voltage-dependent thresholds for the induction
of long-term depression and long-term potentiation in slices of the rat visual cortex. Nature, London,
347, 69–72.
Artola, A. and Singer, W. (1987) Long-term potentiation and NMDA receptors in rat visual cortex. Nature
(London), 330, 649–652.
Artola, A. and Singer, W. (1993) Long-term depression of excitatory synaptic transmission and its relationship to
long-term potentiation. Trends in Neuroscience, 16, 480–487.
Bakin, J.S. and Weinberger, N.M. (1996) Induction of a physiological memory in the cerebral cortex by stimula-
tion of the nucleus basalis. Proceedings of the National Academy of Science, U.S.A., 93, 11219–11224.
Baranyi, A. and Szente, M.B. (1987) Long-lasting potentiation of synaptic transmission requires postsynaptic
modifications in the neocortex. Brain Research, 423, 378–384.
Baskerville, K.A., Schweitzer, J.B. and Herron, P. (1997) Effects of cholinergic depletion on experience-
dependent plasticity in the cortex of the rat. Neuroscience, 80, 1159–1169.
Bear, M.F. (1996) NMDA-receptor-dependent synaptic plasticity in the visual cortex. In: R.R. Mize and R.S.
Erzurumlu (eds), Neural development and plasticity (Progress in Brain Research, Vol. 108). Amsterdam:
Elsevier Science, pp. 205–218.
Bear, M.F. and Abraham, W.C. (1996) Long-term depression in hippocampus. Annual Reviews of Neuroscience,
19, 437–462.
Bienenstock, E.L., Cooper, L.N. and Munro, P.W. (1982) Theory for the development of neuron selectivity,
orientation specifity and binocular interaction in visual cortex. Journal of Neuroscience, 2, 32–48.
Bjordahl, T.S., Dimyan, M.A. and Weinberger, N.M. (1998) Induction of long-term receptive field plasticity in
the auditory cortex of the waking guinea pig by stimulation of the nucleus basalis. Behavioral Neuro-
science, 112, 467–479.
Blakemore, C. (1977) Genetic instructions and developmental plasticity in the kitten’s visual cortex. Philosophical
Transactions of the Royal Society, London, B Biological Sciences, 278, 425–434.

Bliss, T.V.P. and Lømo, T. (1973) Long-lasting potentiation of synaptic transmission in the dentate area of
the anaesthetized rabbit following stimulation of the perforant path. Journal of Physiology (London), 232,
331–356.
Bliss, T.V.P. and Collingridge, G.L. (1993) A synaptic model of memory: long-term potentiation in the hippo-
campus. Nature (London), 361, 31–39.
Boroojerdi, B., Bushara, K.O., Corwell, B., Immisch, I., Battaglia, F., Muellbacher, W. and Cohen, L.G. (2000)
Enhanced excitability of the human visual cortex induced by short-term light deprivation. Cerebral Cortex,
10, 529–534.
Braun, C., Schweizer, R., Elbert, T., Birbaumer, N. and Taub, E. (2000) Differential activation in somatosensory
cortex for different discrimination tasks. Journal of Neuroscience, 20, 446–450.
Buisseret, P., Gary-Bobo, E. and Imbert, M. (1978) Ocular motility and recovery of orientational properties of
visual cortical neurones in dark-reared kittens. Nature (London), 272, 816–817.
Buonomano, D.V. and Merzenich, M.M. (1998) Cortical plasticity: from synapses to maps. Annual Review of
Byl, N.N., Merzenich, M.M. and Jenkins, W.M. (1996) A primate genesis model of focal dystonia and repetitive
strain injury: I. Learning-induced dedifferentiation of the representation of the hand in the primary
somatosensory cortex in adult monkeys. Neurology, 47, 508–520.
Byl, N.N., Merzenich, M.M., Cheung, S., Bedenbaugh, P., Nagarajan, S.S. and Jenkins, W.M. (1997) A primate
model for studying focal dystonia and repetitive strain injury: effects on the primary somatosensory cortex.
Physiological Therapy, 77, 269–284.
Calford, M.B. and Tweedale, R. (1988) Immediate changes in responses of somatosensory cortex in adult flying-
fox after digit amputation. Nature (London), 332, 446–447.
Calford, M.B., Schmid, L.M. and Rosa, M.G. (1999) Monocular focal retinal lesions induce short-term
topographic plasticity in adult cat visual cortex. Proceedings of the Royal Society, London, B, Biological
Sciences, 266, 499–507.
Calford, M.B., Wang, C., Taglianetti, V., Waleszczyk, W.J., Burke, W. and Dreher, B. (2000) Plasticity in adult
cat visual cortex (area 17) following circumscribed monocular lesions of all retinal layers. Journal of
Carew, T.J., Hawkins, R.D., Abrams, T.W. and Kandel, E.R. (1984) A test of Hebb’s postulate at identified
synapses which mediate classical conditioning in Aplysia. Journal of Neuroscience, 4, 1217–1224.
Carroll, R.C., Lissin, D.V., von Zastrow, M., Nicoll, R.A. and Malenka, R.C. (1999) Rapid redistribution of
glutamate receptors contributes to long-term depression in hippocampal cultures. Nature, Neuroscience, 2,
454–460.
Castro-Alamancos, M.A., Donoghue, J.P. and Connors, B.W. (1995) Different forms of synaptic plasticity in
somatosensory and motor areas of the neocortex. Journal of Neuroscience, 15, 5324–5333.
Castro-Alamancos, M.A. and Connors, B.W. (1997) Distinct forms of short-term plasticity at excitatory
synapses of hippocampus and neocortex. Proceedings of the National Academy of Science, U.S.A., 94,
4161–4166.
Chance, F.S., Nelson, S.B. and Abbott, L.F. (1998) Synaptic depression and the temporal response characteris-
tics of V1 cells. Journal of Neuroscience, 18, 4785–4799.
Chino, Y.M., Smith III, E.L., Kaas, J.H., Sasaki, Y. and Cheng, H. (1995) Receptive-field properties of deaffer-
entated visual cortical neurons after topographic map reorganization in adult cats. Journal of Neuroscience,
15, 2417–2433.
Clark, S.A., Allard, T.T., Jenkins, W.M. and Merzenich, M.M. (1988) Receptive fields in the body-surface map
in adult cortex defined by temporally correlated input. Nature (London), 332, 444–445.
Cohen, L.G., Brasil, N., Pascual-Leone, A. and Hallett, M. (1993) Plasticity of cortical motor output organization
following deafferentation, cerebral lesions, and skill acquisition. Advances in Neurology, 63, 187–200.
Cohen, L.G., Celnik, P., Pascual-Leone, A., Corwell, B., Falz, L., Dambrosia, J., Honda, M., Sadato, N.,
Gerloff, C., Catala, M.D. and Hallett, M. (1997) Functional relevance of cross-modal plasticity in blind
humans. Nature, London, 389, 180–183.
Cohen, L.G., Weeks, R.A., Sadato, N., Celnik, P., Ishii, K. and Hallett, M. (1999) Period of susceptibility for
cross-modal plasticity in the blind. Annals of Neurology, 45, 451–460.
Collingridge, G.L., Kehl, S.J. and McLennan, H.J. (1983) Excitatory amino acids in synaptic transmission in the
Schaffer collateral-commisural pathway of the rat hippocampus. Journal of Physiology (London), 334, 33–46.
Collingridge, G.L. and Lester, R.A.J. (1989) Excitatory amino acid receptors in the vertebrate central nervous
system. Pharmacological Reviews, 40, 143–210.
Cormier, R.J., Greenwood, A.C. and Connor, J.A. (2001) Bidirectional synaptic plasticity correlated with the
magnitude of dendritic calcium transients above a threshold. Journal of Neurophysiology, 85, 399–406.
Creutzfeldt, O.D. and Heggelund, P. (1975) Neural plasticity in visual cortex of adult cats after exposure to
visual patterns, Nature (London), 188, 1025–1027.
Crist, R.E., Kapadia, M.K., Westheimer, G. and Gilbert, C.D. (1997) Perceptual learning of spatial localization:
specificity for orientation, position, and context. Journal of Neurophysiology, 78, 2889–2894.

Cruikshank, S.J. and Weinberger, N.M. (1996a) Evidence for the Hebbian hypothesis in experience-dependent
physiological plasticity of neocortex: A critical review. Brain Research Reviews, 22, 191–228.
Cruikshank, S.J. and Weinberger, N.M. (1996b) Receptive-field plasticity in the adult auditory cortex induced by
Hebbian covariance. Journal of Neuroscience, 16, 861–875.
Darian-Smith, C. and Gilbert, C.D. (1995) Topographic reorganization in the striate cortex of the adult cat and
monkey is cortically mediated. Journal of Neuroscience, 15, 1631–1647.
Debanne, D., Guérineau, N.C., Gähwiler, B.H. and Thompson, S.M. (1996) Paired-pulse facilitation and
depression at unitary synapses in rat hippocampus: quantal fluctuation affects subsequent release. Journal
of Physiology (London), 491, 163–176.
Diamond, D.M. and Weinberger, N.M. (1986) Classical conditioning rapidly induces specific changes in
frequency receptive fields of single neurons in secondary and ventral ectosylvian auditory cortical fields.
Diamond, M.E., Armstrong-James, M. and Ebner, F.F. (1993) Experience-dependent plasticity in adult rat barrel
cortex. Proceedings of the National Academy of Science, U.S.A., 90, 2082–2086.
Dinse, H.R. (1994) A time-based approach towards cortical functions: neural mechanisms underlying dynamic
aspects of information processing before and after postontogenetic plastic processes. Physica D, 75, 129–150.
Dinse, H.R., Recanzone, G. and Merzenich, M.M. (1990) Direct observation of neural assemblies during neocortical
representational reorganization. In: R. Eckmiller, G. Hartmann and G. Hauske (eds), Parallel Processing in
Neural Systems and Computers. Amsterdam: Elsevier, pp. 65–70.
Dinse, H.R., Recanzone, G.H. and Merzenich, M.M. (1993) Alterations in correlated activity parallel ICMS-
induced representational plasticity. NeuroReport, 5, 173–176.
Dinse, H.R., Godde, B., Hilger, T., Haupt, S.S., Spengler, F. and Zepka, R. (1997a) Short-term functional plasti-
city of cortical and thalamic sensory representations and its implication for information processing.
Advances in Neurology, 73, 159–178.
Dinse, H.R., Reuter, G., Cords, S.M., Godde, B., Hilger, T. and Lenarz, T. (1997b) Optical imaging of cat auditory
cortical organization after electrical stimulation of a multichannel cochlear implant: differential effects of
acute and chronic stimulation. American Journal of Otology, 18, 6S, 17–18.
Dinse, H.R., Reuter, G., Godde, B., Cords, S.M., Hilger, T., Fischer, M. and Lenarz, T. (1998) Effects of chronic
electrical stimulation of the cochlea on the cochleotopic organization of cat primary auditory cortex. Society
of Neuroscience Abstracts, 24, 905.
Dinse, H.R. and Merzenich, M.M. (2002) Adaptation of Inputs in the Somatosensory System. In: M. Fahle and
T. Poggio (eds), Perceptual Learning. Cambridge: MIT Press, in press.
Dinse, H.R. (2001) The Aging Brain. Introductory remarks. In: G.W. Kreutzberg and N. Elsner (eds), Proceedings
of the 4th Meeting of the German Neuroscience Society. The 28th Göttingen Neurobiology Conference
2001. Stuttgart, New York: Thieme, pp. 356–363.
Donoghue, J.P. (1995) Plasticity of adult sensorimotor representations. Current Opinion in Neurobiology, 5, 749–754.
Dragoi, V., Sharma, J. and Sur, M. (2000) Adaptation-induced plasticity of orientation tuning in adult visual
cortex. Neuron, 28, 287–298.
Edeline, J.M. and Weinberger, N.M. (1993) Receptive field plasticity in the auditory cortex during frequency
discrimination training: selective retuning independent of task difficulty. Behavioral Neuroscience,
107, 82–103.
Edeline, J.M., Hars, B., Maho, C. and Hennevin, E. (1994) Transient and prolonged facilitation of tone-evoked
responses induced by basal forebrain stimulations in the rat auditory cortex. Experimental Brain Research,
97, 373–386.
Edeline, J.M. (1996) Does Hebbian synaptic plasticity explain learning-induced sensory plasticity in adult
mammals? Journal of Physiology (Paris), 90, 271–276.
Elbert, T., Pantev, C., Wienbruch, C., Rockstroh, B. and Taub, E. (1995) Increased cortical representation of the
fingers of the left hand in string players. Science (Washington), 270, 305–307.
Elbert, T., Candia, V., Altenmuller, E., Rau, H., Sterr, A., Rockstroh, B., Pantev, C. and Taub, E. (1998)
Alteration of digital representations in somatosensory cortex in focal hand dystonia. NeuroReport, 9,
3571–3575.
Elgersma, Y. and Silva, A.J. (1999) Molecular mechanisms of synaptic plasticity and memory. Current Opinion
in Neurobiology, 9, 209–213.
Eysel, U.T. (1992) Remodelling receptive fields in sensory cortices. Current Opinion in Neurobiology, 2, 389–391.
Eysel, U.T., Eyding, D. and Schweigart, G. (1998) Repetitive optical stimulation elicits fast receptive field
changes in mature visual cortex. NeuroReport, 9, 949–954.
Faggin, B.M., Nguyen, K.T. and Nicolelis, M.A. (1997) Immediate and simultaneous sensory reorganization at
cortical and subcortical levels of the somatosensory system. Proceedings of the National Academy of
Science, U.S.A., 94, 9428–9433.
Fahle, M. (1997) Specificity of learning curvature, orientation, and vernier discriminations. Vision Research, 37,
1885–1895.
Fahle, M. and Poggio, T. (2002) Perceptual Learning. Cambridge: MIT Press.

Feldman, D.E. (2000) Timing-based LTP and LTD in vertical inputs to layer II/III pyramidal cells in rat barrel
cortex. Neuron, 27, 45–56.
Feldman, D.E., Nicoll, R.A. and Malenka, R.C. (1999) Synaptic plasticity at thalamocortical synapses in devel-
oping rat somatosensory cortex: LTP, LTD, and silent synapses. Journal of Neurobiology, 41, 92–101.
Finnerty, G.T., Roberts, L.S.E. and Connors, B.W. (1999) Sensory experience modifies the short-term dynamics
of neocortical synapses. Nature (London), 400, 367–371.
Fleidervish, I.A. and Gutnick, M.J. (1995) Paired-pulse facilitation of IPSCs in slices of immature and mature
mouse somatosensory neocortex. Journal of Neurophysiology, 73, 2591–2595.
Flor, H., Braun, C., Elbert, T. and Birbaumer, N. (1997) Extensive reorganization of primary somatosensory
cortex in chronic back pain patients. Neuroscience Letters, 224, 5–8.
Flor, H., Elbert, T., Knecht, S., Wienbruch, C., Pantev, C., Birbaumer, N., Larbig, W. and Taub, E. (1995)
Phantom-limb pain as a perceptual correlate of cortical reorganization following arm amputation. Nature
(London), 375, 482–484.
Flor, H., Elbert, T., Mühlnickel, W., Pantev, C., Wienbruch, C. and Taub, E. (1998) Cortical reorganization and
phantom phenomena in congenital and traumatic upper-extremity amputees. Experimental Brain Research,
119, 205–212.
Florence, S.L. and Kaas, J.H. (1995) Large-scale reorganization at multiple levels of the somatosensory pathway
follows therapeutic amputation of the hand in monkeys. Journal of Neuroscience, 15, 8083–8095.
Florence, S.L., Taub, H.B. and Kaas, J.H. (1998) Large-scale sprouting of cortical connections after peripheral
injury in adult macaque monkeys. Science (Washington), 282, 1117–1121.
Fox, K. and Daw, N.W. (1993) Do NMDA receptors have a critical function in visual cortical plasticity? Trends
in Neuroscience, 16, 116–122.
Fregnac, Y., Shulz, D., Thorpe, S. and Bienenstock, E. (1988) A cellular analogue of visual cortical plasticity.
Fregnac, Y., Shulz, D., Thorpe, S. and Bienenstock, E. (1992) Cellular analogs of visual cortical epigenesis.
I. Plasticity of orientation selectivity. Journal of Neuroscience, 12, 1280–1300.
Fuhr, P., Cohen, L.G., Dang, N., Findley, T.W., Haghighi, S., Oro, J. and Hallett, M. (1992) Physiological
analysis of motor reorganization following lower limb amputation. Electroencephalography and Clinical
Galarreta, M. and Hestrin, S. (1998) Frequency-dependent synaptic depression and the balance of excitation and
inhibition in the neocortex. Nature, Neuroscience, 1, 587–594.
Gardner, E.P. and Costanzo, R.M. (1980) Temporal integration of multiple-point stimuli in primary somatosen-
sory cortical receptive fields of alert monkeys. Journal of Neurophysiology, 43, 444–468.
Garraghty, P.E. and Kaas, J.H. (1992) Dynamic features of sensory and motor maps. Current Opinions in
Garraghty, P.E. and Muja, N. (1996) NMDA receptors and plasticity in adult primate somatosensory cortex.
Ghazanfar, A.A., Stambaugh, C.R. and Nicolelis, M.A. (2000) Encoding of tactile stimulus location by somato
sensory thalamocortical ensembles. Journal of Neuroscience, 20, 3761–3775.
Gibson, E.J. (1953) Improvement in perceptual judgements as a function of controlled practice or training. Psycho-
logy B, 50, 401–431.
Gilbert, C.D. (1998) Adult cortical dynamics. Physiological Reviews, 78, 467–485.
Gilbert, C.D. and Wiesel, T.N. (1992) Receptive field dynamics in adult primary visual cortex. Nature (London),
356, 150–152.
Glazewski, S., Chen, C.-M., Silva, A. and Fox, K. (1996) Requirement for a-CaMKII in experience-dependent
plasticity of the barrel cortex. Science (Washington), 272, 421–423.
Godde, B., Spengler, F. and Dinse, H.R. (1996) Associative pairing of tactile stimulation induces somatosensory
cortical reorganization in rats and humans. NeuroReport, 8, 281–285.
Godde, B., Reuter, G. and Dinse, H.R. (1998) Cochlear Implant Stimulation—From Single Cells to Behavior.
Introductory remarks. In: R. Wehner and N. Elsner (eds), New Neuroethology on the move: The 26-th Göttin-
gen Neurobiology Conference 1998, Vol. I. Stuttgart, New York: Thieme, pp. 458–464.
Godde, B. and Dinse, H.R. (1999) Intracortical microstimulation is efficient to induce systematic plastic reorgan-
ization of orientation maps in area 18 of adult cats. Neuroimage, 9, S286.
Godde, B., Stauffenberg, B., Spengler, F. and Dinse, H.R. (2000) Tactile coactivation induced changes in spatial
discrimination performance. Journal of Neuroscience, 20, 1597–1604.
Godde, B., Leonhardt, R., Cords, S. and Dinse, H.R. (2002) Plasticity of orientation preference maps in the
visual cortex of adult cats. Proceedings of the National Academy of Science, U.S.A. in press.
Gödecke, I. and Bonhoeffer, T. (1996) Development of identical orientation maps for two eyes without common
visual experience. Nature (London), 379, 251–254.
Granger, R., Whitson, J., Larson, J. and Lynch, G. (1994) Non-Hebbian properties of long-term potentiation
enable high-capacity encoding of temporal sequences. Proceedings of the National Academy of Science,
U.S.A., 91, 10104–10108.

Greuel, J.M., Luhmann, H.J. and Singer, W. (1988) Pharmacological induction of use-dependent receptive field
modifications in the visual cortex. Science (Washington), 242, 74–77.
Gu, X. and Fortier, P.A. (1996) Early enhancement but no late changes of motor responses induced by intra-
cortical microstimulation in the ketamine-anesthetized rat. Experimental Brain Research, 108, 119–128.
Halligan, P.W., Marshall, J.C., Wade, D.T., Davey, J. and Morrison, D. (1993) Thumb in cheek? Sensory
reorganization and perceptual plasticity after limb amputation. NeuroReport, 4, 233–236.
Hansel, C., Artola, A. and Singer, W. (1997) Relation between dendritic Ca2+ levels and the polarity of synaptic
long-term modifications in rat visual cortex neurons. European Journal of Neuroscience, 9, 2309–2322.
Hebb, D.O. (1949) The Organization of Behavior. New York: J Wiley and Sons.
Hess, G. and Donoghue, J.P. (1994) Long-term potentiation of horizontal connections provides a mechanism to
recognize cortical maps. Journal of Neurophysiology, 71, 2543–2547.
Hess, G. and Donoghue, J.P. (1996) Long-term depression of horizontal connections in rat motor cortex.
European Journal of Neuroscience, 8, 658–665.
Heusler, P., Cebulla, B., Boehmer, G. and Dinse, H.R. (2000) A repetitive intracortical microstimulation pattern
induces long-lasting synaptic depression in brain slices of the rat primary somatosensory cortex. Experimental
Hirata, Y., Kuriki, S. and Pantev, C. (1999) Musicians with absolute pitch show distinct neural activities in the
auditory cortex. NeuroReport, 10, 999–1002.
Hirsch, H.V.B. and Spinelli, D.N. (1970) Visual experience modifies distribution of horizontally and vertically
oriented receptive fields in cats. Science (Washington), 168, 869–871.
Hirsch, J.C. and Crepel, F. (1990) Use-dependent changes in synaptic efficacy in rat prefrontal neurons in vitro.
Journal of Physiology (London), 427, 31–49.
Hirsch, J.C. and Gilbert, C.D. (1993) Long-term changes in synaptic strength along specific intrinsic pathways in
the cat visual cortex. Journal of Physiology (London), 461, 247–262.
Huntley, G.W., Vickers, J.C. and Morrison, J.H. (1994) Cellular and synaptic localization of NMDA and non-
NMDA receptor subunits in neocortex: organizational features related to cortical circuitry, function and
disease. Trends in Neuroscience, 17, 536–543.
Imamura, K., Kasamatsu, T., Shirokawa, T. and Ohashi, T. (1999) Restoration of ocular dominance plasticity
mediated by adenosine 3′,5′-monophosphate in adult visual cortex. Proceedings of the Royal Society,
London, B Biological Sciences, 266, 1507–1516.
Isaac, J.T., Nicoll, R.A. and Malenka, R.C. (1995) Evidence for silent synapses: implications for the expression
of LTP. Neuron, 15, 427–434.
Jain, N., Florence, S.L., Qi, H.X. and Kaas, J.H. (2000) Growth of new brainstem connections in adult monkeys
with massive sensory loss. Proceedings of the National Academy of Science, U.S.A., 97, 5546–5550.
James, W. (1890) Psychology: Brief Course. Cambridge: Harvard University Press.
Jastreboff, P.J. (1990) Phantom auditory perception (tinnitus): mechanisms of generation and perception. Neuro-
science Research, 8, 221–254.
Johnston, D., Williams, S., Jaffe, D. and Gray, R. (1992) NMDA-receptor-independent long-term potentiation.
Annual Reviews of Physiology, 54, 489–505.
Joublin, F., Spengler, F., Wacquant, S. and Dinse, H.R. (1996) A columnar model of somatosensory reorganiza-
tional plasticity based on Hebbian and non-Hebbian learning rules. Biological Cybernetics, 74, 275–286.
Jürgens, M. and Dinse, H.R. (1995) Spatial and temporal integration properties of cortical somatosensory
neurons in aged rats—Lack of age-related cortical changes in behaviorally unimpaired individuals of high
age. Society of Neuroscience Abstracts, 21, 197.
Jürgens, M. and Dinse, H.R. (1997a) Differential effects of the Ca2+-influx blocker nimodipine on receptive field
properties and response latencies of somatosensory cortical neurons in aged rats. Institut für Neuroinformatik,
Ruhr-University Bochum, Internal Report, 97–10, 1–23.
Jürgens, M. and Dinse, H.R. (1997b) Use-dependent plasticity of SI cortical hindpaw neurons induced by
modification of walking in adult rats: a model for age related alterations. Society of Neuroscience Abstracts,
23, 1800.
Kaas, J.H. (1991) Plasticity of sensory and motor maps in adult mammals. Annual Review of Neuroscience, 14,
137–167.
Kaas, J.H., Krubitzer, L.A., Chino, Y.M., Langston, A.L., Polley, E.H. and Blair, N. (1990) Reorganization of
retinotopic cortical maps in adult mammals after lesions of the retina. Science (Washington), 248, 229–231.
Kaas, J.H. and Florence, S.L. (1997) Mechanisms of reorganization in sensory systems of primates after peripheral
nerve injury. Advances in Neurology, 73, 147–158.
Kaas, J.H., Florence, S.L. and Jain, N. (1999) Subcortical contributions to massive cortical reorganizations.
Neuron, 22, 657–660.
Kaczmarek, L., Kossut, M. and Skangiel-Kramska, J. (1997) Glutamate receptors in cortical plasticity: molecular
and cellular biology. Physiological Reviews, 77, 217–255.
Kakigi, A., Hirakawa, H., Harel, N., Mount, R.J. and Harrison, R.V. (2000) Tonotopic mapping in auditory cortex
of the adult chinchilla with amikacin-induced cochlear lesions. Audiology, 39, 153–160.

Kalaska, J. and Pomeranz, B. (1979) Chronic paw denervation causes an age-dependent appearance of novel
responses from forearm in “paw cortex” of kittens and adult cats. Journal of Neurophysiology, 42, 618–633.
Kapadia, M.K., Gilbert, C.D. and Westheimer, G. (1994) A quantitative measure for short-term cortical plasticity
in human vision. Journal of Neuroscience, 14, 451–457.
Karni, A. and Sagi, D. (1991) Where practice makes perfect in texture discrimination: evidence for primary
visual cortex plasticity. Proceedings of the National Academy of Science, U.S.A., 88, 4966–4970.
Kasamatsu, T., Pettigrew, J.D. and Ary, M. (1979) Restoration of visual cortical plasticity by local microperfusion
of norepinephrine. Journal of Comparative Neurology, 185, 163–181.
Kasten, E., Wust, S., Behrens-Baumann, W. and Sabel, B.A. (1998) Computer-based training for the treatment of
partial blindness. Nature (Medicine), 4, 1083–1087.
Kawakami, Y., Miyata, M. and Oshima, T. (2001) Mechanical vibratory stimulation of feline forepaw skin induces
long-lasting potentiation in the secondary somatosensory cortex. European Journal of Neuroscience, 13, 71–178.
Kelahan, A.M., Ray, R.H., Carson, L.V., Massey, C.E. and Doetsch, G.S. (1981) Functional reorganization of adult
raccoon somatosensory cerebral cortex following neonatal digit amputation. Brain Research, 223, 152–159.
Kew, J.J., Ridding, M.C., Rothwell, J.C., Passingham, R.E., Leigh P.N., Sooriakumaran, S., Frackowiak, R.S.
and Brooks, D.J. (1994) Reorganization of cortical blood flow and transcranial magnetic stimulation maps
in human subjects after upper limb amputation. Journal of Neurophysiology, 72, 2517–2524.
Kilgard, M.P. and Merzenich, M.M. (1998a) Cortical map reorganization enabled by nucleus basalis activity.
Kilgard, M.P. and Merzenich, M.M. (1998b) Plasticity of temporal information processing in the primary auditory
Kimura, A., Melis, F. and Asanuma, H. (1996) Long-lasting changes of neuronal activity in the motor cortex of
cats. NeuroReport, 22, 869–872.
Kirkwood, A., Dudek, S.D., Gold, J.T., Aizenman, C.D. and Bear, M.F. (1993) Common forms of synaptic
plasticity in hippocampus and neocortex in vitro. Science (Washington), 260, 1518–1521.
Kirkwood, A. and Bear, M.F. (1994a) Hebbian synapses in visual cortex. Journal of Neuroscience, 14, 3404–3412.
Kirkwood, A. and Bear, M.F. (1994b) Homosynaptic long-term depression in the visual cortex. Journal of
Kirkwood, A., Rioult, M.G. and Bear, M.F. (1996) Experience-dependent modification of synaptic plasticity in
visual cortex. Nature (London), 381, 526–528.
Kitagawa, H., Nishimura, Y., Yoshioka, K., Lin, M. and Yamamoto, T. (1997) Long-term potentiation and
depression in layer III and V pyramidal neurons in the cat sensorimotor cortex in vitro. Brain Research, 751,
339–343.
Knecht, S., Henningsen, H., Hohling, C., Elbert, T., Flor, H., Pantev, C. and Taub, E. (1998) Plasticity of
plasticity? Changes in the pattern of perceptual correlates of reorganization after amputation. Brain, 121,
717–724.
Komatsu, Y. (1996) GABAB receptors, monoamine receptors, and postsynaptic inositol triphosphate-induced
Ca2+ release are involved in the induction of long-term potentiation at visual cortical inhibitory synapses.
Komatsu, Y. and Iwakiri, M. (1993) Long-term modification of inhibitory synaptic transmission in developing
visual cortex. NeuroReport, 4, 907–910.
Kudoh, M. and Shibuki, K. (1994) Long-term potentiation in the auditory cortex of adult rats. Neuroscience
Letters, 171, 21–23.
Kudoh, M. and Shibuki, K. (1997) Importance of polysynaptic inputs and horizontal connectivity in the generation
of tetanus-induced long-term potentiation in the rat auditory cortex. Journal of Neuroscience, 17, 9458–9465.
Laubach, M., Wessberg, J. and Nicolelis, M.A. (2000) Cortical ensemble activity increasingly predicts behaviour
outcomes during learning of a motor task. Nature (London), 405, 567–571.
Lee, C.J. and Whitsel, B.L. (1992) Mechanisms underlying somatosensory cortical dynamics: I. In vivo studies.
Lee, D.S., Lee, J.S., Oh, S.H., Kim, S.K., Kim, J.W., Chung, J.K., Lee, M.C. and Kim, C.S. (2000) Cross-modal
plasticity and cochlear implants. Nature (London), 409, 149–150.
Lee, S.M., Weisskopf, M.G. and Ebner, F.F. (1991) Horizontal long-term potentiation of responses in rat somato-
sensory cortex. Brain Research, 544, 303–310.
Leonhardt, R., Spengler, F. and Dinse, H.R. (1997) Low reorganizational efficiency of intracortical microstimula-
tion (ICMS) in primary visual cortex of pigmented rats. Society of Neuroscience Abstracts, 23, 2363.
Leonhard, R., Churs, L., Spengler, F. and Dinse, H.R. (1998) Intracortical microstimulation induces plastic
changes in the time-structure of neuronal responses in adult rat area 17. Forum of European Neuroscience,
European Journal of Neuroscience, 10, S10, pp. 227.
Liepert, J., Tegenthoff, M. and Malin, J.P. (1995) Changes of cortical motor area size during immobilization.
Electroencephalography and Clinical Neurophysiology, 97, 382–386.
Liepert, J., Terborg, C. and Weiller, C. (1999) Motor plasticity induced by synchronized thumb and foot move-
ments. Experimental Brain Research, 125, 435–439.

Linden, D.E., Kallenbach, U., Heinecke, A., Singer, W. and Goebel, R. (1999) The myth of upright vision.
A psychophysical and functional imaging study of adaptation to inverting spectacles. Perception,
28, 469–481.
Lockwood, A.H., Salvi, R.J., Coad, M.L., Towsley, M.L., Wack, D.S. and Murphy, B.W. (1998) The functional
neuroanatomy of tinnitus: evidence for limbic system links and neural plasticity. Neurology, 50, 114–120.
Maalouf, M., Dykes, R.W. and Miasnikov, A.A. (1998) Effects of D-AP5 and NMDA microiontophoresis on
associative learning in the barrel cortex of awake rats. Brain Research, 793, 149–168.
101–104.
Maldonado, P.E. and Gerstein, G.L. (1996a) Reorganization in the auditory cortex of the rat induced by intra-
cortical microstimulation: a multiple single-unit study. Experimental Brain Research, 112, 420–430.
Maldonado, P.E. and Gerstein, G.L. (1996b) Neuronal assembly dynamics in the rat auditory cortex during
reorganization induced by intracortical microstimulation. Experimental Brain Research, 112, 431–441.
Maldonado, P.E., Altman, J.A. and Gerstein, G.L. (1998) Neuron discharges in the rat auditory cortex during
electrical intracortical stimulation. Neuroscience and Behavioral Physiology, 28, 48–59.
Malenka, R.C. and Nicoll, R.A. (1993) NMDA-receptor-dependent synaptic plasticity: multiple forms and
mechanisms. Trends in Neurosciences, 16, 521–527.
Malenka, R.C. (1995) Synaptic plasticity in hippocampus and neocortex: a comparison. In: M.J. Gutnick and
I. Mody (eds), The Cortical Neuron. New York: Oxford University Press, pp. 98–110.
Markram, H. and Tsodyks, M. (1996) Redistribution of synaptic efficacy between neocortical pyramidal
neurons. Nature (London), 382, 807–810.
McLean, J. and Palmer, L.A. (1998) Plasticity of neuronal response properties in adult cat striate cortex. Visual
Menning, H., Roberts, L.E. and Pantev, C. (2000) Plastic changes in the auditory cortex induced by intensive
frequency discrimination training. NeuroReport, 11, 817–822.
Merzenich, M.M., Kaas, J.H., Wall, J.T., Sur, M., Nelson, R.J. and Felleman, D.J. (1983) Topographic reorgan-
ization of somatosensory coertical areas 3b and 1 in adult monkeys following restricted deafferentation.
Merzenich, M.M., Nelson, R.J., Stryker, M.P., Cynader, M.S., Schoppmann, A. and Zook, J.M. (1984) Somato-
sensory cortical map changes following digit amputation in adult monkeys. Journal of Comparative
Neurology, 224, 591–605.
Merzenich, M.M., Recanzone, G., Jenkins, W.M., Allard, T.T. and Nudo, R.J. (1988) Cortical representational
plasticity. In: P. Rakic and W. Singer (eds), Neurobiology of Neocortex. New York: Wiley, pp. 41–67.
Merzenich, M.M., Schreiner, C., Jenkins, W. and Wang, X. (1993) Neural mechanisms underlying temporal
integration, segmentation, and input sequence representation: some implications for the origin of learning
disabilities. Annals of the NewYork Academy of Science, 682, 1–22.
Merzenich, M.M., Jenkins, W.M., Johnston, P., Schreiner, C., Miller, S.L. and Tallal, P. (1996) Temporal
processing deficits of language-learning impaired children ameliorated by training. Science (Washington),
271, 77–81.
Metherate, R. and Weinberger, N.M. (1990) Cholinergic modulation of responses to single tones produces tone-
specific receptive field alterations in cat auditory cortex. Synapse, 6, 133–145.
Miller, K.D., Erwin, E. and Kayser, A. (1999) Is the development of orientation selectivity instructed by
activity? Journal of Neurobiology, 41, 44–57.
Mogilner, A., Grossman, J.A., Ribary, U., Joliot, M., Volkmann, J., Rapaport, D., Beasley, R.W. and Llinas, R.R.
(1993) Somatosensory cortical plasticity in adult humans revealed by magnetoencephalography.
Proceedings of the National Academy of Science, U.S.A., 90, 3593–3597.
Montague, P.R. and Sejnowski, T.J. (1994) The predictive brain: Temporal coincidence and temporal order in
synaptic learning mechanisms. Learning and Memory, 1, 1–33.
Mühlnickel, W., Elbert, T., Taub, E. and Flor, H. (1998) Reorganization of auditory cortex in tinnitus. Proceed-
ings of the National Academy of Science, U.S.A., 95, 10340–10343.
Mulkey, R.M. and Malenka, J.C. (1992) Mechanisms underlying homosynaptic long-term depression in area
CA1 of the hippocampus. Neuron, 9, 967–975.
Murayama, Y., Fujita, I. and Kato, M. (1997) Contrasting forms of synaptic plasticity in monkey inferotemporal
and primary visual cortices. NeuroReport, 8, 1503–1508.
Nicolelis, M.A., Katz, D. and Krupa, D.J. (1998a) Potential circuit mechanisms underlying concurrent thalamic
and cortical plasticity. Reviews in Neuroscience, 9, 213–224.
Nicolelis, M.A.L. (1999) Methods in Neural Ensemble Recordings. New York: CRC Press.
Nudo, R.J., Jenkins, W.M. and Merzenich, M.M. (1990) Repetitive microstimulation alters the cortical
representation of movements in adult rats. Somatosensory and Motor Research, 7, 463–483.
Nudo, R.J., Milliken, G.W., Jenkins, W.M. and Merzenich, M.M. (1996) Use-dependent alterations of
movement representations in primary motor cortex of adult squirrel monkeys. Journal of Neuroscience,
16, 785–807.

Ohl, F.W. and Scheich, H. (1996) Differential frequency conditioning enhances spectral contrast sensitivity
of units in auditory cortex (field Al) of the alert Mongolian gerbil. European Journal of Neuroscience,
8, 1001–1017.
Pantev, C., Oostenveld, R., Engelien, A., Ross, B., Roberts, L.E. and Hoke, M. (1998) Increased auditory cortical
representation in musicians. Nature (London), 392, 811–814.
Pantev, C., Roberts, L.E., Schulz, M., Engelien, A. and Ross, B. (2001) Timbre-specific enhancement of audi-
tory cortical representations in musicians. NeuroReport, 12, 169–174.
Pascual-Leone, A. and Torres, F. (1993) Plasticity of the sensorimotor cortex representation of the reading finger
in Braille readers. Brain, 116, 39–52.
Paulsen, O. and Sejnowski, T.J. (2000) Natural patterns of activity and long-term synaptic plasticity. Current
Opinion in Neurobiology, 10, 172–179.
Plautz, E.J., Milliken, G.W. and Nudo, R.J. (2000) Effects of repetitive motor training on movement representa-
tions in adult squirrel monkeys: role of use versus learning. Neurobiology of Learning and Memory, 74, 27–55.
Pleger, B., Dinse, H.R., Ragert, P., Schwenkreis, P., Malin, J.P. and Tegenthoff, M. (2001) Shifts in cortical
representations predict human discrimation improvement. Proceedings of the National Academy of Science
U.S.A., 98, 12255–12260.
Polley, D.B., Chen-Bee, C.H. and Frostig, R.D. (1999a) Varying the degree of single-whisker stimulation differ-
entially affects phases of intrinsic signals in rat barrel cortex. Journal of Neurophysiology, 81, 692–701.
Polley, D.B., Chen-Bee, C.H. and Frostig, R.D. (1999b) Two directions of plasticity in the sensory-deprived
adult cortex. Neuron, 24, 623–637.
Pons, T.P., Garraghty, P.E., Ommaya, A.K., Kaas, J.H., Taub, E. and Mishkin, M. (1991) Massive cortical
reorganization after sensory deafferenatation in adult macaques. Science (Washington), 252, 1857–1860.
Ponton, C.W., Don, M., Eggermont, J.J., Waring, M.D., Kwong, B. and Masuda, A. (1996) Auditory system
plasticity in children after long periods of complete deafness. NeuroReport, 8, 61–65.
Pöppel, E., Stoerig, P., Logothetis, N., Fries, W., Boergen, K.P., Oertel, W. and Zihl, J. (1987) Plasticity and rigidity
in the representation of the human visual field. Experimental Brain Research, 68, 445–448.
Qian, N. and Matthews, N. (1999) A physiological theory for visual perceptual learning of orientation discrimination.
Society of Neuroscience Abstracts, 25, 1316.
Raggio, M.W. and Schreiner, C.E. (1999) Neuronal responses in cat primary auditory cortex to electrical
cochlear stimulation. III. Activation patterns in short- and long-term deafness. Journal of Neurophysiology,
82, 3506–3526.
Rajan, R. and Irvine, D.R. (1998) Absence of plasticity of the frequency map in dorsal cochlear nucleus of adult
cats after unilateral partial cochlear lesions. Journal of Comparative Neurology, 399, 35–46.
Rajan, R., Irvine, D.R., Wise, L.Z. and Heil, P. (1993) Effect of unilateral partial cochlear lesions in adult cats on
the representation of lesioned and unlesioned cochleas in primary auditory cortex. Journal of Comparative
Neurology, 338, 17–49.
Ramachandran, V.S., Stewart, M. and Rogers-Ramachandran, D.C. (1992) Perceptual correlates of massive
cortical reorganization. NeuroReport, 3, 583–586.
Rasmusson, D.D. (1982) Reorganization of raccoon somatosensory cortex following removal of the fifth digit.
Rasmusson, D.D. and Dykes, R.W. (1988) Long-term enhancement of evoked potentials in cat somatosensory
cortex produced by co-activation of the basal forebrain and cutaneous receptors. Experimental Brain
Research, 70, 276–286.
Rauschecker, J.P. (1999) Auditory cortical plasticity: a comparison with other sensory systems. Trends in
Recanzone, G. (2000) Cerebral cortex plasticity: perception and skill acquisition. In: M.S. Gazzaniga (ed.),
The New Cognitive Neurosciences. Boston: MIT-Press, pp. 237–250.
Recanzone, G.H., Jenkins, W.M., Hradek, G.T. and Merzenich, M.M. (1992a) Progressive improvement in
discriminative abilities in adult owl monkeys performing a tactile frequency discrimination task. Journal of
Recanzone, G.H., Merzenich, M.M., Jenkins, W.M., Grajski, K. and Dinse, H.R. (1992b) Topographic reorgan-
ization of the hand representation in cortical area 3b of owl monkeys trained in a frequency discrimination
task. Journal of Neurophysiology, 67, 1031–1056.
Recanzone, G.H., Merzenich, M.M. and Schreiner, C.E. (1992c) Changes in the distributed temporal response
properties of SI cortical neurons reflect improvements in performance on a temporally-based tactile
diskrimination task. Journal of Neurophysiology, 67, 1071–1091.
Recanzone, G.H., Merzenich, M.M. and Dinse, H.R. (1992d) Expansion of the cortical representation of a specific
skin field in primary somatosensory cortex by intracortical microstimulation. Cerebral Cortex, 2, 181–196.
Recanzone, G.H., Schreiner, C.E. and Merzenich, M.M. (1993) Plasticity in the frequency representation of primary
auditory cortex following discrimination training in adult owl monkeys. Journal of Neuroscience, 13, 87–103.
Rioult-Pedotti, M.-S., Friedman, D., Hess, G. and Donoghue, J.P. (1998) Strengthening of horizontal cortical
connections following skill learning. Nature, Neuroscience, 1, 230–234.

Robertson, D. and Irvine, D.R. (1989) Plasticity of frequency organization in auditory cortex of guinea pigs with
partial unilateral deafness. Journal of Comparative Neurology, 282, 456–471.
Röder, B., Teder-Sälejärvi, W., Sterr, A., Rösler, F., Hillyard, S.A. and Neville, H.J. (1999) Improved auditory
spatial tuning in blind humans. Nature (London), 400, 162–166.
Rosa, M.G., Schmid, L.M. and Calford, M.B. (1995) Responsiveness of cat area 17 after monocular inactivation:
limitation of topographic plasticity in adult cortex. Journal of Physiology (London), 482, 589–608.
Rumpel, S., Hatt, H. and Gottmann, K. (1998) Silent synapses in the developing rat visual cortex: evidence for
postsynaptic expression of synaptic plasticity. Journal of Neuroscience, 18, 8863–8874.
Sachdev, R.N., Lu, S.M., Wiley, R.G. and Ebner, F.F. (1998) Role of the basal forebrain cholinergic projection
in somatosensory cortical plasticity. Journal of Neurophysiology, 79, 3216–3228.
Sachdev, R.N., Egli, M., Stonecypher, M., Wiley, R.G. and Ebner, F.F. (2000) Enhancement of cortical plasticity
by behavioral training in acetylcholine-depleted adult rats. Journal of Neurophysiology, 84, 1971–1981.
Sadato, N., Pascual-Leone, A., Grafman, J., Ibanez, V., Deiber, M.P., Dold, G. and Hallett, M. (1996) Activation
of the primary visual cortex by Braille reading in blind subjects. Nature (London), 380, 526–528.
Sakai, M. and Suga, N. (2001) Plasticity of the cochleotopic (frequency) map in specialized and nonspecialized
auditory cortices. Proceedings of the National Academy of Science, U.S.A., 98, 3507–3512.
Sameshima, K. and Merzenich, M.M. (1993) Cortical plasticity and memory. Current Opinion Neurobiol., 3,
187–196.
Sanes, J.N. and Donoghue, J.P. (1997) Static and dynamic organization of motor cortex. Advances in Neurology,
73, 277–296.
Scannevin, R.H. and Huganir, R.L. (2000) Postsynaptic organization and regulation of excitatory synapses.
Nature Reviews Neuroscience, 1, 133–141.
Scheich, H. (1991) Auditory cortex: comparative aspects of maps and plasticity. Currrent Opinion in Neurobiology,
1, 236–247.
Schmid, L.M., Rosa, M.G., Calford, M.B. and Ambler, J.S. (1996) Visuotopic reorganization in the primary
visual cortex of adult cats following monocular and binocular retinal lesions. Cerebral Cortex, 6, 388–405.
Schoups, A., Vogel, R., Qian, N. and Orban, G. (2001) Practising orientation identification improves orientation
coding in VI neurons. Nature (London), 412, 549–553.
Schoups, A.A., Vogels, R. and Orban, G.A. (1995) Human perceptual learning in identifying the oblique orientation:
retinotopy, orientation specificity and monocularity. Journal of Physiology (London), 483, 797–810.
Schoups, A.A., Vogels, R. and Orban, G.A. (2000) Orientation tuning changes after perceptual leaning in
orientation identification. Society of Neuroscience Abstracts, 26, 1082.
Sengpiel, F., Godecke, I., Stawinski, P., Hübener, M., Lowel, S. and Bonhoeffer, T. (1998) Intrinsic and
environmental factors in the development of functional maps in cat visual cortex. Neuropharmacology, 37,
607–621.
Sengpiel, F., Stawinski, P. and Bonhoeffer, T. (1999) Influence of experience on orientation maps in cat visual
Shulz, D.E., Sosnik, R., Ego, V., Haidarliu, S. and Ahissar, E. (2000) A neuronal analogue of state-dependent
learning. Nature (London), 403, 549–553.
Sil’kis, I.G. and Rapoport, S.S. (1995) Plastic reorganizations of the receptive fields of neurons of the auditory
cortex and the medial geniculate body induced by microstimulation of the auditory cortex. Neuroscience
and Behavioral Physiology, 25, 322–339.
Spengler, F. and Dinse, H.R. (1994) Reversible relocation of representational boundaries of adult rats by intra-
cortical microstimulation (ICMS). NeuroReport, 5, 949–953.
Spengler, F., Godde, B. and Dinse, H.R. (1995) Effects of aging on topographic organization of somatosensory
cortex. NeuroReport, 6, 469–473.
Spengler, F., Roberts, T., Poeppel, D., Byl, N., Wang, X. and Merzenich, M.M. (1997) Cortical reorganization
and transfer in humans trained in tactile discrimination. Neuroscience Letters, 232, 151–154.
Spinelli, D.N., Hirsch, H.V., Phelps, R.W. and Metzler, J. (1972) Visual experience as a determinant of the
response characteristics of cortical receptive fields in cats. Experimental Brain Research, 15, 289–304.
Stanton, S.G and Harrison, R.V. (1996) Abnormal cochleotopic organization in the auditory cortex of cats reared
in a frequency augmented environment. Auditory Neuroscience, 2, 97–108.
Sterr, A., Muller, M.M., Elbert, T., Rockstroh, B., Pantev, C. and Taub, E. (1998a) Perceptual correlates of
changes in cortical representation of fingers in blind multifinger Braille readers. Journal of Neuroscience,
18, 4417–4423.
Sterr, A., Muller, M.M., Elbert, T., Rockstroh, B., Pantev, C. and Taub, E. (1998b) Changed perceptions in
Braille readers. Nature (London), 391, 134–135.
Stoney, Jr., S.D., Thompson, W.D. and Asanuma, H. (1968) Excitation of pyramidal tract cells by intracortical
microstimulation: effective extent of stimulating current. Journal of Neurophysiology, 31, 659–669.
Suedfeld, P. (1975) The benefits of boredom: sensory deprivation reconsidered. American Scientist, 63, 60–69.
Sugita, Y. (1996) Global plasticity in adult visual cortex following reversal of visual input. Nature, Neuroscience,
380, 523–526.

Swindale, N.V., Shoham, D., Grinvald, A., Bonhoeffer, T. and Hübener, M. (2000) Visual cortex maps are
optimized for uniform coverage. Nature, Neuroscience, 3, 822–826.
Tallal, P., Miller, S. and Fitch, R.H. (1993) Neurobiological basis of speech: a case for the preeminence of
temporal processing. Annals of the New York Academy of Science, 682, 27–47.
Tallal, P., Miller, S.L., Bedi, G., Byma, G., Wang, X., Nagarajan, S.S., Schreiner, C., Jenkins, W.M. and Merzenich,
M.M. (1996) Language comprehension in language-learning impaired children improved with acoustically
modified speech. Science (Washington), 271, 81–84.
Thomson, A.M., Deuchars, J. and West, D.C. (1993) Large, deep layer pyramid-pyramid single axon EPSPs in
slices of rat motor cortex display paired pulse and frequency-dependent depression, mediated presynaptically
and self-facilitation, mediated postsynaptically. Journal of Neurophysiology, 70, 2354–2369.
Tommerdahl, M., Delemos, K.A., Favorov, O.V., Metz, C.B., Vierck, Jr., C.J. and Whitsel, B.L. (1998) Response
of anterior parietal cortex to different modes of same-site skin stimulation. Journal of Neurophysiology,
80, 3272–3283.
Tsumoto, T. and Yasuda, H. (1996) A switching role of postsynaptic calcium in the induction of long-term
potentiation or long-term depression in visual cortex. Seminars in the Neurosciences, 8, 311–319.
Turrigiano, G.G., Leslie, K.R., Desai, N.S., Rutherford, L.C. and Nelson, S.B. (1998) Activity-dependent scaling
of quantal amplitude in neocortical neurons. Nature (London), 391, 892–896.
Varela, J.A., Song, S., Turrigiano, G.G. and Nelson, S.B. (1999) Differential depression at excitatory and inhibitory
synapses in visual cortex. Journal of Neuroscience, 19, 4293–4304.
Vogels, R. and Orban, G.A. (1985) The effect of practice on the oblique effect in line orientation judgements.
Vision Research, 25, 1679–1687.
Vogels, B. and Orban, G.A. (1994) Does practice in orientation discrimination lead to changes in the response
properties of macaque inferior temporal neurons? European Journal of Neuroscience, 6, 1680–1690.
Volgushev, M., Voronin, L.L., Chistiakova, M. and Singer, W. (1997) Relations between long-term synaptic
modifications and paired-pulse interactions in the rat neocortex. European Journal of Neuroscience,
9, 1656–1665.
Voronin, L.L., Volgushev, M., Chistiakova, M., Kuhnt, U. and Singer, W. (1996) Involvement of silent synapses
in the induction of long-term potentiation and long-term depression in neocortical and hippocampal
neurons. Neuroscience, 74, 323–330.
Wall, J.T. and Cusick, C.G. (1984) Cutaneous responsiveness in primary somatosensory (S-I) hindpaw cortex
before and after partial hindpaw deafferentation in adult rats. Journal of Neuroscience, 4, 1499–1515.
Wallhäusser-Franke, E., Braun, S. and Langner, G. (1996) Salicylate alters 2-DG uptake in the auditory system:
a model for tinnitus? NeuroReport, 7, 1585–1588.
Wang, X., Merzenich, M.M., Sameshima, K. and Jenkins, W.M. (1995) Remodelling of hand representation in
adult cortex determined by timing of tactile stimulation. Nature (London), 378, 71–75.
Watt, A.J., van Rossum, M.C.W., MacLeod, K.M., Nelson, S.B. and Turrigiano, G.G. (2000) Activity co-regulates
quantal AMPA and NMDA currents at neocortical synapses. Neuron, 26, 659–670.
Weeks, R., Horwitz, B., Aziz-Sultan, A., Tian, B., Wessinger, C.M., Cohen, L.G., Hallett, M. and Rauschecker,
J.P. (2000) A positron emission tomographic study of auditory localization in the congenitally blind.
Weinberger, N.M. (1995) Dynamic regulation of receptive fields and maps in the adult sensory cortex. Annual
Weinberger, N.M., Ashe, J.H., Metherate, R., McKenna, T.M., Diamond, D.M. and Bakin, J. (1990) Retuning
auditory cortex by learning: a preliminary model of receptive field plasticity. Concepts in Neuroscience,
1, 91–132.
Whitaker, D. and McGraw, P.V. (2000) Long-term visual experience recalibrates human orientation perception.
Nature, Neuroscience, 3, 13.
Wilson, B.S., Finley, C.C., Lawson, D.T., Wolford, R.D., Eddington, D.K. and Rabinowitz, W.M. (1991) Better
speech recognition with cochlear implants. Nature (London), 352, 236–238.
Wolf, F. and Geisel, T. (1998) Spontaneous pinwheel annihilation during visual development. Nature (London),
395, 73–78.
Xerri, C., Stern, J.M. and Merzenich, M.M. (1994) Alterations of the cortical representation of the rat ventrum
induced by nursing behavior. Journal of Neuroscience, 14, 1710–1721.
Xing, J. and Gerstein, G.L. (1996) Networks with lateral connectivity. III. Plasticity and reorganization of
somatosensory cortex. Journal of Neurophysiology, 75, 217–232.
Yang, T.T., Gallen, C.C., Cobb, S., Schwartz, B.J. and Bloom, F.E. (1994) Noninvasive detection of cerebral
plasticity in adult human somatosensory cortex. NeuroReport, 5, 701–704.
Zucker, R.S. (1989) Short-term synaptic plasticity. Annual Reviews of Neuroscience, 12, 13–31.

Part V
MORPHOLOGICAL SUBSTRATES OF
SEGREGATION AND INTEGRATION

15 Connectional Organisation and Function in the Macaque
Cerebral Cortex
Malcolm P. Young
Neural Systems Group, Department of Psychology, Claremont Place
Newcastle upon Tyne, NE1 7RU, United Kingdom
Tel: 0044-191-222-7525; FAX: 0044-191-222-5622; e-mail: m.p.young@ncl.ac.uk
Experimental neuroanatomy has revealed a very numerous and complex set of connections between
many discriminable brain structures. These data are uniquely important to defining the organisation
of brain systems, and so represent the raw material for an unmistakable step toward understanding
brain function. However, the interpretation of these data in terms of brain organisation has previ-
ously suffered from a widespread failure to appreciate two things. First, the data are sufficiently
numerous and complex that principled analysis of the data is necessary, before reliable conclusions
about organisation can be drawn from them, exactly as in all other areas of science where data are
complex and numerous. Second, the organisational principles at the neural systems level do not
emerge from informal inspection of the primary data, no matter how eminent the inspector. It is now
almost universally acknowledged that neuroinformatics, the computer-based collation, management
and analysis of these data, is a necessary step in drawing scientifically justified conclusions about
the organisation of neural systems. This chapter reviews results from the “first generation” of neuro-
informatics, and shows that fairly simple organisational principles and regularities underlie the
complexity of the connections, and that that these principles together define the first reliable repre-
sentations of the organisation of central brain systems. The “second generation” of neuroinformatics
is presently more concerned with the interpretation of these organising principles and regularities in
terms of their meaning for brain function. Some progress in this latter area is also reviewed in the
context of the primate visual system, and arguments are presented that the functional architecture
implied by the connection data is in an important sense the opposite of the feed-forward network
widely supposed. The chapter ends by exploring whether a single theoretical net can be thrown over
results from a wide range of neuroscience disciplines, if this net has as its central tenet that process-
ing is principally inferential in nature.
KEYWORDS: analysis of connectivity, Bayesian inference, connectivity, feedback, feedforward,

neuroinformatics, receptive field
1. NEUROANATOMY AND NEUROINFORMATICS
Almost all neuroscientists consider that understanding brain structure will aid understanding
of brain function. This assumption is reproduced in studies at every level of the nervous
system, and of every presumed functional subsystem. At the level of the whole cortex, for
example, many researchers assume that information processing is closely determined by
the inputs, the internal connectivity and computations, and the outputs of the network of
areas and nuclei that make up the brain. Defining the connectivity of brain structures has
therefore been a primary research focus for neuroanatomists over many decades.
351
352 Malcolm P. Young
Their enterprise has been enormously successful, perhaps among the most successful in
all of biology. However, this success has brought with it a problem that is very similar to
that faced by some other biological disciplines: the quantity and complexity of connection
data, and their dispersion through an extensive and idiosyncratic literature make it very
difficult indeed to derive reliable conclusions about the information they collectively bear
about the organisation of the system. An example of the scale of this problem is provided
by noting that more than 14,000 individual reports of connections between different gross
structures of the rat brain have been made in the last 20 years (Burns and Young, 2000).
Data so numerous and complex provide excellent opportunities for the derivation of false
conclusions if examined only informally, simply through the ease by which inconvenient
data can be overlooked. Similarly, for the macaque visual system, V1 is known to be con-
nected to more than 50 other structures (e.g. Young et al., 1995). More than 300 ipsilateral
cortico-cortical connections have been described between at least 30 differentiable visual
processing regions (e.g. Felleman and Van Essen, 1991). These connections, together with
the connections that visual areas make with other cortical regions, constitute a cortical
network defined by almost a thousand gross connections (e.g. Young, 1993, 1995). Further-
more, a plethora of callosal and other commissural connections link the two hemispheres;
and the cortical visual systems stand upon a thalamo-cortical network of almost equal
complexity (e.g. Scannell et al., 1999).
Quite surprisingly, given the complexities revealed by neuroanatomists’ experiments, it
has only recently become widely acknowledged that statements about the organisation of
Figure 15.1. This diagram re-emphasizes the distinction between, on the one hand, primary information about
those brain structures which are connected, and, on the other hand, principles of organisation of the neural
systems which are defined by these connections. All connections collated from the neuroanatomical literature in
Young (1993) are plotted, and so a very great deal of information about the connectivity between brain struc-
tures is represented. However, the positions of the points representing brain structures in the diagram have
simply been placed at random, and so simulate complete innocence of the organisation of the systems defined by
these connections. Organisational principles at the neural systems level do not emerge from informal inspection
of primary data, no matter how eminent the inspector. Principled and detailed data analysis is necessary to untangle
the connections, and so reveal the organisational principles of neural systems in the brain.

Connectivity and Function in the Cortex 353
neural systems need to be supported by analyses of these data. Most areas of science
already employ methods of statistical data analysis to substantiate experimenters’ inter-
pretations of their data. No study in neurophysiology, for instance, would be published
without detailed analysis of the data, the results of which constrain rigorously the conclu-
sions that can reliably be drawn from them. Areas that have not routinely employed data
analysis methods have most often not done so because the experimental data have not
appeared to be tractable for data analysis, or because the benefits of data analysis have not
been particularly clear. Both these considerations seem to have previously applied to the
application of methods of data analysis to data on connectional neuroanatomy. However,
it is instructive to consider the kinds of issue that raw, un-analysed data can inform, and
the kinds of issue that can only be informed by the results of analyses. Consider, for
instance, that an experimental study revealed the carriage of retrograde label from V4 to
MT 4. This datum would be sufficient (assuming no trans-neuronal labelling) to conclude
that MT is connected to V4. A problem arises, however, when conclusions about the
organisation of the system are made on the basis of individual data. For example, it has
been argued that there cannot be two streams in visual cortex because V4 (the prototypical
ventral-stream area) and MT (the prototypical dorsal-stream area) are reciprocally inter-
connected (see Young, 1995). In this case, a conclusion about the organisation of the
visual system—that it is not organised into two streams—is based on readily replicable,
uncontentious data. The problem arises because the organisation of the system is defined
by many hundreds of connections (e.g. Felleman and Van Essen, 1991; Young, 1992), of
which the connections mentioned are only two. It is clearly not possible to draw reliable
conclusions about something defined by hundreds of data on the basis of only two data.
In exactly the same way, a few spikes fired to null stimuli could not be used to argue that
a neurone is not directionally tuned. Because large numbers of connections define neural
systems, conclusions about their organisation unequivocally require the support of analyses
(Young, 1995). For these reasons, both analysis of neuroanatomical data and neuro-
anatomical experiments are necessary before reliable conclusions about the organisation
of neural systems can be drawn. Also, both experimental and data-analytic work are
required to further refine knowledge of neural organisation.
The complexity of connection data, and their often qualitative rather than quantitative
nature, present a problem recognisable to any statistician: the problem of finding implicit
structure or order in badly-behaved real-world data. A frequent analogy for this problem
concerns the fable of the elephant in the dark room (e.g. McDonald, 1986). Many hands
are required to feel around the mysterious creature in the dark. The coherence and agreement
between these independent analyses then determines the degree to which belief should be
invested in the results (Young et al., 1995). Accordingly, a rather wide variety of different
data-analytic methods have now been applied to several different varieties of connection
data. These have included the following methods: (i) modelling data on the quantitative
distributions of connection strength by statistical geometry (Young et al., 1995); (ii) com-
putational hierarchical analysis of laminar origin and termination data (e.g. Hilgetag et al.,
1996, 2000b); (iii) seriation (Young et al., 1994, 1995); (iv) Optimal Set Analysis
(Hilgetag et al., 2000a); (v) cluster analysis (Hilgetag et al., 2000a); and (v) non-metric
multidimensional scaling analyses (NMDS: Young, 1992, 1993; Young et al., 1994,
4
For full list of abbreviations used in this chapter, see the Appendix (p. 371).

1995). Data for method (i) are unfortunately very rare. For the other methods data are very
numerous. Happily, these many independent analyses of several different kinds of data
produce a largely self-consistent picture of neural system organisation, of which, even
more fortunately, rather few aggregate characteristics can be stated (Young et al., 1995;
Hilgetag et al., 2000a). This chapter reviews some of these analyses of the connectional
organisation of cortical systms in the macaque, with particular emphasis on the visual
system, and attempts to place some interpretation on what the results might mean for how
the brain mediates behaviour.
2. THE MACAQUE VISUAL SYSTEM
The cortical visual system appears to occupy a little more than half the area of the macaque’s
cerebral cortex (Felleman and Van Essen, 1991). It is composed of (the order of) a billion
neurones, which reveal complex patterns of visual feature preferences to the microelectrode,
and are distributed in more than thirty discriminable visual processing compartments or
areas (e.g. Felleman and Van Essen, 1991; Zeki and Shipp, 1988), whose identities and
borders continue to be debated. These visual cortical areas are interconnected by hundreds
of ipsi- and contra-lateral cortico-cortical connections, as well as by a very rich subcortical
network (Kaas and Huerta, 1988; Young, 1992, 1995; Young et al., 1995).
Felleman and Van Essen (1991) collated neuroanatomical data on visual structures accord-
ing to a detailed parcellation scheme, and we analysed their collation, with a number of
small changes (see Young, 1992; Young et al., 1995).
2.1. Analysis of the Macaque Visual System Data by Non-Metric

Multidimensional Scaling (NMDS)
Consider that the function of any brain structure is constrained by its inputs and outputs.
Its inputs determine the kinds of information it can process, and its outputs determine
which other structures it can inform directly about its computations. The pattern of con-
nections that any brain area makes and receives therefore will be a key determinant of its
function. One can reason further that the more similar are the patterns of connections of
any two brain areas, the more similar will be their functions. Hence, a very simple
approach to the analysis of connectivity is to find some spatial configuration of points
representing brain areas that optimally reflects the similarities and differences in these
areas’ connection patterns. Brain areas with similar patterns of connection should be
placed close together, while brain areas with very different patterns of connections should
be placed far apart. In this way a system’s “functional architecture” should be simply read
off the configuration using the developed capabilities of the human visual system. There
are many methods available to perform such an analysis of connectivity data, but the
readiest to hand, and in fact the first employed in this task, is non-metric multidimensional
scaling (NMDS) (Young, 1992; Young et al., 1995).
Figure 15.2 shows the results of such an analysis for the primate visual system. The
points of the structure are concentrated into an annular region of the “space”, as is
expected from considering the quantitative aspects of connection data (see Young et al.,
1995). The dimensions of the solution correspond approximately to the anterior-posterior
and dorsal-ventral distribution of the areas as they are placed within the brain. Parietal

Figure 15.2. The structure derived from NMDS analysis of the macaque cortical visual system matrix. The
features of this structure and the means by which it was derived are described in Young (1992), and in much
more detail in Young et al. (1995). Briefly, the structure implies that the visual system is divided into two gross
streams, both of which are hierarchically organized, and which reconverge in areas of the temporal and frontal
lobes. The figure is reprinted with permission from Young et al. (1995).
areas, for example, are placed toward the top of the diagram, whereas infero-temporal
areas are placed toward the bottom. If it is remembered that data on area-to-area con-
nectivity were the only type of information that entered the analysis—no information
regarding the spatial arrangement of the areas on the cortical sheet entered the analysis—
this aspect of the configuration implies that the spatial location of an area predicts to
a degree the areas to which that area is likely to be connected. A possible explanation is
that nearby areas tend to exchange connections with one another (Young, 1992; Young
and Scannell, 1996; see also Cowey, 1979; Mitchison, 1991). A different explanation, that
brain areas have migrated to come into positions conducive to economical wiring volume,
has been suggested (Cherniak, 1994), but we have argued that this alternative explanation
is not biologically plausible (Young and Scannell, 1996).
The primary visual cortex (V1) is located farthest left in Figure 15.2. A group of pre-
striate areas including V2, V3, VP, V4t, V3A, MT, and, surprisingly, as it is a posterior
parietal area, PIP, are placed close to V1. MT and V3A are placed further from V1 than
other members of this group. MT is further distinguishable from its topological neigh-
bours by its weak projection to frontal cortex area 46 and its apparently stronger projection
to the frontal eye fields (FEF) (see Felleman and Van Essen, 1991). Every area in the

prestriate group projects to a further set of areas, which consists of areas of the posterior
parietal cortex and the caudal superior temporal sulcus, namely FST, MSTd, MSTl, VIP,
PO, LIP and DP. These areas then project to FEF, area 7a, the posterior region of the
superior temporal polysensory area (STPp), and to frontal area 46 and the anterior STP
(STPa) (Young, 1992; Young et al., 1995).
Returning to V1, but now concentrating on the lower part of the structure, V1 projects
to V4, while V2 and VP project to VOT. Signals are relayed from V4 and VOT into the
areas of the inferotemporal (IT) cortex. The IT areas appear to be rather serially organized,
with more anterior areas generally being placed successively further toward the right of
the diagram (except AITv), and so further away from the sensory periphery. The areas of
IT at the greatest remove from the visual periphery, that is, the highest order areas accord-
ing to this analysis, are associated with areas TF and TH of the parahippocampal cortex.
The topologically higher-order IT areas project to and receive from STPa and area 46
(Young, 1992, 1995; Young et al., 1995).
It is a feature of the structure that relatively few connections pass across the central
region between the parietal and inferotemporal groupings of areas, by comparison to the
number that pass around the rim. Hence, the analysis suggests that there are two relatively
distinct sets of areas in the macaque cortical visual system, which are much more pro-
fusely interconnected within groupings than between them. These two sets of areas cor-
respond in straightforward manner to the dorsal and ventral streams of visual processing,
which were proposed most clearly by Ungerleider and Mishkin (1982) on the basis of the
behavioural effects of lesions. The higher-order areas of both streams are interconnected.
This feature of the structure implies that there are opportunities for the reconvergence of
processed visual information using “feed-forward” pathways in the rostral parts of the
temporal lobe and in the frontal lobe (Young, 1992; Young et al., 1995).
These aspects of the structure that results from NMDS analysis of the macaque cortical
visual system indicate that, at the gross level of connections between brain areas, four
principles underlie its organization. (i) Neighbouring areas tend to exchange connections;
(ii) the system is dichotomized into two streams; (iii) both streams are broadly hierarch-
ical, and (iv) the streams may reconverge in temporal and frontal areas (Young, 1992;
Young et al., 1995).
How robust are these conclusions? The grossest possible perturbation of the matrix is to
assume that all unreported connections exist (Young, 1992). We analysed a matrix derived
from coding all unreported connections as existing. Connections explicitly reported as
having been looked for and found absent remained “0”. This “control” analysis yields
a configuration in many respects similar to Figure 15.2. In both structures, the parietal and
infero-temporal structures are segregated, with V1 and prestriate areas between them at
one side, and STP and area 46 between them at the other. Quantitatively, the relation
between the two structures is characterised by 76% of the variance in the one structure being
explained by the other (i.e. the two structures are 76% the same in quantitative terms:
Young, 1992, 1995). The “control” structure implies the same gross organizing principles
as the structure in Figure 15.2. Even when the grossest possible perturbation is applied to the
dataset, therefore, the gross conclusions are not perturbed. The solutions are similar
because a sufficiently large number of connections have been confirmed absent (see
Felleman and Van Essen, 1991). The main difference between the structures lies in shifts
of the positions of less well-studied areas, such as VOT and V4t. These areas have their
positions shifted toward the centre of the structure by their acquiring a large number of

hypothetical new connections, which correspond to possible connections that have not yet
been explicitly ruled out. It would be very surprising if these structures were in reality to
possess such rich connectivity. In any case, the similarity between these structures suggests
that the organizational conclusions are workably robust against future changes to the
possible connections concerning which neuroanatomists have not thus far reported.
A further question about robustness concerns the question of whether the NMDS solu-
tion for the visual system data in Figure 15.2 systematically reflects the data structure of
the matrix. This question arises because where the badness-of-fit of an NMDS solution is
very high (or very low), an annular structure can emerge in the solution for entirely spurious
reasons (see Young et al., 1995, for a detailed discussion). Randomly ordered data, for
example, give rise to very high badness-of-fit, and to annular configurations, in which the
ordering of the points within the annulus is random. However, some data structures are
themselves annular, as for example in the celebrated colour circle (Shepard, 1962). The
traditional means of deciding between cases of “systematic-annularity” and “spurious-
annularity” is to compare the badness-of-fit of a solution with analyses of comparable
random data (in this case, a matrix of connectivity re-connected at random), thereby deter-
mining whether the solution is drawn from the distribution in which very low fit could
have produced annular artefact (Stenson and Knoll, 1969). The probability of the badness-
of-fit of the solution in Figure 15.2 falling within the distribution of random data with very
low fit was less than 10–30, a probability that corresponded to a very large z-score for the
real solution’s badness-of-fit (Young et al., 1994; 1995). Hence, there is no ground for
believing that the NMDS solutions for the visual system data should not be trusted to be
a systematic reflection of the connection data’s structure (cf. Simmen et al., 1994).
2.2. Analysis of the Visual System Data using Data Conditioning Methods
and NMDS
We have previously shown (e.g. Young et al., 1995) that several data conditioning methods,
when coupled with NMDS, are very successful in recovering known parameters from test
data at the same level of measurement as anatomical connection data. Data conditioning
methods “wdsm1” (weighted dissimilarity transform-one) and “pth1” (path-length transform-
one) were particularly successful (for mathematical definition of these transforms please
see Young et al., 1995). Analyses of visual system data by these data transformations and
NMDS are of interest as they are a means to ensure that as few aspects of data structure as
possible are obscured by the connection data’s genuine sparsity, in the small number of
dimensions required in an output configuration. Figure 15.3 shows the solutions derived by
wdsm1-NMDS and pth1-NMDS for the visual system, with the solution at the top being
that for wdsm1. Both solutions are similar, despite the very different algorithms by which
the data were transformed in each case. In both configurations, all the “dorsal stream”
areas are concentrated in the top portion of each diagram, while all the “ventral stream”
areas are concentrated in the bottom part. The two streams originate in a number of
occipital visual areas, including V1, which are placed at the left, and appear to reconverge
via STP and area 46, which are placed at the right.
The wdsm1 and pth1 structures both share 92% of their variability with the untrans-
formed structure, in Figure 15.2. The wdsm1 and pth1 configurations share 93% of their
variability with one-another. They are both therefore very similar to one another, and very

Figure 15.3. Configurations derived by submitting the macaque visual system matrix to the wdsm1 (top) and
pth1 (bottom) data-conditioning routines, and then to analysis by untied NMDS (see Young et al., 1995, for full
details). The same conclusions about the gross organization of the system would be drawn from these solutions
as from the untransformed data analysis. The figure is reprinted with permission from Young et al. (1995).
similar to the untransformed structure. The probabilities that these correspondences could
come about by chance, according to approximate randomization tests, are all less than 1 in
a million (Young et al., 1995). Hence, the same conclusions would be drawn concerning
the gross organization of the system from these solutions as from Figure 15.2.
The wdsm1 and pth1 solutions suggest, however, that there may be a further division of
labour within the two visual streams. This is particularly apparent for areas in the dorsal
stream, for which there seems to be a further division that draws areas PO, MSTl and VIP
away from their associates. Similarly, a possible distinction between V4 and TF and the
other ventral stream areas is suggested.

2.3. Simple Statistical Properties of the Connection Patterns

A simple further means of testing whether these organizing principles hold is to employ a
χ2 test to determine whether the observed connections are those expected under the vari-
ous hypotheses. One important dispute is about whether the visual system is segregated
into distinct processing streams. On the one hand, analysis of the behavioural effects of
cortical lesions (e.g. Ungerleider and Mishkin, 1982) and all analyses of connectivity so
far undertaken (Young, 1992; Young et al., 1995; Hilgetag et al., 2000a) and, to some
extent, the physiological evidence (e.g. Merigan and Maunsell, 1993) suggest that the
system is divided into two gross streams. On the other hand, it is maintained by some that
the system is not internally segregated into parallel streams (e.g. Martin, 1992; Goodhill
et al., 1994). These hypotheses about the organization of the macaque visual system are
easy to test decisively. If the system is dichotomized into streams, then the elements of the
dorsal stream should be significantly more connected with their dorsal associates than
with the elements of the ventral stream, and vice versa. Similarly, there should be signifi-
cantly more connections that have been confirmed absent between the dorsal and ventral
areas than within each of these groupings. If these comparisons were to fail to reach signi-
ficance, then the null hypothesis, that the areas are not segregated, could not be rejected.
We divided all the macaque cortical visual areas into four sets, which corresponded to
“early”, “late”, “dorsal” and “ventral” groupings (Young et al., 1995). The “early” set of
areas contained V1, V2, V3, VP, V3A, PIP and V4t; the “late” set contained areas FEF,
46, STPa, STPp, TF and TH; the “dorsal” group contained MT, MSTd, MSTl, FST, PO,
LIP, VIP, DP and 7a; the “ventral” group contained V4, VOT, PITd, PITv, CITd, CITv,
AITd and AITv. The analysis of reported connections showed the dorsal and ventral areas
are much more connected internally than the null hypothesis predicts (χ2 = 17.2,
p < 0.00004; Young et al., 1995). The analysis of connections that have been demonstrated
absent showed the dorsal and ventral areas exchange many fewer connections than the null
hypothesis predicts (χ2 = 18.6, p < 0.00002; Young et al., 1995). The hypothesis that the
visual areas are distributed without segregation into a dorsal and a ventral stream is hence
rejected at a very high level of statistical significance (Young et al., 1995).
Exactly comparable tests of the hypothesis that the visual areas are serially ordered
have also been undertaken (Young et al., 1995). The analysis shows that early and late
visual areas are significantly more connected internally within each group than would be
expected from the Null hypothesis (χ2 = 9.8, p < 0.002; Young et al., 1995). Early and late
areas are significantly less connected between each other than would be expected from the
Null hypothesis (χ2 = 10.9, p < 0.001) (Young et al., 1995).
The analyses of the “confirmed absent” entries in the connection matrix illustrate the
fact that important information about the organization of the system is carried by “connec-
tions” that are not present, and therefore that it would be helpful if these absences were
more often reported explicitly by experimental neuroanatomists.
2.4. Optimal Set Analysis of Visual System Data

The χ2 tests described above show that the incidence of connections agrees with predic-
tions, for example by the pre-classification of structures into two streams. However, the
analysis does not explore the possibility that better classifications of the structures into
other clusters could be found. Hence, χ2 analysis cannot itself rule out the possibility that

there are more than two streams within the system. We explored whether clusters of areas
could be found by computation that fit optimally the explicit criteria, both in the macaque
visual system and in cortical systems in general. The requirement to define (for the above
χ2 analyses) the way in which real data might quantitatively support or refute hypotheses
about neural organisation forced us to specify an explicit definition of a connectionally
differentiated stream (or cluster or system) (Hilgetag et al., 2000a). According to our
definition, a cluster is a set of structures that are more connected among themselves than
they are with any other structures, and more disconnected from other structures than they
are among themselves. This definition is quite generally applicable. We developed a com-
putational method, Optimal Set Analysis (OSA), to find clusters of areas that optimally fit
the explicit criteria, using an optimisation algorithm with an explicit cost function formalised
from our definition of what a cluster is, as above (Hilgetag et al., 2000a). This process was
implemented in a network processor (Hilgetag et al., 2000a).
For the macaque visual system, the processor obtained optimal (lowest-cost) solutions,
all of which showed a clear separation of the visual areas into two main clusters. The first
group comprises areas MSTl, MT, LIP, V2, V3, VP, V3A, PIP, V4, V4t, FST, VIP, DP,
PO, FEF, and MSTd, and the second STPp, PITd, TF, TH, AITd, PITv, CITd, 46, AITv,
CITv, STPa, and 7a. Primary cortical area V1 was associated with the first group in one
third of the solutions and formed a separate cluster in the others. The only area with an
alternating preference for the two groups was area VOT, which appeared in the first
cluster in one third of the solutions and in the second cluster in the remaining solutions.
The keen eye will notice that the assignments of area 7a and V4 in these solutions are
the reverse of those usually suspected to obtain. Area 7a would be expected to cluster with
its dorsal stream associates and not with ventral stream stations, and V4 should cluster
with its ventral stream associates and not dorsal stations. However, a specific aspect of
data structure gives rise to the apparent misassignment. For the χ2 test (above) and in
Young et al. (1995), the ventral stream was defined as V4, VOT, PITd, PITv, CITd, CITv,
AITd, AITv and the dorsal as MT, MSTd, MSTl, FST, PO, LIP, VIP, DP, 7a. All other
areas were classified as either “early” or “late”. We re-examined the connectivity between
the groupings of Young et al. (1995) and found that they represented the optimal (and
unique) arrangement for these areas according to the OSA cost functions. This arrange-
ment had a “cost” of 30, while the same arrangement with V4 and 7a swapped carried
a cost of 51. V4 and 7a swapped into the “incongruous” assignments derived by OSA only
when the connections of all 32 areas were included. When all connections are included,
the cost moves from 135 for our apparently paradoxical arrangement to 141, when V4 is
placed in the ventral and 7a in the dorsal stream. Hence, adding the “early” and the “late”
areas of Young et al. (1995) causes V4 and 7a to move to the anomalous groupings. The
explanation is that the earlier visual areas tend to cluster with the dorsal stream stations.
Many of these areas share many interconnections with V4, and so draw V4 toward the
dorsal cluster when all areas are included. Correspondingly, all the “late” areas except
FEF tend to cluster with the ventral stream stations, and share many connections with 7a,
and hence draw this area toward the ventral grouping. Thus, the apparent misassignments
of these areas reflect a bona-fide feature of data structure: the dorsal-ventral dichotomy is
not orthogonal to the early-late organisation of areas. The dorsal stream tends to be
“lower”, and the ventral stream “higher”. The positions of V4 and 7a in the latter respect
affect their clustering assignments. V4 is lower than 7a, and so tends to cluster with its
early associates, while the opposite is true for 7a.

2.5. Seriation Analysis of the Macaque Cortical Visual System Data

Seriation analysis has previously been used to examine the chronological order of archae-
ological grave site data (Wilkinson, 1971; Laporte and Taillefer, 1987), but has recently
been suggested as a means of investigating serial ordering in connection data (Simmen
et al., 1994). The method attempts to find the uni-dimensional ordering of the elements of
a matrix, amongst the n! (factorial n) possible orderings, that minimizes a measure of dis-
tance between the elements. It does this by finding permutations of the rows that minimize
the cumulative mismatch between all rows. The row order then indicates the serial ordering
of the rows. In the case of connection data, this ordering then specifies the optimal serial
ordering of the brain structures. We implemented a seriation algorithm using simulated
annealing to find optimal orderings (Young et al., 1994, 1995).
Several organizational features were common to optimal length orderings. Parietal and
IT areas were always segregated as far apart as possible, being joined at the one side by
V1 and the prestriate areas, and at the other side by area 46 and the parcellations of STP.
Again, then, very similar organisational features emerged for the visual system from this
analysis as for others. We used Procrustes rotation to provide a quantitative measure of the
closeness of relation between the results of the seriation algorithm and that of NMDS.
Coefficients were distributed about a mean of 0.9 (p < 0.000001) (Young et al., 1995).
2.6. Hierarchical Analysis of the Macaque Cortical Visual System

Hierarchical analysis (Rockland and Pandya, 1979; Maunsell and Van Essen, 1983;
Felleman and Van Essen, 1991) represents a widely acknowledged approach to finding
organisational features in connection data. This approach examines the patterns of laminar
origins and terminations of projections between cortical areas, using a framework of some
simple rules. Connections from one area to another are classified as ascending if they
originate from supragranular layers of the cortex, or bilaminarly from superficial and
deeper layers, and terminate mainly in layer 4. Descending connections can arise in bilam-
inar fashion from upper and deeper layers, or in a unilaminar pattern from infragranular
layers, but tend to terminate in laminae other than layer 4. Connections classified as lateral
originate in superficial and infragranular layers, and terminate in a columnar pattern
throughout the cortical mantle. This classification scheme fails to accommodate about
10% of the connections in the monkey visual system for which some laminar information
is known. These rules thus determine whether projections should be classified as “ascend-
ing”, “descending” or “lateral”. Application of these rules to experimental data yields a set
of constraints that defines pair-wise hierarchical relations between areas. It is then
possible to arrange the cortical areas into largely consistent hierarchies, which involve few
violations of the pair-wise hierarchical constraints. Hence, this analysis demonstrates that
some sensory systems are hierarchically organised overall, and it gives insight into the
ordering of structures in each hierarchy (e.g. Felleman and Van Essen, 1991).
Re-analysing the issue of hierarchical classification of data from the monkey visual
system, however, we calculated that the total number of possible hierarchies for these
areas is greater than 1037 (Hilgetag et al., 1996, 2000b). Considering the huge numbers of
possible hierarchies, the incompleteness and partial inconsistency of the experimental
data, and the use of a discrete cost function (number of violated rule constraints), we
greatly doubted that a uniquely optimal solution could be obtained by hand, despite such

orderings having been reported in the literature. Accordingly, we constructed a computer

program that could perform hierarchical analysis automatically. To do this, we caused
a simulated network’s structure itself, rather than just the activity of its elements (as would
be the case in a conventional neural network), to reflect the relations between the cortical
areas (Hilgetag et al., 2000b). An algorithm then manipulated the structure of this network
by simulated annealing until its structure optimally reflected the input constraints.
The processor found more than 150,000 different solutions, all of which had the same
optimally-small number of rule violations (Hilgetag et al., 1996, 2000b). All the computed
hierarchies possessed a cost of 6 violated rules. In all cases, this involved two symmetrical
pairs of three anatomical connection constraints that could not be satisfied for any par-
ticular hierarchy. This is a smaller number of violations than for any previous manually-
obtained solution. The familiar Felleman and Van Essen scheme of the visual cortical
hierarchy (Felleman and Van Essen, 1991), for instance, possesses 8 constraint violations,
when violations are counted in the same way. This cost is impressive given the informal
methods used, but it is unlikely that the hierarchy of Felleman and Van Essen (1991) is
found in the top million hierarchies (Hilgetag et al., 2000b). The number of rule violations
in our computed solutions was very small by comparison to optimal solutions derived from
randomly shuffled tables, demonstrating that the regularities captured experimentally are
surprisingly systematic.
The number of levels in the optimal hierarchies ranged between 13 and 24 (Hilgetag
et al., 2000b). Hence, optimal solutions possess markedly more hierarchical stages than
Figure 15.4. Frequency distribution of optimal hierarchies for visual areas (Hilgetag et al., 2000b). The boxes
are shaded according to the relative occurrence of an area at a particular level in all 152 803 computed solutions.
The main peaks of the area frequencies are denoted by frames in thicker lines, and the ordering of the peak
solution also represents an optimal hierarchy. Violations of the hierarchical constraints are as follows: for the
constraint FST £ MST, relative occurrence of violations, in all solutions circa: 17%; for FST < STPp: 14%; for
LIP = PITv: 12%; for LIP £ MSTd: 2%; for FST ≥ TF: 2%; for MSTd < PITv: 2%; for FST ≥ PITd: <<1%; and
for MSTd < PITd: <<1% (together with their corresponding symmetrical rules: e.g. MSTd ≥ FST). The first
three violations, together with their counterparts, are the violations for the peak solution. Reprinted with permis-
sion from Hilgetag et al. (2000b).

have been apparent hitherto. Because there are so many equally well-fitting optimal hier-
archies, it is misleading to single out one particular hierarchy from the large solution set,
and impractical to show all of the optimal hierarchies. A statistical summary of the many
solutions may therefore be the most appropriate format in which to represent the hierarch-
ical structure of the primate visual system revealed by current data and rules. Figure 15.4
reflects the frequency with which one area appears on a particular hierarchical level,
taking into account all 152,803 different solutions.
The hierarchical constraints were sufficient to fix only V1 and V2 uniquely. These areas
were always found on levels 1 and 2, respectively. The placement of all other areas
depended on the structure of a particular solution. The positions of some areas were strongly
associated with those of others, since these areas had an identical frequency distribution.
This indicates a fixed position relative to each other in all the solutions. These relation-
ships concerned, aside from V1 and V2, areas V4t and MT (same level), MSTd and VIP
(same level), and CITv, CITd, and STPp (same level).
We compared these results from hierarchical analysis of laminar data with that from
NMDS analysis of area-to-area data. In lieu of performing 150,000 Procrustes rotations of
the new and optimal orderings against the NMDS solution in Figure 15.2, we compared
the median hierarchical solution with the NMDS solution. The median hierarchical model,
which possessed only one dimension, nonetheless accounted for 41% of the variability of
the two-dimensional NMDS configuration. Hence, these completely independent meth-
ods, deployed on completely different types of data, show remarkable agreement about the
ordering of cortical stations in the visual system.
2.7. Summary of Analyses of the Macaque Cortical Visual System

All the various different analyses described in the foregoing sections agree, quantitatively,
and significantly (in statistical terms). Recently, analysis of the propagation of activity
through the visual system after experimental disinhibition has also concurred with the
organisational principles derived above (Stephan et al., 2000). Hence, three propositions
account for results from six independent methods of analysis of three different types of
neurobiological data. These are: (i) that the macaque visual system is hierarchically orga-
nized; (ii) It is divided into gross streams; (iii) It provides opportunities for processed
visual signals to reconverge. We suppose, therefore, that NMDS, χ2 analysis, seriation,
Optimal Set Analysis, and computational hierarchical analysis have faithfully extracted
underlying aspects of the structure from their particular types of data, and that these data
themselves captured real aspects of a self-consistent neural system. Indeed, it becomes
increasingly difficult to find explanations for the concordances between the different
results that do not acknowledge that they have captured real aspects of the organisation of
the visual system. Aspects of the functional interpretation of these purely structural results
are examined in a later section.
3. MACAQUE CORTICAL SYSTEMS
In the above account, analysis of connections between visual cortical areas proceeded by
different methods and operated on several different kinds of connectivity data. Nonethe-
less, the results of all these different approaches concurred in their conclusions about the

organisation of the cortical visual system. This mutual corroboration prompts the hope that
these approaches faithfully recover many aspects of the underlying structure of connectiv-
ity data, and that, consequently, they may be used to inform wider questions about the
organisation of the whole macaque cerebral cortex. I turn now, therefore, to a brief review of
the application of the same approach to the connections of the entire cerebral cortex,
including the other major sensory systems in the macaque brain.
Analysis of connectional data begins by identifying cortical regions of interest, and
continues by collating information on the projections between these areas. This informa-
tion can be represented in connection matrices for analysis by methods such as NMDS and
seriation, and those aspects of it that bear on laminar patterns of connectivity may be
represented in a table of constraints for analysis by hierarchical optimization. For the
primate cerebral cortex, we based the division of the cortical sheet into areas principally
on the parcellation developed by Felleman and Van Essen (1991), except for the areas
of the superior temporal cortex for which we followed the parcellation by Pandya and
Yeterian (1985). The neuroanatomical literature was examined for connections between
the 72 areas of this parcellation, and reported connections were collated into a form tract-
able for data analysis (Young, 1993).
We previously divided the connectivity matrix into subsets of areas that corresponded,
according to conventional knowledge, to the cortical auditory and somatosensory-motor
systems, and we also examined the whole matrix (Young, 1993; Young et al., 1994).
These connection matrices were analysed using the full battery of methods from neuro-
informatics, employed as above, for the visual structures. I turn first to results of these
analyses for the central sensory systems, and then to those for the whole cortical system as
collated.
3.1. Analysis of Macaque Auditory Cortex

Analysis of data from the primate auditory cortex, showed a clustering of secondary areas,
including paAc, reit, paAl, proA and Tpt, around the primary auditory area, the auditory
koniocortical area (KA). Area paAr appeared to be less peripheral than its neighbours.
Afferent signals would appear to be processed somewhat successively along the superior
temporal gyrus (STG) from, for example, area Tpt, the caudalmost field of the STG, through
TS3, TS2 to TS1, and thence through the dorsal temporal polar cortex, TGd. The auditory
areas of the rostral STG interact with elements of the limbic system, namely the entorhinal
cortex, perirhinal cortex (35) and, through these structures, the hippocampus, which are all
positioned at the greatest connectional remove from the auditory periphery.
The primate cortical auditory system hence appears to be hierarchically organized, with
a primary area, secondary areas and auditory association areas progressively distant topo-
logically from the auditory brainstem and periphery. Unlike the organization of the primate
cortical visual system, however, the auditory system presents itself in the NMDS solution as
a single hierarchy, since “horizontal” connections between areas at a similar remove from
the primary area are as numerous as those between stations above and below each structure.
3.2. Analysis of the Macaque Somatosensory-Motor Cortex

Analysis of data from the primate somatosensory-motor cortex shows the post-central
primary somatosensory cortex in areas 3a, 3b, 1 and 2 to be grouped. These relatively

peripheral areas are then associated with areas 5 and SII, and these areas with area 7b,
retroinsular (Ri) and insular cortex (Ig and Id), which thus appear to comprise the higher-
order somatosensory fields. These areas interact with limbic structures, such as the
entorhinal cortex, perirhinal cortex (area 35), and these structures contact the hippocam-
pus. These features suggest that the cortical somatosensory system is organized as a single
hierarchy, somewhat as the cortical auditory system (Young, 1993): “horizontal” connec-
tions between areas placed at a similar remove from the primary area are as numerous as
those between stations above and below each. One feature of this system, however, that is
rather unlike either the visual or auditory systems, is that part of the system comprises the
cortical motor system. This part includes the medial supplementary motor area (SMA), the
premotor cortex (area 6), and the primary motor cortex (area 4). These motor areas are
arranged in what appears to be a hierarchy, in which the primary motor area is associated
with primary somatosensory areas, the premotor area with higher somatosensory areas,
and SMA with still higher ones. Primary motor cortex (area 4) seems not to be connected
at the cortical level to any area outside the somatosensory-motor system, which may imply
that, as well as being a sensory hierarchy, this system plays an important role in the integ-
ration of cortical motor signals (Young et al., 1994).
3.3. Analysis of the Connectivity of the Whole Primate Cortex

Figure15.5 represents 834 connections between 72 structures in the visual, auditory,
somatosensory-motor, frontal and limbic cortex of the macaque monkey (see Young,
1993; Young et al., 1995). Despite the complexity of the solution structure, features of the
gross organization of the systems are apparent in it, and in general every organizational
feature apparent has been corroborated by subsequent analyses with other methods (e.g.
Hilgetag et al., 2000a; Stephan et al., 2000; Young et al., 1995).
Propitiously, areas within each of the visual, auditory and somatosensory-motor systems
are clustered together. Each cluster of central sensory structures is topologically distant
from the other clusters of sensory areas. Visual cortical stations are placed at the left, with
earlier visual areas located toward the bottom, and successively higher visual areas located
progressively toward the top left of the structure. The two streams of the visual system are
clearly separated, with elements of the dorsal stream (e.g. FEF, LIP, 7a) positioned
relatively further toward the centre of the diagram, where they are drawn by their greater
connectivity outside the visual system, chiefly with the somatosensory-motor system.
Auditory cortical areas are positioned at the top right of the diagram, with primary and
secondary areas placed at the extreme right, and the areas of the higher auditory cortex
located progressively further toward the top left. The somatosensory-motor cortex is
located at the bottom right of the configuration, with the primary areas at the very bottom.
The primary motor cortex (area 4) is placed in the centre of this group of areas and is
positioned furthest from the centre of the structure. Area 7b, as well as being strongly
connected to the other areas of the somatosensory-motor system, has rich interconnections
with the other sensory systems and has been drawn nearer to the centre of the structure by
these connections (Young, 1993).
Structures in which the most elaborated sensory signals are likely to be processed are
connected to a further cluster of areas positioned at the top left of the diagram. This cluster
of areas is composed of the limbic system (the entorhinal, perirhinal and cingulate cortex
and the hippocampus), in association with some of the areas of the frontal and prefrontal

Figure 15.5. The topological organization of the entire primate cortical processing system as represented in the
collated connection data (Young, 1993; Young et al., 1994, 1995). Aspects of the organisation of these neural
systems are described in the text. Reprinted with permission from Young (1993).
cortex (e.g. areas 13, 10 and 12). The amygdala is placed near the geometric centre of the
configuration, mainly because of a very rich output connectivity: the basal nuclei of the
amygdala project in varying degrees of density to all but 8 of the areas of the cortical
parcellation. Its inputs, however, arise selectively in the higher order sensory processing
areas. The “fronto-limbic complex”, at the top left of the diagram, is the cluster of areas
furthest from the sensory-motor periphery of the cortical processing system, which is rep-
resented at the bottom and right-hand edge of the structure. This result confirms the idea
that the limbic system is topologically central, but also shows the topological association
between limbic structures and elements of frontal cortex. Frontal cortex (in contrast to
occipital cortex) however, may not be a connectionally homogeneous set of areas, and
therefore not functionally homogeneous either, since some frontal cortical areas are
associated with the limbic system and the fronto-limbic complex, while others are more
associated with one of the sensory modalities (e.g. areas 8 [FEF] and 46 with vision, and
area 45 with the somatosensory-motor system). Areas 14, 32 and 25 appear in some
analyses to form a somewhat distinct grouping, which is increasingly difficult to impute
solely to the low number of connections reported for area 14 (cf. Young et al., 1995).
The anterior superior temporal polysensory area (STPa) appears to be a central area in
the topology of the cortical processing system, and the relative positioning of this area in

the system of connections may be related to the elaborate stimulus preferences of cells in
that area (e.g. Perrett et al., 1990; Young and Yamane, 1992).
The somatosensory-motor system is located between the clusters of areas associated
with processing the other two sensory modalities: it appears connectionally closer to the
visual and auditory systems than vision and audition are to each other. Cortical communic-
ation between the auditory and visual systems, except within the fronto-limbic complex,
seems to be limited to interactions mediated through the frontal eye fields (FEF). Inter-
actions between the somatosensory-motor system and audition are more abundant, and
particularly involve the connections of areas 7b and Tpt. Interactions of the visual and
somatosensory-motor systems are also rich, especially those mediated by areas 7b, FEF,
46 and LIP.
3.4. Summary of Analyses of the Macaque Cortical Systems

At the aggregate level, the analyses of connectivity reviewed above suggest a division of
the cortical areas into four major topological clusters of areas. These clusters correspond
to the visual, auditory, somatosensory-motor systems, and the fronto-limbic complex.
Fronto-limbic
complex
m
ste
sy
A syst
ud em
al
su
ito
Vi
ry
Somato-motor
system
Auditory
Visual input
input
Somatosensory
Motor
input
output
Figure 15.6. Summary diagram showing gross features of the organization of the macaque cerebral cortex.
The sensory-motor periphery is represented as being at the bottom of the diagram. Three hierarchical sensory
systems, the visual, auditory and somatosensory-motor systems, connect the periphery with a fourth system, the
fronto-limbic complex, which is at the greatest connectional remove from the periphery. There is restricted
cross-talk between the sensory hierarchies outside the fronto-limbic complex. Amended from Young (1993).

The analyses suggest that all the major cortical sensory systems are organized in a hier-
archical manner, and that the fronto-limbic complex is situated at the furthest remove from
the sensory-motor periphery, due to its being connected mainly with the higher-order areas
of each sensory system. The primate visual system is unlike the other cortical sensory sys-
tems in that it is larger and more complex, and in that it, alone of any central cortical sensory
system, possesses a clear division into two internal streams of processing. These features
of the organization of the primate cortical systems are summarized in Figure 15.6.
4. WHAT DOES NEUROANATOMICAL ORGANISATION MEAN?
Thus far, I have reviewed a variety of analyses of purely structural data. It has been a hope
of long standing that better knowledge of brain organisation would aid improved under-
standing of brain function, and so I now turn to a number of functional interpretations of
these structural results. This entails trying to forge explicit links between structure and
function.
It is already apparent that structure and function are sufficiently closely linked at the
systems level to allow successful predictions from aspects of organisation to aspects of
function. For example, the mystery of where cells with particular complex motion prefer-
ences might be found in the cat brain was resolved by scrutinizing structural diagrams of
exactly the sort described above (Scannell et al., 1996). The strategy was to employ the
patchy information available from neurophysiology on the properties of some stations,
together with the structural diagrams of that system, to predict the most likely location for
cells with the sought-after property. Such locations were then studied neurophysiologic-
ally. This strategy has succeeded twice, in the two studies in which it has been explored
(Scannell et al., 1996, 1997; cf. Merabet et al., 1998). Doubtless, some other search strategy
might have succeeded, but in fact none did. Similarly, this strategy might not have
succeeded, but in fact it did. Another example of an explicit structure-function relation-
ship is that the patterns of activation over the cortex following experimental disinhibition
could be predicted using connectivity matrices (Young and Scannell, 2000; Stephan et al.,
2000; Kötter and Sommer, 2000). In this case, it is particularly interesting that exclusion
of the many indirect pathways by which activity might propagate from one station to
another diminishes the goodness-of-fit between structure and observed activity propagation
(Kötter and Sommer, 2000).
It is instructive also to compare those possible structure-function links that have led to
successful predictions from those that have not. One informative example of a “failure”
comes from the limbic system. Even though the prelimbic cortex of the rat (PL) receives
a direct connection from CA1, electrophysiological experiments that attempted to find
so-called “head-direction” or “place” cells in PL reported a null result (Poucet, 1997; Jung
et al., 1998). Similarly, V4 and MT exchange quite robust projections, but each is the
home of neurones with very different stimulus selectivities.
An interpretation of these various successes and “failures” in exploring possible structure-
function relationships is that specific individual projections have a rather lower functional
significance than many might have expected. What appears to matter most is the pattern
of connections. This is reflected in the striking correspondence between the positions
taken by particular cortical stations in the configurations from analysis of connectivity—
which reflect the patterns of connections they make and receive—and the similarities or

differences in physiological properties of neurones in these areas (Young, 1992). Areas

containing cells with similar response properties tend to have similar patterns of connec-
tions, and to be placed close together in the configurations. This correspondence implies
that local connectivity, and perhaps also the biophysics of cortical cells, may not vary very
greatly across the cortex (Mountcastle, 1982), so that the “location” of an area in the pattern
of cortical macro-circuitry may determine in large part the area’s functional properties.
The correspondence between patterns of connection and function has proved to be an
aid to prediction, but the finding that patterns of connections are more important to func-
tion than individual projections cuts across widely held, though often implicit, views of
cortical function. To illustrate this, and explain why there should be a correspondence of
this kind and no other, I return to considering the cortical visual system.
I believe there exists a very widely credited and largely unchallenged notion of how the
visual system functions. This is the view that vision consists of hierarchical analysis or
filtration of the retinal input generated by a scene or an image. This vision-as-analysis
model supposes that most of the useful information about the visual world is present in the
immediate input from the retina. Feed-forward connections then convey this information
at high gain and high fidelity to (and through) a hierarchically organised visual cortex,
information being successively extracted by increasingly sophisticated receptive fields,
each effectively a filter, situated at each successive stage. The goal of this process is to
provide representations of objects, surfaces and spatial relationships, each being extracted
from the input information in the visual subsystem to which such computations are
imputed (e.g. Lennie, 1998). In this framework, the content of the scene and its relation-
ship to the observer’s knowledge of the world are of little importance, although feedback
and lateral connections are thought to play some contextual role. In an extreme form, it
has been suggested that the visual system is simply a stacked series of competitive learn-
ing networks, together forming a kind of “bagatelle board”, in which an input vector falls in
at the eye, is fed forward through the system, and an output vector, possessing the virtues of
invariance, emerges at the other end, presumably to inform frontal and limbic structures.
The finding that patterns of connections are more important to function than individual
projections is inimical to vision-as-analysis in the following way. If the visual system is
designed to analyse retinal input, signals derived from this input should be relayed with
high gain and fidelity throughout the system. Correspondingly, it is widely assumed that
a large proportion of the synapses in V1 come from neurones in the LGN, the principal
relay for signals from the eye to the cortex in primates. But this is not the case. Quantitative
neuroanatomy shows that axons from the geniculate account for no more than 5% of the
total excitatory synapses on the average pyramidal neurone in the layers in V1 receiving
from the geniculate (e.g. Peters and Payne, 1993). Consequently, over 95% of the excitat-
ory synapses, even in geniculo-recipient layers in primary visual cortex, are made by neu-
rones that originate, not in the LGN, but from other parts of V1, other cortical areas, and
other thalamic nuclei. Similarly, individual cortico-cortical projections take up only a low
proportion of the synapses in their targets. The projection from V1 to MT provides fewer
than 5% of the excitatory synapses in MT (Anderson et al., 1998), and the projection from
V2 to V1 provides less than 6% of the excitatory synapses in V1 (Budd, 1998). However,
while individual extrinsic thalamo-cortical or cortico-cortical projections only rarely
contribute more than 5% of the synapses in a given area, each cortical area in this system
typically receives a large number of inputs from other areas and thalamic nuclei (27 on
average; Scannell and Young, 2000). When all the connections are considered (Scannell

and Young, 2000; Kennedy and Barone, 1999), perhaps 30% of the synapses within a
given volume of visual cortex typically come from distant cortical areas or non-geniculate
thalamic nuclei.
The system, therefore, seems not to be nearly as concerned with transmission of retinal
input with high gain and fidelity as one would expect, if analysis of the input were the
principal computational goal. Retinal input appears not to be the primary or only informa-
tion source, even at the very first cortical stage. Also, long-range connections reach into
local circuitry and form a substantial part of it. Local computations will hence be mark-
edly affected by information washing over them from a very wide variety of sources,
almost all of them internal. In addition, because this cross-sharing of synapses occurs in
every station, description of the dynamics of the system will involve interaction terms—
which will be large—and so both local computation and global aspects of the activity in
the whole visual system will reflect highly interactive dynamics. Hence, a highly dynamic
“information soup”, in which retinal input plays only a part, seems a much more likely
model than bottom-up sequential extraction of retinal information. This conclusion reso-
nates with earlier ones from neuroanatomy, which I believe deserve greater attention than
they have received (e.g. Braitenberg and Schüz, 1991; Miller, 1991). Furthermore, the
inputs to an area are akin to a veritable “chorus”. They are a much more numerous influ-
ence than the strongest single input. Thus the overall pattern of connections will be an
important determinant of the functional properties of cortical areas, explaining the above
mentioned correspondence between the patterns of connections reflected in analyses of
connectivity and aspects of function visible to neurophysiologists.
Neuroanatomical, neurophysiological and psychophysical evidence now suggests that it
may be better to think in terms of vision as inference rather than as analysis (see e.g. Knill
and Richards, 1996). On this inferential model, the majority of useful information is
present in the system and not in the immediate retinal input (Scannell and Young, 2000).
The findings from neuroanatomy and neuroinformatics which I have reviewed reflect an
architecture rather inconsistent with bottom-up analysis, and favourable for inference. The
high degree of cross-sharing of synapses between stations, for example, suggests strongly
interactive dynamics at the systems level, which might generate a dynamic and transitory
global consensus in the system. Cross-sharing of synapses in many visual stations is pre-
cisely the architecture expected for a network implementing inference, because it provides
for local computation of a wide range of information on which to base inferences.
In the inferential framework, neurones signal the probability that some feature constel-
lation is present at some external location, and do so on the basis of knowledge of the
statistical structure of the visual world and of any other information available (e.g. memory
for recent events). Hence, in this framework, a neurophysiologically-mapped field represents
the neurone’s projection of an inferred probability onto the world. It does not represent the
result of simple analysis of the local features within the classical RF. However, electro-
physiological effects made apparent with improbable stimuli do not dissociate inference
from analysis, or projective from receptive fields. If the prior probabilities of stimuli
(“priors” hereafter) are flat, the function of afferent likelihood maps directly onto the
posterior probability, so that inference reduces to analysis. Improbable stimuli hence may
themselves suggest, potentially misleadingly, that vision is bottom-up analysis. When
priors are not flat, as for example during the processing of normal visual scenes, neurones
should behave in predictably different ways, and the two models will dissociate experi-
mentally. Indeed, neurophysiology already suggests that neurones do behave in different

ways when confronted with traditional experimental stimuli compared to normal vision
(Baddeley et al., 1997; Gallant et al., 1998; Rieke et al., 1997), and that visual neurones
generate tuned responses that can be only weakly dependent on analysis of the local image
region corresponding to their classical RF (Grosof et al., 1993; Peterhans and von der
Heydt, 1989; Treue and Maunsell, 1996). In general, priors should be made available to
local computation via connections that are remodelled by activity-dependent processes,
and so come to reflect properties of the visual diet and visual digestion. Extensively
remodelled connections include “feedack” projections and local connectivity (Barone et al.,
1995; Kennedy et al., 1989). This may explain why the effects of feedback projections on
their targets have been so elusive physiologically: traditional improbable stimuli would
not engage these projections in the computational role they enjoy in normal vision.
One of the ramifications of these suspicions about inferential processes in the visual
cortex is that stimulus realism should matter. Natural, or at least, probable, stimuli may be
required if normal vision is to be understood. The old certainties, among them that one can
employ a bar or grating in a denuded visual scene and hope to understand normal vision,
are beginning to give way. At present, there is insufficient information from well-designed
experiments to conclude that visual computation is actually inferential, but the clearly
different predictions that the traditional and inferential models make for neuronal process-
ing when priors are peaked suggest that visual neurophysiologists may again be entering,
as it is said in Chinese, interesting times. The fact that a single theoretical net can be cast
over results from neuroanatomy, neuroinformatics, neurophysiology, and psychophysics
suggests that brain organisation and function actually are now beginning to inform one-
another, even if the particular theory proves to be in error.
APPENDIX
List of Abbreviations
V1 primary visual cortex, area 17
V2 second visual area, part of area 18
V3 third visual area
VP ventral posterior visual area
V3A visual area 3A
V4 the fourth visual area
VOT visual occipito-temporal area
V4t transitional zone of V4 abutting MT
MT middle temporal area
MSTd dorsal middle superior temporal area
MSTl lateral/ventral middle superior temporal area
FST floor of superior temporal
PITd posterior inferotemporal, dorsal
PITv posterior inferotemporal, ventral
CITd central inferotemporal, dorsal
CITv central inferotemporal, ventral
AITd anterior inferotemporal, dorsal
AITv anterior inferotemporal, ventral

STPp superior temporal polysensory, posterior

STPa superior temporal polysensory, anterior
TF area TF of the parahippocampal cortex
TH area TH of the parahippocampal cortex
PO parieto-occipital visual area
PIP posterior intraparietal
LIP lateral intraparietal
VIP ventral intraparietal
DP dorsal prelunate
7A area 7a
FEF area 8, frontal eye fields
46 frontal area 46
TGv ventral temporal polar cortex
ER entorhinal cortex
HIPP hippocampus
3a primary somatosensory cortex area 3a
3b primary somatosensory cortex area 3b
1 primary somatosensory cortex area 1
2 primary somatosensory cortex area 2
5 area 5 of the somatosensory cortex
Ri retroinsular cortex
SII second somatosensory area
7b parietal area 7b
Ig insula granular
Id insula dysgranular
35 perirhinal cortex
4 primary motor cortex
6 premotor cortex
SMA supplementary motor cortex
30 area 30
23 posterior cingulate
24 anterior cingulate
9 prefrontal area 9
32 area 32
25 area 25
14 prefrontal area 14
10 area 10
15 area 15
12 area 12
11 area 11
13 area 13
G gustatory cortex
PaAr auditory parakoniocortical, rostral
PaAl auditory parakoniocortical, lateral
PaAc auditory parakoniocortical, caudal
KA auditory koniocortical, primary
proA auditory prokoniocortex

reit auditory retroinsular temporal cortex

TGd dorsal temporal polar cortex
TS1 superior temporal auditory area 1
Tpt auditory area Tpt
REFERENCES
Anderson, J.C., Binzegger, T., Martin, K.A.C. and Rockland, K.S. (1998) The connection from cortical area of
V1 to V5: a light and electron microscopic study. Journal of Neuroscience, 18, 10525–10540.
Baddeley, R., Abbot, L.F., Booth, M.C.A., Sengpiel, F., Freeman, T., Wakeman, E.A. and Rolls, E.T. (1997)
Responses of neurons in primary and inferior temporal visual cortices to natural scenes. Proceedings of the
Royal Society, series B, 264, 1775–1783.
Barone, P., Dehay, C., Berland, M., Bullier, J. and Kennedy, H. (1995) Developmental remodelling of primate
visual cortical pathways. Cerebral Cortex, 5, 22–38.
Braitenberg, V. and Schüz, A. (1991) Anatomy of the Cortex: Statistics and Geometry of Neuronal Connectivity
(Studies in Brain Functio series, Vol. 18). Berlin: Springer-Verlag.
Budd, J.M.L. (1998) Extrastriate feedback to primary visual cortex in primates: a quantitative analysis of
connectivity. Proceedings of the Royal Society, series B, 265, 1037–1044.
Burns, G.A.P.C. and Young, M.P. (2000) Analysis of the connectional organisation of neural systems associated
with the hippocampus in rats. Philosophical Transactions of the Royal Society: Biological Sciences, 355,
55–70.
Cherniak, C. (1994) Component placement optimization in the brain. Journal of Neuroscience, 14, 2418–2427.
Cowey, A. (1979) Cortical maps and visual perception. Quarterly Journal of Experimental Psychology, 31, 1–17.
Gallant, J.L., Connor, C.E. and Van Essen, D.C. (1998) Neural activity in areas V1, V2, and V4 during free
viewing of natural scenes compared to controlled viewing. Neuroreport, 9, 85–90.
Goodhill, G.J., Simmen, M.W. and Willshaw D.J. (1994) An evaluation of the use of multidimensional scaling
for understanding brain connectivity. Philosophical Transactions of the Royal Society of London—Series B:
Biological Sciences, 348, 265–280.
Grosof, D.H., Shapley, R.M. and Hawken, M.J. (1993) Macaque V1 neurons can signal illusory contours.
Hilgetag, C.-C., O’Neill, M.A. and Young, M.P. (1996) Indeterminate organization of the visual hierarchy.
Hilgetag, C.-C., O’Neill, M.A. and Young, M.P. (2000a) Hierarchical organisation of macaque and cat cortical
sensory systems explored with a novel network processor. Philosophical Transactions of the Royal Society:
Hilgetag, C.-C., Stephan, K., Burns, G., O’Neill, M.A. and Young, M.P. (2000b) Clustered organisation of
cortex and connectivity in macaque monkey and cat. Philosophical Transactions of the Royal Society:
Jung, M., Qin, Y., McNaughton, B. and Barnes, C. (1998) Firing characteristics of deep layer neurons in pre-
frontal cortex in rats performing spatial working memory tasks. Cerebral Cortex, 8, 437–450.
Kaas, J.H. and Huerta, M.F. (1988) The subcortical visual system of primates. In: H.D. Steklis and J. Erwin
(eds), Comparative Primate Biology, Vol. 4: Neurosciences. New York: A.R. Liss, pp. 327–391.
Kennedy, H. and Barone, P. (1999) Relative contribution of cortical projections to areas V1 and V4. Society for
Neuroscience Abstracts, 25, 572.18.
Kennedy, H., Bullier, J. and Dehay, C. (1989) Transient projections from the superior temporal sulcus to area 17
in the newborn macaque monkey. Proceedings of the National Academy of Science, U.S.A., 86, 8093–8097.
Knill, D.C. and Richards, W. (1996) Perception as Bayesian inference. Cambridge: Cambridge University Press.
Kötter, R. and Sommer, F.T. (2000) Global relationship between anatomical connectivity and activity propaga-
tion in the cerebral cortex. Philosophical Transactions of the Royal Society: Biological Sciences, 355,
127–134.
Laporte, G. and Taillefer, S. (1987) An efficient interchange procedure for the Archaeological seriation problem.
Journal of Archaeological Science, 14, 283–289.
Lennie, P. (1998) Single units and visual cortical organization. Perception, 27, 889–935.
Martin, K.A.C. (1992) Parallel pathways converge. Current Biology, 2, 555–557.

Maunsell, J.H. and Van Essen, D.C. (1993) The connections of the middle temporal visual area (MT) and their
relationship to a cortical hierarchy in the macaque monkey. Journal of Neuroscience, 3, 2563–2586.
McDonald, R.P. (1986) Describing the elephant: structure and function in multivariate data. Psychometrika, 51,
513–534.
Merabet, L., Desautels, A., Minville, K. and Casanova, C. (1998) Motion integration in a thalamic visual
nucleus. Nature (London), 396, 265–268.
Merigan, W.H. and Maunsell, J.H.R. (1993) How parallel are the primate visual pathways? Annual Review of
Miller, R. (1991) Cortico-hippocampal interplay and the representation of contexts in the brain. (Studies of
Brain Function series, Vol. 17), Berlin: Springer-Verlag.
Mitchison, G. (1991) Neuronal branching patterns and the economy of cortical wiring”. Proceedings of the Royal
Society, series B, 245, 151–158.
Mountcastle, V.B. (1978) An organizing principle for cerebral function: the unit module and the distributed
system In: G.M. Edelman and V.B. Mountcastle (eds), The Mindful Brain. Cambridge, Mass: MIT Press,
pp. 7–50.
Pandya, D.N. and Yeterian, E.H. (1985) Architecture and connections of cortical association areas. In: A. Peters
and E.G. Jones (eds), Cerebral Cortex, Vol. 4. New York and London: Plenum Press, pp. 3–61.
Perrett, D.I., Harries, M.H., Benson, P.J., Chitty, A.J. and Mistlin A.J. (1990) Neurones responsive to faces in the
temporal cortex: studies of functional organization, sensitivity to identity and relation to perception. Human
Peterhans, E. and von der Heydt, R. (1989) Mechanisms of contour perception in monkey visual cortex: I.
contours bridging gaps. Journal of Neuroscience, 9, 1749–1764.
Peters, A. and Payne, B.R. (1993) Numerical relationships between geniculocortical afferents and pyramidal cell
modules in cat primary visual cortex. Cerebral Cortex, 1, 69–78.
Poucet, B. (1997) Searching for spatial unit firing in the prelimbic area of the rat medial prefrontal cortex.
Behavioural Brain Research, 84, 151–159.
Rieke, F., Warland, D., de Ruyter van Steveninck, R. and Bialek, W. (1997) Spikes. Exploring the neural code.
Cambridge, Mass.: MIT Press.
Rockland, K.S. and Pandya, D.N. (1979) Laminar origins and terminations of cortical connections of the
occipital lobe in the rhesus monkey. Brain Research, 179, 3–20.
Scannell, J.W. and Young, M.P. (2000) Primary visual cortex within the cortico-cortico-thalamic network. In:
A. Peters, E.G. Jones and B.R. Payne (eds), Cerebral Cortex, Vol. 15. Cat Primary Visual Cortex. New
York: Plenum, in Press.
Scannell, J.W., Sengpiel, F., Benson, P.J., Tovee, M.J., Blakemore, C. and Young, M.P. (1996) Visual
motion processing processing in the anterior ectosylvian sulcus of the cat. Journal of Neurophysiology,
76, 895–907.
Scannell, J.W., Burns, G., O’Neill, M.A., Hilgetag, C.-C. and Young, M. (1997) The organization of the
thalamo-cortical network of the cat. Society for Neuroscience Abstracts, 23, 514.12.
Scannell, J.W., Burns, G., O’Neill, M.A. and Young, M.P. (1999) The connectional organisation of the thalamo-
cortico-cortical system of the cat. Cerebral Cortex, 9, 277–299.
Shepard, R.N. (1962) The analysis of proximities: multidimensional scaling with an unknown distance function.
I. Psychometrika, 27, 219–246.
Simmen, M.W., Goodhill, G.J. and Willshaw, D.J. (1994) Scaling and brain connectivity. Nature (London), 369,
448–50.
Stenson, H.H. and Knoll, R.L. (1969) Goodness of fit for random rankings in Kruskal’s nonmetric scaling
procedure. Psychological Bulletin, 71, 122–126.
Stephan, K.E., Hilgetag, C.-C., Burns, G.A.P.C., O’Neill, M.A., Young, M.P. and Kötter, R. (2000) Computa-
tional analysis of global functional connectivity between areas of primate cerebral cortex. Philosophical
Transactions of the Royal Society: Biological Sciences, 355, 111–126.
Treue, S. and Maunsell, J.H.R. (1996) Attentional modulation of visual motion processing in cortical areas MT
and MST. Nature (London), 382, 539–541
Ungerleider, L.G. and Mishkin, M. (1982) Two cortical visual systems. In: D.G. Ingle, M.A. Goodale and
R.J.Q. Mansfield (eds), Analysis of Visual Behavior. Cambridge MA: MIT Press, pp. 549–586.
Wilkinson, E.M. (1971) Archaeological seriation and the travelling salesman problem. In: F.R. Hodson,
D.G. Kendall and P. Tautu (eds), Mathematics in the Archaeological and Historical Sciences. Edinburgh:
Edinburgh University Press, pp. 276–283.
Young M.P. (1992) Objective analysis of the topological organization of the primate cortical visual system.
Young, M.P. (1993) The organization of neural systems in the primate cerebral cortex. Proceedings of the Royal
Society, series B, 252, 13–18.
Young, M.P. (1995) Open questions about the neural mechanisms of visual pattern recognition. In: M.S. Gazzaniga
(ed.), The Cognitive Neurosciences. London: MIT Press, pp. 463–474.

Young, M.P. and Scannell, J.W. (1996) Component placement optimization in the brain. Trends in Neuro-
sciences, 19, 413–414.
Young, M.P., Scannell, J.W. and Burns, G. (1994) The Analysis of Cortical Connectivity. New York: R.G.
Landes; Heidelberg: Springer.
Young, M.P., Scannell, J.W., O’Neill, M.A., Hilgetag, C.C., Burns, G. and Blakemore, C. (1995) Non-metric
multidimensional scaling in the analysis of neuroanatomical connection data and the organization of the
primate cortical visual system. Philosophical Transactions of the Royal Society, Biological Sciences, 348,
281–308.
Young, M.P. and Yamane, S. (1992) Sparse population coding of faces in the inferotemporal cortex. Science
(Washington), 256, 1327–1331.
Young, M.P. and Scannel, J.W. (2000) Brain structure-function relationships: advances from neuroinformatics.
Philosophical Transactions of the Royal Society, Biological Sciences, 355, 3–6.
Zeki, S. and Shipp, S. (1988) The functional logic of cortical connections. Nature (London), 335, 311–317.

16 The Human Cortical White Matter: Quantitative
Aspects of Cortico-Cortical Long-Range Connectivity
Almut Schüz1,* and Valentino Braitenberg1,2,3
1
Max-Planck-Institut für biologische Kybernetik, Spemannstr. 38 72076 Tübingen, Germany
2
Institut für medizinische Psychologie der Universität Tübingen Gartenstr. 29
72072 Tübingen, Germany
3
Laboratorio di Scienze Cognitive, Università di Trento, v.Tartarotti 7, Rovereto (Italy)
*Correspondence: Max-Planck-Institut für biologische Kybernetik, Spemannstr. 38,
72076 Tübingen, Germany, Tel: 0049-7071-601 544; FAX: 0049-7071-601 577
e-mail: almut.schuez@tuebingen.mpg.de; valentino.braitenberg@tuebingen.mpg.de
We investigated the human cortical white matter in order to get insights into quantitative aspects of
connectivity between cortical regions. We dissected the long-range bundles in the depth of the white
matter which run over large distances and connect the cortical lobes to each other. Measuring the
cross-sectional areas of these bundles and multiplying them by the assumed density of fibers we
could estimate the number of fibres in these bundles. It turns out that the total number of fibers in the
intrahemispheric long-range bundles is only about 2% of the total number of cortico-cortical fibres,
and is of the same order as the number of fibers in the callosal system. Evidently, the vast majority
of cortico-cortical fibers are of shorter range, connecting cortical areas within one lobe or neighbour-
ing areas belonging to different lobes. As a rough rule, the number of fibres of a certain range of
lengths is inversely proportional to their length. The results are discussed in relation to information
processing within and between modalities.
KEYWORDS: cell assembly, cortical areas, cortical hierarchy, cortical lobes, fascicle, fibre length
1. INTRODUCTION
When we talk about the cerebral cortex, what we usually have in mind is the sheet of
laminated grey matter surrounding the forebrain. This is where the phenomenon of areal
diversification manifests itself most evidently. However, the white matter underneath it is
an equally important constituent of the telencephalic cortex: hardly any other structural
feature characterizes the cortex as much as the mass of fibres underlying it, mostly com-
posed of axons of cortical neurones. The vast majority of these fibres connect the cortex to
itself as is evident in the large amount of cortical white matter as compared to the thick-
ness of the various subcortical fibre tracts (Braitenberg, 1974; Seitz, this volume). While
the intrinsic connectivity within an area seems to be provided to a large extent by collat-
erals running within the grey matter (e.g. Yoshioka et al., 1992; Amir et al., 1993), the
traffic between different areas is mainly the responsibility of the main axons, running
through the white matter. This is certainly true for the majority of fibres in large brains, in
spite of the fact that axon collaterals freely cross the borders between neighbouring
cortical areas (e.g. DeFelipe et al., 1986). The spatial separation between the two systems
377
378 Almut Schüz and Valentino Braitenberg
is less distinct in small brains such as that of the mouse, in which areas measure no more
than a few millimeters across.
2. WHITE MATTER AND BRAIN SIZE
A large amount of white matter is by no means unique to the human brain, but common to
large brains with a folded cortex. In contrast, in small, lissencephalic brains the white
matter is confined to a thin sheet. The volume of white matter increases from about 6% in
small insectivores to about 42% of the total neocortical volume (grey + white matter) in
humans (Frahm et al., 1982). This is so for obvious combinatorial reasons, and in addition
because with larger brains axons become longer on average (Braitenberg, 2001) and, to
a certain degree, also thicker (Ringo, 1991; Jerison, 1992; Schüz and Preißl, 1996). The
large amount of white matter in large brains does not necessarily mean that the cortex is
more “self-reflexive” in larger brains, but may actually be a consequence of a common
allometric principle (Braitenberg, 1978).
3. COMPOSITION OF THE HUMAN CORTICAL WHITE MATTER
At first sight, the white matter has the appearance of a homogeneous mass. However, as
has been known for a long time (Déjérine, 1895), fibres having the same origin and des-
tination tend to stick together, forming sheets and fascicles, which can be recognized on
myelin preparations and can be isolated by blunt dissection as practiced traditionally in
laboratories of human anatomy. Cortico-cortical fibres which run over long distances
within the same hemisphere are located in the depth of the white matter, while fibres of
shorter range are located more superficially, the most superficial ones forming the
so-called U-fibre system, and staying mostly within the compass of a single gyrus or sulcus.
Also, the callosal and the afferent and efferent systems tend to run in sheets. The various
systems are interdigitated in places where they cross each other, but in spite of this they
can be visualized more or less separately at various locations, and their dissection can be
facilitated by repeated freezing and unfreezing of the formalin-fixed tissue, as shown in
the beautiful atlas by Gluhbegovic and Williams (1980).
In the core of the white matter, five main fascicles can be discerned which connect the
different cortical lobes with each other: the superior and inferior occipitofrontal fascicle,
the superior and inferior longitudinal fascicle and the uncinate fascicle (Figure 16.1).
4. QUANTITATIVE ASPECTS
We have made an estimate of the number of fibres in these fascicles, in order to see how
large a part of the cortico-cortical long-range system they represent. We did this by first
assessing the cross-sectional areas of the various fascicles. For this we used both, brain
atlases (Gluhbegovic and Williams, 1980; Montemurro and Bruni, 1981) and our own
dissection material. For the measurements, we used the central portion of the course of the
fasicles where they can be assumed to contain a substantial proportion of the fibres which

The Human Cortical White Matter 379
Figure 16.1. Overview over the intrahemispheric long-range bundles in the depth of the white matter.
1. Superior occipitofrontal fascicle, 2. Location of the Corona radiata, 3–5. Superior longitudinal fascicle,
6. Outline of the insula, 7. Inferior occipitofrontal fascicle, 8. Inferior longitudinal fascicle, 9. Location of the
anterior commissure, 10. Uncinate fascicle. From Nieuwenhuys et al. (1991). Copyright: Springer Verlag,
with permission.
they collect, on a segment where they were most compact and not loosened by crossing
fibres from other systems.
The measured cross-sectional areas were then multiplied by the density of fibres, in
order to assess the number of fibres in the fascicles. Since we could not find any data on
the density of fibres in these fascicles, we used the density of fibers in the corpus callosum
for which data are available (see collection of data in Blinkov and Glezer, 1968; Tomasch,
1954; Aboitiz et al., 1992). Since the corpus callosum is part of cortical white matter, and
since it connects the cortex of both hemispheres over similar distances as the fascicles do
between the lobes within one hemisphere, fibre thickness in these two systems may be
assumed to be similar. The density of fibres in the human corpus callosum is about
380 000/mm2 (Aboitiz et al., 1992) including a few percent unmyelinated fibres.
We also included in our measurements of the intrahemispheric long-range bundles the
cingulum, which connects the hippocampal formation to the cingulate gyrus, as well as
connecting portions of the cingulate gyrus to each other (Brodal, 1981).
The cross-sectional areas of the cingulum and of the inferior longitudinal fascicle are of
similar size, about 19.6 mm2 (slightly more than the diameter of the optic nerve). The
uncinate fascicle and the inferior occipitofrontal fascicle form a compact bundle at their
transition between temporal and frontal lobe where they have together a cross-sectional

Table 16.1.
Density of axons in the corpus callosum (Aboitiz et al., 1992) 380 000/mm2
Total number of axons in the corpus callosum 2–3 × 108
(Blinkov and Glezer, 1968; Aboitiz et al., 1992)
Number of axons in the
Cingulum 7.4 × 106
Superior longitudinal fascicle 4.3 × 107
Uncinate fascicle + inferior occipitofrontal fascicle 3.1 × 107
Inferior longitudinal fascicle 7.4 × 106
Superior occipitofrontal Fascicle between 6 × 106 and 1.2 × 107
Total number of axons in the far-reaching fascicles about 108
in one hemisphere
area of about 81 mm2. The superior occipitofrontal fascicle is difficult to measure but has
a cross-sectional area one- to two-times that of the optic nerve, i.e. 16 to 32 mm2. The largest
bundle, the superior longitudinal fascicle, has a cross-sectional area of about 114 mm2.
Table 16.1 shows the results of our estimates. The number of fibres ranges within the
orders of magnitude of 106 and 107 in the various fascicles. The total number of fibres in
all of them together in one hemisphere may therefore reach the order of 108. Interestingly,
this is the same order as that of the number of fibres in the corpus callosum.
We then wondered how large a part of the total cortico-cortical system of one hemi-
sphere is contained in these long intrahemispheric fascicles. The calculation is shown in
Table 16.2. For this we need the total number of neurones within the cortex of one
hemisphere, which is, according to Haug (1986), 7.5 × 109. From this number, we have to
subtract the following: (i) the non-pyramidal cells (which make only local connections),
(ii) the callosal neurones projecting to the other hemisphere, and (iii) the efferent neurones
projecting to other parts of the brain. The third of these is the number about which least is
known. Our guess is of the order of 108 for one hemisphere. This is based on the assump-
tion that the cortical neurones projecting to the thalamus are not much more than the
thalamo-cortical neurones for which an upper limit is given by the number of thalamic
neurones (at most 108 for one hemisphere, see data in Blinkov and Glezer, 1968), and on
the assumption that cortico-striatal neurones are roughly of the same number, and neu-
rones projecting to the brain stem are much fewer (of the order of 107 to the main output
structure of the brain stem, the pontine nuclei [Brodal, 1981; Glickstein, 1987]).
Table 16.2.
Neurones in 1 hemisphere
(Haug, 1986) 1.5 × 1010: 2 = 7.5 × 109
minus 15% non-pyramidal cells
(Braak and Braak, 1986) – 1.1 × 109
minus callosal neurones in 1 hemisphere
(see Table 16.1) – 1 to 1.5 × 108
minus efferent neurones –2 × 108
= Cortico-cortical ipsilateral = 6 × 109
⇒about 2% of those (10 ) are in the long fascicles
8

With the estimates given in Table 16.2, one ends up with a total number of about 6 × 103
neurons making cortico-cortical projections within the same hemisphere (80% of all the
neurones in the cortex of one hemisphere). In spite of all the uncertainties contained in
these estimates, it is evident that the 108 fibers in the deep fascicles constitute only about
2% of the cortico-cortical fibres in the white matter.
5. CONSIDERATIONS ON THE DISTRIBUTION OF FIBRES OF DIFFERENT

LENGTH IN THE PYRAMIDAL-TO-PYRAMIDAL CELL SYSTEM
What one would ultimately like to know in the context of a theory of the global operation
of the cortex, is the distribution of fibre lengths in the system of fibres connecting the cor-
tex to itself. Our estimate of a 2% share of each of the two long-range systems, the corpus
callosum and the intrahemispheric bundles, is but a small contribution to this. Tentatively,
we may go toward a complete picture in a rough way by comparing estimates of the
numbers of (a) very short (intracortical) tangential connections, (b) longer fibers, linking
the cortical grey matter as U-fibers and (c) long fibers in the deep bundles and the callosal
system (Figure 16.2).
The range of intracortical connections, provided by the horizontal collaterals of
descending axons is known only for the cortex of smaller primates (a few millimeters in
the macaque; e.g. Fisken et al., 1975; DeFelipe et al., 1986; Amir et al., 1993; Lund et al.,
1993; Levitt et al., 1993; Levitt and Lund, this volume; Valverde et al., this volume) and
Figure 16.2. Number of fibers (in both hemispheres together) vs range (log–log plot) in the three compart-
ments described in the text. Compartment A is assumed to contain intracortical horizontal fibres up to a length
of 3 mm, compartment B corresponds to the fibres in the U-fibre system of the white matter, and compartment C
to cortico-cortical fibres of longer range, which do not follow the pattern of cortical folding. The latter are
assumed to include the deep fascicles, the more superficial bundles and the callosal system. The inset on the
upper right shows a sketch of the 3 fibre systems quantified here.

can be inferred only indirectly for humans. In Figure 16.3 we assume a maximum hori-
zontal distance of 3 mm, spanned by the longest axon collaterals within the grey matter.
The number of horizontal axon collaterals is a few times the number of pyramidal cells
(column A in Figure 16.2).
The next component for which an estimate is possible, are the fibres leaving the cortex
for the white substance and re-entering it at a distance of one or a few centimeters. (We
assumed up to 30 mm). They are largely contained in the fibre layer which underlies the
cortex, following its convoluted shape. This layer, the so-called U-fibre system, has a thick-
ness of about 1.5 mm, as measured beneath the sulci. The number of fibres can be esti-
mated on the basis of the volume of this layer and the (estimated) thickness and length of
the fibres. To estimate the volume, we take the surface of the cortex (about 1600 cm2 for
both hemispheres together; see Blinkov and Glezer, 1968; Henery and Mayhew, 1989;
Sisodiya et al., 1996), multiplied by the thickness of the U-fibre layer. We take 2 µm2 as
the cross-section area of the fibres, slightly less than that of the fibres in the human corpus
callosum, according to Blinkov and Chernyshev (see Table 16.1). The average length of
the fibres is estimated as 15 mm.
In order to assess the number of cortico-cortical fibres in the third compartment (C in
Figure 16.2), we made an estimate of the volume occupied by these fibres in the white
matter. The volume of the cortical white matter in both hemispheres together (including
the corpus callosum, but excluding the internal capsule) is about 420 cm3 (Frahm et al.,
1982). From this we subtract the U-fiber system as measured above (240 cm3), as well as
an estimated volume for the efferent and afferent fibres. According to the assumptions
made in the previous section, their number may reach 6 × 108 in both hemispheres
together. Assuming an average length of 50 mm for these fibres, and a thickness of 2 µm2,
we end up with a volume of 60 cm3. This leaves a volume of 120 cm3 for the longer
cortico-cortical fibres in the intrahemispheric and commissural systems. These fibres may
reach a length of 15 cm; for the average we took a length of 60 mm. With a fibre thickness
of 2.6 µm2 corresponding to the fibre thickness in the corpus callosum, we end up with
a number of about 8 × 108 in compartment C in Figure 16.2.
The number of fibres in the compartments B and C in Figure 16.2 do not quite add up to
the total number of pyramidal cells assumed in the previous section. This is not astonish-
ing in view of the uncertainties inherent in these estimates: These include uncertainties in
the total number of cortical neurones, in the percentage of pyramidal cells that send an
axon to the white substance, in the range of volumes given for the white matter (Blinkov
and Glezer, 1968) and the uncertain quantitative data on the efferent and afferent systems.
Also, the fibre thickness in the various compartments of the white substance is known
only in an approximate way.
In spite of this, orders of magnitude are probably correct. Roughly, the relation between
range and number of cortico-cortical fibres is such that for an increase of one order of
magnitude in length, their number goes down by one order of magnitude.
6. OVERALL ORGANIZATION OF THE LONG-RANGE SYSTEM

AND HEBB’S THEORY
We end up with the following picture: (1) The two far-reaching fibre systems of the
cortical white matter, the callosal system connecting both hemispheres and the ipsilateral

deep fascicles connecting the various lobes of one hemisphere with each other are of
a similar magnitude. (2) Both of them together contain only a small percentage (no more
than 4%) of the total number of cortico-cortical fibres in the white matter. The over-
whelming majority of cortico-cortical fibres must be of shorter range, connecting
areas within one lobe, or neighbouring areas belonging to different lobes. Most of
these shorter fibres of the white substance are contained in the thin layer of so-called
U-fibers lining the deep surface of the cortex. Even larger is the number of intracortical
fibers which connect pyramidal cells of one neighbourhood without entering the white
substance.
The preponderance of fibres not exceeding a few centimeters in length agrees well
with the results reported by the group of Young on the connectivity between cortical
areas in cat and monkey, though based on a completely different approach (see Young,
this volume). Bringing together all the connections which have been reported in the lit-
erature as obtained by tracer studies, it turned out that the connectional vicinity between
cortical areas reflects—to an astonishing degree—their topographic vicinity in the
cortex (Young et al., 1994, 1995). Almost half of the connections between areas are
“nearest-neighbour-or-next-door-but-one” connections (47.8% in the cat; Scannell et al.,
1995).
This sheds an interesting light on the representation of concepts in the cortex. Accord-
ing to Hebb’s theory (1949), briefly sketched in the introductory chapter, a concept is
represented in the cortex by a group of neurones binding together the various features
pertaining to it. This leads us to ask at what level in the hierarchy of connections this
occurs. Since direct connections between primary sensory areas of different modalities
seem not to exist, it cannot be at the lowest level, at least as far as concepts which involve
more than one modality are concerned. In principle, it is still possible that neurones in
areas quite close to the primary sensory areas are connected into a multimodal cell
assembly. As is evident from the scheme of connections between cortical areas in the
monkey (Young, 1993), direct connections between areas of different modalities already
exist two levels above the primary sensory areas of the visual, acoustic and somatosenory
system. In addition, there are connections between the first level above the primary sens-
ory area of one modality and the second level of the other modalities for all three sensory
systems.
However, the majority of connections seems not to be involved in this kind of low-
level integration between different sensory systems which enter the cortex far from
each other in the different cortical lobes. Obviously, the greater part of cortico-cortical
fibres are involved in the processing of information between the areas of one neigh-
bourhood, i.e. to a large extent within the same modality. This suggests that, normally,
a great deal of preprocessing happens below the level at which cell assemblies repre-
senting a multimodal or abstract concept are assembled. This preprocessing could be
both in the service of invariance of shape, size, etc. (Zeki, 1993), or of the establish-
ment of repertoires of meaningful features (e.g. Tanaka et al., 1991; 1997; Oram and
Perrett, 1994).
To the neuroanatomist the surprizing result of our survey is probably that most of the
cortico-cortical fibres in the white substance of the hemisphere are contained in the
narrow layer just beneath the cortex. The brain theorist may find food for thought in the
inverse relation of number and length of fibres, which may remind him of the well-known
Zipf-law of linguistics.

ACKNOWLEDGEMENT
We are grateful to Prof. Wagner who gave us the possibility to use dissection material
from the Anatomical Institute at the University of Tübingen.
REFERENCES
Aboitiz, F., Scheibel, A.B., Fisher, R.S. and Zaidel, E. (1992) Fiber composition of the human Corpus callosum.
Blinkov, S.M. and Glezer, I.I. (1968) The Central Nervous system in figures and tables, German edition of the
Russian original (1964), Jena: VEB Fischer Verlag, pp. 140–148, p. 164.
Blinkov, S.M. and Chernyshev, A.S. (1935) The variability of the human Corpus callosum (in Russian), Tr. In-ta
mozga I, 175–237, Moscow, quoted in Blinkov, S.M. and Glezer, I.I. (1968).
Braak, H. and Braak, E. (1986) Ratio of pyramidal cells versus non-pyramidal cells in the human frontal isocor-
tex and changes in ratio with ageing and Alzheimer’s disease. In: D.F. Swaab, E. Fliers, M. Mirmiram,
W.A. van Gool and F. van Haaren (eds), Aging of the brain and Alzheimer’s disease (Progress in Brain
Research, Vol. 70). Amsterdam: Elsevier, pp. 185–212.
Braitenberg, V. (1974) Thoughts on the cerebral cortex. Journal of Theoretical Biology, 46, 421–447.
Braitenberg, V. (1978) Cortical architectonics: general and areal. In: M.A.B. Brazier and H. Petsche Architecton-
ics of the cerebral cortex, New York: Raven Press, pp. 443–465.
Braitenberg, V. (2001) Brain size and the number of neurons: an exercise in synthetic neuroanatomy. Journal of
Computational Neurocience, 10, 71–77.
Brodal, A. (1981) Neurological Anatomy in Relation to Clinical Medicine, 3rd edn. New York, Oxford: Oxford
University Press, p. 669.
DeFelipe, J. Conley, M. and Jones, E.G. (1986) Long-range focal collateralization of axons arising from cortico-
cortical cells in the monkey sensory-motor cortex. Journal of Neuroscience, 6, 3749–3766.
Déjérine, J. (1895) Anatomie des Centre Nerveux, Vol.1, edn. of 1980. Paris, New York: Masson.
Fisken, R.A., Garey, L.J. and Powell, T.P.S. (1975) The intrinsic, association and commissural connections of
area 17 of the visual cortex. Philosophical Transactions of the Royal Society of London, B, 272, 487–536.
Frahm, H.D., Stephan, H. and Stephan, M. (1982) Comparison of brain structures in insectivores and primates. I.
Neocortex. Journal für Hirnforschung, 23, 375–389.
Glickstein, M., May III, J.G. and Mercier, B.E. (1985) Corticopontine projection in the Macaque: The distribution
of labelled cortical cells after large injections of horseradish peroxidase in the pontine nuclei. Journal of
Gluhbegovic, N. and Williams, T.H. (1980) The Human Brain, a photographic guide. Hagerstown: Harper and
Row, Publishers.
Haug, H. (1986) History of neuromorphometry. Journal of Neuroscience Methods, 18, 1–17 (Special Issue:
H.B.M. Uylings, R.W.H. Verwer and J. van Pelt [eds], Morphometry and Stereology in Neurosciences).
Hebb, D.O. (1949) Organization of Behaviour. A Neuropsychological Theory, 2nd edn. 1961. New York: Wiley
and Sons.
Henery, C.C. and Mayhew, T.M. (1989) The cerebrum and cerebellum of the fixed human brain: efficient and
unbiased estimates of volumes and cortical surface areas. Journal of Anatomy, 167, 167–180.
Jerison, H. (1991) Brain size and the evolution of mind. 59th James Arthur Lecture. New York: American
Museum of Natural History.
Levitt, J.B., Lewis, D.A., Yoshioka, T. and Lund, J.S. (1993) Topography of pyramidal neuron intrinsic connections in
macaque monkey prefrontal cortex (Areas 9 and 46). Journal of Comparative Neurology, 338, 360–376.
Montemurro, D.G. and Bruni, J.E. (1981) The Human Brain in Dissection. Philadelphia, London: W.B. Saunders
Company.
Nieuwenhuys, R., Voogd, J. and van Huijzen, Chr. (1991) The Human Central Nervous System—A synopsis and
Atlas, 2nd German edn. Berlin, Heidelberg: Springer-Verlag.
Oram, M.W. and Perrett, D.I. (1994) Modeling visual recognition from neurobiological constraints. Neural
Networks, 7, 945–972.
Ringo, J.L. (1991) Neuronal interconnections as a function of brain size. Brain Behavior and Evolution, 38, 1–6.
Scannell, M.P., Blakemore, C. and Young, M.P. (1995) Analysis of connectivity in the cat cerebral cortex.
The Journal of Neuroscience, 15, 1463–1483.

Schüz, A. and Preißl, H. (1996) Basic connectivity of the cerebral cortex and some considerations on the Corpus
callosum. Neuroscience and Biobehavioral Reviews, 20, 567–570.
Sisodiya, S., Free, S., Fish, D. and Shorvon, S. (1996) MRI-based surface area estimates in the normal adult
human brain: evidence for structural organisation. Journal of Anatomy, 188, 425–438.
Tanaka, M. (1997) Mechanisms of visual object recognition: monkey and human studies. Current Opinion in
Tanaka, K., Fukada, S.H. and Moriya, M. (1991) Coding visual images of objects in the inferotemporal cortex of
the macaque monkey. Journal of Neurophysiology, 66, 170–189.
Tomasch, J. (1954) Size, distribution and number of fibers in the human Corpus callosum. Anatomical Record,
119, 119–135.
Yoshioka, T., Levitt, J.B. and Lund, J. (1992) Intrinsic lattice connections of macaque monkey visual cortical
area 4. Journal of Neuroscience, 12, 2785–2802.
Young, M.P. (1993) The organization of neural systems in the primate cerebral cortex. Proceedings of the Royal
Society, London, B, 252, 13–18.
Young, M.P., Scannell, J.W., Burns, G.A.P.C. and Blakemore, C. (1994) Analysis of connectivity: neural
systems in the cerebral cortex. Reviews in the Neursciences, 5, 227–249.
Young, M.P., Scannell, J.W. and Burns, G. (1995) The Analysis of Cortical Connectivity. New York, Berlin:
Springer.
Zeki, S. (1993) A vision of the brain. Oxford: Blackwell Scientific Publications.

17 Fundamentals of Association Cortex
Stewart Shipp
Wellcome Department of Cognitive Neurology, University College, London
Correspondence: S. Shipp, Wellcome Department of Cognitive Neurology,
University College, Gower Street, London WC1E 6BT, U.K.
Tel: (+44)207-679-3910; FAX: (+44)207-679-7316; e-mail: s.shipp@ucl.ac.uk
There are many instances of parallelism in the neural architecture of the visual system. This review
deals with three, expressed by various forms of anatomical segregation at progressively higher
levels: (a) the three output channels of the retina (M, P and K); (b) the three ‘metabolic’ pathways at
the level of V1 and V2 (blob-thin stripe, interblob-interstripe and layer 4B-thick stripe), and (c) the
dual dorsal and ventral pathways, feeding the parietal and temporal lobes of the brain. Only if these
three parallel systems dovetail with each other could the global architecture of the visual system be
truly described as parallel, but this is not so: there is abundant cross-talk at the interface of (a) with
(b), and (b) with (c)—and internally within each system. Cross-talk reflects integrative functions that
tend to be neglected by parallel models. The ethos of the article is that cross-talk serves co-operative
interactions that facilitate the operational goals of a particular pathway, rather than as acting to dilute
its specific, specialised character. In that sense, the functional organisation of the visual system is
more parallel than its underlying circuitry.
1. INTRODUCTION
In the history of thinking about the brain, localisation of function is one of the oldest
thoughts. As the resolution of anatomical studies has increased, it has stayed with us, to be
applied to structures across the cortical sheet, ranging in scale from lobes and architectural
fields, through areas, to compartments and columns. To give it another name, localisation
can be thought of as segregation, the idea that any structurally (or neurochemically, or
genetically) distinct module of tissue is likely to have a corresponding functional identity,
a specialisation that distinguishes it from neighbouring structures. To anyone willing to
accept the label “neuroanatomist”, this notion is probably axiomatic.
Naturally enough, there are comparable levels of functional classification to accompany
the varying scales of anatomical segregation. Within vision, we can identify the basic attributes
of seeing colour, form (shape, depth and texture) and motion. At a higher level, we find the
recruitment of these modalities to serve different types of action: grasping and manipulating
an object, as opposed to verbal description of it (or perhaps just silent contemplation). At
a subordinate level, it is common to identify, for example, neural selectivity for particular
hues, orientations, or disparities. Clearly, any global theory of “how the brain works” has
to match the functions to the anatomical structures. This is exactly what the most generally
accepted theory of visual brain function, the dogma of two visual pathways, undertakes.
It cites the mass of evidence for the divergence in function of the parietal and temporal
387
388 Stewart Shipp
lobes of the primate brain (allying human neuropsychology and functional imaging with
monkey neurophysiology and behaviour) and links it to anatomical evidence that these two
centres are fed by two separate pathways arising from the interconnections of prestriate
areas and V1. Thus the dorsal (“WHERE”) pathway is held to terminate in the parietal lobe
and to subserve location of objects in space, and the ventral (“WHAT”) pathway is thought
to terminate in the temporal lobe, subserving object recognition (Mishkin et al., 1983; Haxby
et al., 1991). This dogma has been extended to stress the role of the dorsal pathway in motor
control, and of the ventral pathway in perceptual awareness (Goodale and Milner, 1992).
For a passing moment, it had seemed possible that the cortical “two pathway” dogma
could be extended still further downwards, to incorporate the fundamental distinction between
the magno- (M) and parvocellular (P) retinogeniculate pathways, a kind of grand unification
of two-pathway theories (Livingstone and Hubel, 1987b). The idea was that the dorsal
pathway might be fed exclusively by the M system, and the ventral pathway by the P system,
but this view began to unravel almost before it could be proposed: the P and M systems
are not so completely segregated. In general, the problem is that the segregations existing
at one stage do not necessarily exist in 1:1 relationship with those at the next. This feature
(cross-talk, or cross-connectivity) is an omnipresent feature of cortical organisation, and it
also applies to the post-striate levels of the two pathway dogma.
More on that below. To complete the introduction, it remains to point out that cross-talk
is a reflection of the other fundamental property of association cortex, integration. Like
segregation, integration is a broad term, referring to anatomical relationships and functional
processes over a range of different levels. The historical defining characteristic of association
cortex is the association of sensations arising from different sensory modalities. The analogous
property, within vision, is the binding of visual attributes recognised to belong to the same
object, yet processed along separable pathways. Another example might be the synthesis of
one attribute out of another, for instance form-from-motion. The element of integration is
one that is neglected by parallel models of brain function.
Segregation and integration are the twin fundamentals of cortical organisation. In an
elementary sense, the latter cannot exist without the former. It is easy to cast them as oppon-
ent processes—but this, surely, is to miss the evident reality that they must cooperate to
subserve the goals of cortical processing. The aim of what follows is, first, to ask precisely
what is segregated in early visual analysis, and then to examine the greyer territory of where,
how and why it is brought back together—if at all.
2. GENICULO-CORTICAL PATHWAYS
There are three classes of relay neurone in the lateral geniculate nucleus (LGN): magno-
cellular (M) and parvocellular (P) cells, whose properties are well documented, and konio-
cellular (K) cells, which are not. That they are not is because K cells have tiny cell bodies
and are scattered throughout the intercalated layers of the LGN, so that they are extremely
difficult to study (Hendry and Yoshioka, 1994). All three classes seem to inherit their
response properties from equivalent classes of retinal ganglion cells, avoiding cross-talk
between channels. M and P cells have similar concentric receptive fields (ON-centre,
OFF-surround or vice versa), although only P cells have cone opponent inputs (Wiesel and
Hubel, 1966; Dreher et al., 1976; Schiller and Malpeli, 1978). The two channels are
conceived as spatio-temporal filters with overlapping ranges of responsivity, although

Fundamentals of Association Cortex 389
these response ranges are notably dependent on the nature of stimulus contrast (i.e. luminance
or chromatic). M cells have the higher absolute sensitivity to luminance contrast, and they
respond to a higher range of temporal frequencies (Shapley et al., 1981; Hicks et al., 1983;
Derrington and Lennie, 1984). With larger calibre axons, they also have faster signal
conduction velocities. These properties give a higher resolution for timing of events, so
that M cells are the natural substrate for coding dynamic changes in an image. P cells, with
smaller receptive fields, can respond to a higher range of spatial frequencies (Derrington
and Lennie, 1984); furthermore, P cells outnumber M cells by 10:1 in the LGN, and by
a still greater ratio in the fovea (Connolly and Van Essen, 1984), so they are the prime source
of fine-grain spatial representation. This picture changes considerably for chromatic contrast.
The M system is “colour blind” in the sense that it provides no information about colour,
and its spectral sensitivity is equivalent to that of the V(λ) human luminous efficiency
curve; using the technique of heterochromatic flicker photometry, activity of the M system
reaches a minimum at the point of minimal perceived flicker, which defines the psycho-
physical state of isoluminance between the tested pair of coloured lights (Lee et al., 1988).
In P cells, chromatic contrast (tested with isoluminant stimuli) has the effect of shifting the
response range to low-pass for spatial frequencies (DeValois et al., 1977; Derrington
et al., 1984). This means that, with everyday visual images, P cells extract a chromatic signal
from low spatial frequencies, and a multiplexed signal that is increasingly dominated by
achromatic contrast, from higher spatial frequencies (Ingling and Martinez-Uriegas, 1983).
Lesions of specific layers in the LGN support this picture (Schiller et al., 1990; Merigan
et al., 1991a,b). Lesions in the M layers create deficits in seeing flicker and motion.
Lesions in the P layers abolish colour vision, and severely impair detailed form vision and
stereopsis. However the capacity to make certain coarse discriminations of form, brightness
and size is unaffected by lesions in either layer (although abolished by wholesale LGN
destruction) meaning that one channel can substitute for another in performing these
simple tasks. The lesion studies are a more reliable method of distinguishing the roles of
the M and P channels than purely psychophysical procedures using isoluminant coloured
stimuli to “neutralise” the M system. Because, at isoluminance, the M system is not
exactly silent, and because the P response is also of different character (Schiller and Colby,
1983; Lee et al., 1988; Hubel and Livingstone, 1990), it is unreliable to ascribe all residual
perceptual capacities at isoluminance to the P system, and all deficits in performance to
inactivity of the M system (Logothetis et al., 1990). These conclusions may be correct with
regard to residual motion perception, which may seem slowed or jumpy, i.e. lacking
smoothness (Livingstone and Hubel, 1987b; Gegenfurtner and Hawken, 1996). However,
they are incorrect, for instance with regard to isoluminant deficits in stereovision (Livingstone
and Hubel, 1987b): the lesion experiments suggest that stereopsis is primarily supported by
the P system (Schiller et al., 1990).
The demonstration that different percepts exploit different subcortical channels reveals
nothing about how these channels are processed cortically. Thus, one must ask how far
centrally the segregation between the M and P systems is maintained beyond their initial
relay through layer 4C of V1. Layer 4B provides an immediate example of a layer with
evident magnocellular character. It is driven by M input from 4Ca (Lund et al., 1994;
Callaway, 1998), and has neurones with high contrast sensitivity, typical of the M system
(Blasdel and Fitzpatrick, 1984; Hawken and Parker, 1984). Layer 4B has several other
unique characteristics: it is the site of the dense lateral fibre plexus (the stria of Gennari, after
which striate cortex is named); it is populated by large spiny stellate projection neurones

390 Stewart Shipp
(like pyramidal cells shorn of their apical dendrites), whose dendrites are thus confined to
4B, but whose axons leap several rungs of the cortical hierarchy to terminate in area V5
(Shipp and Zeki, 1989a); and it is the earliest site in the primate visual system where
direction-selective cells are located (Dow, 1974; Orban et al., 1986). Hence it is but four
synapses from the retina to area V5/MT, and this is generally recognised as a fast pathway
devoted to motion analysis. The dominance of the M system in V5 can be demonstrated by
selective inactivation of the LGN: blockade of the M layers essentially eliminates activity in
V5, apart from a few sites where P-driven responses can be detected (Maunsell et al., 1990).
The relationship of the M and P systems turns out to be asymmetrical, however. Pro-
gressive accounts differ in detail, but it seems clear that the domain of the P system (layers
4A, 3 and 2 of V1) is substantially invaded by M input. Selective blockade of the LGN, for
instance, suggests relatively equal contributions from both systems to these layers, and
convergence at the level of single neurones (Nealey and Maunsell, 1994). A detailed
anatomical picture is provided by Figure 17.1. In brief, layer 4A is itself a direct target of
P geniculocortical axons, which are reinforced by further relays from 4Cb, extending
through 4A and lower layer 3. There is no detectable bias in these relays either toward or
away from the characteristic cytochrome oxidase blobs. M input to interblobs derives from
cells situated around the border zone of 4Ca and 4Cb, whose dendrites should be able to
sample both M and P input. The weight of M input to layers 2/3 may be biased toward
blobs, however, as cells in the upper part of 4Ca, and also in 4B, have axonal arbors that
are focused within blobs. This concurs with evidence that contrast sensitivity is higher in blobs
than interblobs (Tootell et al., 1988a; Edwards et al., 1995).
The cytochrome oxidase blobs and interblobs of layer 2/3 can be pictured as de-multiplex-
ing the P signal in different ways: blobs extract the low spatial frequency colour informa-
tion, and interblobs extract the high frequency pattern information, using it to construct
K system: input to V1 and superior colliculus

M & P input and intrinsic connections of area V1
M system: extrageniculate relays via pulvinar
striate cortex (V1) extrastriate cortex

1 1 1
2 2 2
3 3
4A 4A 3
4B V2 4B
4C a 4C a 4
4C b 4C b
V5
5
. V3
5 5
6
. V2 6 6
LGN - K . . .
PARVO .. . . . . PULVINAR
. ..
PARVO MAGNO
RELAYS RELAYS
MAGNO K M
K
LGN
SUPERIOR
MIXED PARVO &
COLLICULUS
MAGNO RELAYS
M&K
RETINA
Figure 17.1. Left: magnocellular and parvocellular inputs to V1; intrinsic relays from layer 4; outputs to
prestriate cortex. Right: koniocellular inputs to V1; extrageniculate relays via superior colliculus and pulvinar.

orientation selective cells (Livingstone and Hubel, 1984a; Zeki and Shipp, 1988). Given
that orientation tuning is also found in 4B (and in 4Ca) M input can clearly support the latter
process (Blasdel and Fitzpatrick, 1984). Similarly, the contrast sensitivity of the M system
may be exploited by blobs in synthesising a 3D colour space (typically represented by ortho-
gonal luminance (black/white), red/green, and blue/yellow axes. It is known that there is an
preponderance of red/green units amongst P cells. Input from shortwave (blue) cones is
rare in this system, but relatively more prominent amongst K cells (Martin et al., 1997).
To complete the picture, the K system also provides a specific input to blobs (Casagrande,
1994); indeed, perhaps it is K input that constructs blobs, as all direct geniculate targets in
V1 are zones of elevated metabolism.
3. THE NATURE OF SPECIALISATION
The simple message of the above is that P, M and K input is mixed, as appropriate, into
three distinct internal compartments in V1—blobs, interblobs and layer 4B—with special-
isations related to colour, form and motion respectively. It is, however, potentially mislead-
ing to employ perceptual labels for such an early stage of analysis. At the level of the LGN,
for instance, one could state that the M system is suitable for motion analysis, but not that
it is specialised for motion. Although it has now developed direction and orientation tuning,
the same could still be said of the M system at the level of layer 4B. It is only the component
of motion orthogonal to the contour that is signalled, and there is no sensitivity to “pattern
motion”, as found at higher levels (Movshon et al., 1985; Rodman and Albright, 1989).
Furthermore, the activity of non-directional cells in 4B, signalling the orientation of moving
contours, could be considered a specialisation for “dynamic form” (Zeki and Shipp, 1988).
Finally, it is reported that layer 4B is rich in neurones that are tuned for stereo disparity
(Hubel and Livingstone, 1990). This portfolio of properties in 4B is probably dictated by
factors to do with economy of visual processing, and it is not satisfactorily encompassed
by any simple perceptual category. With this thought in mind, it is also worth examining the
specialisations of the other compartments more closely.
The orientation selective cells of the interblobs are better suited for analysis of static
patterns, at higher resolution, so this looks like the inception of texture and shape processing.
Broadband spectral tuning is explicable by convergence of P relays of different wave-
length sensitivity. This would also enable extraction of the maximum spatial information
from P signals. Such convergence may not be universal, as some interblob cells display
both orientation and colour tuning (Livingstone and Hubel, 1984a). Nevertheless, even
when there is no overt sign of colour tuning, some orientation tuned cells do still respond to
oriented contours defined by isoluminant colour borders, i.e. the response is not dependent
on the sign of colour-contrast (Gouras and Krüger, 1979; Thorell et al., 1984; Hubel and
Livingstone, 1990). Both the former “colour-signed” and latter “colour-unsigned” signals
could thus serve “form-from-colour”, corresponding to the fact that recognisable forms still
exist in uniformly isoluminant pictures.
There is psychophysical and single-unit evidence for both signed and unsigned chromatic
contributions to motion perception (Dobkins and Albright, 1993, 1994; Gegenfurtner and
Hawken, 1996). Does this imply that motion is not served only by the M system? Residual
responses of the M system to moving isoluminant stimuli could in theory account for the
unsigned colour contribution, but there is actually direct evidence from the LGN lesion

392 Stewart Shipp
technique for a P contribution to motion: the same studies that reveal a motion deficit after
M lesions also demonstrate a residual capacity to discriminate direction, if only at a sub-
stantially higher luminance contrast threshold (Merigan et al., 1991a). This, in turn could
mean two things: either that the motion pathway does recruit some P input, or that what is
labelled the motion pathway is not the exclusive domain of all motion-related processing.
Such a question cannot be satisfactorily resolved without looking beyond V1. It is worth
pointing out, however, that although signals from the P system are not directly relayed to
layer 4B, they can be sampled by pyramidal output neurones in this layer whose apical
dendrites ascend into layers 4A and 3 (Sawatari and Callaway, 1996).
Is there less ambiguity concerning the role of blobs in colour vision? It is a common
observation that blobs are present in the nocturnal owl monkey, an animal with a single
type of cone and rudimentary colour vision (Jacobs et al., 1993). However, even if owl
monkeys see only shades of grey, this lightness perception is akin to colour vision, and is
a function likely to be served by blobs (Allman and Zucker, 1990). The chromatic analogue
of relative lightness for shades of grey is colour constancy. Notably, the responses of V1
colour cells correlate with the intensity and spectral quality of their local stimulation, but not
with the colour percepts of a human observer, which arise from the global pattern of
stimulation (Zeki, 1983); thus the mechanisms achieving constancy are partly, at least, the
properties of higher areas (e.g. V4).
There is analogous evidence relating to the role of V1 in form vision; namely, the relative
activity of V1 orientation cells does not correlate with the momentary percept of orientation,
in an experimental set-up where orthogonal gratings, presented dichoptically, produce a state
of retinal rivalry (Leopold and Logothetis, 1996). In general terms, the corpus of the
“neural correlate of consciousness” is certainly larger than V1, and it may even exclude V1
(Crick and Koch, 1995). Thus, the three compartments at the level of V1 represent inter-
mediate stages in the synthesis of percepts, a synthesis that depends on a degree of mixing
of M, P and K inputs. The next question is whether the V1 output channels serve as a more
stable basis for parallel (“two visual pathway”) models of the visual brain, or whether
they, in turn, are desegregated and reconstituted/recombined.
4. METABOLIC PATHWAYS
It is perhaps fairly well known—certainly well reviewed (DeYoe and Van Essen, 1988;
Livingstone and Hubel, 1988; Zeki and Shipp, 1988)—that the three compartments of V1
lead on to functionally comparable metabolic subdivisions in area V2 (thick stripes, thin
stripes and paler interstripes), and then to areas V4 and V5, the latter originally portrayed
as the roots of the ventral and dorsal visual pathways (Mishkin et al., 1983). Two, three or
whatever number, a parallel pathway model implies separable entities that are not strongly
interlinked. The presence of cross-talk is always acknowledged, but how much cross-talk
can there be without a network model becoming preferable to a parallel one? There is no
certain answer to this question, despite the wealth of relevant anatomical and physiolo-
gical data.
The model of Figure 17.2 is cast in the traditional framework, with accentuated detail
at the early levels of the pathway to complement this part of the article. The ventral
pathway is the route from blobs and interblobs to thin stripes and interstripes, and then on to
inferior temporal cortex, via area V4. The dorsal pathway is traditionally the route from

Figure 17.2. Origins and higher stages of the interblob-interstripe (Ib-Is), blob-thin stripe (B-N*) and layer
4B-thick stripe (4B-K*) pathways (small, medium and large arrows respectively at the output from V1, input to
V2, and inputs to V3, V4 and V5). Input to V3 from thin stripes and interstripes is less certain than input from thick
stripes (arrows in parentheses). V4 is shown with two types of internal module; input from interstripes to one of
these is uncertain. The diagram is configured to suggest a relationship between the 4B-Ks and dorsal pathways,
and between the B-Ns/Ib-Is and ventral pathways, but there is also substantial cross-talk: internally within V1 and
V2 (pale arrows); in the distributed outputs of V3; and in cross-connectivity of the parietal and temporal lobes.
* For these abbreviations, see note 6, p. 401.

394 Stewart Shipp
layer 4B to V5, via the thick stripes of V2 as well as directly, and from V5 on to areas of
parietal cortex. However the status of V3, which feeds both these routes, is ambivalent,
and there are other forms of cross-connection, at higher stages. To put it simply, the ana-
tomical basis of a parallel, two pathway model is not its greatest strength. A parallel model
is better served by the title “two visual streams”, if “stream” is understood to index both the
properties of an area as well as its connectivity. The notion requires that the statistical
weight of the internal links within a pathway exceeds the weight of the external links with
the other pathway (Young, 1992). Each stream may therefore preserve a functional homo-
geneity despite receiving alien inputs. As we proceed, this idea will be examined in more
detail, at successive levels in the visual hierarchy.
4.1. Diversity of Function Across the Metabolic Compartments

4.1.1. Segregated pathways?
The projections from blobs to thin stripes, interblobs to interstripes and from layer 4B to
thick stripes appear to be direct and non-divergent (e.g. no cross-connections from blobs
to thick stripes—(Livingstone and Hubel, 1983, 1987a; Levitt et al., 1994b; Malach et al.,
1994). Any cross-connectivity at this level must therefore be intrinsic to V1 and/or V2.
Lateral connections in V1 are 80% specific, in that they link blob to blob, and interblob to
interblob (Livingstone and Hubel, 1984b; Yoshioka et al., 1996). Also, blobs are not
watertight compartments in that dendrites of some blob cells do extend into the interblob
matrix, and vice versa (Hübener and Bolz, 1992). Finally there are inputs from layer 4B to
the blobs. These form a vertical column of ascending input; the blobs do not provide reci-
procal axonal output back to layer 4B (Callaway, 1998). Within V2, by contrast, reciprocal
cross-talk appears to be the rule. Intrinsic connections between V2 stripes were initially
described as being specific (i.e. thin stripe to thin stripe—(Livingstone and Hubel, 1984a)
but subsequent studies have found that non-specific links are also common: each type of
stripe communicates with each other type of stripe (Levitt et al., 1994a; Malach et al.,
1994).
What function do these cross-connections serve? The exact nature of the specialisa-
tion across stripes in V2 bears discussion in detail, as the drift of several recent reports
has been to downgrade segregation, and to emphasise a similarity in functional charac-
teristics across stripes (Levitt et al., 1994a,b; Gegenfurtner et al., 1996; Tamura et al.,
1996; Kiper et al., 1997). One may thus ask whether (a) the stripes are less diverse than
the compartments in V1, this trend toward homogeneity being brought about by intrinsic
connectivity in V2? Or (b) the diversity of function within V1 has also been exagger-
ated? (Leventhal et al., 1995). Or, (c) the case against diversity and segregation has
been overstated?
4.1.2. Segregated functions—form and colour?

Overall, the body of single unit work on area V2 is ambivalent on the properties of each
compartment (see Table 17.1); there is no shortage of potential explanations for these
inconsistencies, although it is hard to know which applies where. Variations in stimuli,
criteria for response selectivity, anaesthetic level, electrode characteristics/neural
sampling, track reconstruction and criteria for stripe identification are all potential

Table 17.1.
Thick Thin Inter report n
% orientation selective 1. 72 1. 21 1. 67 1. 667

2. 51 2. 20 2. 17 2. 86
3. 87 3. 64 3. 82 3. 390
4. 92 4. 48 4. 78 4. 83
5. 86 5. 38 5. 84 5. 426
6. 85 6. 73 6. 96 6. 100
7. 21* 7. 21* 7. 38 7. 55
MEDIAN 85% 38% 78%
% direction selective 1. 16 1. 1 1. 0 1. 667
2. 19 2. 7 2. 0 2. 62
3. 30 3. 21 3. 34 3. 190
4. 50 4. 4 4. 13 4. 83
6. 28 6. 9 6. 25 6. 100
7. 18* 7. 18* 7. 31 7. 55
MEDIAN 24% 8% 19%
% colour selective 1. 24 1. 55 1. 24 1. 660
2. 16 2. 86 2. 64 2. 81
4. 10 4. 27 4. 27 4. 111
5. 16 5. 75 5. 12 5. 426
6. 39 6. 65 6. 33 6. 72
7. 28* 7. 28* 7. 63 7. 55
MEDIAN 20% 60% 30%
% disparity selective 2. 68 2. 33 2. 22 2. 91
3. 38 3. 21 3. 15 3. 390
[2. = binocular interaction] 5. 77 5. 10 5. 1 5. 26
MEDIAN 68% 21% 15%
% incidence end-stopping 1. 30 1. 49 1. 22 1. 601
[1. includes length antagonism 3. 20 3. 22 3. 30 3. 390
within RF] 4. 17 4. 30 4. 18 4. 111
5. 42 5. 16 5. 37 5. 426
6. 13 6. 19 6. 41 6. 76
MEDIAN 20% 22% 30%
Notes:
Summary of the incidence of visual selectivity across stripes, as reported in seven studies.
1. Shipp (2002) Visual Neuroscience (in press).
2. DeYoe and Van Essen (1985); 3. Peterhans and von der Heydt (1993); 4. Levitt et al. (1994); 5. Roe
and Ts’o (1995); 6. Gegenfurtner et al. (1996); 7. Tamura et al. (1996).
*—study 7 did not distinguish between thick and thin dark stripes.
factors—plus the fact that stripe borders cannot be localised to better than ±100 µm owing
to the diffuse staining characteristics of cytochrome oxidase histology. Thus, with rela-
tively small populations of neurones in some studies, plus the acknowledged clustering of
similar neurones within some form of stripe sub-structure, there is ample scope for the
observed variability in the reported incidence of response selectivities across stripes. In the
original formulation, thin stripes were characterised by colour selectivity, interstripes by orien-
tation selectivity accompanied by end-stopping, and thick stripes by orientation select-
ivity accompanied by disparity selectivity (Hubel and Livingstone, 1987). Groups of these
cells were found in the appropriate sequence as the electrode traversed one stripe after

396 Stewart Shipp
another, but the authors (perhaps wisely) seem to have avoided defining stripe boundaries
precisely, and refrained from tabulating the exact proportions of each cell class to be found
in each stripe. Most other authors have done the decent thing, and provided exact percent-
age incidences; their work is summarised in Table 17.1.
Going by the median % figures, colour and lack of orientation selectivity emerge from
Table 17.1 as consensus characteristics for thin stripes, but medians are hardly a very
satisfactory resolution of the discrepancies apparent across studies. Fortunately, there is
much extra data to hand, in the form of functional imaging studies (2-deoxyglucose labelling
and optical imaging). These techniques generate maps of functional architecture that may
be compared to several cycles of the stripe pattern, and their conclusions have generally
proven to be more consistent with each other, and concordant across V1 and V2, than the
results of electrophysiological studies. For instance, imaging techniques show that the thick
and thin stripes of V2 are responsive to a lower range of spatial frequencies than the inter-
stripes, and the blobs and interblobs of V1 can be similarly distinguished (Tootell et al., 1988c;
Tootell and Hamilton, 1989). Single unit findings agree that in V1, blobs respond to a lower
spatial frequency than interblobs (Tootell and Born, 1991; Edwards et al., 1995) and, in
V2, that low-pass units occur only within thin stripes (Levitt et al., 1994a)—yet, in this
very study, the optimal spatial frequency across V2 compartments was assessed to be stat-
istically indistinguishable (Levitt et al., 1994a).
How about selectivity for colour? Using low frequency isoluminant chromatic stimuli,
in place of luminance contrast, only V1 blobs and V2 thin stripes appear in the imaging
data (Tootell et al., 1988b; Tootell and Hamilton, 1989). These low frequencies include
diffuse fields of zero spatial structure, so it is fair to conclude that the colour specific
response is related to hue discrimination, rather than any kind of chromatic form vision
(call it “colour-from-colour”). This result holds up to frequencies of 2 cycles/deg in central
visual field. The introduction of higher spatial frequencies into isoluminant stimuli
activates all three stripes in V2, and both blobs and interblobs in V1 (the relative intensity
across compartments depending on the interplay of the frequency content, and eccentricity
of the stimulus). Thus, the whole of V2 can be responsive to isoluminant chromatic stimuli.
However optical imaging reveals that the thin stripes are the only location where the
magnitude of the isoluminant response can actually exceed that of a spatially equivalent
luminance stimulus (Roe and Ts’o, 1997).
The relative response of single neurones to luminance contrast, and to isoluminant
chromatic contrast (call it “chrominance”) is well known to depend on spatial frequency.
For red-green P units of the LGN, as mentioned above, chrominance alone is effective at
zero spatial frequency, but the cell’s response is increasingly dominated by luminance at
higher spatial frequencies (Derrington et al., 1984). Cortical units in V1 develop far tighter
tuning for spatial frequency. Hence colour selective cells (i.e. those in which the chrominance
responses are most dominant) are also tuned to the lowest ranges of spatial frequency
(Lennie et al., 1990; Leventhal et al., 1995). Perversely, the latter two studies are just
those which have questioned whether such colour selective cells are present in significant
numbers within layers 2 and 3 of V1, or are restricted to blobs in their distribution. Both
used sinsusoidal luminance and chrominance gratings as stimuli; the negative findings with
regard to blobs conflict with the imaging work using similar stimuli, and with the previous
single-unit work, using traditional stimulation by flashed coloured spots (Livingstone and
Hubel, 1984a; Ts’o and Gilbert, 1988). The latter was based on substantially larger samples
of cells, and demonstrates a high incidence of colour coding within V1 blobs using

tangential sections with good histological definition of the darker-staining cytochrome

oxidase elements.
A third variable associated with blobs and thin stripes is lack of orientation selectivity,
yet this has been questioned, both in V2 (Levitt et al., 1994a; Gegenfurtner et al., 1996),
and V1 (Leventhal et al., 1995): the latter study indicates that sinusoidal gratings are
a very sensitive probe for resolving weak orientation preferences and, uniquely, reports
orientation selectivity within layer 4Cb. Yet again, functional imaging data supports the
conventional picture, since, in V1, blobs are known to be situated upon fractures in the
orientation map of V1, i.e. regions across which there is both a rapid change in preferred
orientation and the broadest tuning for orientation (Bartfeld and Grinvald, 1992; Blasdel,
1992). In V2, zones of minimal selectivity to orientation appear to coincide directly with
thin stripes (Malach et al., 1994; Roe and Ts’o, 1997). As noted above, the minimal
orientation tuning in blobs and thin stripes hints that the contribution of these structures to
colour vision lies in seeing attributes such as hue, saturation and brightness. Outside blobs
and thin stripes, chromatic selectivity is more likely to be accompanied by tuning for
orientation and higher spatial frequencies, and hence to be subserving ‘form-from-colour’.
In a similar vein, only simple oriented cells can provide a signed-colour signal (one that
modulates in phase with the periodicity of a chromatic grating); complex cells produce an
unsigned signal, for although they may be selective for modulation along a particular axis
in colour space (e.g. red-green), the firing rate is independent of the phase of the grating
(Thorell et al., 1984; Lennie et al., 1990). In V2 roughly 75% of cells may be complex
(Levitt et al., 1994a); if responsive to chrominance gratings, they may show spectral
tuning narrower than that of V1 cells and no bias toward thin stripes (Kiper et al., 1997)—
yet an unsigned-colour signal cannot be subserving colour-from-colour. Thus, as a more
general point, the discovery of certain selectivities across all compartments in V2 does not
necessarily undermine the case for functional segregation, although it might compromise
the simple labels commonly attached to the segregated functions.
4.1.3. Direction and disparity

Direction and disparity selectivity are characteristic properties of the network with which
thick stripes are associated, namely areas V5, V3 and layer 4B of V1, but the evidence for
V2 (from Table 17.1) is equivocal, at least for direction; unfortunately, there is no
available evidence from functional imaging to break this deadlock. It can be noted that
responses contingent upon coherent motion of stimuli are most common in thick stripes,
a specialisation related to form-from-motion (dynamic form) (Peterhans and von der
Heydt, 1993). There is some tension between the view that the thick stripes are the most
prominent compartment for disparity selectivity, and the conclusion from LGN lesion
work that fine stereovision depends more on the P system than the M system (Schiller
et al., 1990). One explanation is that P signals may reach thick stripes via dendrites of
some layer 4B cells that stick into 4A and 3, although this may be more pertinent for the
minority of colour sensitive cells in the thick stripes (median incidence = 16%), than for
the more prevalent property of disparity selectivity (median incidence = 68%). Could there
be hidden disparity tuning in interstripes? Note that the typical test for neural disparity
tuning equates to psychophysical procedures for measuring absolute disparity sensitivity
in human observers (i.e. judging the distance of a single isolated object); however it is
a well documented result that sensitivity to relative disparity (the difference in distance of

398 Stewart Shipp
two objects) is far higher than the absolute disparity sensitivity (Westheimer, 1979). If thick
stripes’ stereo properties do derive from the M system, it is a fair proposition that the P
system may feed a stereo capacity on the part of interstripes that is much more proficient
than so far demonstrated and which (unlike V1 [Cumming and Parker, 1999]) is sensitive
to relative disparity.
The distinct feature of interstripes was proposed to be end-stopping (Hubel and Living-
stone, 1987), but this characteristic may be less singular than first suggested (Table 17.1).
Otherwise, interstripes resemble thick stripes in orientation selectivity, and sensitivity to
contrast. (Only the direction-selective cells of thick stripes show the high contrast sensitiv-
ity that is typical of the M system—Levitt et al., 1994a.) Nevertheless, despite the similar
appearance of the orientation signals emanating from thick stripes and interstripes, some
element of distinction is implied by the fact that they participate in different networks of
higher areas. More on this later.
4.1.4. Single neurones with multiple selectivities

Another observation meriting consideration is the combination of selectivities for different
attributes in single neurones. If, for instance, colour and direction selectivity were found to
be positively associated, each could not be characteristic of a separate compartment. Or, if
colour and direction selectivity indices are independently assorted across neurones, as
reported by Gegenfurtner et al. (1996) it is necessary to consider their stripe distribution
more closely. A segregation model would require, minimally, that the majority of the
neurones with the highest colour or direction indices would not be found in the “wrong”
compartment. This was not examined, but it was commented that the highest indices of
selectivity for colour or motion were mutually exclusive (Gegenfurtner et al., 1996). Data
from the author’s laboratory (Table 17.1 and Figure 17.3) support the idea that the neur-
ones most selective for colour or motion are tightly restricted to thin stripes and thick
stripes respectively5. Another consideration is that a joint colour- and direction-tuned cell,
if found in a thick stripe, could be serving motion-from-colour (the converse notion, for
the same cell in a thin stripe to be serving “colour-from-motion”, is clearly nonsensical).
Another study finds that neurones with joint colour and direction selectivities are sub-
stantially more frequent in V2 than V1 (Tamura et al., 1996). The underlying idea here is
that these cells may be created within V2, by intrinsic connections. This is an important
notion, and it is instructive to pursue its ramifications. Suppose that the association neur-
one receives convergent colour and direction selective signals, and consider the outcome
if the integration were to follow either logical AND or logical OR rules. In the former case
the spatial properties of both inputs must be concordant (e.g. both are driven by spots) and
a successful stimulus must combine both the preferred colour and the preferred direction
of motion, or else the association neurone will not be activated. In this respect it would
resemble a direction selective unit that had been directly constructed from signed-colour
inputs. By contrast, OR integration would produce a very different “animal”. A successful
stimulus need possess just one of the requisite features, and the form selectivities need not
be concordant. This may seem a bizarre conception, but it may be closer to the experimental
5
The overall proportion of units classified as direction-selective or biased is far smaller than in other studies,
implying a stricter criterion; hence these must be the units with the most extreme selectivities, and they were
virtually all within thick stripes.

Fundamentals of Association Cortex
Figure 17.3. Distribution of orientation and wavelength selectivity across layers and stripes of area V2. Top left: orientation across stripes (K: ThicK stripes; N: ThiN stripes;
I: Interstripes; K:I and N:I represent border zones at thick/inter and thin/inter junctions respectively). Top right: wavelength selectivity across stripes. Bottom left: wavelength
selectivity across layers of thick stripes (“3.0” indicates mid layer 3; “3.5” indicates layer 3/4 border zone, etc). Bottom right: wavelength selectivity across layers of thin stripes.
399
Wavelength-related responses classified as selective or biased for spectral stimuli, “dark” indicates preference for dark stimulus on light background.

400 Stewart Shipp
picture: the paper illustrates a superficial V2 neurone driven by either (a) rightward motion
of an achromatic slit, or (b) flashed presentation of a relatively large, blue spot; the effect
of amalgamating these features in a single stimulus is not mentioned (Tamura et al.,
1996). It is perplexing to think how the ambiguity inherent in the signal from this cell
might be interpreted by higher visual centres.
Carrying the interpretation further is bedevilled by a lack of evidence, but it is worth
sketching an outline hypothesis. Suppose that the OR association neurone does not have
an output to other cortical areas, and that its connections are all local intrinsic ones. Note
that output cells, projecting from the thick stripes to higher areas such as V5, reside mainly
in layer 3B (Shipp and Zeki, 1989b). Recordings from patches of such cells (efferent to V5)
show that they are heavily direction selective, i.e. they conform to the thick stripes’ particular
character (DeYoe and Van Essen, 1985). By contrast, the author’s data (Figure 17.3)
shows that the majority of colour cells in thick stripes are found more superficially, in layers
2 or 3A. So perhaps it is cells in these layers, i.e. non-output cells, that show the less
characteristic (multiple) response selectivities. What could their function be? It must be
some form of multimodal integration, or binding (encoding the fact that attributes processed
along separate channels belong to the same object). Evidently, building neurones with
multiple selectivities is one way of doing this—a simple tactic, but one with a recognised
fundamental flaw, often dignified as the “combinatorial explosion”: there are more potential
combinations of features than neurones available to represent them. A more recent idea is
that binding is achieved by synchronisation of the neurones activated by a given object
(Engel et al., 1997). The hypothetical synchronisation of colour and direction cells has yet
to be examined experimentally. Nonetheless, the author’s speculation is that superficial,
multimodal association neurones act as part of the local machinery in V2 that induces sets
of functionally dissimilar (unimodal) projection neurones to adopt correlated rates of firing:
putatively, the projection neurones in a stripe embody its functional characteristics, and
the role of local neurones is to represent not features themselves, but their contingencies.
4.1.5. Interim summary

Of the three choices listed above (4.1.1.), regarding the status of functional segregation in
V2, the viewpoint on offer here is option (c): that functional architecture of V2 is most
effectively summarised as a set of three segregated pathways linking V1 to distant prestriate
areas. Indications to the contrary, such as evidence for the “wrong” kind of selectivity in
each stripe, could be counted (1) as quirks of histological analysis (or any other factor
responsible for the diversity apparent in Table 17.1); or (2) an alternative means of repre-
senting the variable of interest in a particular stripe; or even (3) as part of mechanism for
achieving binding-by-synchrony. The possibility of (1–3) means that discordant physiolo-
gical evidence cannot yet be held to overturn a segregationist model of V2 that is strongly
supported by anatomical and functional imaging data. A potential development of the
model is the idea that segregation is more robust for output cells, and less prominent for
local neurones. This builds on the earlier notion that, although V2 is a vehicle for the
continuation of three compartments established by V1, it does also play an important role
in facilitating cross-talk between these channels (Shipp and Zeki, 1989b; Roe and Ts’o,
1995). Such cross-talk is not pictured as weakening the identity of the three segregated
pathways, but as co-ordinating their mutual activity at a stage just before they diverge into
the wildnerness of prestriate cortex.

5. PARALLEL PATHWAYS FROM V2
If the metabolic compartments maintain their identity in the passage through V2, what
happens next? Further diversification and segregation? Reconvergence and desegrega-
tion? Or, somewhere between these two, maintenance of a three-fold parallel process?
A generalised answer to this question is complicated by two facts: that V2 projects to at least
ten additional cortical areas, nominally up to four (or more) rungs above V2 in the cortical
hierarchy (Felleman and Van Essen, 1991; Gattass et al., 1997); and that the source of
most of these outputs has yet to be specified at the level of stripes. So, whatever the rule, it
is virtually guaranteed that future exceptions will be found (and the connections of V3 seem
likely to afford an exception on the basis of presently available evidence). However, the
best documented pathways are those leading to areas V4 and V5, and their best description
does appear to be “parallel”: thick stripes lead to V5 (MT), thin stripes and interstripes
lead to V4 (DeYoe and Van Essen, 1985; Shipp and Zeki, 1985). Cross-talk, e.g. output
from thick stripes to V4, is not unknown, but it is not prevalent (Zeki and Shipp, 1989;
Nakamura et al., 1993). Furthermore, the ascending outputs from V4 and V5 do not
themselves show a great deal of convergence, even though some of them reach into similar
zones of cortex. On this basis it is possible to carve out groups of higher areas served by
thick stripes, or by thin stripes/interstripes (see Figure 17.4), that partially equate to the
dorsal and ventral streams. The aim of this part is to describe the outer limits of par-
allelism in the visual cortex; ultimately there is reconvergence, and this forms the subject
of the final part.
5.1. The Output of Thin Stripes and Interstripes to V4, and Beyond
Having created the separate blob-thin stripe (B-Ns)6 and interblob-interstripe (Ib-Is) pathways,
it is something of a mystery why the visual system requires them to stay in close proximity
as they relay through V4, to inferotemporal cortex. How high up this chain do they retain
their separate identities? It is difficult to resolve this question because metabolic compart-
ments are not visible in V4, and no other marker has yet been found to directly visualise
the modular substructure of V4. Yet V4-modules are believed to exist, because localised
deposits of retrograde tracer, made blindly into V4, reveal two distinct patterns of output
from V2 (Shipp and Zeki, 1985; Zeki and Shipp, 1989; Munk et al., 1995). The source
cells may be found only within interstripes; or, they are within thin stripes bracketed by
interstripes. The former suggests a type of module in V4 that is the exclusive target of
interstripes. The latter suggests one of two things: (a) a second type of module that
receives input from both thin stripes and interstripes; or (b) a module that receives input
only from thin stripes, but has smaller dimensions than the Ib-Is recipient module, such
that tracers injected into it tend to overflow into the adjoining Ib-Is recipient territory. The
second option, (b)—implying continued segregation in V4 of the B-Ns and Ib-Is pathways—
is the one that has been favoured by subsequent work (Felleman et al., 1997), and especially
by the most recent study, directly visualising separate axonal fields in V4 arising from
a pair of adjacent inter- and thin stripes (Xiao et al., 1999).
6
The abbreviation ‘Ns’ for thiN stripe uses its last letter, to avoid ambiguity with thicK stripes. The latter are
abbreviated ‘Ks’, to avoid confusion with ‘K’ for koniocellular.

402 Stewart Shipp
Figure 17.4. Ascending pathways from V2 divided into two domains: left (dark ground) pathways derived
from thick stripes; right (light ground) from interstripes/thin stripes. Dashed paths represent cross-talk between
domains. Not all known pathways, or areas are illustrated. Pairs or triplets of areas grouped in boxes are mutually
connected (arrows not shown). Arrow-heads terminating (or arrow-stems commencing) on outline of box represent
a connection to all the boxed areas; arrow-heads (or -stems) penetrating box are specific to the indicated area:
this system is used, for instance, to indicate specific input to V3 from thick stripes, but unspecific/undetermined
input to V3/V3A/PIP from V2 generally. Indicated at right are ranks in the cortical hierarchy—NB that, at level
6+, relative elevation within the diagram ceases to be an indication of relative rank.
The same anatomical experiments that reveal the sources, within V2, of ascending
inputs to V4 also show the sources of descending inputs to V4 (and, assuming the connec-
tions to be reciprocal, the likely targets of ascending projections from V4) (DeYoe et al.,
1994; Shipp and Zeki, 1995; Felleman et al., 1997). These are multiple sites in IT cortex
(e.g. areas TEO and TE) and also in STS (e.g. V4A, PITd and PITv). This means that a V4
module can be identified by its input from V2, and further characterised by the distribution
of its output. So far, the putative modules in V4 do not differ appreciably in the general
distribution of their outputs, as they all seem to connect with virtually the same set of
areas, in IT cortex and elsewhere. There is, however, some distinctiveness at a local level,
such that ascending connections of different modules tend to be segregated from each
other (note that, experimentally, this requires the use of dual tracers at nearby sites in V4)
(DeYoe et al., 1994; Felleman et al., 1997). In other words, it is as if the Ib-Is and B-Ns
pathways retain some elements of a separate identity not only through V4 but in higher

areas also. Yet the potential for mixing is by no means eliminated. As noted above, it will
be necessary to delineate the suspected modules by some means other than their connec-
tivity in order to trace the evolution of the Ib-Is and B-Ns pathways with greater certainty.
5.2. Evidence from Human Neuropathology

There are two contrasting clinical conditions—arising from occipital lobe lesions in the
region of the fusiform gyrus, or from carbon monoxide (CO) poisoning—whose associated
syndromes involve colour vision: There is either selective loss (achromatopsia) in the first
case, or selective preservation (chromatopsia) in the second (Zeki, 1993). If the B-Ns
pathway is the principal agency for extracting attributes of hue, brightness and saturation
from P signals (dubbed “colour from colour”), then the clinical observations would seem
to speak to the issue of the separability of the B-Ns pathway. Do these two syndromes
provide evidence that the B-Ns pathway does achieve a fuller form of isolation, at least in
human visual cortex?
Let us first consider the more general problem of assimilating primate data across
species. Monkey and human psychophysical data are highly similar, and yet the human
brain is not simply a scaled up version of a monkey brain with a 1:1 homology between
all identifiable areas and pathways. Somewhere in the system, there may be psychophys-
ical correlates of structural dissimilarities. With regard to the pathways under con-
sideration, human V1 and V2 are known to possess blobs and some form of stripes,
respectively (Wong-Riley et al., 1993), but even at this early stage there is at least one
difference in the metabolic signature of these areas, as human V1 lacks layer 4A (Horton
and Hedley-White, 1984). It is awkward to go further, for the details of human connectivity
are unknown. So, the tactic in monkey-human comparisons must be to assess the poten-
tial insights that might be gained on the assumption of structural identity across species
brains, with a common pool of functional data. Let us suppose, therefore, that monkeys
can talk and describe human-like perceptual consequences of focal, and diffuse, brain
damage.
Take the latter first. CO pathophysiology is not totally understood, but, if residual
colour vision is present in several cases (Milner et al., 1991; Sparr et al., 1991), this seems
likely to hinge on the coincidence of the B-Ns pathway with metabolically active regions.
In V1, layer 4Cb and the blobs of the upper layers are known to be more vascular than the
surrounding tissue (Zheng et al., 1991). In normal conditions this sustains an elevated rate
of firing and, in the presence of CO, provides a slightly greater cushion against the effects
of hypoxia. Since higher areas do not differ substantially in their overall metabolism, the
selective preservation of colour might result from differential survival of the B-Ns
pathway at an early level (i.e. V1 and perhaps V2); there need not be a strong implication
for selective sparing of a higher area that is the preserve of the B-Ns pathway and devoted
to “colour-from-colour”. However, the support for this notion from the evidence of cerebral
achromatopsia is stronger, especially from those cases where relatively small lesions give
rise to a relatively pure colour deficit (Damasio et al., 1980; Kolmel, 1988). These patients
see only black and white (and shades of grey), but otherwise their vision is described as close
to normal. The inference from these cases of pure achromatopsia for a “colour-from-colour”
area that is selectively damaged, is hard to resist—especially when specific tests reveal
preserved motion- and form-from-colour capacities (Barbur et al., 1994; Cavanagh et al.,
1998; Heywood et al., 1998).

404 Stewart Shipp
But could such an area still receive signals relating to contours (i.e. surface boundaries)
from colour-signed (or even colour-unsigned) components of the Ib-Is pathway? Perceptu-
ally, coloured surfaces are filled-in between visible borders (Krauskopf, 1963). It is as if
the colour contrast signal at the border is more important than the colour content sampled
from points in the surface’s interior. The Cornsweet illusion (in achromatic and chromatic
versions), is another demonstration of this principle (Ware and Cowan, 1983; Purves et al.,
1999). It is not only surface colour (including brightness) but also texture that can be filled
in (Ramachandran and Gregory, 1991). A powerful theoretical formulation describes
neural circuitry in terms of a “boundary contour system”, and a “feature contour system”,
one determining the boundaries of segmented objects, the other filling in their surface
properties (Grossberg, 1994). It is natural to equate these with the Ib-Is and B-Ns
pathways. The theoretical account posits repeated feedback interactions between the two
systems at successive stages in order to account for diverse perceptual phenomena. This is
a fairly accurate analogue for the anatomical status of the Ib-Is and B-Ns pathways.
In short, there is no firm evidence for large scale divergence of the Ib-Is and B-Ns
pathways. The two may remain tightly integrated over a succession of relays, whilst retain-
ing some form of separate identity. If so, perhaps they truly deserve the tag “parallel
pathways”: never far apart, they merge only at infinity!
5.3. Pathways from the Thick Stripes

The early stages of the motion pathway constitute a fully interconnected network of four
stations: layer 4B of V1, the thick stripes of V2, and total areas V3 and V5 (the “4B-Ks”
pathway). All four are characterised by directional selectivity, but V5, three rungs above
V1 in hierarchical status, has the highest incidence, verging on 100% (Zeki, 1974). There
is a wealth of additional evidence linking primate area V5 to motion processing and
perception, and satellite areas such as MST to optic flow (Orban, 1997). Area V5 is held to
be a pivotal area in the dorsal pathway—so that initially, the higher reaches of the pathway
were virtually defined to be those sites receiving ascending inputs, directly or indirectly,
from V5 (Mishkin et al., 1983; Van Essen and Maunsell, 1983; Ungerleider and Desi-
mone, 1986). These sites include areas VIP, LIP and 7a of the inferior parietal lobule, and
V6/PO and V6A in the superior lobule. The functions of these areas are related to visual
control of reaching, in near or far personal space (VIP and V6A—Duhamel et al., 1998;
Shipp et al., 1998) or saccadic eye movements (7a/LIP). More recently, a branch of V5
output leading to mid-temporal areas has also been recognised (Boussaoud et al., 1990).
Should it be considered a ventral limb of the dorsal pathway, or as part of the ventral
consortium, or (as the authors choose) a third pathway?
Whatever the verdict, this degree of distribution is ultimately the undoing of a totally
parallel model of visual connectivity. The parallelism of the two visual pathways model
originates with separate outputs from V2 stripes to V4 and V5, and relies on the ascending
outputs of V4 (also widely distributed) avoiding those from V5. But, crucially, outputs
from V4 and V5 very nearly come together in the middle temporal and inferior parietal
lobes. Their fields of connections are adjacent, with restricted instances of direct overlap—
a pattern that has been termed “juxtaconvergence” (Shipp and Zeki, 1995)—see Figure
17.5. The ultimate convergence of the two pathways is but one (or at most two) stages
away—see below. In this sense the overall set of connections is not parallel, because the
highways passing via V4 and V5 do eventually amalgamate. Furthermore, the element of

parallelism is unduly magnified by restricting attention to V4 and V5: other, earlier

instances of convergence can be seen in examining the connections of V3.
V3 certainly receives input from the thick stripes. Contributions from the thin stripes
and interstripes are also likely, but appear with greater variability in the author’s own
material. Like V5, V3 has many direction-selective cells, at 40% not as many as V5, but
also some with the advanced property of pattern motion (Felleman and Van Essen, 1987;
Gegenfurtner et al., 1997). The functional distinction between V3 and V5 is addressed by
the notion that V3 serves dynamic form, that is, the analysis of the orientation of moving
contours, perhaps deriving contours from motion contrast (“form-from-motion”). This
rationalises its output to V4 (as well as V5), the dynamic form information presumably
complementing static form information in the relays to IT cortex subserving object recog-
nition. V3 also projects to V3A, and the caudal intraparietal sulcus where neurones are
sensitive to the 3D orientation of objects, very likely as part of a pathway for controlling
the hand in grasping movements (Sakata et al., 1998).
Another of V3’s specialities could be the analysis of slowly moving coloured patterns.
There is psychophysical evidence that these are detected by the colour opponent (P)
system and that V5, although activated by isoluminant, colour-signed signals, is insufficiently
sensitive to be the neural basis (Gegenfurtner and Hawken, 1996). V3, by contrast, has
more abundant neurones with colour tuning and lower optimal speeds (Gegenfurtner
et al., 1997). Furthermore, if V3 is the site for such processing, it does not appear to be
handed on to V5; so there is almost evidence for an alternative motion pathway arising in
V3, although whether it proceeds via V3A, or V4 (or both) is totally unknown.
These latter examples are potential instances of convergence between the outputs from
V2 compartments, but the circuitry is insufficiently determined to be sure. For instance,
there is a hypothetical possibility that V3 (like V4) might possess hidden modules, and
that its output to V4 emerges from a module not receiving from the thick stripes. Details
of the output from V2 stripes to other prestriate areas (e.g. V3A/PIP, V6/PO) are equally
scarce. There is some likelihood that they will be dominated by the M system/thick stripes,
like other routes to parietal cortex, but perhaps not to the same degree as the V5 motion
pathway. This, as stated at the beginning, is the cleanest known example of a segregated,
specialised pathway, synthesising luminant motion signals from basic M inputs. It was also
one of the earliest to be traced—perhaps it has been over-influential. And finally, even the
motion pathway must find its end within association cortex.
6. CONVERGENCE AND INTEGRATION
Cross-connections between pathways exist at all levels, but following our earlier analysis
(Zeki and Shipp, 1988) can be divided into three categories: ascending, lateral and
descending. To these can be added a fourth, namely re-entrant subcortical loops. These
linkages serve a variety of functions of which few, if any are properly understood. Yet all
could be considered, in one way or another, as forms of integration.
6.1. Integration through Cortico-Thalamo-Cortical Circuitry?

This topic is considered in detail elsewhere in this volume, so the treatment here is brief.
A general principle that has emerged from anatomical studies is that if two cortical areas

406 Stewart Shipp
are linked directly (i.e. by cortico-cortical connections), then they are also likely to be linked
via the thalamus (i.e. via the pulvinar nucleus for interactions between visual association
areas). Connections between the cortex and thalamus are reciprocal, and retain elements
of both retinal and cortical topography. However the topography is far from precise—and,
evidently, there must be some topographic jitter to permit non-adjacent cortical areas to
inter-communicate via thalamic circuitry. A single site in the pulvinar thus tends to
connect with a rather broad expanse of cortex, spanning several separate areas (Sherman and
Guillery, 1996).
It is something of a surprise, therefore, to find that the pulvinar projections of V5 and
V4 are basically separate from each other (Shipp and Zeki, 1995); it is experimentally
difficult to exclude all possibility of overlap, but certainly the pulvinar circuitry does not
look as if it is engineered to promote intercommunication of these areas via the thalamus.
Even so, if the projection fields of V5 and V4 do not overlap in the physical sense, they
are both in register with the retinotopic map of the pulvinar (or to be specific, the ventral
pulvinar, containing what has traditionally been termed the inferior nucleus [Bender,
1981]). To understand how this can be, imagine that the pulvinar map is expressed in
a particular anatomical plane, and replicated many times along the axis orthogonal to that
plane (e.g. like cortical maps extending through the layers of the cortex). The retinotopic
order seen in the V5 and V4 projection fields is simply expressed within separate replicas
of the pulvinar map.
This organisation also gives rise to a simple functional hypothesis. Some inputs to the
ventral pulvinar, for instance from V1 or from the superior colliculus, traverse many or all
of these replica maps. Thus focal activity arising in V1 or the colliculus would be
transmitted, via the pulvinar maps, to retinotopically equivalent sites in the maps of areas
V5 and V4 (and several other prestriate areas with similarly organised connections). The
resultant activity in this set of prestriate areas would be spatially co-ordinated, pointing to
a role in exogenously driven spatial attention. In summary, the thalamic circuitry of V4
and V5 may not act as a form of communication between the two areas, but it may be
instrumental in co-ordinating activity within their respective spatial maps, perhaps facilit-
ating integration within other circuits.
6.2. Integration by Ascending Connections

6.2.1. Temporal cortex
Inferotemporal (IT) cortex is the traditional mainstay of the ventral pathway, and receives
widespread input from V4, but the status of mid-temporal visual cortex (i.e. all the
temporal lobe cortex buried within the superior temporal sulcus [STS]), is less clear-cut.
It could be portrayed as a second limb of the dorsal pathway. This is because V5 has been
found to project, directly and indirectly, to a series of sites along the STS. The direct
projections are to a pair of areas MST and FST (Boussaoud et al., 1990), and these in turn
project to an ill-defined conglomerate area, STP (superior temporal polysensory) with
components known as areas TPO, PGa and IPa, respectively occupying the anterior bank,
fundus and lower posterior bank of the STS (Seltzer and Pandya, 1978). V4 projections
target a succession of IT areas that dip into the lower STS bank—PITd, CITd and AITd
(Van Essen et al., 1990)—often subsumed within areas TEO and TE. If projections of
both V4 and V5 are examined concurrently, their respective fields within the STS are

seen, as expected, to be largely non-overlapping (Shipp and Zeki, 1995). However these
STS and IT areas are massively interconnected, as demonstrated by tracers deposited in
TEO or TE (Morel and Bullier, 1990; Baizer et al., 1991; Distler et al., 1993), or within
the STS (Seltzer and Pandya, 1978, 1994; Boussaoud et al., 1990). Why should this be so?
IT cortex is responsible for object recognition in general, and face recognition in particular.
Face sensitive cells are found in clusters throughout IT cortex, on the IT gyral surface and
within the STS (e.g. Tanaka et al., 1991; Perrett et al., 1992). Neighbouring clusters of
cells in the STS have very different properties, such as selectivity for static views of the
head (Perrett et al., 1991), or shape-indeterminate motion selectivity (Oram et al., 1993).
Nonetheless, as these authors point out, the tuning in either category is clustered around
the same body-centred axes. Intriguingly, there are also cells responsive to dynamic views
of particular bodily movements (e.g. front or rear view of a standing, sitting, or walking
action—Perrett et al., 1990). These types of cells are thought to subserve percepts of
biological motion, the characteristic gait and limb movements produced by primate loco-
motion, a function that can also be identified at comparable locations of human cortex
(Bonda et al., 1996). In other words, what is being identified is not an object so much as
a particular pattern of motion—but this is still a form of recognition. The juxtaposition of
biological motion and face sensitivity makes sense in terms of a general role for this part
of the brain in determining the intentions and behaviour of other individuals in a social
setting (Brothers and Ring, 1993). There are clearly conspecific gestures that require both
facial identity, expression and body movement to be faithfully interpreted—requiring the
ultimate convergence of static form and motion processing.
6.2.2. Inferior parietal cortex

Inferior parietal cortex is part of the dorsal pathway, yet it does receive a strong input from
V4, as well as V5. On closer examination, the two projections are found to be largely
separate. V5 projects to the lower, ventral part of the lateral bank of the intraparietal
sulcus, a region of relatively heavy myelination that probably includes areas LIP and VIP
(Lateral and Ventral Intra-Parietal areas) (Maunsell and Van Essen, 1983; Ungerleider and
Desimone, 1986). V4 projects mainly to the lightly myelinated dorsal part, LIPd (Blatt
et al., 1990; Morel and Bullier, 1990). When both projections are examined concurrently,
the V4 and V5 fields are seen to be contiguous, with small zones of overlap (Shipp and
Zeki, 1995); an additional complication is that both of these fields are discontinuous,
being segmented into parallel bands that run orthogonal to the axis of the sulcus (see
Figure 17.5). Analogous perhaps to the situation in temporal cortex, convergence between
the 4B-Ks and B-Ns/Ib-Is pathways is then achieved by the dense interconnectedness of
LIPv and LIPd7 (Andersen et al., 1990; Blatt et al., 1990; Morel and Bullier, 1990; Baizer
et al., 1991).
What does LIP do? The area was first identified as a parietal region heavily connected
with the frontal eye field (Andersen et al., 1985), and is best known for its presaccadic
7
A technical limitation is noteworthy: a large proportion of experiments with tracers placed in LIP and demon-
strating intrinsic connections, simultaneously reveal inputs to LIP from both V5 and V4—suggesting that the
tracer placements are neither confined to LIPv nor LIPd, but involve both. It is therefore the minority of experi-
ments that fail to label one of V5 or V4 that provide stronger evidence of interconnections between LIPd and
LIPv.

408 Stewart Shipp
Figure 17.5. Distribution of connections with areas V4 (red) and V5 (green) within parietal and temporal
lobes. Paired injections of separate tracers were made into retinotopically corresponding sites in areas V4 and
V5 within a single hemisphere. Fields of connectivity are shown on a flattened cortical surface, reconstructed
from serial horizontal sections at 300 µm intervals. Injection sites appear as solid and dashed lines (core and halo
respectively). Top: lateral bank of IPS; the limits of the sulcus are shown by + symbols; also shown is the distri-
bution of interhemispheric connections (“callosal degeneration”). Bottom: superior temporal sulcus, with sections
aligned at the point of maximal convexity in the fundus of the sulcus; red (V4) labelled regions outside the left
hand margin of the sulcus lie on the surfaces of the prelunate and inferior temporal gyri. (see Color Plate 6)

activity (Barash et al., 1991); subsequent studies have not unearthed any regional vari-
ation that might correspond to the anatomical subdivisions LIPd and LIPv. Visual
responses of LIP neurones are activated only by salient stimuli (Gottlieb et al., 1998)—
likely targets for a saccadic eye movement (or any other behavioural response)—and evi-
dently stimuli can be salient with respect to either of the sensory domains associated with
V4 or V5. However, the particular quality responsible for salience is not itself explicitly
represented in LIP. LIP seems to act as a working memory spatial map (for locations of
salient objects); the map is updated after each eye movement, and even starts to update
itself before the movement is initiated (Colby et al., 1995). However, since activity of LIP
neurones is not irrevocably followed by an eye movement, there are good arguments for
attributing LIP with a more general function, namely that it is an important centre for
control of the locus of spatial attention (Colby and Goldberg, 1999). This concurs with the
“premotor theory” of attention, which holds that the cortical system for directing attention
within space evolved from the mechanisms controlling eye movements, and may share
much of the same circuitry (Rizzolatti et al., 1987; Sheliga et al., 1995); and there is per-
suasive evidence from human imaging work, showing that fronto-parietal activations
induced by either ocular or attentional shifts are massively overlapping (Corbetta, 1998;
Corbetta et al., 1998).
From this viewpoint, LIP would be the source (or at least one important source) of an
attentional “spotlight”, whose effects are felt elsewhere in the visual system. There are
well documented neural effects related to spatial attention (both facilitatory and suppress-
ive) in areas such as V4 and IT cortex (Moran and Desimone, 1985; Motter, 1993; Connor
et al., 1997; Luck et al., 1997). These studies found only minor effects in V1, but there is
now rapidly accumulating evidence from human functional imaging for significant effects
of spatial attention in V1, as well as a broad spectrum of extrastriate areas (Brefczynski
and DeYoe, 1999; Gandhi et al., 1999; Martinez et al., 1999; Somers et al., 1999).
For present purposes, it is useful to enquire about the circuitry that could mediate these
effects, especially vis-a-vis the likelihood of convergence, either across the tripartite
channels stemming from V2, or across the dual dorsal and ventral processing streams.
First of all, there are direct reciprocal connections from LIPd back to V4, and these obviously
act to link up the dorsal and ventral streams, as do additional outputs from LIP to TEO and
TE. Importantly, the latter derive from LIPv as well as LIPd (Blatt et al., 1990; Morel and
Bullier, 1990; Baizer et al., 1991; Distler et al., 1993; Webster et al., 1994), so cross-talk
from the 4B-Ks pathway that feeds LIPv is also substantial. Descending connections from
LIP reach as far as V3 and V3A, but not to V2, nor V1. Thus, although a spatial attention
“signal” might reach V1 and V2 indirectly, via a cortical relay, it is also interesting to
consider subcortical pathways that could be more potent.
Some blending of the B-Ns/Ib-Is and 4B-Ks pathways within subcortical output of LIP
seems unavoidable. Outputs to the thalamus from LIPv and LIPd are convergent, and the
thalamo-cortical loop returns to LIP as well as neighbouring areas (Asanuma et al., 1985;
Schmahmann and Pandya, 1990; Hardy and Lynch, 1992). The circuitry does not, how-
ever, invade V4 or IT cortex to a great extent, as it has been demonstrated that LIP and IT
cortex connect with largely distinct thalamic realms, centred in dorsal and ventral pulvinar
respectively (Baleydier and Morel, 1992; Baizer et al., 1993). A potentially far more
pervasive re-entrant loop operates via the superior colliculus: LIP has a very dense output
to the deep layers of the colliculus (Lynch et al., 1985)—this is the circuitry that mediates
LIP’s control of eye movements. Moreover, as well as the descending output to brainstem

410 Stewart Shipp
oculomotor nuclei, the colliculus has ascending outputs to the thalamus, including
substantial parts of the pulvinar (Harting et al., 1980). The loop is completed by pulvino-
cortical projections which, maximally, include the entire visual cortex (Sherman and
Guillery, 1996). The re-entry into V4 and IT cortex establishes convergence between the
parietal and temporal streams, and this is notably asymmetric, since there is no equivalent
subcortical route from IT to LIP.
Colliculo-pulvino-cortical circuitry was mentioned above in the context of exogenously
driven (bottom-up) spatial attention, whereas spatial attention effects mediated via LIP are
more likely to be endogenous (top-down). There may be some separation of the respective
circuits, since retinal input to the colliculus targets its superficial layers, LIP input is
directed to its intermediate/deep layers, and prestriate areas like V4 and V5 (and TEO)
have convergent projections to the superficial/intermediate layers; furthermore, there is
some evidence that different layers of the colliculus project to different parts of the pulvinar
(Benevento and Standage, 1983). Unfortunately, the available anatomical data is insuffi-
cient to resolve the issue. It is nonetheless worth mentioning two experiments that imply
participation of the colliculus in endogenous spatial attention: (a) small collicular lesions
can interfere with a colour discrimination task, if the colour-target stimulus is distin-
guished from a distractor by a spatial cue, and the cue (and target) are placed into the
lesioned field location (Desimone et al., 1990); (b) the saccadic vector induced by
microstimulation of the colliculus is found to be deflected by a concurrent, covert shift of
attention, even if the attended location is indicated by a symbolic cue and the trained
behavioural response does not call for an eye movement (Kustov and Robinson, 1996).
6.2.3. Ventral-to-dorsal communication

The general conclusion from the above is that a blend of 4B-Ks with B-Ns/Ib-Is outputs is
broadcast from the dorsal stream to the ventral stream (perhaps principally via a subcortical
route)—thus integrating all these modalities of the parallel pathways. There is good reason to
believe that this dorsal-to-ventral communication, if it sustains focal spatial attention, must
play a rather fundamental role in the normal operation of the ventral pathway. By contrast,
the reverse interaction (ventral-to-dorsal) is rather less obtrusive. The supporting evidence
is not so abundant, as much of it derives from careful neuropsychological assessment of
a single patient.
This famous patient is DF, who became agnosic following CO poisoning and ventral-
occipital cortical damage. She has largely intact visuomotor ability, successfully reaching
and grasping objects whose size and orientation she is unable to report verbally (Milner,
1997). This is held to represent independent (but unconscious) processing on the part of an
intact dorsal pathway. Following this formula, any demonstrable visuomotor impairment
in DF must reflect the absence of contributions from the ventral pathway to dorsal
pathway functions, and this has now been demonstrated in several respects. For instance,
DF is unusually reliant on stereoscopic cues: under monocular viewing her wrist orienta-
tion and grasp size are severely impaired compared to normals, who are presumed to make
use of monocular pictorial cues (e.g. perspective) in order to achieve accuracy in grasping
movements (Dijkerman et al., 1996; Marotta et al., 1997). For another example, note that
DF is good at grasping tools, but does not always take hold of the correct bit (i.e. the handle)
(Carey et al., 1996). The first example shows how ventral pathway contributions may, in
certain circumstances, abet visuomotor control. The second, by contrast, is essential—there

Figure 17.6. Schema for dorsal pathway (oblongs), ventral pathway (ovals), and damaged cortex (diagonal
shading) in patient DF. Partial overlap of the two pathways is depicted at all levels; in striate cortex this reflects
greater allocation of peripheral field to the dorsal pathway, and of central visual field to the ventral pathway;
overlap decreases as the pathways ascend through prestriate cortices to the parietal and occipito-temporal lobes.
Areas of damage in occipito-temporal (and parts of ventral prestriate?) cortex are largely conjectural. Spared
parts of the ventral pathway must mediate residual colour and pattern vision. The conscious realm is diminished
in DF (she is agnostic for object identity, size and orientation); the “zone of lost consciousness” may exceed the
region of physical damage due to loss of feedback from destroyed higher centres. Hence the “realm of the
unconscious” is depicted to include some components of the ventral pathway.
has to be ventral-to-dorsal communication to select the correct target of a grasping action8.

There being no analogous neuropsychological data for monkeys, it is somewhat ground-
less to specify pathways, except to underline the existence of reciprocal parieto-temporal
connections, and to note the likelihood of participation by circuits involving frontal/pre-
motor cortex in any kind of planned action, such as picking up a tool by its handle.
The usual interpretation of case DF is that it not only demonstrates the operation of
a separate dorsal pathway for visuomotor control, but also that the pathway incorporates
a surprisingly autonomous level of form processing that is unavailable to the ventral path-
way, and is not consciously perceived (Goodale and Milner, 1992; Milner, 1997). This
may be true, but it is not unquestionably so. Figure 17.6 provides a (pitifully) crude
schematic depiction. Firstly, it is unlikely that the diffuse damage caused by CO anoxia
could restrict itself to the ventral pathway, when, as we have seen, the ventral pathway is
so difficult to pin down anatomically. Perhaps some overlapping elements of the dorsal
pathway are also damaged—at the very least, areas processing pictorial cues that seem
capable of contributing to both perception and motor control, as noted above. Equally,
8
A degree of shape selectivity intrinsic to the dorsal pathway has been reported (Sereno and Maunsell, 1998)—
but in the limit, this could not suffice in all circumstances, unless it were fully to duplicate the object identifying
capacity of IT cortex.

412 Stewart Shipp
some areas of spared cortex could also possess dual functionality. These are shown within
the “non-conscious” realm in Figure 17.6, in accordance with the fact that DF cannot
report any form percepts, either verbally or, in analogue form, by gestures. However this
does not mean that these areas are indubitably outside the NCC (neural correlate of
consciousness); they may only have been rendered so by the effect of the lesion to higher
areas, depriving them of feedback. There is a considerable body of thought implying the
operation of feedback loops within the normal framework of perception (e.g. Pollen,
1999).
None of this is to gainsay the fact that the processes subserving visuo-motor control and
visual cognition may differ in many respects: e.g. egocentric versus allocentric spatial
coding; residence in short term memory; relationship to a stored knowledge base, etc.
Ultimately, the underlying circuitry must reflect this schism, and it is important properly
to understand its evolution. Figure 17.6 could be viewed as a network, with progressively
individuating modes that never achieve total separation. It would be ironic if the two
pathway formulation, adhered to too tightly, were to become a roadblock to better under-
standing.
6.3. Integration through Non-specific Circuitry

The circuitry that is the subject matter for this concluding section is ‘non-specific’ in two
senses, either: (a) in the sense that it is anatomically diffuse, terminating freely across
functionally specialised compartments and/or areas, and with a modulatory synaptic
action; or (b) it underpins certain integrative functional phenomena that clearly demand
some form of cross-modal interaction, and yet whose level and locale within the visual
network cannot be specified with any degree of reliability.
Included in (a) are all feedback connections, anatomically defined as axonal termina-
tions focused upon layers 1 and 6; although excitatory, these probably avoid unstable
positive feedback effects by acting through voltage dependent synapses that modulate the
response to ascending (feed-forward) inputs (e.g. Crick and Koch, 1998). Additional
circuitries in this category may be components of “intrinsic” and “lateral” connections
(connections internal to an area, and between areas of equal hierarchical status) whose
laminar patterns are a hybrid of the feedback and forward patterns (Felleman and Van
Essen, 1991). Feedback connections are more diffusely organised than forward connec-
tions (Zeki and Shipp, 1988; Krubitzer and Kaas, 1989; Rockland et al., 1994). For
instance, the feedback from areas V4 and V5 to V2 is partly reciprocal, in that it princip-
ally targets the appropriate source stripe (thin/inter or thick respectively), and partly
diffuse, in that all the intervening territory in V2 also receives some degree of feedback,
to layers 1 and 6 (Shipp and Zeki, 1989b; Zeki and Shipp, 1989).
Under consideration in (b) are some phenomena related to forms of attention. Although
spatial attention can be locked to an empty position in space (Posner, 1980), it is often
synonymous with attention to an object present at a particular location: the perceptual
resolution of this object is enhanced, whilst that of other nearby objects is attenuated (e.g.
Lee et al., 1999). At the neural level, these features of attention are captured by models of
competition (Desimone, 1998; Lee et al., 1999). Evidently there is a role here for local
intrinsic connections, the competition being effected between neurones with overlapping
receptive fields. There is a reduced response to non-attended items (Moran and Desimone,
1985): essentially, if there is an underlying plasticity in the selection of items to which

a neurone can respond, attention is a process of focusing local neural processing resources
upon a single dominant stimulus (Luck et al., 1997; McAdams and Maunsell, 1999;
Reynolds et al., 1999; Treue and Martinez Trujillo, 1999; Treue and Maunsell, 1999).
The global integrative element in this dynamic allocation of processing resources
becomes apparent when it is considered that attention tends to spread to all the attributes
of the object at the target location: or that the (visual) “network tends to cascade into a state
in which the same object is dominant throughout” (Duncan, 1998). This means not only
that diverse object attributes, e.g. colour, form and motion, are better resolved but also that,
in a crowded scene, the individual attributes of the attended object are tied together. This
is known as binding (Treisman, 1996, 1998)9. For example, watching a game of snooker
(or pool), the colour and velocity of only one attended ball is known; awareness of the
remainder is for balls of diverse colour moving in diverse directions, but there is little
knowledge of specific colour-motion pairings. Evidence for the role of attention in binding
comes from psychological studies (Treisman, 1998), and the apparent dramatic rise in fail-
ures of binding (“illusory conjunctions”) suffered by a parietal lesion case demonstrating
the classic features of Balint’s syndrome, including simultanagnosia (Friedman-Hill et al.,
1995). It is thus obvious that binding (or concurrent, object-based, multimodal feature
attention) is a prime example of large scale integrative action across the visual system.
The neural basis of binding is far from understood, but there is hopeful progress in
hypotheses relating to the role of neural synchronisation in feature integration (Engel
et al., 1997). Synchronised activity can be demonstrated transcortically (e.g. between
V1 and V2: Nowak et al., 1999; Roe and Ts’o, 1999). Importantly, some synchronised
responses are contingent upon simple stimuli, e.g. oriented bars, moving coherently, as if
they were part of the same object (Kreiter and Singer, 1996). As yet there is no comparable
demonstration that synchronisation can support cross-modal feature binding, nor that
synchronisation is contingent upon focal attention. Nonetheless, synchronisation has the
promise of coding for the partnership of an object’s features processed along separate
pathways. In other words, it has the capacity to code for transient associations between all
possible feature combinations (brown and upward, red and leftward etc., in the snooker
example). Hence the circuitry sustaining this capacity would in all likelihood be diffusely
organised, including all those connections that are the usual suspects for mediating the
construction of dual selectivity. Models of neural networks find that synchronisation is an
emergent property of networks that are richly connected across levels, to model forward,
backward and lateral connectivity (Tononi et al., 1992; Schillen and König, 1994; Lumer
et al., 1997).
The cross-connections between stripes within V2, and other examples of lateral connec-
tivity, for instance V4/V5, or LIP/TEO, could all contribute to establishing transient states
of synchrony between neurones responding to different features of the same object. In
addition, there is the propensity of feedback connections to terminate diffusely, such that
the feedback from V3, V4 and V5 spreads across all pathways at the level of V2. This
contrasts to the situation in V1, where the feedback from areas V4 and V5 is more tightly
reciprocal (Zeki and Shipp, 1988). This lends indirect support for the teleological view
that cross-channel communication is a particular attribute of V2, its real raison d’être, as if
9
Readers of Duncan (1998) and Treisman (1998) will note that the former’s competition model and the latter’s
feature integration theory are not quite so readily reconciled as this passage implies.

414 Stewart Shipp
the later divergence of the pathways within prestriate cortex is delayed in order to allow
their more effective integration. Finally, it is worth noting that the relevant circuitry is
unlikely to be confined to the visual cortex, as interactions with the frontal lobe are bound
to be an integral component of all attentive phenomena.
7. CONCLUSION
Although many different localities in the visual system display a parallel architecture, it is
not clear that the global organisation of the system is best described as “parallel”. Take for
instance the three distinct retinal outputs (M, P and K); these retain their identity through
a couple of synapses, but are then variously remixed into three new cortical channels
(B-Ns, Ib-Is and 4B-Ks). These cortical channels retain their separation to a degree
(depending where you look) but again, they certainly do not function in splendid isolation.
Cortical areas select from these channels the appropriate mix from which to construct their
specialised representations; they may, perhaps, be even less discriminating in sampling
each others’ output in order to co-ordinate their activities.
Nevertheless, the functional specialisms of some areas are sufficiently distinct to
identify separate dorsal and ventral “pathways”—each an assembly of particular areas and
networks of connections. It is an interesting question whether this is a casual or a causal
distinction. A casual distinction is one of convenience, a way of breaking down a complex
system for ease of description. There may be an underlying continuum of function across
the cortical mantle, but a partition enables certain generalisations to be made, just as
a dividing line around the globe (the equator) enables North–South distinctions of macro-
economic activity. The simple tags for function (What vs Where) and location (parietal
lobe vs temporal lobe) aid and abet this distinction. A causal distinction would reflect
some deeper dualistic factor determining cerebral organisation—perhaps a duality
between the goals of visual processing, a physical representation to guide immediate
motor acts, and a categorical representation to permit cognitive manipulations. In the latter
case there should be some observable discontinuity of the functional continuum across the
parietal/temporal divide.
This question is not resolved here. The burden of this article is that even if the dorsal/
ventral distinction is a fundamental one, it poorly encapsulates the architecture of the
visual system, whose functional integrity depends on interactions between these two
systems. Binding of stimulus attributes into coherent object properties is one notional
function, that seems to make sense of cross-talk at all levels. This should be a reasonably
symmetrical relationship: for instance, an object’s location and velocity can be important
factors in knowing what it is, just as object identity is an important determinant in for-
mulating how (and whether) to reach out and manipulate it. But there is also one very
significant asymmetrical relationship, which concerns spatial attention (itself not
unrelated to binding). It seems very likely that the control of spatial attention has evolved
from the mechanisms that direct the eyes toward items of behavioural significance,
a crucial dorsal pathway function. But the circuitry has diversified and, through cortical
and subcortical loops, infiltrates the ventral pathway at all levels. It would seem that the
dorsal system wields an intimate influence over its twin. The operational details of this
neural cookery are obscure. Not surprisingly, it is easier to provide a list of ingredients
than a recipe.

REFERENCES
Allman, J. and Zucker, S. (1990) Cytochrome oxidase and functional coding in primate striate cortex: a hypo-
thesis. Cold Spring Harbor Symposia on Quantitative Biology, 15, 979–982.
Andersen, R.A., Asanuma, A. and Cowan, W.M. (1985) Callosal and prefrontal associational projecting cell
populations in area 7A of the macaque monkey: a study using retrogradely transported fluorescent dyes.
Andersen, R.A., Asanuma, A., Essick, G. and Siegel, R.M. (1990) Corticocortical connections of anatomically
and physiologically defined subdivisions within the inferior parietal lobule. Journal of Comparative Neuro-
logy, 296, 65–113.
Asanuma, A., Andersen, A. and Cowan, W.M. (1985) The thalamic relations of the caudal inferior parietal
lobule and the lateral prefrontal cortex in monkeys: divergent cortical projections from cell clusters in the
medial pulvinar nucleus. Journal of Comparative Neurology, 241, 357–381.
Baizer, J.S., Desimone, R. and Ungerleider, L.G. (1991) Organization of visual inputs to the inferior temporal
and posterior parietal cortex in macaques. Journal of Neuroscience, 11, 168–190.
Baizer, J.S., Desimone, R. and Ungerleider, L.G. (1993) Comparison of subcortical connections of inferior
temporal and posterior parietal cortex in monkeys. Visual Neuroscience, 10, 59–72.
Baleydier, C. and Morel, A. (1992) Segregated thalamocortical pathways to inferior parietal and inferotemporal
cortex in macaque monkey. Visual Neuroscience, 8, 391–405.
Barash, S., Bracewell, R.M., Fogassi, L., Gnadt, J.W. and Andersen, R.A. (1991) Saccade-related activity in
the intraparietal area I. Temporal properties; comparison with area 7a. Journal of Neurophysiology, 66,
1095–1108.
Barbur, J.L., Harlow, A.J. and Plant, G.T. (1994) Insights into the different exploits of color in the visual-cortex.
Proceedings of the Royal Society (London) series B, 258, 327–334.
Bartfeld, E. and Grinvald, A. (1992) Relationships between orientation-preference pinwheels, cytochrome
oxidase blobs, and ocular-dominance columns in primate striate cortex. Proceedings of the National
Academy of Sciences of the U.S.A., 89, 11905–11909.
Bender, D.B. (1981) Retinotopic organization of macaque pulvinar. Journal of Neurophysiology, 46, 672–693.
Benevento, L.A. and Standage, G.P. (1983) The organization of projections of the retinorecipient and nonretino-
recipient nuclei of the pretectal complex and layers of the superior colliculus to the lateral pulvinar and
medial pulvinar in the macaque monkey. Journal of Comparative Neurology, 217, 307–336.
Blasdel, G.G. (1992) Orientation selectivity, preference, and continuity in monkey striate cortex. Journal of
Blasdel, G.G. and Fitzpatrick, D. (1984) Physiological organization of layer 4 in macaque striate cortex. Journal
Blatt, G.J., Andersen, R.A. and Stoner, G.R. (1990) Visual receptive field organization and cortico-cortical
connections of the lateral intraparietal area (area LIP) in the macaque. Journal of Comparative Neurology,
299, 421–445.
Bonda, E., Petrides, M. and Evans, A. (1996) Specific involvement of human parietal systems and the amygdala
in the perception of biological motion. Journal of Neuroscience, 16, 3737–3744.
Boussaoud, D., Ungerleider, L.G. and Desimone, R. (1990) Pathways for motion analysis: cortical connections
of the medial superior temporal and fundus of the superior temporal visual areas in the macaque. Journal of
Brefczynski, J.A. and DeYoe, E.A. (1999) A physiological correlate of the ‘spotlight’ of visual attention. Nature
Brothers, L. and Ring, B. (1993) Mesial temporal neurons in the macaque monkey with responses selective for
aspects of social stimuli. Behavioural Brain Research, 57, 53–61.
Callaway, E.M. (1998) Local circuits in primary visual cortex of the macaque monkey. Annual Review Neuro-
science, 21, 47–74.
Carey, D.P., Harvey, M. and Milner, A.D. (1996) Visuomotor sensitivity for shape and orientation in a patient
with visual form agnosia. Neuropsychologia, 34, 329–337.
Casagrande, V.A. (1994) A third parallel visual pathway to primate area V1. Trends in Neuroscience, 17, 305–310.
Cavanagh, P., Henaff, M.A., Michel, F., Landis, T., Troscianko, T. and Intriligator, J. (1998) Complete
sparing of high-contrast color input to motion perception in cortical color blindness. Nature Neuroscience,
1, 242–247.
Colby, C.L., Duhamel, J.R. and Goldberg, M.E. (1995) Oculocentric spatial representation in parietal cortex.
Colby, C.L. and Goldberg, M.E. (1999) Space and attention in parietal cortex. Annual Review of Neuroscience,
22, 319–349.
Connolly, M. and Van Essen, D.C. (1984) The representation of the visual field in parvicellular and magnocellular
layers of the lateral geniculate nucleus in the macaque monkey. Journal of Comparative Neurology, 226,
544–564.

416 Stewart Shipp
Connor, C.E., Preddie, D.C., Gallant, J.L. and Van Essen, D.C. (1997) Spatial attention effects in macaque area
V4. Journal of Neuroscience, 17, 3201–3214.
Corbetta, M. (1998) Frontoparietal networks for directing attention and the eye to visual locations: Identical,
independent or overlapping neural systems. Proceedings of the National Academy of Sciences of the U.S.A.,
95, 831–838.
Corbetta, M., Akbudak, E., Conturo, T.E., Snyder, A.Z., Ollinger, J.M., Drury H.A. et al. (1998) A common
network of functional areas for attention and eye movements. Neuron, 21, 761–773.
Crick, F. and Koch, C. (1995) Are we aware of neural activity in primary visual cortex? Nature (London), 375,
121–123.
Crick, F. and Koch, C. (1998) Constraints on cortical and thalamic projections: the no-strong-loops hypothesis.
Cumming, B.G. and Parker, A.J. (1999) Binocular neurons in V1 of awake monkeys are selective for absolute,
not relative, disparity. Journal of Neuroscience, 19, 5602–5618.
Damasio, A., Yamada, T., Damasio, H., Corbett, J. and McKee, J. (1980) Central achromatopsia: behavioral,
anatomic, and physiologic aspects. Neurology, 30, 1064–1071.
Derrington, A.M., Krauskopf, J. and Lennie, P. (1984) Chromatic mechanisms in lateral geniculate nucleus of
macaque. Journal of Physiology (London), 357, 241–265.
Derrington, A.M. and Lennie, P. (1984) Spatial and temporal contrast sensitivities of neurones in lateral genicu-
late nucleus of macaque. Journal of Physiology (London), 357, 219–240.
Desimone, R. (1998) Visual attention mediated by biased competition in extrastriate visual cortex. Philosophical
Transactions of the Royal Society of London, series B, 353, 1245–1255.
Desimone, R., Wessinger, M., Thomas, L. and Schneider, W. (1990) Attentional control of visual perception:
cortical and subcortical mechanisms. Cold Spring Harbor Symposia on Quantitative Biology, 55, 963–971.
DeValois, R.L., Snodderly, D.M., Yund, E.W. and Hepler, N.K. (1977) Responses of macaque lateral geniculate
cells to luminance and color figures. Sensory Processes, 1, 244–259.
DeYoe, E.A., Felleman, D.J., Van Essen, D.C. and McClendon, E. (1994) Multiple processing streams in occipi-
totemporal visual cortex. Nature (London), 371, 151–154.
DeYoe, E.A. and Van Essen, D.C. (1985) Segregation of efferent connections and receptive field properties in
visual area 2 of the macaque. Nature (London), 317, 58–61.
DeYoe, E.A. and Van Essen, D.C. (1988) Concurrent processing streams in monkey visual cortex. Trends in
Dijkerman, H.C., Milner, A.D. and Carey, D.P. (1996) The perception and prehension of objects oriented in the
depth plane. I. Effects of visual form agnosia. Experimental Brain Research, 112, 442–451.
Distler, C., Boussaoud, D., Desimone, R. and Ungerleider, L.G. (1993) Cortical connections of inferior temporal
area TEO in macaque monkeys. Journal of Comparative Neurology, 334, 125–150.
Dobkins, K.R. and Albright, T.D. (1993) What happens if it changes color when it moves?: psychophysical
experiments on the nature of chromatic input to motion detectors. Vision Research, 33, 1019–1036.
Dobkins, K.R. and Albright, T.D. (1994) What happens if it changes color when it moves?: the nature of
chromatic input to macaque visual area MT. Journal of Neuroscience, 14, 4854–4870.
Dow, B.M. (1974) Functional classes of cells and their laminar distribution in monkey visual cortex. Journal of
Dreher, B., Fukada, Y. and Rodieck, R.W. (1976) Identification, classification and anatomical segregation of
cells with X-like and Y-like properties in the lateral geniculate nucleus of Old-World primates. Journal of
Duhamel, J.-R., Colby, C.L. and Goldberg, M.E. (1998) Ventral intraparietal area of the macaque: congruent
visual and somatic response properties. Journal of Neurophysiology, 79, 126–136.
Duncan, J. (1998) Converging levels of analysis in the cognitive neuroscience of visual attention. Philosophical
Transactions of the Royal Society of London, series B, 353, 1307–1317.
Edwards, D.P., Purpura, K.P. and Kaplan, E. (1995) Contrast sensitivity and spatial frequency response
of primate cortical neurons in and around the cytochrome oxidase blobs. Vision Research, 35,
1501–1523.
Engel, A.K., Roelfsema, P.R., Fries, P., Brecht, M. and Singer, W. (1997) Role of the temporal domain for
response selection and perceptual binding. Cerebral Cortex, 7, 571–582.
Felleman, D.J. and Van Essen, D.C. (1987) Receptive field properties of neurons in area V3 of macaque monkey
extrastriate cortex. Journal of Neurophysiology, 57, 889–920.
Felleman, D.J., Xiao, Y. and McClendon, E. (1997) Modular organization of occipito-temporal pathways:
cortical connections between visual area 4 and visual area 2 and posterior inferotemporal ventral area in
macaque monkeys. Journal of Neuroscience, 17, 3185–3200.
Friedman-Hill, S.R., Robertson, L.C. and Treisman, A. (1995) Parietal contributions to visual feature binding:
evidence from a patient with bilateral lesions. Science (Washington), 269, 853–855.

Gandhi, S.P., Heeger, D.J. and Boynton, G.M. (1999) Spatial attention affects brain activity in human primary
visual cortex. Proceedings of the National Academy of Sciences of the USA, 96, 3314–3319.
Gattass, R., Sousa, A.B.P., Mishkin, M. and Ungerleider, L.G. (1997) Cortical projections of area V2 in the
macaque. Cerebral Cortex, 7, 110–129.
Gegenfurtner, K.R. and Hawken, M.J. (1996) Interaction of motion and colour in the visual pathways. Trends in
Neurosciences, 19, 394–401.
Gegenfurtner, K.R., Kiper, D.C. and Fenstemaker, S.B. (1996) Processing of colour, form and motion in
macaque area V2. Visual Neuroscience, 13, 161–172.
Gegenfurtner, K.R., Kiper, D.C. and Levitt, J.B. (1997) Functional properties of neurons in macaque area V3.
Goodale, M.A. and Milner, A.D. (1992) Separate visual pathways for perception and action. Trends in
Neurosciences, 15, 20–25.
Gottlieb, J.P., Kusunoki, M. and Goldberg, M.E. (1998) The representation of visual salience in monkey parietal
Gouras, P. and Krüger, J. (1979) Responses of cells in foveal visual cortex of the monkey to pure color contrast.
Grossberg, S. (1994) 3-D vision and figure ground separation by visual cortex. Perception and Psychophysics,
55, 48–120.
Hardy, S.G.P. and Lynch, J.C. (1992) The spatial distribution of pulvinar neurons that project to two subregions
of the inferior parietal lobule in the macaque. Cerebral Cortex, 2, 217–230.
Harting, J.K., Huerta, M.F., Frankfurter, A.J., Strominger, N.L. and Royce, G.J. (1980) Ascending pathways
from the monkey superior colliculus: an autoradiographic study. Journal of Comparative Neurology,
192, 853–882.
Hawken, M.J. and Parker, A.J. (1984) Contrast sensitivity and orientation selectivity in lamina IV of the striate
cortex of Old World primates. Experimental Brain Research, 54, 367–372.
Haxby, J.V., Grady, C.L., Horwitz, B., Ungerleider, L.G., Mishkin, M., Carson, R.E., Herscovitch, P.,
Schapiro, M.P. and Rapoport, S.I. (1991) Dissociation of object and spatial processing pathways in
human extrastriate cortex. Proceedings of the National Academy of Sciences of the USA, 88, 1621–1625.
Hendry, S.H.C. and Yoshioka, T. (1994) A neurochemically distinct third channel in the macaque dorsal lateral
geniculate nucleus. Science (Washington), 264, 575–577.
Heywood, C.A., Kentridge, R.W. and Cowey, A. (1998) Form and motion from colour in cerebral achromatop-
sia. Experimental Brain Research, 123, 145–153.
Hicks, T.P., Lee, B.B. and Vidyasagar, T.R. (1983) The responses of cells in macaque lateral geniculate nucleus
to sinusoidal gratings. Journal of Physiology (London), 337, 183–200.
Horton, J.C. and Hedley-White, E.T. (1984) Mapping of cytochrome oxidase patches and ocular dominance
coulumns in human visual cortex. Philosophical Transactions of the Royal Society of London, series B, 304,
255–272.
Hubel, D.H. and Livingstone, M.S. (1987) Segregation of form, color and stereopsis in primate area 18. Journal
Hubel, D.H. and Livingstone, M.S. (1990) Color and contrast sensitivity in the lateral geniculate body and
primary visual cortex of the macaque monkey. Journal of Neuroscience, 10, 2223–2237.
Hübener, M. and Bolz, J. (1992) Relationships between dendritic morphology and cytochrome oxidase compart-
ments in monkey striate cortex. Journal of Comparative Neurology, 324, 67–80.
Ingling, Jr., C.R. and Martinez-Uriegas, E. (1983) The relationship between spectral sensitivity and spatial
sensitivity for the primate r-g X-channel. Vision Research, 23, 1495–1500.
Jacobs, G.H., Deegan, J.F., Deegan, D., Neitz, J., Crognale, M.A. and Neitz, M. (1993) Photopigments and color
vision in the nocturnal monkey, Aotus. Vision Research, 33, 1773–1783.
Kiper, D.C., Fenstemaker, S.B. and Gegenfurtner, K.R. (1997) Chromatic properties of neurons in macaque area
V2. Visual Neuroscience, 14, 1061–1072.
Kolmel, H.W. (1988) Pure homonymous hemiachromatopsia: findings with neuro-ophthalmolgic examination
and imaging procedures. European Archives of Psychiatry and Neurological sciences, 237, 237–243.
Krauskopf, J. (1963) Effect of retinal image stabilization on the appearance of heterochromatic targets. Journal
of the Optical Society of America, 53, 741–744.
Kreiter, A.K. and Singer, W. (1996) Stimulus-dependent synchronization of neuronal responses in the visual
cortex of the awake macaque monkey. Journal of Neuroscience, 16, 2381–2396.
Krubitzer, L.A. and Kaas, J.H. (1989) Cortical integration of parallel pathways in the visual system of primates.
Kustov, A.A. and Robinson, D.L. (1996) Shared neural control of attentional shifts and eye movements. Nature
(London), 384, 74–77.
Lee, B.B., Martin, P.R. and Valberg, A. (1988) The physiological basis of heterochromatic flicker photo-
metry demonstrated in the ganglion cells of the macaque retina. Journal of Physiology (London), 404,
323–347.

418 Stewart Shipp
Lee, D.K., Itti, L., Koch, C. and Braun, J. (1999) Attention activates winner-take-all competition among visual
filters. Nature Neuroscience, 2, 375–381.
Lennie, P., Krauskopf, J. and Sclar, G. (1990) Chromatic mechanisms in striate cortex of macaque. Journal of
Leopold, D.A. and Logothetis, N.K. (1996) Activity changes in early visual cortex reflect monkeys’ percepts
during binocular rivalry. Nature (London), 379, 549–553.
Leventhal, A.G., Thompson, K.G., Liu, D., Zhou, Y. and Ault, S.J. (1995) Concomitant sensitivity to orientation,
direction, and color of cells in layers 2, 3 and 4 of monkey striate cortex. Journal of Neuroscience, 15,
1808–1818.
Levitt, J.B., Kiper, D.C. and Movshon, J.A. (1994a) Receptive fields and functional architecture of macaque V2.
Levitt, J.B., Yoshioka, T. and Lund, J.S. (1994b) Intrinsic cortical connections in macaque visual area V2:
evidence for interaction between different functional streams. Journal of Comparative Neurology, 342,
551–570.
Livingstone, M.S. and Hubel, D.H. (1983) Specificity of cortico-cortical connections in monkey visual system.
Livingstone, M.S. and Hubel, D.H. (1984a) Anatomy and physiology of a color system in the primate visual
Livingstone, M.S. and Hubel, D.H. (1984b) Specificity of intrinsic connections in primate primary visual cortex.
Livingstone, M.S. and Hubel, D.H. (1987a) Connections between layer 4B of area 17 and the thick cytochrome
oxidase stripes of area 18 in the squirrel monkey. Journal of Neuroscience, 7, 3371–3377.
Livingstone, M.S. and Hubel, D.H. (1987b) Psychophysical evidence for separate channels for the perception of
form, color, movement, and depth. Journal of Neuroscience, 7, 3416–3468.
Livingstone, M.S. and Hubel, D.H. (1988) Segregation of form, color, movement, and depth: anatomy,
physiology, and perception. Science (Washington), 240, 740–749.
Logothetis, N.K., Schiller, P.H., Charles, E.R. and Hurlbert, A.C. (1990) Perceptual deficits and the activity of
the color-opponent and broad-band pathways at isoluminance. Science (Washington), 247, 214–217.
Luck, S.J., Chelazzi, L., Hillyard, S.A. and Desimone, R. (1997) Neural mechanisms of spatial selective
attention in areas V1, V2 and V4 of macaque visual cortex. Journal of Neurophysiology, 77, 24–42.
Lumer, E.D., Edelman, G.M. and Tononi, G. (1997) Neural dynamics in a model of the thalamocortical system.
I. Layers, loops and the emergence of fast synchronous rhythms. Cerebral Cortex, 7, 207–227.
Lund, S., Yoshioka, T. and Levitt, B. (1994) Substrates for interlaminar connections in area V1 of macaque
monkey visual cortex. In: A. Peters and K.S. Rockland (eds), Primary Visual Cortex in Primates. New
York: Plenum, pp. 37–60.
Lynch, J.C., Graybiel, A.M. and Lobeck, L.J. (1985) The differential projection of two cytoarchitectonic
subregions of the inferior parietal lobule of macaque upon the deep layers of the superior colliculus. Journal
Malach, R., Tootell, R.B.H. and Malonek, D. (1994) Relationship between orientation domains, cytochrome oxi-
dase stripes, and intrinsic horizontal connections in Squirrel monkey area V2. Cerebral Cortex, 4, 151–165.
Marotta, J.J., Behrmann, M. and Goodale, M.A. (1997) The removal of binocular cues disrupts the calibration of
grasping in patients with visual form agnosia. Experimental Brain Research, 116, 113–121.
Martin, P.R., White, A.J., Goodchild, A.K., Wilder, H.D. and Sefton, A.E. (1997) Evidence that blue-on cells are
part of the third geniculocortical pathway in primates. European Journal of Neuroscience, 9, 1536–1541.
Martinez, A., Anllo-Vento, L., Sereno, M.I., Frank, L.R., Buxton, R.B., Dubowitz, D.J. et al. (1999) Involve-
ment of striate and extrastriate visual cortical areas in spatial attention. Nature Neuroscience, 2, 364–369.
Maunsell, J.H.R., Nealey, T.A. and DePriest, D.D. (1990) Magnocellular and parvocellular contributions to
3323–3334.
Maunsell, J.H.R. and Van Essen, D.C. (1983) The connections of the middle temporal area and their relationship
to a cortical hierarchy in the macaque monkey. Journal of Neuroscience, 3, 2563–2586.
McAdams, C.J. and Maunsell, J.H.R. (1999) Effects of attention on orientation-tuning functions of single
neurons in macaque cortical area V4. Journal of Neuroscience, 19, 431–441.
Merigan, W.H., Byrne, C.E. and Maunsell, J.H.R. (1991a) Does primate motion perception depend on the
magnocellular pathway? Journal of Neuroscience, 11, 3422–3429.
Merigan, W.H., Katz, L.M. and Maunsell, J.H. (1991b) The effects of parvocellular lateral geniculate lesions on
the acuity and contrast sensitivity of macaque monkeys. Journal of Neuroscience, 11, 994–1001.
Milner, A.D. (1997) Vision without knowledge. Philosophical Transactions of the Royal Society of London,
series B, 352, 1249–1256.
Milner, A.D., Perrett, D.I., Johnston, R.S., Benson, P.J., Jordan, T.R., Heeley, D.W., Beltucci, D., Mortara, F.,
Mutani, R., Terazzi, E. and Davidson, D.L.W. (1991) Perception and action in ‘visual form agnosia’. Brain,
114, 405–428.

Mishkin, M., Ungerleider, L.G. and Macko, K.A. (1983) Object vision and spatial vision: two cortical pathways.
Moran, J. and Desimone, R. (1985) Selective attention gates visual processing in the extrastriate cortex. Science
(Washington), 229, 782–784.
Morel, A. and Bullier, J. (1990) Anatomical segregation of two cortical visual pathways. Visual Neuroscience, 4,
555–578.
Motter, B.C. (1993) Focal attention produces spatially selective processing in visual cortical areas V1, V2, and
V4 in the presence of competing stimuli. Journal of Neurophysiology, 70, 909–919.
Movshon, J.A., Adelson, E.H., Gizzi, M.S. and Newsome, W.T. (1985) The analysis of moving visual patterns.
In: C. Chages, R. Gattass and C.G. Gross (eds), Pattern recognition mechanisms. Vatican: Pontifical
Academy of Science, pp. 117–151.
Munk, M.H., Nowak, L.G., Girard, P., Chounlamountri, N. and Bullier, J. (1995) Visual latencies in cytochrome
oxidase bands of macaque area V2. Proceedings of the National Academy of Sciences of the USA, 92,
988–992.
Nakamura, M., Gattass, R., Desimone, R. and Ungerleider, L.G. (1993) The modular organization of projections
from areas V1 and V2 to areas V4 and TEO in macaques. Journal of Neuroscience, 13, 3681–3691.
Nealey, T.A. and Maunsell, J.H.R. (1994) Magnocellular and parvocellular contributions to the responses of
neurons in macaque striate cortex. Journal of Neuroscience, 14, 2069–2079.
Nowak, L.G., Munk, M.H., James, A.C., Girard, P. and Bullier, J. (1999) Cross-correlation study of the
temporal interactions between areas V1 and V2 of the macaque monkey. Journal of Neurophysiology, 81,
1057–1074.
Oram, M.W., Perrett, D.I. and Hietanen, J.K. (1993) Directional tuning of motion-sensitive cells in the anterior
superior temporal polysensory area of the macaque. Experimental Brain Research, 97, 274–294.
Orban, G.A. (1997) Visual Processing in Macaque Area MT/V5 and its Satellites (MSTd and MSTv). In:
K.S. Rockland, J.H. Kaas and A. Peters (eds), Cerebral Cortex Vol. 12: Extrastriate Visual Cortex in
Primates. New York: Plenum, pp. 359–434.
Orban, G.A., Kennedy, H. and Bullier, J. (1986) Velocity and direction selectivity of neurons in areas V1 and V2
of the monkey: influence of eccentricity. Journal of Neurophysiology, 56, 462–480.
Perrett, D.I., Harries, M.H., Chitty, A.J. and Mistlin, A.J. (1990) Three stages in the classification of body
movements by visual neurons. In: H.B. Barlow, C. Blakemore and M. Weston-Smith (eds), Images and
Understanding. Cambridge, UK: Cambridge University Press, pp. 94–107.
Perrett, D.I., Hietanen, J.K., Oram, M.W. and Benson, P.J. (1992) Organization and functions of cells responsive to
faces in the temporal cortex. Philosophical Transactions of the Royal Society of London, series B, 335, 23–30.
Perrett, D.I., Oram, M.W., Harries, M.H., Bevan, R., Hietanen, J.K., Benson, P.J. and Thomas, S. (1991)
Viewer-centred and object-centred coding of heads in the macaque temporal cortex. Experimental Brain
Research, 86, 159–173.
Peterhans, E. and von der Heydt, R. (1993) Functional organization of area V2 in the alert macaque. European
Pollen, D.A. (1999) On the neural correlates of visual perception. Cerebral Cortex, 9, 4–19.
Posner, M.I. (1980) Orienting of attention. Quarterly Journal of Experimental Psychology, 32, 3–25.
Purves, D., Shimpi, A. and Lotto, R.B. (1999) An empirical explanation of the cornsweet effect. Journal of
Ramachandran, S. and Gregory, R.L. (1991) Perceptual filling in of artificially induced scotomas in human
vision. Nature (London), 350, 699–702.
Reynolds, J.H., Chelazzi, L. and Desimone, R. (1999) Competitive mechanisms subserve attention in macaque
areas V2 and V4. Journal of Neuroscience, 19, 1736–1753.
Rizzolatti, G., Riggio, L., Dascola, I. and Umilta, C. (1987) Reorienting attention across the horizontal and
vertical meridians: evidence in favor of a premotor theory of attention. Neuropsychologia, 25, 31–40.
Rockland, K.S., Saleem, K.S. and Tanaka, K. (1994) Divergent feedback conections from areas V4 and TEO in
the macaque. Visual Neuroscience, 11, 579–600.
Rodman, H.R. and Albright, T.D. (1989) Single-unit analysis of pattern-motion selective properties in the middle
temporal visual area. Experimental Brain Research, 75, 53–64.
Roe, A.W. and Ts’o, D.Y. (1995) Visual topography in primate V2: multiple representation across functional
stripes. Journal of Neuroscience, 15, 3689–3715.
Roe, A.W. and Ts’o, D.Y. (1997) The Functional Architecture of Area V2 in the Macaque Monkey. In:
K.S. Rockland, J.H. Kaas and A. Peters (eds), Cerebral Cortex Vol. 12: Extrastriate Visual Cortex in
Primates. New York: Plenum Press, pp. 295–333.
Roe, A.W. and Ts’o, D.Y. (1999) Specificity of color connectivity between primate V1 and V2. Journal of
Sakata, H., Taira, M., Kusunoki, M., Murata, A., Tanaka, Y. and Tsutsui, K. (1998) Neural coding of 3D features
of objects for hand action in the parietal cortex of the monkey. Philosophical Transactions of the Royal
Society of London, series B, 353, 1363–1373.

420 Stewart Shipp
Sawatari, A. and Callaway, E.M. (1996) Convergence of magno- and parvocellular pathways in layer 4B of
macaque primary visual cortex. Nature (London), 380, 442–446.
Schillen, T.B. and König, P. (1994) Binding by temporal structure in multiple feature domains of an oscillatory
neuronal network. Biological Cybernetics, 70, 397–405.
Schiller, P.H. and Colby, C.L. (1983) The responses of single cells in the lateral geniculate nucleus of the rhesus
monkey to color and luminance contrast. Vision Research, 23, 1631–1641.
Schiller, P.H., Logothetis, N.K. and Charles E.R. (1990) Role of colour-opponent and broad-band channels in
vision. Visual Neuroscience, 5, 321–346.
Schiller, P.H. and Malpeli, J.G. (1978) Functional specificity of lateral geniculate nucleus laminae of the rhesus
monkey. Journal of Neurophysiology, 41, 788–797.
Schmahmann, J.D. and Pandya, D.N. (1990) Anatomical investigation of projections from thalamus to posterior
parietal cortex in the rhesus monkey: a WGA-HRP and fluorescent tracer study. Journal of Comparative
Neurology, 237, 408–426.
Seltzer, B. and Pandya, D.N. (1978) Afferent cortical connections and architectonics of the superior temporal
sulcus and surrounding cortex in the rhesus monkey. Brain Research, 149, 1–24.
Seltzer, B. and Pandya, D.N. (1994) Parietal, temporal, and occipital projections to cortex of the superior temporal
sulcus in the rhesus monkey: a retrograde tracer study. Journal of Comparative Neurology, 343, 445–463.
Sereno, A.B. and Maunsell, J.H. (1998) Shape selectivity in primate lateral intraparietal cortex. Nature (London),
395, 500–503.
Shapley, R.M., Kaplan, E. and Soodak, R. (1981) Spatial summation and contrast sensitivity of X and Y cells in
the lateral geniculate nucleus of the macaque. Nature (London), 292, 543–545.
Sheliga, B.M., Riggio, L. and Rizzolatti, G. (1995) Spatial attention and eye movements. Experimental Brain
Research, 105, 261–275.
Sherman, S.M. and Guillery, R.W. (1996) Functional organization of thalamocortical relays. Journal of
Shipp, S., Blanton, M. and Zeki, S. (1998) A visuo-somatomotor pathway through superior parietal cortex in
the macaque monkey: cortical connections of areas V6 and V6A. European Journal of Neuroscience, 10,
3171–3193.
Shipp, S. and Zeki, S. (1985) Segregation of pathways leading from area V2 to areas V4 and V5 of macaque
monkey visual cortex. Nature (London), 315, 322–325.
Shipp, S. and Zeki, S. (1989a) The organization of connections between areas V5 and V1 in macaque monkey
Shipp, S. and Zeki, S. (1989b) The organization of connections between areas V5 and V2 in macaque monkey
Shipp, S. and Zeki, S. (1995) Segregation and convergence of specialised pathways in macaque monkey visual
cortex. Journal of Anatomy, 187, 547–562.
Somers, D.C., Dale, A.M., Seiffert, A.E. and Tootell, R.B. (1999) Functional MRI reveals spatially specific
attentional modulation in human primary visual cortex. Proceedings of the National Academy of Sciences of
the USA, 96, 1663–1668.
Sparr, S.A., Jay, M., Drislane, F.W. and Venna, N. (1991) A historic case of visual agnosia revisited after 40
years. Brain, 114, 789–800.
Tamura, H., Sato, H., Katsuyama, N., Hata, Y. and Tsumoto, T. (1996) Less segregated processing of visual
information in V2 than V1 of the monkey visual cortex. European Journal of Neuroscience, 8, 300–309.
Tanaka, K., Saito, H., Fukada, Y. and Moriya, M. (1991) Coding visual images of objects in the inferotemporal
cortex of the macaque monkey. Journal of Neurophysiology, 66, 170–189.
Thorell, L.G., Devalois, R.L. and Albrecht, D.G. (1984) Spatial mapping of monkey V1 cells with pure color and
luminance stimuli. Vision Research, 24, 751–769.
Tononi, G., Sporns, O. and Edelman, G.M. (1992) Reentry and the problem of integrating multiple cortical areas:
simulation of dynamic integration in the visual system. Cerebral Cortex, 2, 310–335.
Tootell, R.B.H. and Born, R.T. (1991) Spatial frequency tuning of units in macaque supragranular striate cortex.
Proceedings of the National Academy of Sciences of the USA, 88, 7066–7070.
Tootell, R.B.H. and Hamilton, S.L. (1989) Functional anatomy of the second visual area (V2) in the macaque.
Tootell, R.B.H., Hamilton, S.L. and Switkes, E. (1988a) Functional anatomy of macaque striate cortex. IV.
Contrast and magno parvo streams. Journal of Neuroscience, 8, 1594–1609.
Tootell, R.B.H., Silverman, M.S., Hamilton, S.L., DeValois, R.L. and Switkes, E. (1988b) Functional anatomy
of macaque striate cortex. III. Color. Journal of Neuroscience, 8, 1569–1593.
Tootell, R.B.H., Silverman, M.S., Hamilton, S.L., Switkes, E. and DeValois, R.L. (1988c) Functional anatomy
of macaque striate cortex. V. Spatial frequency. Journal of Neuroscience, 8, 1610–1624.
Treisman, A. (1996) The binding problem. Current Opinion in Neurobiology, 6, 171–178.
Treisman, A. (1998) Feature binding, attention and object perception. Philosophical Transactions of the Royal
Society of London, series B, 353, 1295–1306.

Treue, S. and Martinez Trujillo, J.C. (1999) Feature-based attention influences motion processing gain in
macaque visual cortex. Nature (London), 399, 575–579.
Treue, S. and Maunsell, J.H. (1999) Effects of attention on the processing of motion in macaque middle temporal
and medial superior temporal visual cortical areas. Journal of Neuroscience, 19, 7591–7602.
Ts’o, D.Y. and Gilbert, C.D. (1988) The organization of chromatic and spatial interactions in the primate striate
Ungerleider, L.G. and Desimone, R. (1986) Cortical connections of visual area MT in the macaque. Journal of
Van Essen, D.C., Felleman, D.J., DeYoe, E.A., Olavarria, J. and Knierim, J. (1990) Modular and hierarchical
organization of extrastriate visual cortex in the macaque monkey. Cold Spring Harbor Symposia on
Quantitative Biology, 55, 679–696.
Van Essen, D.C. and Maunsell, J.H.R. (1983) Hierarchical Organization and functional streams in the visual
cortex. Trends in Neurosciences, 6, 370–375.
Ware, C. and Cowan, W.B. (1983) The chromatic Cornsweet effect. Vision Research, 10, 1075–1077.
Webster, M.J., Bachevalier, J. and Ungerleider, L.G. (1994) Connections of inferior temporal areas TEO and TE
with parietal and frontal cortex in macaque monkeys. Cerebral Cortex, 4, 470–483.
Westheimer, G. (1979) Coperative neural processes involved in stereoscopic acuity. Experimental Brain
Research, 36, 585–597.
Wiesel, T.N. and Hubel, D.H. (1966) Spatial and chromatic interactions in the lateral geniculate body of the
rhesus monkey. Journal of Neurophysiology, 29, 1115–1156.
Wong-Riley, M.T., Hevner, R.F., Cutlan, R., Earnest, M., Egan, R., Frost, J. and Nguyen,T. (1993) Cytochrome
oxidase in the human visual cortex: distribution in the developing and the adult brain. Visual Neuroscience,
10, 41–58.
Xiao, Y., Zych, A. and Felleman, D.J. (1999) Segregation and convergence of functionally defined V2 thin stripe
and interstripe compartment projections to area V4 of macaques. Cerebral Cortex, 9, 792–804.
Yoshioka, T., Blasdel, G.G., Levitt, J.B. and Lund, J.S. (1996) Relation between patterns of intrinsic lateral
connectivity, ocular dominance, and cytochrome oxidase-reactive regions in macaque monkey striate
cortex. Cerebral Cortex, 6, 297–310.
Young, M.P. (1992) Objective analysis of the topological organization of the primate visual system. Nature
(London), 358, 152–154.
Zeki, S. (1983) Colour coding in the cerebral cortex: the reaction of cells in monkey visual cortex to wavelengths
and colours. Neuroscience, 9, 741–765.
Zeki, S. (1993) A Vision of the Brain. Oxford: Blackwell.
Zeki, S. and Shipp, S. (1988) The functional logic of cortical connections. Nature, 335, 311–317.
Zeki, S. and Shipp, S. (1989) Modular connections between areas V2 and V4 of macaque monkey visual cortex.
European Journal of Neuroscience, 1, 494–506.
Zeki, S.M. (1974) Functional organization of a visual area in the posterior bank of the superior temporal sulcus
of the rhesus monkey. Journal of Physiology (London), 236, 549–573.
Zheng, D., LaMantia, A.S. and Purves, D. (1991) Specialized vascularization of the primate visual cortex.

18 Wheels within Wheels: Circuits for Integration
of Neural Assemblies on Small and Large Scales
Robert Miller
Otago Centre for Theoretical Studies in Psychiatry and Neuroscience,
c/o Department of Anatomy and Structural Biology, School of Medical Science,
University of Otago, P.O. Box 913, Dunedin, New Zealand
Tel: 0064-3-4797357; FAX: 0064-3-479-7254; e-mail: robert.miller@stonebow.otago.ac.nz
The environments in which we live include an infinite variety of possible combinations of information,
and yet, within this, contain considerable redundancy. The mammalian forebrain has developed
a system of staggering complexity and beauty for representing such environments. Within the
cerebral cortex, the basic repository of information is a network with rather low levels of activity, so
that the tendency to uncontrolled “explosions” is reduced to safe levels. Nevertheless, priming of
this rather quiet “library” by more active parts of the forebrain (e.g. cortical lamina V, and thalamic
projection neurones) can, in the waking state, set into activity assemblies of nerve cells which
represent the fine spatial and temporal structure of the environment. By interaction between cortex
and basal ganglia the temporal structure on a larger scale can be represented as sequences, when
exact temporal representation is not possible. Ambiguities of representation, which are inherent in
the cortical network can be resolved by cortico-hippocampal interplay, given that a global repres-
entation of the current environment has been established. Two sorts of cell assemblies are needed,
representing respectively spatial and temporal structure. They can both form in a realistic cortical
network, but cannot both operate at the same time in such a network. This may be the reason for
hemispheric specialization, and should apply throughout the mammalian kingdom. “Omniconnection”
as a substrate for consciousness is not strictly realised (although it is approximated). In view of the
redundancy of information in the environment, strict omniconnection is not necessary.
KEYWORDS: axonal conduction, cell assemblies, hemispheric specialization, cortico-basal ganglionic

loops, exact timing, neocortical laminae, cortico-hippocampal interplay, “omniconnection”, sequencing,
thalamo-cortical relations
1. INTRODUCTION
The mammalian cerebral cortex is a vast network of interconnecting neurones, in which

about 89% of synaptic links are excitatory (Braitenberg and Schüz, 1991, 1998). The view
is gaining increasing support that the basic principle for functional organization of the
cerebral cortex is the cell assembly (or neural assembly). This concept was first advanced
by Hebb (1949) and has been developed in many ways subsequently (Braitenberg, 1978;
Palm, 1982; Miller, 1991, 1996a,b,c; Miller and Wickens, 1991; Sakurai, 1996; Wickens and
Miller, 1997).
In essence, a cell assembly is a distributed group of cortical neurones, with stronger
functional connections amongst themselves than with the surrounding majority of other
neurones. The strong connections required to form a cell assembly are envisaged to arise by
423
424 Robert Miller
synaptic strengthening, according to the well-known rule, also originally proposed by Hebb
(1949). Once formed, cell assemblies constitute a static store of information coded in the con-
figuration of very many strengthened connections. That information may be brought to an
active form only if the cell assembly is “ignited” (Braitenberg, 1978). This occurs when
member neurones of the assembly are activated from extrinsic sources above a threshold
level, at which point reverberating activity can spread to all members of the assembly (Wickens
and Miller, 1997). Since the member neurones which constitute an assembly may be widely
distributed, any one neurone can be part of several different assemblies, and assemblies can
represent meaningful information in a very versatile manner; and since the unit of information
storage is the strengthened synapse rather than the neurone, this means of storage is a very
economical use of the anatomically-defined cortical network. However, there is a serious
problem with the cell assembly theory, which has at present been resolved only partially.
In a network consisting mainly of excitatory connections, there is always a danger that
excitation will spread well beyond the boundaries of a specific cell assembly. Once activity
in the network as a whole exceeds a certain level, overall activity levels would tend to
increase in a rapidly-accelerating fashion, until an “explosion” of maximal activity occurs,
limited only by pathological effects such as neuronal fatigue. If this occurs, the possibility
of selective representation of specific categories of information would be totally lost. The
fact that various forms of epileptic seizure can occur in the cortex indicates that this is not
just a theoretical possibility in the cortex, but an ever-present danger, requiring powerful
safeguards if normal cortical functioning is to prevail.
There are several possible processes within the cortex which could serve to limit the
uncontrolled spread of excitation. Most obviously, excitatory links are not the only way in
which cortical neurones can interact. There are many inhibitory interneurones in the cortex,
utilising GABA as a transmitter, and it is estimated that approximately 11% of cortical
synapses are GABAergic (Braitenberg and Schüz, 1998). However, another aspect of the
design of the cortex to allow effective operation of neural assemblies has rarely been
considered: the layers of the cortical mantle where most of the cortico-cortical connections
have their origin and termination, and which are thus dominant for neural assembly function,
are relatively quiescent. Under many circumstances, the amount of neural traffic carried by
neurones in these layers is rather small. Perhaps, therefore the problem of stable operation
of cell assemblies should be viewed from the opposite perspective, captured by the following
questions: How can such relatively inert layers of cortical network tissue ever be activated
sufficiently to ignite cell assemblies? How can they be activated to such a degree that cell
assemblies form in the first place? The aim of this chapter is to present evidence and argu-
ments that there is in fact a series of interlocking circuits, some very local and small scale,
others more global and all-encompassing, whose role is to bring selected member neurones
of the cortical network to the level of activity at which cell assemblies can form and perform
their informational functions. The scale of the mechanisms varies, and as it does so, the scale
of integration achieved by cell assemblies also varies, from the very local to the global.
2. DIFFERENCES BETWEEN ACTIVITY LEVELS ACROSS

CORTICAL LAMINAE
A crucial fact in elucidating the operation of cell assemblies in the cortex is the laminar
differences in firing rate (discussed also in Miller, 1996a). According to Swadlow (1988,

Neural Assemblies on Small and Large Scales 425
1989, 1994) in the awake rabbit the firing rate of pyramidal neurones in laminae II and III
of various cortical areas ranges from 0.1–2 Hz. Many additional neurones in these laminae
may be essentially silent, and so escape detection. Due to such a sampling bias (discussed
by Miller, 1996c) the true range of firing rates may be even lower than in the samples
observed by Swadlow. According to Swadlow, pyramidal neurones in lamina VI also have
such low levels of spontaneous activity. In lamina V, however, the median rate of pyramidal
cell firing is considerably higher. This is a common observation in anaesthetized prepara-
tions, and is also found in the waking state. In the waking rabbits studied by Swadlow, spon-
taneous firing rate in such neurones ranged between 4 and 8 Hz. Revishchin (1985) also
finds that spontaneous firing rates are much higher for neurones in lamina V of rabbit
visual cortex, than for laminae II and III. From older data there are indications that these
conclusions apply to other species, although there are hints that there is some variation in
the finding. Schiller et al. (1976a) found, in monkey striate cortex that one class of neurone
(i.e. those whose excitatory and inhibitory field areas were coextensive) commonly had
higher rates of spontaneous firing if they were located in lamina V or VI than in laminae II
to IV. Gilbert (1977) studying cat primary visual cortex found two bands of high spontaneous
activity, namely lower lamina III and lamina V. Mangini and Pearlman (1980) and Lemmon
and Pearlman (1981) show, in mouse, that corticotectal units (which are located exclusively
in lamina V) have higher spontaneous firing rates than other classes of neurone.
Some studies also document the fact that pyramidal neurones in laminae II and III are
less able to sustain high firing rates in response to sensory stimuli than those in lamina V.
McKenna et al. (1984) made recordings from superficial laminae of the cat primary soma-
tosensory cortex. Although the evidence is confined to upper laminae, their comparison
with other studies suggests that many neurones in these laminae have response properties
different from those in deep laminae of the same column. In particular, in responding to
repeated stimuli, neurones in superficial laminae prefer infrequent repetition (<0.5/s
stimulation): repetition at slow rates leads to enhancement of the response, while repetition
at fast rates leads to a decrease of the response, and then to silencing of spontaneous activity.
Similarly, Chapin (1986), finds that neurones in laminae II, III and VI in S1 cortex of the rat
responded more slowly and weakly to sensory stimuli than those in lamina IV. (No specific
statement is made about those in lamina V). In addition, Swadlow (1990) finds for the
primary somatosensory cortex of awake rabbits that a lower proportion of neurones in
laminae II, III or VI than in lamina V could show a sustained response.
3. IMPLICATIONS FOR CELL ASSEMBLY IGNITION,

MAINTENANCE AND FORMATION
What are the implications of such findings for cell assembly function? Envisage the circum-
stances in which a sensory stimulus activates the cortex, and there is the possibility of an
already-existing assembly being ignited. What determines whether it will actually do so?
Computations of this were conducted by Wickens and Miller (1998) based on assumptions
which were empirically plausible for laminae II and III of the cortex, and referring to a small
(1 mm3) block of cortical tissue. The assumptions included those for cell assembly size,
connectivity ratios, neuronal firing rate, EPSP size, threshold, and hence convergence
ratios needed to produce suprathreshold excitation. It was shown that ignition of a localized
neural assembly from a small fraction of its members could be achieved over a wide range

426 Robert Miller
of plausible parameter values. For instance, with the probability of connection between
neighbouring pyramidal cells assigned the value of 0.1, assembly size of 1000, and con-
vergence ratio of 10 (i.e. near-maximally strengthened synapses), it required only about 50
of the 1000 members of the assembly to be activated by a stimulus, in order for excitation
to spread throughout the assembly (Wickens and Miller, 1998). It is quite plausible that
a stimulus could activate this number of neurones in the assembly. For networks with weaker
synapses, larger numbers of neurones would need to be activated, which may not be
achieved in practice.
However, when one considers the conditions for maintenance of assemblies (that is the
maintenance of sufficient synaptic strength in the connections between member neurones),
a severe problem arises. According to recent views on synaptic plasticity (Bienenstock
et al., 1982; Dudek and Bear, 1992), conjunction of incident EPSPs is not in itself adequate
to ensure synaptic strengthening. The conjunctions themselves must occur close enough in
time—that is within about 100 msec of each other—if strengthening is to occur. For
conjunctions further separated in time than 100 msec synaptic depression occurs. Using
plausible assumptions about the firing rate in maximally-activated neurones in laminae II
and III (obtained from Swadlow, 1989, 1994), the probability of interconjunction intervals
less than 100 msec was computed by Wickens and Miller (1998). It was found to be far too
low for synaptic strengthening to outweigh synaptic depression. To put it another way,
average firing rate of activated assembly neurones would need to be several times what is
actually observed, if the assembly is to maintain the synaptic strength in its interconnections.
Overall, if one considers laminae II and III in isolation one can conclude that pyramidal
neurones in these laminae may have activity levels high enough to permit cell assembly
ignition, but can neither maintain the integrity of cell assembly structure over time nor
form new assemblies from the “uncommitted” state of the neuropil.
4. LOCAL SPATIOTEMPORAL INTEGRATION OF CORTICAL ACTIVITY
4.1. Scheme for Interplay between Laminae II/III and Lamina V

The above calculations were based on the connectivity ratios and physiological properties
of pyramidal neurones in laminae II and III. Pyramidal neurones in lamina V receive
inputs from the local collaterals of lamina II/III pyramidal cells (by synaptic contacts
either on their apical dendrites as they penetrate the superficial laminae, or on their basal
dendrites with terminals from descending collaterals of superficial cells). In addition, lamina V
pyramidal cells regularly give local axon collaterals ascending to the superficial laminae,
where they are likely to make synaptic contact with pyramidal cells (discussed in Miller,
1996a). Thus, in morphological terms, lamina V pyramidal cells can be regarded as inter-
mediate stations in indirect pathways from one superficial pyramidal cell to another. We
can represent this as a series of triangular configurations, with single lamina V cells at the
apex of each triangle (see Figure 18.1). There are no quantitative data on the probability of
connection between superficial pyramidal cells and those in lamina V, although qualitative
studies (Burkhalter, 1989 [their Figures 4–6 and 9]) suggest they make up a substantial
proportion of all local connections derived from lamina II/III. Thus each lamina V pyramidal
cell presumably receives synapses from a variety of superficial cells, and also gives
ascending collaterals back to a number of superficial cells. Thus, it is likely that each

Figure 18.1. Triangular configurations of neurones, with “base” as a direct link between two neurones in
lamina II/III, and “apex” as a neurone in lamina V, receiving from one of the neurones in lamina II/III and
transmitting to the other. (A) One lamina V neurone can be the apex of many triangles based in lamina II/III.
(B) One pair of lamina II/III neurones can form the base for many triangle with different apices in lamina V.
lamina V pyramidal cell could in principle serve as an intermediate station between the
members of a large number of pairs of superficial cells, thus forming the apex for a large
number of such “interlaminar triangles” (see Figure 18.1A). In terms of anatomical connectiv-
ity, it is also likely that a single direct synaptic link between two lamina II/III cells could form
the “base” of a number of triangular configurations involving, as their apices, different
neurones in lamina V (see Figure 18.1B).
Furthermore, as noted above, lamina V pyramidal cells have on-going impulse frequency
one or two orders of magnitude higher than in pyramidal cells in laminae II and III. The
higher levels of activity in lamina V pyramidal cells, compared with those in other
laminae suggests that in the waking state the lamina V cells have membrane potentials

428 Robert Miller
held closer to threshold than those in other laminae. Swadlow (1992) has produced more
direct evidence that this is actually the case. Cells were identified by antidromic stimulation
from various distant sites, and, in addition, current pulses could be delivered by the
recording electrode in the vicinity of the neuronal soma. With such juxta-somal current
pulses, all neurones could be made to respond directly, at similar threshold current intensities,
regardless of the lamina, presumably as a result of direct activation of the neuronal mem-
brane. However, many neurones in lamina V also gave responses of a different type, with
longer latency and greater latency “jitter”, superimposed on the shock artefact. These
responses were interpreted as indirect responses, produced by synaptic activation of the
recorded cell rather than by direct somal activation. With near-threshold current pulses,
such responses were found in 20% of lamina V neurones, but extremely rarely in neurones
in lamina II, III and VI. With more intense current pulses (10 µA) 80% of lamina V cells
showed indirect responses, while only 5% of those in the other lamina did so. Clearly in the
situation of normal waking, lamina V cells are held much closer to threshold for synaptic
excitation than those in the superficial lamina.
It is also known that in urethane-anaesthetized animals the membrane potential of
cortical pyramidal cells is usually bistable, and can undergo a state transition from a
“down state” with membrane potentials in the region –65 to –70 mV to an “up” state with
membrane potentials between –50 and –55 mV (Metherate and Ashe, 1993). Cowan and
Wilson (1994) and Stern et al. (1997) also document a bistability for corticostriatal
neurones in lamina V in urethanized rats, between a “down state” and “up state”. Illustra-
tions show that, in the up-state, membrane potential fluctuates by less than 5 mV, and is
within 5 mV of threshold. Similar behaviour is illustrated by Inubushi et al., 1978 [their
Figure 3]), in unanaesthetized brain preparations, at a time when the EEG showed slow-
wave activity and spindle bursts. Scrutiny of figures in this and later papers by this group
(Ezure and Oshima, 1981a,b), reveals that electrographic arousal was accompanied by
excitation of neurones, with membrane potential undergoing depolarization of 5–10 mV,
and becoming stabilized within a few mV of spike threshold. At present it is uncertain
whether such bistability of membrane potential applies equally to pyramidal cells in all
cortical laminae, or shows laminar differences. The implication of the higher firing rate in
lamina V cells than in superficial pyramidal cells may be that the former more commonly
hold the “up” state than the latter. However, detailed information on laminar differences
in this feature are not available.
Given that lamina V cells are often held closer to firing threshold than those in laminae
II/III, one can regard the indirect pathway between two lamina II/III cells (via one or more
lamina V cells) as a more secure route than the direct pathway. Whereas convergence
from many co-active synapses is required to activate a neurone in lamina II or III, activa-
tion of a lamina V neurone should require only a low degree of convergence, or even just one
or two unitary synaptic activations. Since we are considering only local interactions
(defined above as a 1 mm3 block of cortical tissue), synaptic and conduction delays will
not be sufficient to produce temporal dispersion beyond the neuronal integration time. All
events are thus effectively instantaneous as far as integration within single neurones is
concerned. Given this, every time one lamina II/III neurone elicits an EPSP directly in
another neurone in the same laminae, it is likely that there will be a number of other
closely-coinciding EPSPs generated indirectly via lamina V neurones in the same recipient
neurone in lamina II/III. The direct synaptic connection by itself can only produce
subthreshold activation; but when combined with indirectly-generated EPSPs via several

interlaminar triangles, activation above threshold may occur. The indirect pathways can
then be seen as “priming” the direct pathway, adding synaptic activations to the recipient
neurone in lamina II/III to bring its membrane potential above threshold for firing (Figure
18.1B). There is actually some evidence for such a relationship: Hess et al. (1996), using
slice preparations, found that stimulation of horizontal pathways in lamina II/III would not
normally produce long term potentiation, unless the tonic GABA activity was eliminated
with a GABAa blocking agent (bicucculine). However, LTP could be induced in lamina II/
III neurones, without bicucculine, if vertical pathways ascending to the recording site
were tetanized at the same time as the horizontal ones in lamina II/III.
In this scheme, each lamina V cell could perform its priming function for a large
number of pairs of superficial cells (Figure 18.1A), and each pair of superficial cells
would be primed by a number of lamina V cells (Figure 18.1B). Each lamina V cell would
thus not be critical in defining the information content of links between lamina II/III cells,
this being specified by the direct link. Indeed, indirect disynaptic links between two
cortical pyramidal cells are inherently ambiguous in information content, for reasons
considered in section 7, dealing with the role of the hippocampus. Nevertheless, when the
superficial laminae operate in association with the more excitable lamina V, it may become
possible to overcome the problem defined above, arising from the relative inexcitability of
the superficial lamina. This may enable local cell assemblies (defined principally by local
lamina II/III connections) not only to ignite on appropriate occasions, but also to form
from the uncommitted neural network and maintain synaptic strength, given stimulation at
empirically-plausible levels.
4.2. Receptive Field Evidence for Interlaminar Interplay

The schema outlined above cannot yet be expressed in rigorous formal terms, because
a number of quantitative data are needed, such as the convergence ratio needed to activate
each class of cells and the probability of synaptic connections being made in either
direction between pyramidal cells in laminae II/III and lamina V. Such data are not yet
available. However, other empirical evidence provides support for this scheme, in the
absence of formal theoretical reasoning. If the above scheme of laminar interaction is correct,
one would expect that the pyramidal cells in lamina V involved in cell assembly operation
would have a facilitating role with respect to many pairs of superficial cells, and therefore
for many different assemblies. One would therefore expect lamina V pyramidal cells to
have lower information selectivity than those in superficial laminae. In many areas of cortex,
the precise implications of this are difficult to specify. However, in sensory areas, one would
predict that lamina V cells would have lower selectivity of sensory responsiveness than
those in laminae II and III.
From the rich evidence available on receptive field properties of sensory cortical
neurones, there is much evidence supporting this prediction. The paragraphs below give
some of the clearest examples. Most of the evidence comes from the primary visual cortex,
although there are several relevant papers from somatosensory areas. The best evidence
comes from lower mammals rather than from primate species. (It is becoming clear that
the primary sensory—especially visual—cortex in primates is rather specialized, not only
in comparison with other cortical areas, but in relation to sensory areas of lower species.)
The prediction that lamina V neurones have low selectivity for sensory features might be
capable of expansion beyond the examples given below, but all the details of sensory

430 Robert Miller
convergence upon neurones in lamina V cannot be specified without a better understand-

ing of which are the basic features extracted in the visual cortex, and which are derived by
combination.
In the primary visual cortex, neuronal receptive fields are generally larger (i.e. spatially
less specific) in lamina V than in other laminae. This has been found in rodents by
Mangini and Pearlman (1980), Lemmon and Pearlman (1981), Swadlow (1988), and
by Metin et al. (1988). In cats, Palmer and Rosenquist (1974) found that identified corti-
cotectal cells (which are located in lamina V) typically had large fields. Schiller et al.
(1976a) made a distinction between cells with spatially separate excitatory and inhibitory
subfields, and those where the excitatory and inhibitory areas were coextensive. The former
showed little laminar difference in field size or spontaneous activity, but the latter
included a portion with larger field sizes and higher spontaneous rates in laminae V/VI
than in laminae II-IV. Leventhal and Hirsch (1978) found that both classes of visual
cortical cells as defined by Schiller et al. (1976a) tended to include ones with larger
receptive fields in laminae V/VI than in laminae II-IV. Gilbert (1977) found, for “standard
complex” cells (that is, ones showing summation of response as stimulus length was
increased) cells in lamina V had a greater receptive field area than those in lamina II-IV,
while those in lamina VI had receptive field areas greater even than in lamina V. In
monkeys, however, the low spatial selectivity in lamina V is not found. Snodderly and Gur
(1995) found that the width of receptive fields for laminae V cells was small, comparable
to those of laminae II/III cells and lower than those of lamina IVa and IVc cells. Spontan-
eous activity in lamina V units was also as low as in units in laminae II/III, and much
lower than in IVa, or IVc.
Many studies show that units in lamina V of the visual cortex generally have a lower degree
of specificity for one or other eye (i.e. they more commonly have a binocular input), than
those in other laminae. This was found in rodents by Metin et al. (1988) and Swadlow
(1988), in cat by Palmer and Rosenquist (1974), Schiller et al. (1976b), Gilbert (1977), and
Ferster (1981), in the mink by LeVay et al. (1987), and in a noctural primate, the bush
baby, by DeBruyn et al. (1993). Leventhal and Hirsch (1978) found that the preponderance
of binocular cells in lamina V was restricted to those cells whose excitatory and inhibitory
fields were co-extensive, while Berman et al. (1982) failed to finding any excess of binocular
cells in lamina V.
Several studies show that orientation tuning is less selective for lamina V cells than for
those in other laminae. This was found in rodents by Metin et al. (1988) and Swadlow
(1988), although this was not found by Revishchin (1985). Mangini and Pearlman (1980)
found in mouse visual cortex that a non-oriented class of cells was located exclusively in
lamina V. Similar findings have been reported in cats by Leventhal and Hirsch (1978, for
cells with co-extensive excitatory and inhibitory fields), in the bush baby by DeBruyn
et al. (1993), but not by Schiller et al. (1976b), or Gilbert (1977) in cats.
Ferster (1981), in the cat, finds that lamina V cells in area 17 are almost all insensitive
to ocular disparity, whereas in other laminae (II-IV and VI) cells form a variety of classes
with differing disparity sensitivities. DeBruyn et al. (1993) finds that lamina V cells in the
bush baby have the least specific spatial frequency tuning of all layers. In these papers,
a variety of other parameters of visual responsiveness are reported to vary between laminae,
such as direction sensitivity, optimal spatial frequency selectivity, optimal temporal
frequency and temporal frequency cut-off, contrast sensitivity, peak velocity sensitivities
and high-velocity cut-off.

In the somatosensory areas of cortex, several studies find, as in the visual cortex, that
receptive fields are larger in lamina V than in other layers. In rodents, this was found by
Ito (1985), Chapin (1986) and Swadlow (1990, 1991, 1994), in cat by McKenna et al.
(1984), but the finding was not confirmed in monkeys (Sur et al., 1985). Swadlow (1990)
also found, in rabbits, that when units responding to deep and superficial cutaneous stimuli
were recorded in the same laminae in the same electrode penetration, the receptive fields of
units for “deep” and “superficial” sensory modalities were likely to be spatially congruent
for units in lamina II/III and VI, but could be far from congruent for units in lamina V.
While the exact interpretation of this is uncertain, the result suggests that the pyramidal
cells in lamina V sample a wider extent of the sensory surface than do those in other
laminae. Uhr and Chapin (1983) confirmed that lamina V cells in the barrel cortex have
extended vibrissal receptive fields, and investigated the interaction by which lamina V cells
came to have larger receptive fields than those in other laminae. In the barrel cortex of anaes-
thetized rats, intracortical stimulation at sites were identified which could produce activations
of recorded neurones across a distance of a few mm. When such stimulation sites were
destroyed with an electrolytic microlesion, the extended vibrissal receptive fields of lam-
ina V cells were reduced in size, losing input from the vibrissae represented in the region
destroyed. While this experiment does not positively identify inputs from superficial
laminae as providing the input, it does show that cortico-cortical inputs converging from
cortex some distance from the lamina V neurone provide it with its extended receptive field.
Despite the interpretation of some of the above findings being uncertain, there is consider-
able evidence supporting the prediction that lamina V cell have lower selectivity for sensory
parameters than those in superficial laminae. It is plausible to suggest that this laminar
difference reflects the fact that lamina V cells receive convergence from cortical cells in other
laminae with a variety of different selectivities, and so have themselves a low selectivity.
5. SPATIOTEMPORAL INTEGRATION OF ACTIVITY IN MORE

WIDESPREAD ASSEMBLIES
5.1. Constraints on Non-Local Operation of Cell Assemblies

The original intuition leading to the cell assembly hypothesis envisaged that the member
neurones of a cell assembly could be widely spread over the cortical mantle. For instance,
Hebb (1949) considered assemblies with member neurones in each of several visual areas
of the cortex. However, in terms of real networks of cortical nerve cells, integration of
neuronal activity, such as is proposed for cell assemblies, becomes more problematical
both in the spatial and temporal dimensions when that integration is non-local. The reasons
for this are easily explained. In the spatial dimension, the probability of connection between
two neurones is maximal –0.8—for nearest neighbours, falls to a value of 0.1 and below for
more distantly-related pairs of neurones within a 1 mm3 block (Braitenberg and Schüz, 1998;
Hellwig, 2000), and falls even further for more distantly located pairs of neurones. Thus
one neurone may be able to make direct synaptic contact with about 10% of neurones
within a distance of a few hundred microns (Braitenberg and Schüz, 1991, 1998). However,
if one considers an area of cortex of dimensions 10 × 10 mm, centred on a chosen neurone,
there are of the order of 10 million pyramidal cells within this area, giving an overall ratio
for direct connectivity of less than 1/1000.

432 Robert Miller
In the temporal dimension, for local interactions, axonal conduction time could be ignored,
because it was never likely to be more than the neuronal integration time. When assemblies
are distributed over and between cytoarchitectonic areas, even if morphological connec-
tivity is compatible with effective convergence, the probability of functionally-effective
convergence would fall due to temporal dispersion of signals. For instance, suppose that
five-fold convergence of afferent activity from a transmitting cortical locus was required to
activate a recipient neurone in another locus. If all afferents to the neurone in question had
conduction times less than 10 msec (an assumed value for the neuronal integration time), loss
of convergence due to temporal dispersion would be insignificant. However, for distances of
axonal conduction of 10 mm, conduction time could range from 3 to 33 msec (corresponding
to conduction velocities of 3 to 0.3 m/sec, a range of values corresponding to those found
in Swadlow’s studies) (see also Swadlow, 2000). Longer conduction times would be
expected for axonal trajectories longer than this (which certainly exist). This would lead to
a reduction of at least 3-fold (and more, for the longer trajectories) in the degree of con-
vergence occurring in a typical recipient neurone. Such reduction of amplification appears
to be a severe constraint on large-scale integration of cortical information processing.
Overall, despite the fact that there is a very rich array of long cortico-cortical connec-
tions, a neurone can only synapse with a tiny fraction of those located in other cortical
areas, and the possibility of suprathreshold convergence is also reduced by temporal
dispersion of signals. However, this constraint may be overcome if recent data and con-
ceptual developments about cortical connectivity and corticothalamic interaction are taken
into account. There appear to be mechanisms for “concentration” both in space and time,
of the influences of sufficient single units, to make suprathreshold activation more prob-
able than would be expected on the basis of random spatiotemporal connectivity alone.
5.2. Cortical Connectivity and Spatial Concentration of Neural Influences

In the spatial dimension, there is much evidence that long cortico-cortical connections
are not randomly distributed within the territory they innervate. Instead they are distributed
in patches with high local connection density, with intervening regions having few connec-
tions (e.g. Rockland and Lund, 1983; Amir et al., 1993; Levitt et al., 1993, 1994; review:
Malach, 1994). In a projection from one cortical area to another, connections arising in
adjacent loci may each have patchy distribution of terminals in the recipient area, without
overlap of the respective patches. It has been pointed out by Schüz (1994), that this
increases the probability of convergence for long-distance connectivity. This principle
applies not only within an individual cytoarchitectonic area, but also over much longer
distances, since a patchy distribution of connections is also found for such long con-
nections. In spatial terms (i.e. strictly anatomical aspects of connectivity), the cortex as
a whole is likely to be integrated better in this way, than with more uniform random
connectivity. However, the patchy distribution does mean that some associative links across
the cortex are not direct monosynaptic ones, but polysynaptic. This raises further problems
for global cortical integration, considered in section 7, on the role of the hippocampus.
In addition, in functional terms, the scheme of long cortico-cortical connections just
described may nevertheless still be insecure, especially if we consider the reduction of
temporal convergence due to temporal dispersion along long pathways. However, when
we consider circuits between cortex and thalamus additional mechanisms can be suggested
which may promote convergence in recipient cortical loci, not only in terms of spatial

aspects of connectivity, but also with respect to temporal aspects. Together, these mechan-
isms may permit long-distance connectivity to be functionally effective in building widely-
distributed cell assemblies.
5.3. The Role of Corticothalamic Interplay in Spatiotemporal Integration

on the Medium Scale
5.3.1. Evidence constraining ideas of corticothalamic interplay
The simplest, and most often-repeated account of the thalamus is that it is a relay station
on the way to the cerebral cortex. This account arises from the strong emphasis in past
decades on sensory processing in the cortex. However, it ignores several secure facts:
Morphologically identified synapses upon thalamic projection neurones are of two main
types, different in size and in other features. Quantitatively, the number of morphologically-
defined inputs to these neurones from the cortex (identified as small round synaptic
profiles) outnumber by far those ascending in sensory pathways from the brain stem, or
other systems of the brain (identified as larger synaptic profiles) (Tseng and Royce, 1986;
Sawyer et al., 1991; Wilson et al., 1984; Liu et al., 1995). Some nuclei of the thalamus
have no major “ascending” inputs, and so receive both types of synapse from cortico-
thalamic inputs (Mathers, 1972; Schwartz et al., 1991; Kuroda et al., 1992). In functional
terms, impulse activity in thalamic principle neurones is under the control of cortical activity,
as well that of pathways ascending from lower parts of the neuraxis. This statement is true
both for the “spontaneous activity” in thalamic neurones (Bures et al., 1963; Waller and
Feldman, 1967; Albe-Fessard et al., 1983; Kayama et al., 1984; Villa et al., 1991) and also
for activity induced in response to sensory stimuli, since sensory response properties of
single thalamic neurones change in subtle ways when the corresponding region of cortex is
inactivated or ablated (Schmielau and Singer, 1977; Chow et al., 1978; Vidyasagar and
Urbas, 1982; Molotchnikoff et al., 1984; Varela and Singer, 1987; Gulyas et al., 1990; Funke
and Eysel, 1992; Villa, 2000; Ghazanfar and Nicolelis, 2000).
Several clues are available to help define this interplay. First one should consider the
spatial distribution of cortico-thalamic and thalamo-cortical connections. It has traditionally
been believed that these two sets of connections have a strictly reciprocal distribution.
This assumption was taken to apply generally in a recent comprehensive analysis of
thalamo-cortical connections in cat (Scannell et al., 1999). However, this conclusion has
been derived from separate studies of the two pathways in different animals, and averaging
the resultant maps across animals (inevitably with low spatial resolution). Reciprocality
generally referred to a relatively large scale (cortical area to thalamic nucleus and vice
versa) rather than within subdivisions of individual areas and nuclei. In addition, precise
evaluation of the concept of reciprocality was not possible until it was recognized that
cortico-thalamic connections are of two types, with rather different synapses, laminae of
origin and local terminal distribution (Steriade et al., 1990). It is now becoming clear that
regions of termination of cortico-thalamic axons do not correspond exactly with regions of
origin of thalamo-cortical connections returning to the corresponding cortical region.
Winer and Larne (1987) injected axonal tracers into the auditory cortex of the rat, and thus
produced small regions of both anterogradely labelled terminals and retrogradely labelled
projection cell bodies within the medial geniculate nucleus. Regions of cell body and
terminal labelling were not entirely co-extensive, the cortico-thalamic terminal areas being

434 Robert Miller
generally more extensive than the zones of retrograde neuronal labelling. This principle
has more recently been shown to apply in cats, for a subnucleus of the VPL thalamic complex
and for the pulvinar, using a cocktail cortical injection including both an anterograde and
a retrograde tracer (Darian-Smith et al., 1999). Since both sets of connections were
labelled in the same animal, it was possible to show that, on a fine scale, there was sub-
stantial non-reciprocality of connections, especially for the fine calibre cortico-thalamic
connections arising in cortical lamina VI. Sherman and Guillery (1996) also raise the
possibility that thalamic nuclei may be important in determining transmission from one
cortical area to another, and rather similar possibilities are implied by Deschenes et al. (1998).
Non-reciprocality of connections at the cortical side of cortico-thalamic loops has at
present only fragmentary support. However, an electrophysiological study of cortico-thalamic
projections to the lateral geniculate nucleus of the cat by Lindstrom and Wrobel (1990)
included, as an “unpublished observation” the remark that “individual principal cells receive
convergent excitation from cortico-geniculate neurones in a larger area of cortex than the
termination zone of their axons.”
While the generality of such findings is uncertain at present, they suggest that each
thalamic nucleus is not so much a relay station on the way to the cortex, but a “junction”
where signals originating in one cortical region can be redistributed not only to that region,
but also to other regions. The degree of divergence possible in such a scheme can only
be guessed at this stage; but since non-reciprocality seems to apply to the connections of the
pulvinar, which has widespread, but diffuse projections to the cortical mantle, it is probable
that some thalamic nuclei could re-route signals in a manner permitting wide divergence
across the cortex. In addition, from the wide-ranging recent analysis of thalamo-cortical
connectivity of Scannell et al. (1999), it is clear that most thalamic nuclei project to several
(or many cortical areas). If thalamic neurones projecting to different areas were closely
intermingled within a nucleus, the possibility of re-routing cortico-thalamic signals to other
areas of cortex could be very great.
The second clue to thalamo-cortical interplay is the fact that one class of cortico-
thalamic axonal projections—those originating in lamina VI—have quite slow conduction
velocities (Harvey, 1980; Swadlow and Weyand, 1981, 1987; Ferster and Lindström,
1983; Swadlow, 1988, 1990, 1991, 1994). This conclusion has been corroborated in cross-
correlation studies of interlinked pairs of neurones in thalamic and cortex (Tsumoto et al.,
1978; Johnson and Alloway, 1994). As a result, conduction time from cortex to thalamus
in single axons can be very long, (up to several tens of msec in rabbits, according to
Swadlow) and can span quite a wide range of delays, across a population of such axons,
from just a few msec up to as much as 50 msec. Consequently, these cortico-thalamic axons
are likely to impose a degree of temporal dispersion on a signal originating in a cortical
locus matched only by that in the long cortico-cortical connections.
The third clue to thalamo-cortical interplay is that, during waking, thalamic projection
cells appear to have membrane potentials poised close below threshold. The evidence for
this is admittedly somewhat indirect because of the technical and ethical difficulties of
recording intracellularly from structures deep in the brain in waking animals. Coenen and
Vendrik (1972) made quasi-intracellular records from the visual thalamus of paralysed
cats during transitions from drowsiness to waking. The ratio of spikes to EPSPs was as
high as 0.9–1.0 during waking, implying membrane potentials close to threshold, but was
substantially lower during sleep. Hirsch et al. (1983) recorded intracellularly from cat visual
thalamus during rapid eye movement sleep (a state akin, in many ways, to waking). They

illustrate an impaled neurone with stable membrane potential upon which are superimposed
small irregularly-timed EPSPs, almost all of which give rise to single action potentials.
Intracellular records from thalamic nucleus VL have been published more recently by
Contreras and Steriade (1997), which appear similar to those of Hirsch et al. (1983),
although recorded from cats anaesthetized with zylazine/ketamine combinations. Such
records were obtained at times when the electrocorticogram displayed desynchronization
similar to that during waking. Single projection neurones were depolarized to between 55
and 60 mV, and showed irregular non-bursty, single spikes, each related to small EPSPs.
Sometimes such spikes were locked to fast (gamma range) EEG rhythms (Steriade, 1997).
Deepening of anaesthesia led to increased negativity in the membrane potential of single
cells, and to rhythmic activity in them, and in the electrocorticogram. The authors write “the
high sensitivity of the EEG pattern to the fluctuations in the degree of cellular correlation
and rhythmic behaviour leads us to believe that essentially the same basic principle underlies
intercellular relations during the changes in the EEG that accompany shifts in natural
states of vigilance.”
5.3.2. Hypothesis of Medium-scale Spatiotemporal Integration by

Corticothalamic Interplay
The implication seems to be that thalamic projection neurones, like cortical pyramidal
cells of lamina V, have periods when membrane potential is stabilized just below threshold,
so that only one incident EPSP (or only a very few coincident ones) are required to
produce an action potential. Given that these projection cells are junction points in two-
way cortico-thalamic relays, the relation between them and active cortical assemblies
appears in some ways similar to that between lamina V cortical cells and lamina II/III
cells (which, it was suggested, are the primary substrate for cell assembly formation) (see
Figure 18.2A, B). By analogy with the latter arrangement (and with preceding arguments),
one can suggest that the indirect cortico-thalamo-cortical pathway is more secure than the
direct cortico-cortical pathway between distant cortical loci; and thus two-way cortico-
thalamic interplay can “prime” activity in direct, albeit long, cortico-cortical connections
(see Miller, 1996b).
There are however some important differences between these two sets of inter-neuronal
relationship, arising from the differences in scales of integration, in both the spatial and
the temporal dimension. In the spatial dimension, use of the indirect (cortico-thalamo-
cortical) pathway to ensure security of transmission in the direct (long cortico-cortical)
pathway depends on the condition that indirect thalamo-cortical pathways activated by
a transmitting cortical locus have sufficiently dense projections to the same cortical loci as
to receive direct cortical connections from the transmitting cortical locus. Evidence com-
patible with this is that thalamo-cortical projections, from at least some thalamic nuclei,
have a patchy distribution in the cortical target region (e.g. Giguere and Goldman-Rakic,
1988). More incisive evidence comes from Baleydier and Maugiere (1987) for relationships
between parietal cortical area 7a, cingulate cortical area 23, and medial pulvinar. Within
the patchy distribution of long connections between the two cortical areas, distant loci
which were reciprocally connected also had thalamic projections from common thalamic
loci in the pulvinar; whereas non-connected cortical loci in the same areas received thalamic
projections from separated loci in the pulvinar. In itself, this finding does not show that
a cortical locus which projects directly to another distant cortical locus also sends

436 Robert Miller
Figure 18.2. Schemes for thalamocortical interplay. (A) Two cortical loci (each containing sets of triangular
configurations, as in Figure 18.1) are connected by long connections (dashed lines). Such connections are
supported by equivalent indirect connections via a thalamic nucleus which receives signals from one area and
transmits them to the other. (B) Larger version of Figure 18.2A, in which assemblies—spread across several
cortical areas—are integrated with the support of several thalamic nuclei. Note that in this scheme each
thalamic nucleus has reciprocal connections with one area, but also connects to another area, thus allowing
it to serve as a junction point between two areas.
supporting connections to the same locus indirectly via the thalamus. This is not explicitly
proven because the thalamic loci were identified only by simultaneous retrograde transport
from the two cortical areas, not by combined retrograde/anterograde injections. However,
from Winer and Larne (1987) and from Darian-Smith et al. (1999), it is clear that the

distribution of fine cortico-thalamic terminals is quite extensive and includes both the
thalamic regions which give reciprocal thalamo-cortical connections and other surround-
ing regions which do not. Hence, in the experiments of Baleydier and Maugiere (1987), it
is likely that the thalamic loci which project in common to the two directly-interconnected
cortical loci also receive connections from each of these loci, and thus constitute junction
points for indirect relay from one cortical locus to the other. Thus, one has some tentative
evidence that distributed subsystems exist, in which dense direct cortico-cortical projections
are matched by a correspondingly dense indirect projections via the thalamus.
This seems to imply that the density of the combined connectivity in a recipient cortical
locus received directly from a transmitting locus and indirectly from such a locus via the
thalamus may be much higher than would be expected in a randomly-connnected cortical
network. It may even approach the connection density found locally within each cortical
locus. If so, the main problem becomes that of the temporal dispersion of signals transmitted
from one cortical locus to a distant one. However, since cortico-thalamic connections
appear to have the similar long conduction times, and the similar dispersion of conduction
times across a populations of axons as the long cortico-cortical connections, it is plausible
to suggest that any cortical neurone which receives indirect, and delayed synaptic influences
from a transmitting cortical locus is likely to receive indirect synaptic influences via the
thalamus, some of which have delays matching those of the direct influences. If so, the direct
and indirect synaptic influences may coincide within the same neuronal integration times,
and can therefore summate. Hence, despite the temporal dispersion of signals due to the
distance between the two cortical loci, the indirect pathway can still act to support or prime
the direct cortico-cortical pathway.
Recent evidence, while not directly proving such a relationship, is readily explained by
this hypothesis, and is otherwise rather difficult to explain. Simultaneous recording of spike
trains from several cortical neurones at the same time shows precise temporal structure in
the time relations between spikes in different neurones. In early demonstrations of this (e.g.
Abeles, 1981), the different neurones were recorded very close to each other (from the same
electrode, or from separate electrodes on a single electrode stem). Therefore, the temporal
structure so demonstrated was explained in terms of a model relying on strictly local inter-
actions (Abeles, 1991). However, more recent work has demonstrated temporal structure
across multiple spike trains recorded further apart in the cortex, from spatially-separate
electrodes (Villa, 2000). Thus temporal structure of cortical neuronal firing is not just a local
phenomenon. Even more striking is the demonstration of temporal structure across multiple
spike trains recorded from thalamic projection neurones (in the same, or sometimes in
different nuclei) (Villa and Abeles, 1990; Villa, 2000). This result is particularly dramatic,
because thalamic projection neurones have no direct collateral connections (Steriade et al.,
1990), and so cannot be influencing each other directly. The most likely explanation of these
results is that the temporal structure is determined by thalamo-cortico-thalamic connections,
with the precision and delays inherent in the temporal structure being determined mainly
by conduction delays in cortico-thalamic and/or cortico-cortical axonal connections.
Once again the detailed theory of the interaction proposed cannot be given, for lack of
quantitative data on connectivity ratios, convergence ratios, etc. It is possible that the scheme
outlined above, if taken in isolation, will still fail for lack of sufficient signal amplification.
However, there are yet other tiers to be added to the interactions described, at a yet larger
scale, which may confer greater operational robustness on the patterns of interlaminar and
cortico-thalamic interplay described above.

438 Robert Miller
Given the analogous relations outlined above, on the one hand between neurones in lamina V
and those in lamina II/III, and on the other between cortical assemblies and thalamic
projection neurones, there are implications for mechanisms of widespread stability control
in the cortex. The essential neurones in both such arrangements—the lamina V neurones,
and the thalamic projection neurones—become critical points at which global control of
brain dynamics can be exerted. Quite apart from detailed informational inputs to these
neurones (e.g. sensory inputs) excitatory controlling signals exerted upon such neurones
can be expected to have powerful global influences on activity levels and stability of neural
activity in cortical networks.
6. CORTICO-BASAL GANGLIONIC LOOPS: SELECTIVE ATTENTION,

AND SEQUENCING WITHOUT EXACT TIMING
So far this chapter has presented ideas about the function of cell assemblies which represent
concepts and percepts either containing no temporal structure (being effectively synchronous),
or having a detailed temporal structure which can be coded in a brain, with an exact temporal
metric determined by axonal conduction time. There is a third alternative to consider—
that the brain represents relationships between successive concepts or percepts in terms of
sequence, but not of exact timing. The environment contains many examples of such sequen-
tial, but inexactly-timed relationships, which undoubtedly the brain can come to represent.
The relationship between lightning and thunder is a good example of sequence without
accurate timing. In mammalian psychology, lightning creates an expectation of thunder,
without specifying the time of the thunder. Likewise, in many examples of instrumental
behaviour, the relationship between an emitted piece of behaviour and the perception of its
consequences is not precisely timed, although it is a predictable sequence. How are such
sequential relationships represented in the brain?
Several sorts of neural activity are known which code for the expectation or anticipation
of an event which is forthcoming a short time ahead (up to a few seconds). The “contingent
negative variation” (CNV: Walter et al., 1964) is a potential recorded from the scalp
(especially in frontal regions) which grows in magnitude during the interval between signals
in typical “delay” tasks. The “Bereitschaftspotential” or Readiness Potential (RP: Kornhuber
and Deecke, 1965; Deecke et al., 1969) is a similar negative potential change which
appears in frontal regions in anticipation of spontaneous (untriggered) voluntary move-
ments, and can also be seen in anticipation of well-attended stimuli which require no
response. In typical delayed response tasks, single units can maintain firing throughout the
delay between a warning (or informational) signal, and the imperative signal a few seconds
later which triggers an actual response (Fuster, 1980; Komatsu, 1982; Joseph and Barone,
1987; Okano and Tanji, 1987). Such maintained firing during a delay is probably the
neural event at the single unit level whose signature on the mass level is detected as the
CNV or RP. In such examples, neural activity represents relationships of sequence rather
than accurate timing. (In fact in typical delayed response tasks, training is accomplished
by starting with a short delay, and then moving progressively to longer delays, for which
successful task performance would be impossible without prior training on the easier
short-delay version of the task).
The fact that such neural activity is maintained during a delay, when there is no explicit
stimulus which can directly sustain the activity suggests that some sort of recursive loop,

involving positive feedback is involved. The predominant location of anticipatory neural

activity in frontal regions gives an indication that the circuitry of the basal ganglia might
be involved, since the cortical connections of these structures are established mainly with
anterior cortical regions. It is well known that the basal ganglia are involved, with the
anterior regions of cortex, in complex arrays of connectional loops (Alexander et al.,
1986; Joel and Weiner, 2000). The simplest of these loops involve a cortico-striato-
pallido-thalamo-cortical circuit, but there are others involving in addition the subthalamus
and pars reticulata of the substantia nigra. The basal ganglia are also well known to be
involved in aspects of motor control, and in reward-mediated learning. It is appropriate
that the latter function is dealt with in circuitry more generally concerned with expectation
of events occurring in sequence, since (as mentioned above) the relation between emitted
behaviour and subsequent rewarding consequences is an important variety of such patterns
of events. Pharmacological and clinical evidence, especially that obtained from parkinsonian
patients supports the view that the basal ganglia are critical in generating anticipatory
activity in the cortex. For instance, the RP and CNV are reduced in amplitude in Parkinson’s
disease (Deecke et al., 1977; Simpson and Khurabait, 1986; Dick et al., 1987, 1989); and
their amplitude is restored towards normal during L-DOPA treatment (Amabile et al.,
1986; Dick et al., 1987). It is likely that the repetition of contingencies of sequence which
give rise to expectation are acquired by synaptic modification in the basal ganglia. Probably
dopamine-mediated strengthening of critical synapses in the striatum, which is thought to
underlie reward-mediated learning (Miller, 1981; Wickens, 1993; Wickens et al., 1996),
also allows the formation of functionally-effective loops of connections, in which activity
can circulate, to maintain activity in cortical and other parts of the loops, during delay
tasks. One possibility which is still sub judice, is that, in striatal neurones, neural activity
which leads to behavioural output leaves its trace, as a “state of readiness” for a short
period after the activity itself has subsided (Miller, 1981; Wickens, 1990). According to
this hypothesis, dopaminergic reinforcement occurring while the state of readiness still
endures can achieve strengthening of the synapses which led to the neural activity. With
such a scheme, the slightly-delayed consequences of behaviour become linked with the
initial stimulus which provoked the piece of behaviour (and corresponding neural activ-
ity), and thus the latter can set up anticipation of the former.
Two features of the circuitry by which the basal ganglia influence the cortex deserve
particular attention. First, the pathway through striatum, pallidum (or pars reticulata of
substantia nigra) to thalamus involved two inhibitory synapses in sequence (Carpenter,
1981) producing a net excitatory (or strictly a disinhibitory) effect (Deniau and Chevalier,
1985) (see Figure 18.3). Thus the pathway from cortex through the basal ganglia to the
thalamus and back to the cortex is, as expected, a positive feedback loop, but, significantly
this is achieved not by arranging that all links be excitatory, but by having an even number
(2) of inhibitory links. Therefore, the signal transmitted from striatum to the cortex which
activates behaviour (or other hidden cognitive processes) consists of release from tonic
neural inhibition, rather than actual excitation. Secondly, there are more complex components
of this circuitry involving sequences of three inhibitory synapses. This might involve
either an additional inhibitory link within the striatum, via collaterals of its inhibitory
principal neurones, or it might involve an additional inhibitory link in the subthalamus
(Carpenter, 1981). (The details of operation in this circuitry are not fully understood.)
These relationships have two larger-scale implications: the fact that the final effect on the
thalamic projection neurones (the critical “switch” for cell assemblies) is inhibition rather

440 Robert Miller
Figure 18.3. Depiction of interplay between neocortex, basal ganglia and thalamus. The cortex (above)
contains a number of active loci, symbolized as triangular configurations (see Figure 18.1) and these are in two-
way interaction with thalamic nuclei (thal: lower right). Basal ganglia (including striatum [str], pallidum [pal]
and substantia nigra [sn]) are depicted in lower left. Inhibitory links in the basal ganglia are shown as filled
circles, other connections (excitatory) depicted as arrow heads. The main positive-feedback circuit through the
basal ganglia is shown as bold connections.
than excitation suggests that the combined cell assembly activity of cortex and thalamus
is not limited by inadequate global levels of neural activity, but (sometimes at least) by
a tendency to overactivity. It therefore needs on-going restraint, at least in its anterior
portions, with periodic release from restraint when executive decisions are made. Secondly,
the fact that there can be both positive feedback loops (with 2 inhibitory links) and negative
feedback loops (with 3 inhibitory links) allows selective control over cortical activity—with
enhancement of one focus of activity and inhibition of others. In itself, this could underlie
(in part) the psychological fact of selective attention—amplification of activity in one
assembly, and suppression of activity in the intermingled neurones (see discussion in
Miller and Wickens, 1991). More germane to the present chapter, these facts have implications
for the dynamics of the combination of cortex and thalamus, upon which the basal ganglia
exert their influence. In section 3 of this chapter, it was suggested that the cortical laminae
which are most essential for cell assembly function are constrained by inadequate ampli-
fication of neural activity. However, the facts just mentioned suggest that the additional
mechanisms considered (interlaminar interplay, cortico-thalamic interplay, and global
control mechanisms exerted at critical points of each set of circuits) may already have
overcome this problem. As a result, cortex and thalamus together exhibit more versatile

dynamics, where there is a large region of “state-space” which is relatively stable, without
uncontrolled departure from the stable region—either uncontrolled excesses or cessation of
activity. Within this region, bidirectional control of global activity levels is both necessary
and possible, to ensure efficient information processing. Sometime this involves activation
of specific cell assemblies, but at other times (for instance, control of thalamus and cortex
via the basal ganglia) it involves inhibitory control.
7. ROLE OF HIPPOCAMPUS
So far we have considered the forebrain with the cerebral neocortex in interaction with the
thalamus and the basal ganglia, but without the hippocampus. On the basis of this model,
we can account for cell assembly operation despite the fact that the principal cortical laminae
in which assemblies are housed have activity levels too low to be effective by themselves.
We can account (plausibly, if not conclusively) for cell assembly operation in the spatial
sense, despite constraints on connection probability which would apply in a randomly
connected network; and we can account for assembly operation in the temporal sense
despite constraints set by likely temporal dispersion of signals in long axonal projections.
We have a forebrain which can represent associations in space, and in exactly-measured
time for intervals up to a few hundred msec; and this model of the forebrain can also represent
relations of sequence, including those in reward-mediated learning, which occur, without
an exact temporal metric, over intervals of time rather longer than are possible with exact
temporal measuring. We appear to have modelled (in informal, qualitative terms, if not in
formal quantitative terms) a brain with a profile of psychological capabilities in which any
stimulus configuration has the potential to elicit any response pattern, and the animal can
acquire the programs for this, given appropriate contingencies of reinforcement of spon-
taneously emitted fragments of behaviour performed in the corresponding sensory circum-
stances.
The profile of psychological capabilities, we have modelled, seems in some ways sim-
ilar to that described in the monograph of O’Keefe and Nadel (1978) for behaviour of ani-
mals (usually rats) in which the hippocampus has been destroyed or damaged on both
sides. The common theme in many of the features of these animals is that behaviour is dom-
inated by immediately-present local stimuli as opposed to aspects of the situation derived
by analysis of it over a longer time period or from a wider sensory perspective (including
aspects such as familiarity/novelty of a stimulus, or the place of its presentation rather
than its intrinsic nature). In addition, given constant stimuli, behaviour tends to consist of
stereotyped units, with loss of the normal internally-generated variability. For instance, in
their review the following conclusions are reached about animals with hippocampal
damage: in discrimination learning there is a tendency for animals to perseverate, or, in
spatial discriminations, for responses to be determined by the item or cue presented, regard-
less of its place. Maze learning tends to rely on an inefficient strategy, based on learning
response sequences (e.g. LRL sequences of turning) rather than on the more efficient
strategy of guidance by place. In reinforcement-mediated learning, reinforcing effects are
not fundamentally abnormal. However, in Skinnerian operant conditioning, response
patterns for lever pressing are more predictable (less variable), but more sensitive to
abrupt shifts in reward contingencies. In “differential reinforcement of low rates” schedule
(where an animal is not reinforced if it responds too frequently), hippocampus-lesioned

442 Robert Miller
rats respond at lower than normal rates, except when the end of each delay is signalled by
an explicit cue. Likewise alternation between two levers cannot easily be learned as an
overall pattern, but can be performed well when cues are used at each step as guidance.
There are problems with reversal of spatial discrimination (due to perseveration in
responses) but not of nonspatial discrimination (i.e. that based on a stimulus itself rather
than its position). There is, overall, an integrity of responses to threatening items, but not
to threatening places. There is an excess of nonexploratory stimulus-elicited motor activity,
which does not habituate over time. There is little true exploration of novel items in the
environment, and apparently little awareness of novelty, and with these, there is a reduction
in the duration of interference of ongoing behaviour by novel stimuli distracting from the
main task (defined in stimulus–response terms). Behaviour may be characterised by micro-
stereotypes, rather than true exploration. In two-choice mazes there is a tendency to reiterate
whichever response initially occurs more commonly, rather than alternation of responses.
The animal without either of its hippocampi resembles to some extent the brain we have
modelled thus far, and is also similar to the simplified profile for mammalian behavioural
considered earlier this century under the heading of “stimulus-response theory” (e.g. Hull,
1943). However, in that early work it became clear, after critical examination, that stimulus-
response theory was inadequate as a full description of mammalian behaviour (Tolman,
1932; Miller, 1991). Stimuli do not dictate responses unconditionally, unless the overall
environmental and motivational context has also been specified. Potentially, any stimulus
can call forth a wide variety of responses, and any response can be called forth by a wide
variety of stimuli. Many overlapping stimulus–response relationships can coexist as part
of an adult animal’s behavioural repertoire, and which one is utilized on a particular
occasion depends critically on the overall context in which the animal finds itself.
But what is meant by the word “context” in the above paragraph? One meaning of “context”
is a configuration of information from the environment (and the animal’s relation to it),
which is essentially a global aspect of the animal and its environment. However, when we
look at the cerebral neocortex and related brain structures we find that “context” can come to
have a quite separate (but parallel) definition, referring, again globally, to aspects of forebrain
dynamics, and their relation to information processing.
To gain this perspective, we must remind ourselves of the connectional nature of the
neocortex, which is the primary repository of the behavioural programs upon which an
animal can call. Every pyramidal neurone has of the order of 5000 input and output
connections in mouse, (and probably far more in primates, according to Braitenberg and
Schüz, 1998) and so is potentially linked to several thousand other cortical pyramidal cells
on both its afferent and its efferent side. Given this, pathways for signal transmission
through the cortex have a very high degree of inherent ambiguity once those pathways
have traversed more than a single synaptic relay. As an example, consider Figure 18.4A.
Level “A” can be thought of as a set of neurones in the cortex which is primarily responsible
for receiving signals in the cortex. Level “C” is related disynaptically to level “A”, all path-
ways between the two levels passing through the neurone at level “B”. Level “C” could
thus be considered as a set of output neurones. (Specifically neurones at level “A” may be
assumed to receive sensory signals from the outside world, those at level “C” to be those
controlling behaviour, although such specific assumptions are not necessary in a more
general account.) Figure 18.4A can be thought of as an example of the long connections
possible in the cortex: a transmitting cortical locus (A) projects to distant neurones in
a directly-linked “patch” of terminal distribution (B), and this in turn projects to a locus

Figure 18.4. Cortico-hippocampal interplay: (A) Microarchitecture of the role of the hippocampus in the neocortex.
The neurone “B” is a junction point, receiving several inputs (“A1–4”) and transmitting to several outputs (“C1–4”).
All neurones require multiple convergence before they will fire; and they receive many other synapses than those
shown, which are the ones critical for defining information processing. These other inputs include ones conveying
theta-rhythmic activity from the hippocampus. The ambiguous role of this network in signal transmission can be
resolved if specific combinations of input and output neurones can be selected by matching of the active phase in their
respective theta rhythms imposed by the hippocampus. This A1 and C4 have connections from a theta rhythm with
period 125 msec, and the disynaptic connection from one to the other, via B imposes a delay of 62.5 msec, which
matches the difference in theta phase in the two loci. (B) (reproduced from Psychobiology, vol. 17, pp. 115–128, with
permission of the Psychonomic Society Inc.) Global scheme for self-organization of phase-locked loops between
hippocampus and neocortex. Master oscillator (hippocampus) has afferent and efferent links with various neocortical
loci, all of which include lines with a range of delays. (Conduction time [ms] for each axonal link shown by numbers
adjacent to each axon.) Selected loops (solid lines) are selected to have total loop time of 150 msec, the same as the
concurrent theta period. Different neocortical loci have different phase relations to the hippocampal rhythm.

444 Robert Miller
near to “B” which does not itself receive direct connections from “A”. In such a simple
scheme, contingencies of reinforcement might be directed, essentially, at relations between
input and output, that is between neurones active in levels “A” and “C”. Under different
contextual cirumstances, secure transmission may be needed between different combina-
tions of specific neurones in “A” and “C”. However, given the possibility of great convergence
and divergence in the intermediate neurone “B”, there is in this model (in anatomical terms
at least), no definite, unambiguous pathway between any neurone in “A” and any in “C”.
This problem has been considered in extenso in a previous work (Miller, 1991). The
solution proposed involves interplay between the cerebral neocortex and the hippo-
campus. Such interplay, it is suggested, provides a mechanism by which a wide range of
specific multisynaptic “subnetwork” pathways through the cortex can be activated selec-
tively, each appropriate to the overall context which obtains. A critical feature of this
hypothetical interaction is the theta rhythm. This rhythm is the most regular of all electro-
graphic rhythms, and is generated in the hippocampus, with the medial septal nucleus
acting as a pacemaker under most circumstances. Its frequency (in rats) ranges from 5–12 Hz,
or, equivalently, has oscillation periods ranging from about 80–200 msec. Another critical
feature of the relation between neocortex and hippocampus is a rich array of reciprocal
axonal connections, linking the hippocampus to the neocortex, usually multisynaptically,
many regions of neocortex being involved, with preference for the associational areas
rather than the primary receiving areas. While not explicitly proven, it is likely that these
reciprocal connections are of fine calibre, and therefore act as slow-conducting delay
lines, including a wide range of delays between the hippocampus and any neocortical
locus, as is known to be the case for other long cortico-cortical connections.
Given this, it is proposed that neural activity underlying the theta rhythm can occur not
only in the hippocampus, but can impinge on the neocortex, which may then become
entrained in activity at the theta frequency. Such entrainment depends on the conjecture
that multisynaptic loops of connections, starting in the hippocampus, relaying in neocortical
loci, and returning to the hippocampus can have total “round-trip” conduction times which
match the period of the concurrent theta rhythm. Such connectional loops, it is envisaged,
are selected from a much larger repertoire of possible connectional loops (with a variety of
conduction times) by the requirement that only when “round-trip” conduction time
matches the theta period can synaptic strengthening occur (i.e. in the next excitatory
phase of the theta rhythm in the hippocampus) (see Figure 18.4B). In different external
contexts, different sets of cortical neurones may be set into activity, thus forming the
points of “reflection” in the neocortex for cortico-hippocampal loops. Diverse sets of loops,
each matched to its external context, are possible, even if the theta frequency is identical in
the different contexts. However, frequency may also vary sometimes, from one context to
another (as depicted in Figure 18.4A, different oscillation periods are transmitted to
“A4/C1” and “A1/C4”). Thus in each situational context, different sets of activated neurones,
firing in precise temporal relationship to each other, provide a cerebral representation of the
context. Referring to Figure 18.4A, one such context may (for example) increase the activa-
tion in neurones A1 and C4 at the activated phase of a 5 Hz theta rhythm, another in A4
and C1 at the activated phase of an 8 Hz theta rhythm, and so on. Thus different combina-
tions of A and C neurones are brought closer to firing threshold in different contexts, with
the result that different pathways between neurones A and C can be facilitated under dif-
ferent circumstances. Overall, because of the role of the hippocampus in providing repeated
signals to the neocortex, phase-locked to the underlying theta rhythm, many large-scale

spatiotemporal regimes of activity are set up in the neocortex, each specifying a different
context. These define a wide variety of exact operational modes for “interpreting” cortical
neural activity in a neocortical network which is certainly, in itself, multifunctional, and
uncommitted, and thus highly ambiguous with regard to the modes which might obtain at
any one time. Each of the operational modes is a global pattern of organization of activity,
pervading the whole of the neocortex, and therefore, indirectly, the whole of the forebrain.
Since periods of regular theta rhythm may last several (even many) seconds, the entrain-
ment of the neocortex in rhythmic activity can define the operational mode for similar
lengths of time. Such entrainment thus constitutes another aspect of the multifaceted
psychological concept of “selective attention”.
From animal experiments, it is clear that the times when entrainment of the neocortex is
most vigorous is not when an animal is performing firmly-established learned behaviour,
but during the phase of most active learning (e.g. Adey et al., 1960; Grastyan et al., 1966;
review: Miller, 1991). Thus it appears that, for overlearned tasks, the neocortex by itself
does not need to resonate with the hippocampus at the theta frequency: this may indicate
that, in this situation, the neocortex can reconstruct the necessary spatiotemporal regimes
without the theta rhythm to “synchronize” it, just as a well-rehearsed orchestra can play
without its conductor, although the conductor is needed at rehearsals. However, there is an
alternative interpretation: many tasks, learned by extensive repetition of stimulus–response-
reinforcement combinations, can be acquired by animals without hippocampi. These are
sometimes distinguished as “habits” rather than “memories”, the form of learning being
referred to as “procedural” rather than “declarative”. In this circumstance, the program is
presumably so strongly stamped into the cortical network, that the subtleties of an exact
spatiotemporal regime set up by the hippocampus are unnecessary. It is conceivable that,
for other tasks, finely-tuned cortical dynamics supervised by the hippocampus are essential
at an early stage of learning but not later when performance becomes an ingrained habit.
The respective importance of these two interpretations is at present difficult to assess
either from neurodynamic theory or from experiment.
This theory of cortico-hippocampal interplay was developed from evidence on the
theta rhythm obtained from animals. Comparable evidence from humans is fragmentary,
largely because the recording of activity from hippocampus, which is a deep structure, is
impossible with scalp electrodes, and is seldom attempted in humans with depth elec-
trodes. In the neocortex, an obvious theta rhythm (such as could be recognized in single
electrographic traces recorded from the scalp) is not usually detectable, even when the
hippocampus shows a vigorous theta rhythm. This failure is probably due to the fact that
scalp electrodes average activity across many cortical loci. Thus the different loci which
are points of reflection of an on-going theta rhythm would not be identifiable, due to
averaging of phase-dispersed rhythms from different loci. Despite these difficulties, a number
of recent electroencephalographic studies in humans, using power spectral analyses of
the EEG, have demonstrated that theta activity in the neocortex occurs most abundantly
in behavioural and cognitive epochs when (from animal experiments) entrainment at the
theta frequency between hippocampus and neocortex would be expected. Klimesch et al.
(1997) report that EEG theta power increases during an episodic memory task only for
words which are subsequently remembered correctly. Yordanova and Kolev (1998) show
phase-locking to stimuli of theta activity recorded from scalp electrodes as single traces,
during an auditory oddball task. Doppelmayr et al. (2000) showed that such phase locking
occurs best in subjects with good episodic memory performance. All these findings would

446 Robert Miller
be predicted from the theory of cortico-hippocampal interplay, but are not specific to that
theory.
An alternative view of cortico-hippocampal interplay, compatible with the above data,
was put forward by Bibbig et al., 1995 (see also Wennekers and Palm, 2000), on the basis
of simulations. In this simulation, regular phases of excitation transmitted from the
hippocampus provided “a source of coherent input to the involved sub-assemblies that will,
in general, reside in different and perhaps only weakly-connected cortical areas.” This
“increases the probability for synchronous firing events in and between those areas, and
thereby the chance for the Hebbian strengthening of synapses between the sub-assemblies.”
However, in their simulations, theta rhythms transmitted to the neocortex from the hippo-
campus do not set up any activity in the neocortex entrained, via phase-locked loops, to the
hippocampal rhythm, and thus do not necessitate exact phase-locked timing between unit
firing in hippocampus and neocortex. The predictions for unit firing patterns derived from
the model of Bibbig et al. (1995) are thus different from those advocated here, in that pre-
cise temporal relations between hippocampal and neocortical neural activity need not occur.
Results of recent experiments by Villa (2000) show that precise temporal structure across
the cortex can have definite correlations to on-going behaviour. In this study, the behaviours
in which temporal patterns in multispike trains are seen most frequently are reminiscent of
the times during various learning experiments when animals show theta activity in the
hippocampus (although the hippocampal EEG was not recorded in these experiments). For
instance, temporal spike patterns were more common in Villa’s experiments during the
delay in a delayed-response task or a GO-NOGO task, preceding correct rather than incorrect
task performance, and prior to cues needed for task performance. From other experiments
(reviewed in Miller, 1991) all of these are circumstances when theta activity is prominent,
and in some of these cases there is further evidence compatible with extension of the theta
rhythm to other structures such as the neocortex.
The most incisive test of the theory of cortico-hippocampal interplay requires simultaneous
recording of unit activity in the neocortex, and of the hippocampal EEG, in free-moving
animals. It is predicted that spike trains recorded from single neurones in neocortical loci
should show a tendency to fire at specific precise phases of the concurrent hippocampal
theta rhythm, the phase of firing probably varying greatly from locus to locus. Such phase-
locked unit firing would form part of the known temporal structure seen amongst spike trains
recorded from the neocortex, although its phase-locking to the hippocampal theta rhythm
could not be detected if that is not recorded. This critical experiment, involving simultaneous
recording of the hippocampal theta rhythm and neocortical unit activity has recently been
undertaken in rats (Siapas et al., 2000). These authors present evidence that “the firing of
prefrontal cortical neurons is phase-locked to hippocampal theta oscillations, with prefrontal
neurons firing at a preferred phase of the hippocampal theta rhythm, even in the absence of
local cortical theta oscillations.” These preliminary reports appear to favour the original theory
of cortico-hippocampal interplay Miller, 1991, rather than the model of Bibbig et al. (1995).
8. BALANCE BETWEEN HEMISPHERES IN HUMANS AND ANIMALS
So far this chapter has dealt with generic aspects of the dynamics of the mammalian
forebrain, but has not touched on an aspect of forebrain functional organization which is
usually thought to be specific to humans. This is the topic of cerebral lateralization of

function. However, there is a small number of morphological papers which document

asymmetries between the hemispheres in animals, for instance in size of the hemisphere or
thickness of the cortex (Diamond et al., 1981; Kolb et al., 1982). More interesting from
a functional point of view is a large body of Russian work from the late V.L. Bianki and
coworkers, little known to Western scientists, which shows many functional asymmetries
in the animal brain. Asymmetries demonstrated in rats using the technique of unilateral
spreading depression to inactivate temporarily one or other hemisphere reveal functional
asymmetries which (mutatis mutandis) are in many ways similar to those seen in humans
(Bianki, 1988, 1993), although it is sometimes necessary to study gender effects at the
same time before asymmetry is clearly revealed (Bianki and Filippova, 2000). Thus cerebral
lateralization may also be a generic property of the mammalian forebrain, rather than
specific to humans. Some comments on this topic are therefore appropriate in the present
account of forebrain function.
Issues of cerebral laterality cannot be related to those of brain dynamics at the neuronal
level unless the psychological evidence for cerebral lateralization is related to a theory at
the neuronal level. A recent work (Miller, 1996c) put forward such a theory, in which
differences between the hemispheres in distributions of conduction times in cortico-
cortical axons were proposed to underlie the psychological and presumed cybernetic
differences between the hemispheres. Specifically, it was proposed that the right hemisphere,
which appears to function best in processing of data all of whose components occur close
together in time, does so because it has a preponderance of rapidly conducting axonal
links; while the left hemisphere, which appears better suited for dealing with patterns of
information which are spread in time over brief intervals (up to a few hundred msec), does
so because its axonal links are overall more slowly-conducting, giving greater temporal
dispersion of signals. The direct test of this central hypothesis by cytological examination
of subcortical white matter in the electron microscope is not possible in humans, because
adequately fixed tissue is extremely difficult to obtain. If, as seems probable, cerebral
laterality of similar nature applies to small experimental animals, such experiments may
be possible in animals. Other indirect tests of the central hypothesis (some at the biological
level, many at the psychological level) are discussed by Miller (1996c) giving substantial,
support to the hypothesis, although the conclusive tests have yet to be conducted.
From foregoing sections of this chapter, cell assemblies appear to operate according to
two quite different types of dynamic. On the one hand, in small cortical regions, assemblies
can form which sustain activity without detailed temporal structure between the firing of
neurones. Presumably such assemblies represent static entities in the environment
(Gestalts) which also lack temporal structure. On the other hand, assemblies which form
across wider spans of cortical tissue necessarily have temporal structure in the firing of
neurones, and thus have the potential to represent events with temporal structure. From the
hypothesis just mentioned about cerebral laterality, the first version of cell assembly
operation should apply more strongly to the right hemisphere (and probably, in that hemi-
sphere, for cell assemblies spread well beyond the local domain considered earlier in this
chapter), while the latter should be more characteristic of the left hemisphere. (These two
modes of operation are schematized in Figure 18.5.) Having said this, however, it is likely
that left and right hemispheres differ only in degree, not in kind (i.e. not in sharp categorical
terms). This conclusion follows from the fact that either hemisphere has to some extent the
ability to perform the operations more characteristic of the other hemisphere, though less
efficiently than the more specialized hemisphere. Thus, in terms of the hypothesis about

448 Robert Miller
Figure 18.5. Schematized “space-time” diagram of the differences between assemblies characteristic of the left- and
right-hemispheres. For both the upper and lower staves of this diagram, neurones and their connections are laid out
as a geometrically-repeating pattern along the time axis. In spatial terms, the density of neurones and their connec-
tions is the same in each, as is the convergence ratio of connections in the neurones of each stave (i.e. three-fold). The
difference between the two is that connections in the upper stave (“left hemisphere”) have longer delays
(i.e. greater length in horizontal “time” axis) than those in the lower stave (“right hemisphere”). Activation of
neurones in the upper stave thus depends more on temporal convergence, that in the lower stave on spatial
convergence.
axonal conduction times, it is likely that subcortical white matter in either hemisphere has
a rich mixture of both slow- and fast-conducting axons, the difference between the hemi-
spheres being an overall bias to the slow or fast part of the conduction velocity distribution
in left and right hemispheres, rather than a complete absence of one or other type.
Given this, the following question arises: can the two dynamic modes of cell assembly
operation, characteristic respectively of left and right hemispheres, function together at the
same time in the same slab of neocortical tissue. The answer is probably “No”, for the
following reason: Neural activity in hypothetical left hemisphere-type networks is subject
to loss of amplification by temporal dispersion. This does not apply to hypothetical right
hemisphere-type networks. Thus left hemisphere-type networks require high overall levels
of activity before activity in a particular assembly/chain can be stably maintained over
time. Right hemisphere-type networks on the other hand can function effectively in their
characteristic manner at lower overall activity levels. Such assemblies can burst quickly
into effective activity, and can then subside equally quickly. Even given that a cortical
network has a rich mixture of both slow- and fast-conducting axonal links, the conditions
required for these two dynamic modes (i.e. suitable overall activity levels) are mutually
exclusive. This may be the real reason for hemispheric specialization. If so, the argument
just presented should apply to the mammalian cortex generally, not just to humans, and
cerebral asymmetry is to be expected in subhuman animals as well as in humans.

If one accepts that cerebral asymmetry arises from cellular differences between the cortex
on the two sides (e.g. distribution of calibres of axonal projections) other questions arise.
Do other forebrain structures such as the basal ganglia and hippocampus exhibit any additional
right/left asymmetry? This is a largely unexplored question, and difficult to answer defini-
tively, since asymmetry arising primarily in the cortex could lead in many ways to secondary
asymmetry in subcortical structures, even though no such asymmetry is really intrinsic to
these structures. While this question is at present incapable of a certain answer, the foregoing
theories of forebrain dynamics lead to a few suggestions about how the cortex may engage
in varieties of interaction with these subcortical structures which differ between right and
left side.
Both the basal ganglia and the hippocampus appear to be involved in sustaining patterns
of activity in the cortex for periods of seconds or more (longer than the exactly-timed
chains of activity considered so far). In terms of large-scale psychological concepts, both
therefore contribute to short-term (or “working”) memory. Both are involved in recursive
loops, providing re-entrant influences upon cortical activity, based on activity efferent
from the cortex which activated the loop at a slightly earlier stage. However, on the left side,
hypothetically, exactly-timed chains of activity are typical of the alert hemisphere, while
on the right, sequences of Hebbian cell assemblies which are continuously active without
exact temporal chaining of activity predominate. Given this scenario, one can envisage
that the subcortical structures can co-ordinate the chaining together of exactly-timed
sequences of activity in the left hemisphere whose temporal length is outside the capability
of the left neocortex by itself (e.g. sentences, musical phrases). The subcortical structures
on the right side, on the other hand, may be important for chaining together activity in
assemblies which themselves do not have exact temporal structure, generically a set of
sequences of Gestalt representations.
The theory behind lateralized cortico-subcortical interactions merits much further
development. However, it is a complex topic, because, to test such hemisphere-specific
interactions, it is usually necessary to use material (e.g. verbal vs spatial) for which one or
other hemisphere is specialized, although the real issue is not about hemisphere-specific
material, but about hemisphere-specific styles of information processing. Thus, memory
deficits after hippocampal damage which are selective to verbal tests for left-sided damage
and to visuospatial tests after right-sided damage do not go beyond what is predicted from
wider knowledge of cerebral asymmetry. To devise experimental designs for comparing
cortico-subcortical interactions between hemispheres in tests which are adequately
matched on all features except the style of information processing is very difficult; and
therefore these issues are only at preliminary stages of investigation at present sent. For
convincing demonstration of the different cortico-subcortical interactions in the two hemi-
spheres, one needs to find examples where the laterality effect expected from the test
material itself using a control experimental design is reversed when used in a design with
specific additional requirements for information processing.
Although this sort of experiment has seldom been considered, two published examples
of such laterality effects selective for the style of information processing (rather than the
material used) can be given. In higher aspects of language processing, for example in
appreciation of metaphors, indirect inferences, and jokes, right hemisphere lesions
produce more severe impairment than left hemisphere ones (Winer and Gardner, 1977;
Wapner et al., 1981; Brownell et al., 1986). These appear to be right hemisphere-preferred
functions, even though the component items, being verbal, should give a preferential left

450 Robert Miller
hemisphere representation. This task, although exemplifying laterality based on processing

style, does not provide evidence for a distinct role of any subcortical structure in such later-
ality. Another example of processing-style-specific asymmetry is a PET study by Fuji
et al. (1997). For verbal memory tasks, activation of the parahippocampal gyrus in a non-
matching version of the task, was, as expected, greater on the left side; but when the task
was a matching one using the same material, the right parahippocampal gyrus was activated.
It is uncertain what the basis of this reversal of asymmetry can be. It could rely either on
a distinct type of asymmetry in the hippocampus itself, or just on the fact that the hippo-
campus (assumed to be symmetrical) can better handle certain styles of information
processing, and thus can produce selective improvements on the side with relatively lower
performance. Full interpretation of these complex results requires further development of
the underlying theory of brain dynamics, and new experimental designs will be required to
adequately test that theory.
9. THE “OMNICONNECTION PRINCIPLE” IN SPACE AND TIME:

APPROXIMATION IN REALITY TO THE METAPHYSICAL
IDEAL OF “CONSCIOUSNESS”
In an earlier work (Miller, 1981) the concept of consciousness was analysed in terms of what
was called “the omniconnection principle”. This term referred to an information-processing
network in which all nodes are connected to all others. This concept was developed from
a philosophical perspective, to explain the mysterious “unity” of consciousness. Initially
omniconnection referred to a network which was conceived as completely connected
spatially—an assumption used in very many computer simulations since 1981, albeit for
small-scale networks, rather than the brain as a whole. However, the concept of omni-
connection was also applied in the temporal dimension: since different cortical loci are
interlinked by slow conducting axons which can function as “delay lines”, and since there
may be a rich repertoire of delays between any two loci, one can consider that pairs of loci
may be completely connected in the temporal dimension as well as in the spatial one. In
other words, sequential activation of the any two loci with any intervening time interval can
be matched by the delay in at least one of the axonal connections linking the two loci.
Of course, this notion of omniconnection is an ideal, like the frictionless surface, or the
infinitely efficient heat engine, postulated by physicists to aid their thinking. It thus has
some value in conceptualizing consciousness in philosophical or metaphysical terms.
However, it is far removed from the real physical substrate for information processing in
the brain, and psychological analysis also shows that the ideal of “all pieces of information
having access to all others” is not strictly realized. Nevertheless, it is useful to compare the
idealized omniconnected network of the 1981 formulation with the various levels of
structural organization and cybernetic operation of the forebrain outlined in this chapter.
At the level of local processing in the hypothetical block of cortical tissue of dimensions
1 mm3, the ideal of omniconnection is almost realized. In the temporal dimension, local
transmission of signals never takes them outside the integration interval of the principal
neurones in that locale. All signals therefore occur effectively in the same instant of time,
in an “eternal present”, which, within the local block, cannot be linked with other times. In
the spatial dimension, the probability of connection between pyramidal cells in laminae II
and III, even within the limits of the 1 mm3 domain is much less than unity (A value of 0.8

was assumed above for immediatley-adjacent pyramidal cells, but this probability falls of
sharply within a few hundred microns). This is a necessary consequence of the fact that
each pyramidal cell has only a few thousand locally-distributed synapses, while the
number of neurones in the territory of distribution approaches 100 thousand. However,
disynaptic connectivity ratios are the square of monosynaptic connectivity ratios. Thus, with
disynaptic connections, it is, in principal, possible for every neurone to be in synaptic
connection with every other. Nevertheless, disynaptic connections (e.g. via a neurone in
lamina V) are in themselves ambiguous, unless disambiguated by the direct connections of
the triangular configurations of Figure 18.1. In this respect, omniconnection does not apply
in a local cortical domain. However, the omniconnection principle does not specify that the
“nodes” in such a scheme are actually single cortical neurones: there may well be some
redundancy at this level, in which case omniconnection might be better applied to cortical loci
or columns, rather than single cells.
For more distant cortico-cortical connections there appears to be a complex mosaic of
strongly interconnected bands or patches of cortical tissue, arranged over the whole neocortex.
Probably neighbouring cortical loci belong to different mosaics which overlap little with
each other. In addition, there are hints that cortical bands which have strong direct links
may be supported by parallel patchy connections received indirectly via thalamic relays.
In any case, it is clear that not every cortical locus has links to every other. Nevertheless, it
is plausible to suggest that every locus has links to the general region of every other locus,
the finer details of connectivity then being left to local connections in the recipient region.
In this sense, cortical connectivity approximates to an omniconnected network. However,
the patchy distribution of long cortico-cortical connections implies that multisynaptic
links are needed for every locus to be in connection with every other. As a result, associations
between one arbitrary cortical locus and any other cannot be made unambiguously and
unconditionally. The dictum “neurones that fire together become strongly connected
together” thus cannot apply in a strict sense, if one considers just the connectivity pattern
of the neocortex.
Both the long cortico-cortical connections and cortico-thalamic connections provide a wide
range of conduction times. This gives some measure of temporal omniconnectedness, well
beyond the neuronal integration time, and perhaps up to 100 or 200 msec in the human
brain (see Miller, 1996c, where this value is taken as the approximate time limit of monosyn-
aptic connections). For intervals greater than this, direct delay-specific association between
cortical loci is likely to break down, and is replaced by representation of sequence rather
than of exact timing, probably achieved by interplay between anterior regions of the cortex
and the basal ganglia.
Both in the local cortical domain and for long-distance links across the cortex, the
approximation to the ideal of omniconnectedness which is actually realized involves
disynaptic or multisynaptic connections. This leads to inherent ambiguity of the network,
so that different neurones or loci which repeatedly fire together cannot be assured of
becoming strongly interconnected. However, the theory of cortico-hippocampal interplay
(Miller, 1991), which has now survived a number of attempts to test it, provides a mechan-
ism where the information from a global context in the environment can be used to define
large-scale regimes of preferred pathways for transcortical interaction. In fact, the
information structure of the environments we inhabit includes a great deal of redundancy.
Some items of information are likely to be associated repeatedly, others will never occur
together. In view of this, a strictly omniconnected network, even if possible, would be

452 Robert Miller
Figure 18.6. Schematic diagram of major interactions in the forebrain co-ordinating the operation of neural
assemblies in the neocortex. The central block represents the neocortex, containing a number of local foci of activity
(depicted as clusters of interlaminar triangles). Overall, these local segments of activity are co-ordinated by interplay
with the hippocampus (“hipp”), such that time delays in transmission from one segment to another become parts
of the phase-locked loops between hippocampus (master oscillator) and neocortex. Interplay between neocortex
and thalamus can provide further support for the spatiotemporal structure of cortical neural activity, using recip-
rocal connections and corticothalamic axons as delay lines. Interaction between parts of the neocortex and the basal
ganglia can enable recursive loops to form, which encode expectancy and sequence, even when the exact temporal
relations are indeterminate.
unnecessary. The role of cortico-hippocampal interplay in development of representation

of each distinct environment allows the multipotent neocortex to use selected subnet-
works, each of which represents all the associations that can occur in the corresponding
environment. Figure 18.6 attempts to depict the interactions possible within the cerebral
neocortex, as organized and co-ordinated by interactions with the hippocampus, the thal-
amus and the basal ganglia.
In terms of cell assembly theory, two sorts of cell assembly are capable of stable operation
in this complex spatiotemporal network: first there are those which lack intrinsic temporal
structure, whose neuronal components are linked together by axons having conduction
times less than the neuronal integration time, and therefore incapable of contributing

temporal information to assembly activity. Such assemblies would be very useful for
representing Gestalts, and other configurations of information, all of whose elements
occur very close together in time, and would sustain activity without representing accurate
timing of events. The second type of assembly is that which does have intrinsic temporal
structure, extended for up to several hundred milliseconds. The elements of such an
assembly are linked by slow-conducting axons (“delay lines”). Interplay with the hippo-
campus may be specially important for ensuring secure transmission along such extended
chains of activation. These chains are likely to be important in representing patterns of
information with exact (although only briefly extended) temporal structure. With the help of
the hippocampus, such chains may be extended further.
The vast population of long cortico-cortical axons which make up subcortical white matter
are a mixture of slow- and fast-conducting axons, in either hemisphere. Thus, potentially,
either hemisphere can operate with either class of assembly. However, the two types of
assembly operate best at different levels of overall cortical arousal, so that they cannot
both operate effectively at the same time. It is likely that the left hemisphere has a richer
repertoire of slow-conducting delay lines, the right hemisphere having a richer array of fast-
conducting axons. This, it is suggested, is the basis for the differences in psychological
processing between right and left hemispheres. The fact that the two sorts of assembly
cannot operate simultaneously in the same cortical network, may be real the reason for
hemispheric specialization; and if so, such specialization should apply throughout the
mammalian kingdom, not just in humans.
REFERENCES
Abeles, M. (1981) Local cortical circuits. Studies in Brain Function series, Springer Verlag, Heidelberg.
Abeles, M. (1991) Corticonics: Neural circuits of the cerebral cortex. Cambridge University Press, Cambridge,
U.K.
Adey, W.R., Dunlop, C.W. and Hendrix, C.E. (1960) Hippocampal slow waves. Distribution and phase relation-
ships in the course of approach learning. Archives of Neurology, 3, 74–90.
Albe-Fessard, D., Condes-Lara, M. and Sanderson, P. (1983) The focal tonic cortical control of intralaminar
thalamic neurons may involve a cortico-thalamic loop. Acta Morphologica Hungarica, 31, 9–26.
Alexander, G.E., DeLong, M.R. and Strick, P.L. (1986) Parallel organization of functionally segregated circuits
linking basal ganglia and cortex. Annual Review of Neuroscience, 9, 357–381.
Amabile, G., Fattaposta, F., Pozzessere, G., Albani, G., Sanarelli, L., Rizzo, P.A. and Morocutti, C. (1986) Parkin-
son’s disease: electrophysiological (CNV) analysis related to pharmacological treatment. Electroencephalo-
graphy and Clinical Neurophysiology, 64, 521–524.
Amir, Y., Harel, M. and Malach, R. (1993) Cortical hierarchy reflected in the organization of intrinsic con-
nections in macaque monkey visual cortex. Journal of Comparative Neurology, 334, 19–46.
Balaydier, C. and Maugiere, F. (1987) Network organization of the connectivity between parietal area 7, posterior
cingulate cortex and medial pulvinar nucleus: a double fluorescent tracer study in monkey. Experimental
Berman, N., Payne, B.R., Labar, D.R. and Murphy, E.H. (1982) Functional organization of neurons in cat striate
cortex: variations in ocular dominance and receptive-field type with cortical laminae and location in visual
field. Journal of Neurophysiology, 48, 1362–1377.
Bianki, V.L. (1988) The right and left hemispheres of the animal brain. Cerebral lateralization of function. Gordon
and Breach, New York and London.
Bianki, V.L. (1993) The mechanisms of brain lateralization. Gordon and Breach, New York, and London.
Bianki, V.L. and Filippova, E.B. (2000) Sex differences in lateralization in the animal brain. (T. Ganf: trans;
R. Miller: ed.), Conceptual Advances in Brain Research series, Harwood Academic Publishers.
Bibbig, A, Wennekers, T. and Palm, G. (1995) A neural network model of corticohippocampal interplay and the
representation of context. Behavioural and Brain Research, 66, 169–175.
Bienenstock, E.L., Cooper, L.N. and Munro, P.W. (1982) Theory for the development of neuron selectivity.

454 Robert Miller
Braitenberg, V. (1978) Cell assemblies in the cerebral cortex. In: R. Heim and G. Palm (eds), Lecture
notes in Biomathematics Vol. 21: Theoretical approaches to complex systems. Heidelberg, Springer Verlag,
pp. 171–188.
Braitenberg, V. and Schüz, A. (1991) Anatomy of the cortex: statistics and geometry. Studies on brain function
No. 18. Heidelberg, Springer Verlag.
Braitenberg, V. and Schüz, A. (1998) Cortex: statistics and geometry of neuronal connectivity (Second, revised
edition of Braitenberg and Schüz, 1991). Heidelberg, Springer.
Brownell, H.H., Potter, H.H. and Bihrle, A.M. (1986) Inference deficits in right brain-damaged patients. Brain
and Language, 27, 310–321.
Bures, J., Buresova, O., Weiss, T. and Fifkova, E. (1963) Excitability changes in non-specific thalamic nuclei
during cortical spreading depression in the rat. Electroencephalography and Clinical Neurophysiology, 15,
73–83.
Burkhalter, A. (1989) Intrinsic connections of rat primary visual cortex: laminar organization of axonal projec-
tions. Journal of Comparative Neurology, 279, 171–186.
Carpenter, M.B. (1981) Anatomy of the corpus striatum and brain stem integrating systems. Handbook of
Physiology, The Nervous System II, Motor control Part 2. American Physiological Society, Bethesda MD,
pp. 947–995.
Chapin, J.K. (1986) Laminar differences in sizes, shapes and response profiles of cutaneous receptive fields in
the rat S1 cortex. Experimental Brain Research, 62, 549–559.
Chow, K.L., Brumback, H.D. and Glanzman, D.L. (1978) Abnormal development of lateral geniculate neurons
in rabbits subjected to either eyelid closure or corticofugal paroxysmal discharge. Brain Research, 146,
151–158.
Coenen, A.M.L. and Vendrik, A.J.H. (1972) Determination of the transfer ratio of cat’s geniculate neurons
through quasi-intracellular recordings and the relation with level of alertness. Experimental Brain Research,
14, 227–242.
Contreras, D. and Steriade, M. (1997) State-dependent fluctuations of low frequency rhythms in corticothalamic
networks. Neuroscience, 76, 25–38.
Cowan, R.L. and Wilson, C.J. (1994) Spontaneous firing patterns and axonal projections of single corticostriatal
neurons in the rat medial agranular cortex. Journal of Neurophysiology, 71, 17–32.
Darian-Smith, C., Tan, A. and Edwards, S. (1999) Comparing thalamocortical and corticothalamic microstruc-
ture and spatial reciprocity in the macaque ventral posterolateral nucleus (VPLc) and medial pulvinar.
DeBruyn, E.J., Casagrande, V.A., Beck, P.D. and Bonds, A.B. (1993) Visual resolution and sensitivity of single
cells in the primary visual cortex (V1) of a noctural primate (bush baby): correlations with cortical layers
and cytochrome oxidase patterns. Journal of Neurophysiology, 69, 3–18.
Deecke, L., Scheid, P. and Kornhuber, H.H. (1969) Distribution of readiness potential, pre-motion positivity and
motor potential of the human cerebral cortex preceding voluntary finger movements. Experimental Brain
Research, 7, 158–168.
Deecke, L., Englitz, H.G., Kornhuber, H.H. and Schmitt, G. (1977) Cerebral potentials preceding voluntary move-
ment in patients with bilateral or unilateral Parkinson akinesia. In: J.E. Desmedt (ed.), Attention, voluntary
contraction and event-related cerebral potentials. (Progress in Clinical Neurophysiology, Vol. 1). Karger,
Basel, pp. 151–163.
Deniau, J.M. and Chevalier, G. (1985) Disinhibition as a basic process in the expression of striatal functions. II.
The striato-nigral influence on thalamocortical cells of the ventromedial thalamic nucleus. Brain Research,
334, 227–233.
Deschenes, M., Veinante, P. and Zhang, Z.-W. (1998) The organization of corticothalamic projections. Recipro-
city versus parity. Brain Research Reviews, 28, 286–308.
Diamond, M.C., Dowling, G.A. and Johnson, R.E. (1981) Morphologic asymmetry in male female rats. Experi-
mental Neurology, 71, 261–268.
Dick, J.P.R., Cantello, R., Buruma, O., Gioux, M., Benecke, R., Day, B.L., Rothwell, J.C., Thompson, P.D. and
Marsden, C.D. (1987) The Bereitschaftspotential, L-DOPA and Parkinson’s disease. Electroencephalo-
graphy and Clinical Neurophysiology, 66, 263–274.
Dick, J.P.R., Rothwell, J.C., Day, B.L., Cantello, R., Buruma, O., Gioux, M., Benecke, R., Berardelli, A.,
Thompson, P.D. and Marsaden, C.D. (1989) The Bereitschaftspotential is abnormal in Parkinson’s disease.
Brain, 112, 233–244.
Doppelmayr, M., Klimesch, W., Schwaiger, J., Stadler, W. and Rohm, D. (2000) The time locked theta response
reflects interindividual differences in human memory performance. Neuroscience Letters, 278, 141–144.
Dudek, S.M. and Bear, M.F. (1992) Homosynaptic long-term depression in area CA1 of hippocampus and
effects of N-methyl-D-aspartate receptor blockade. Proceedings of the National Academy of Sciences,
U.S.A., 89, 4363–4367.
Ezure, K. and Oshima, T. (1981a) Dual activity patterns of fast pyramidal tract cells and their family neurones
during EEG arousal in the cat. Japanese Journal of Physiology, 31, 717–736.

Ezure, K. and Oshima, T. (1981b) Excitation of slow pyramidal tract cells and their family neurones during
phasic and tonic phases of EEG arousal. Japanese Journal of Physiology, 31, 737–748.
Ferster, D. (1981) A comparison of binocular depth mechanisms in areas 17 and 18 of the cat visual cortex.
Ferster, D. and Lindström, S. (1983) An intracellular analysis of geniculo-cortical connectivity in area 17 of the
cat. Journal of Physiology (London), 342, 181–215.
Fuji, T., Okuda, J., Kawashima, R., Yamadori, A., Fukatsu, R., Suzuki, K., Ito, M., Goto, R. and Fukuda, H.
(1997) Different roles of the left and right parahippocampal regions in verbal recognition: a PET study.
Funke, K. and Eysel, U.T. (1992) EEG-dependent modulation of response dynamics of cat dLGN relay cells and
the contribution of corticogeniculate feedback. Brain Research, 573, 217–227.
Fuster, J.M. (1980) The prefrontal cortex. Anatomy, physiology and neuropsychology of the frontal lobe. Raven
Press, New York.
Ghazanfar, A.A. and Nicolelis, M.A.L. (2000) The space-time continuum in mammalian sensory pathways.
In: R. Miller (ed.), Time and the brain. Conceptual Advances in Brain Research series. Harwood Academic
Publishers, Amsterdam.
Giguere, M. and Goldman-Rakic, P.S. (1988) Mediodorsal nucleus: areal, laminar and tangential distribution
of afferents and efferents in the frontal lobe of rhesus monkeys. Journal of Comparative Neurology, 277,
195–213.
Gilbert, C.D. (1977) Laminar differences in receptive field properties of cells in cat primary visual cortex.
Grastyan, E., Karmos, G., Vereczkey, L. and Kellenyi, L. (1966) The hippocampal electrical correlates
of the homeostatic regulation of motivation. Electroencephalography and Clinical Neurophysiology,
21, 35–53.
Gulyas, B., Lagae, L., Eysel, U. and Orban, G.A. (1990) Corticofugal feedback influences the responses of
geniculate neurons to moving stimuli. Experimental Brain Research, 79, 441–446.
Harvey, A.R. (1980) A physiological analysis of subcortical and commissural projections of areas 17 and 18 of
the cat. Journal of Physiology (London), 302, 507–534.
Hebb, D.O. (1949) The organization of behavior: a neuropsychological theory. Wiley, New York.
Hellwig, B. (2000) A quantitative analysis of the local connectivity between pyramidal neurons in layers 2/3 of
the rat visual cortex. Biological Cybernetics, 82, 111–121.
Hess, G., Aizenman, C.D. and Donghue, J.P. (1996) Conditions for the induction of long-term potentiation in
layer II/III horizontal connections of the rat motor cortex. Journal of Neurophysiology, 75, 1765–1778.
Hirsch, J.C., Fourment, A. and Marc, M.E. (1983) Sleep-related variations of membrane potenital in the lateral
geniculate body relay neurons of the cat. Brain Research, 259, 308–312.
Hull, C.L. (1943) Principles of behavior. New York: Appleton-Century-Crofts.
Inubushi, S., Kobayashi, T., Oshima, T. and Torii, S. (1978) Intracellular recordings from the motor cortex
during EEG arousal in unanaesthetized brain preparations of the cat. Japanese Journal of Physiology,
28, 669–688.
Ito, M. (1985) Processing of vibrissa sensory information within the rat neocortex. Journal of Neurophysiology,
54, 479–490.
Joel, D. and Weiner, I. (2000) Striatal contention scheduling and the split-circuit scheme of basal ganglia-
thalamocortical circuitry: from anatomy to behaviour. In: R. Miller and J.R. Wickens (eds), Brain dynamics
and the striatal complex. Conceptual Advances in Brain Research series. Harwood Academic Publishers,
Amsterdam.
Johnson, M.J. and Alloway, K.D. (1994) Sensory modulation of synchronous thalamocortical interactions in the
somatosensory system of the cat. Experimental Brain Research, 102, 181–197.
Joseph, J.P. and Barone, P. (1987) Prefrontal unit activity during a delayed oculomotor task in the monkey.
Kayama, Y., Shosaku, A. and Doty, R.W. (1984) Cryogenic blockade of the visual cortico-thalamic projection in
the rat. Experimental Brain Research, 54, 157–165.
Klimesch, W., Doppelmayr, M., Schimke, H. and Ripper, B. (1997) Theta synchronization and alpha desynchroniza-
tion in a memory task. Psychophysiology, 34, 169–176.
Kolb, B., Sutherland, R.Y., Nonneman, A.Y. and Whishaw, J.G. (1982) Asymmetry in the cerebral hemispheres
of the rat, mouse, rabbit and cat: the right hemisphere is larger. Experimental Neurology, 78, 348–359.
Komatsu, H. (1982) Prefrontal unit activity during a color discrimination task with GO and NO-GO responses in
the monkey. Brain Research, 244, 269–277.
Kornhuber, H.H. and Deecke, L. (1965) Hirnpotentialänderungen bei Wilkürbewegungen und passiven
Bewegungen des Menschen: Bereitschaftspotential und reafferente Potentiale, Pflügers Archiv für Gesamte
Physiologie, 284, 1–17.
Kuroda, M., Murakami, K., Kishi, K. and Price, J.L. (1992) Distribution of the piriform cortical terminals to cells
in the central segment of the mediodorsal thalamic nucleus of the rat. Brain Research, 595, 159–163.

456 Robert Miller
Lemmon, V. and Pearlman, A.L. (1981) Does laminar position determine the receptive field properties of cor-
tical neurons? A study of corticotectal cells in area 17 of the normal mouse and the reeler mouse. Journal of
LeVay, S., McConnell, S.K. and Luskin, M.B. (1987) Functional organization of primary visual cortex in the
mink (Mustela vison) and a comparison with the cat. Journal of Comparative Neurology, 257, 422–441.
Leventhal, A.G. and Hirsch, H.V.B. (1978) Receptive-field properties of neurons in different laminae of visual
cortex of the cat. Journal of Neurophysiology, 41, 948–962.
Levitt, J.B., Lewis, D.A., Yoshioka, T. and Lund, J.S. (1993) Topography of pyramidal neuron intrinsic
connections in macaque monkey profrontal cortex (areas 9 and 46). Journal of Comparative Neurology, 338,
360–376.
Levitt, J.B., Yoshioka, T. and Lund, J.S. (1994) Intrinsic cortical connections in macaque visual area V2:
evidence for interraction between different functional streams. Journal of Comparative Neurology, 342,
551–570.
Lindstrom, S. and Wrobel, A. (1990) Frequency dependant corticofugal excitation of principal cells in the cat’s
dorsal lateral geniculate nucleus. Experimental Brain Research, 79, 313–318.
Liu, X.-B., Honda, C.N. and Jones, E.G. (1995) Distribution of four types of synapse on physiologically
identified relay neurons in the ventral posterior thalamic nucleus of the cat. Journal of Comparative Neurology,
352, 69–91.
Malach, R. (1994) Cortical columns as devices for maximizing neuronal diversity. Trends in Neuroscience,
17, 101–104.
McKenna, T.M., Light, A.R. and Whitsel, B.L. (1984) Neurons with unusual response and receptive-field
properties in upper laminae of cat S1 cortex. Journal of Neurophysiology, 51, 1055–1076.
Mangini, N.J. and Pearlman, A.I. (1980) Laminar distribution of receptive field properties in the primary visual
cortex of the mouse. Journal of Comparative Neurology, 193, 203–222.
Mathers, L.H. (1972) The synaptic organization of the cortical projection to the pulvinar of the squirrel monkey.
Metherate, R. and Ashe, J.H. (1993) Ionic flux contributions to the neocortical slow waves and nucleus basalis-
mediated activation: whole-cell recordings in vivo. Journal of Neuroscience, 13, 5312–5323.
Metin, C., Godement, P. and Imbert, M. (1988) The primary visual cortex in the mouse: receptive field properties
and functional organization. Experimental Brain Research, 69, 594–612.
Miller, R. (1981) Meaning and purpose in the intact brain. Oxford University Press, Oxford.
Miller, R. (1991) Cortico-hippocampal interplay and the representation of contexts in the brain. Springer
Verlag, Heidelberg.
Miller, R. (1996a) Neural assemblies and laminar interaction in the cerebral cortex. Biological Cybernetics, 75,
253–275.
Miller, R. (1996b) Cortico-thalamic interplay and the security of operation of neural assemblies and temporal
chains in the cerebral cortex. Biological Cybernetics, 75, 263–275.
Miller, R. (1996c) Axonal conduction time and human cerebral laterality: A psychobiological theory. Harwood
Academic Publishers, Amsterdam.
Miller, R. and Wickens, J.R. (1991) Corticostriatal cell assemblies in selective attention and in representation of
predictable and controllable events: a general statement of corticostriatal interplay and the role of striatal
dopamine. Concepts in Neuroscience, 2, 65–95.
Molotchnikoff, S., Tremblay, F. and Lepore, F. (1984) The role of the visual cortex in response properties of
lateral geniculate cells in rats. Experimental Brain Research, 53, 223–232.
Okano, K. and Tanji, J. (1987) Neuronal activities in the primate motor fields of the agranular frontal cortex
preceding visually triggered and self-paced movement. Experimental Brain Research, 66, 155–166.
O’Keefe, J. and Nadel, L. (1978) The hippocampus as a cognitive map. Clarendon Press, Oxford.
Palm, G. (1982) Neural assemblies. Springer Verlag.
Palmer, L.A. and Rosenquist, A.C. (1974) Visual receptive fields of single striate cortical units projecting to the
superior colliculus in the cat. Brain Research, 67, 27–42.
Revishchin, A.V. (1985) Laminar distribution of neurons with different types of receptive fields in the visual
cortex of the rabbit. Neirofiziologiia, 17, 19–27.
Rockland, K.S. and Lund, J.S. (1983) Intrinsic laminar lattice connections in primate visual cortex. Journal of
Sakurai, Y. (1996) Population coding by cell assemblies—what it really is in the brain. Neuroscience Research,
26, 1–16.
Sawyer, S.F., Martone, M.E. and Groves, P.M. (1991) A GABA immunocytochemical study of rat motor thalamus:
light and electron microscopic observations. Neuroscience, 42, 103–124.
Scannell, J.W., Burns, G.A.P.C., Hilgetag, C.C., O’Neill, M.A. and Young, M.P. (1999) The connectional organiza-
tion of the cortico-thalamic system of the cat. Cerebral Cortex, 9, 277–299.
Schiller, P., Finlay, B.L. and Volman, S.F. (1976a) Quantitative studies of single-cell properties in monkey stri-
ate cortex. I. Spatiotemporal organization of receptive fields. Journal of Neurophysiology, 39, 1288–1319.

Schiller, P., Finlay, B.L. and Volman, S.F. (1976b) Quantitative studies of single-cell properties in monkey striate
cortex. II. Orientation specificity and ocular dominance. Journal of Neurophysiology, 36, 1320–1333.
Schüz, A. (1994) Patchiness as a means to get the message across. Trends in Neuroscience, 17, 365.
Schmielau, F. and Singer, W. (1977) The role of visual cortex for binocular interactions in the cat lateral genicu-
late nucleus. Brain Research, 120, 354–361.
Schwartz, M.L., Dekker, J.J. and Goldman-Rakic, P.S. (1991) Dual modes of corticothalamic synaptic termina-
tion in the medio-dorsal nucleus of rhesus monkey. Journal of Comparative Neurology, 309, 289–304.
Sherman, S.M. and Guillery, R.W. (1996) Functional organization of thalamocortical relays. Journal of
Siapas, A.G., Lee, A.K., Lubenov, E.V. and Wilson, M.A. (2000) Prefrontal phase-locking to hippocampal theta
oscillations. Society for Neuroscience Abstracts, 26, 467–1.
Simpson, J.A. and Khurabait, A.J. (1986) Readiness potential of cortical area 6 in Parkinson’s disease. Evidence
for a dopaminergic striatal control of postural set involving supplementary motor area. Journal of
Neurology Neurosurgery and Psychiatry, 49, 475.
Snodderly, D.M. and Gur, M. (1995) Organization of striate cortex of alert, trained monkeys (Macaca fascicu-
laris): on-going activity, stimulus selectivity, and widths of receptive field activating regions. Journal of
Steriade, M., Jones, E.G. and Llinas, R.R. (1990) Thalamic oscillations and signaling. New York: Wiley
Interscience.
Steriade, M. (1997) Synchronized activities of coupled oscillators in the cerebral cortex and thalamus at different
levels of vigilance. Cerebral Cortex, 7, 583–604.
Stern, E.A., Kincaid, A.E. and Wilson, C.J. (1997) Spontaneous subthreshold membrane potential fluctuations
and action potential variability of rat corticostriatal and striatal neurons in vivo. Journal of Neurophy-
siology, 77, 1697–1715.
Sur, M., Garraghty, P.E. and Bruce, C.J. (1985) Somatosensory cortex in macaque monkeys: laminar differences
in receptive field size in areas 3b and 1. Brain Research, 342, 391–395.
Swadlow, H.A. (1988) Efferent neurons and suspected interneurons in binocular visual cortex of the awake
rabbit: receptive fields and binocular properties. Journal of Neurophysiology, 59, 1162–1187.
Swadlow, H.A. (1989) Efferent neurons and suspected interneurons in S-1 vibrissa cortex of the awake rabbit:
receptive field and axonal properties. Journal of Neurophysiology, 62, 288–308.
Swadlow, H.A. (1990) Efferent neurons and suspected interneurons in S-1 forelimb representation of the awake
rabbit: receptive fields and axonal properties. Journal of Neurophysiology, 63, 1477–1498.
Swadlow, H.A. (1991) Efferent neurons and suspected interneurons in second somatosensory cortex of the
awake rabbit: receptive fields and axonal properties. Journal of Neurophysiology, 66, 1392–1409.
Swadlow, H.A. (1992) Monitoring the excitability of neocortical efferent neurons to direct activation by
extracellular current pulses. Journal of Neurophysiology, 68, 605–619.
Swadlow, H.A. (1994) Efferent neurons and suspected interneurons in motor cortex of the awake rabbit:
axonal properties, sensory receptive fields and subthreshold synaptic inputs. Journal of Neurophysiology,
71, 437–453.
Swadlow, H.A. (2000) Information flow along neocortical axons. In: R. Miller (ed.), Time and the brain. Concep-
tual Advances in Brain Research series. Harwood Academic Publishers, Amsterdam, pp. 131–155.
Swadlow, H.A. and Weyand, T.G. (1981) Efferent systems of the rabbit visual cortex: laminar distribution of the
cells of origin, axonal conduction velocities and identification of axonal branches. Journal of Comparative
Neurology, 203, 799–822.
Swadlow, H.A. and Weyand, T.G. (1987) Corticogeniculate, corticotectal neurons and suspected interneurons in
visual cortex of awake rabbits: receptive field properties, axonal properties and effects of EEG arousal.
Tolman, E.C. (1932) Purposive behaviour in man and animals. Appleton-Century-Crofts, New York.
Tseng, G.-F. and Royce, G.J. (1986) A Golgi and ultrastructural analysis of the centromedian nucleus of the cat.
Tsumoto, T., Creutzfeldt, O.D. and Legendy, C.R. (1978) Functional organization of the corticofugal system
from visual cortex to lateral geniculate nucleus in the cat (with an appendix on geniculo-cortical mono-
synaptic connections). Experimental Brain Research, 32, 345–364.
Uhr, J.L. and Chapin, J.K. (1983) Contribution of thalamo-cortical and cortico-cortical connections to receptive
field properties in rat S1 cortex. Abstracts of the Society for Neuroscience, 74, 10.
Varela, F.J. and Singer, W. (1987) Neuronal dynamics in the visual corticothalamic pathway revealed through
binocular rivalry. Experimental Brain Research, 66, 10–20.
Vidyasagar, T.R. and Urbas, J.V. (1982) Orientation sensitivity of cat LGN neurones with and without inputs
from visual cortical areas 17 and 18. Experimental Brain Research, 46, 157–190.
Villa, A.E.P. (2000) Empirical Evidence about Temporal Structure in Multi-unit recordings. In: R. Miller
(ed.), Time and the brain. Conceptual Advances in Brain Research series. Harwood Academic Publishers,
Amsterdam, pp. 1–51.

458 Robert Miller
Villa, A.E. and Abeles, M. (1990) Evidence for spatiotemporal firing patterns within the auditory thalamus of the
cat. Brain Research, 509, 325–327.
Villa, A.E.P., Rouiller, E.M., Simm, G.M., Zurita, P., de Ripaupierre, Y. and de Ripaupierre, F. (1991) Cortico-
fugal modulation of the information processing in the auditory thalamus of the cat. Experimental Brain
Research, 86, 506–517.
Waller, H.J. and Feldman, S.M. (1967) Somatosensory thalamic neurons: effects of cortical depression. Science,
157, 1074–1077.
Walter, W.G., Cooper, R., Aldridge, V.J., McCallum, W.C. and Winter, A.L. (1964) Contingent negative variation:
an electric sign of sensorimotor association and expectancy in the human brain. Nature, 203, 380–384.
Wapner, W., Hanby, S. and Gardner, H. (1981) The role of the right hemisphere in the apprehension of complex
linguistic material. Brain and Language, 14, 15–33.
Wennekers, T. and Palm, G. (2000) Cell assemblies, associative memory and temporal structure in brain signals.
In: R. Miller (ed.), Time and the brain. Conceptual Advances in Brain Research series. Harwood Academic
Publishers, Amsterdam, pp. 251–273.
Wickens, J.R. (1990) Striatal dopamine in motor activation and reward-mediated learning: steps towards a unifying
model. Journal of Neural Transmission (General Section), 80, 9–31.
Wickens, J.R. (1993) A theory of the mammalian striatum. Pergamon Press, Oxford.
Wickens, J.R., Begg, A.J. and Arbuthnott, G.W.(1996) Dopamine reverses the depression of rat cortico-striatal
synapses which normally follows high-frequency stimulation of cortex in vitro. Neuroscience, 70, 1–5.
Wickens, J.R. and Miller, R. (1997) A formalization of the neural assembly concept: constraints on neural assembly
size. Biological Cybernetics, 77, 351–358.
Wickens, J.R. and Miller, R. (1998) Requirements for neural assembly formation, ignition and maintenance:
Formalization of traditional concepts. In: N. Kasabov (ed.), Progress in Connectionist-based information
systems. Proceedings of 1997 International Conference on Neural Information Processing and Intelligent
Information Systems. Heidelberg, Springer Verlag, pp. 117–120.
Wilson, J.R., Friedlander, M.J. and Sherman, S.M. (1984). Fine structural morphology of identified X- and Y-cells
in the lateral geniculate nucleus. Proceedings of the Royal Society, Series B, 221, 411–436.
Winer, E. and Gardner, H. (1977) The comprehension of metaphor in brain-damaged patients. Brain, 100, 716–727.
Winer, J.A. and Larne, D.T. (1987) Patterns of reciprocity in auditory thalamocortical and corticothalamic connec-
tions: study with horseradish peroxidase and autoradiographic methods in the rat medial geniculate body.
Yordanova, J. and Kolev, V. (1998) Single sweep analysis of the theta frequency band during an auditory
oddball task. Psychophysiology, 35, 116–120.

19 Discussion Section
Robert Miller and Almut Schüz
This chapter arose in the following way: when chapters had been submitted, we listed a set
of issues for each chapter which would merit further discussion amongst the contributors.
Following this, we drafted individual letters to chapter authors, posing one (or several)
questions raised by their chapters, and invited authors’ comments. From the responses
received, we have compiled the chapter that follows. For most of the issues we raise, we
consider ourselves to be relatively impartial editors, drawing from the contributors their
further insights into the questions we pose. Here we are identified as “Editors”. For a few
topics, however, we are certainly not unbiased, but are protagonists in the debate. For such
issues we identify ourselves as “Schüz” or “Miller”. We have broken the discussion into
sections, starting with detail and finishing with broader scientific and philosophical issues.
1. SPECIFIC EXAMPLES WHERE CYTOARCHITECTURE MIGHT HAVE

A BEARING ON CYBERNETICS OF CORTICAL TISSUE
Editors: In the chapter of Hellwig, in a variety of areas, the derivation of myelin patterns
from cytoarchitectonic data was elegant and convincing. However, the principles used
concerned only the stripes of Baillarger (i.e. horizontal collaterals). We asked Hellwig if
he had any suggestions about the fibre systems which could contribute to the other differ-
ences in myeloarchitectonics between areas, such as those in the total amount of myelin,
or the less spectacular differences in the horizontal stripes higher up in the cortex (layers
I and III). In particular, we asked if myelination in long cortico-cortical connections afferent
to an area could also make a contribution to the observed pattern of myelination in an area.
Hellwig: I think that the overall amount of myelin depends on both horizontal and vertical
fibres, the latter consisting mainly of afferent and efferent cortico-cortical connections.
As to myelin in upper layers (layer 1 or the stripe of Kaes-Bechterew in upper layer III),
I simply do not know where it is derived from.
Editors: We also had a somewhat technical question regarding the observer-independent

cytoarchitectonic mapping of Amunts et al. In one of the papers cited in their chapter
(Amunts et al., 1999), the statistical measure used to describe the laminar profile of each
area, documented by automated methods, showed greater intersubject variability within an
area than interarea variability across subjects. We asked Amunts/Zilles, if this is correct,
and also asked if, in any way, it weakens the robustness of the automated definition of
areas, and the borders between them.
459
460 Robert Miller and Almut Schüz
Amunts/Zilles: In some cases of non-primary cortical areas, we have found that inter-
subject variability in cytoarchitecture of one cortical area (e.g. areas 44) can be greater
than the differences in cytoarchitecture between this and a neighbouring area of one and
the same brain. This was shown for areas 44 and 45 which are very similar with respect to
their cytoarchitecture in layers I-III and V-VI: both have large pyramidal cells in deep
layer III, and large pyramidal cells in layer V (somewhat smaller in layer V than in layer
III). Both show a pronounced columnar arrangement of cells. The major difference
between them is that area 44 is a dysgranular region (i.e. layer IV is not clearly visible
because pyramidal cells of layers III and V intermingle with granular cells of layer IV),
whereas area 45 is a granular area with a distinct inner granular layer (layer IV). Our auto-
mated approach is sensitive enough to detect such subtle differences in cytoarchitecture.
Thus, the automated detection of borders is robust. Despite the intersubject variability, the
borders between areas 44 and 45 were found in all sections and brains which we have
analysed.
The variability in cytoarchitecture of one area across subjects may be considerable, but
the difference to neighbouring areas is still visible and detectable. Such intersubject differ-
ences in cytoarchitecture may reflect not only “true”, biological variability, but are also
influenced by the methodology (e.g. post mortem delay and conditions, histological
proceeding, staining, etc.). All these sources of variability influence the cytoarchitecture
of a single cortical area across subjects, but do not impair the detection of areal borders in
a single brain.
Editors: The chapter of Jacobs and Scheibel raised interesting questions about quan-
titative anatomical relationships within different cortical areas. In principle, the total
length of the dendritic arbor of a neurone can vary either by increasing the length of
dendrites or by making more branches. We asked whether the two measures covaried, in
the areas these authors have investigated. Alternatively, did some of the areas stand out
by having complex dendritic arbors, and others by loosely but widely ramifying den-
dritic trees?
Jacobs: Overall dendritic length can increase by changes in length and/or in the number
of branches. In the cortical areas we have examined, dendritic segment length and number
do not reveal differential patterns for specific cortical regions. In the pyramidal cell
populations we have examined, each subsequent branch order (up until the 5th) tends to
increase by an average of about 20 µm in length, such that 1st order basal dendrites are
around 20 µm in length, 2nd order average 40 µm, 3rd order average 60 µm, and so on.
Fifth order and higher branches average approximately 80–100 µm. In terms of segment
number, 1st order branches average around 5 per cell, 2nd order about 10, 3rd and 4th
around 15, 5th decreasing to about 10, with subsequent decreases thereafter. This pattern
appears to hold regardless of cortical region, with regional differences being essentially
quantitative in nature (i.e. slightly longer and more numerous branches expressed con-
sistently throughout the dendritic envelope) rather than qualitative (i.e. differential
branching patterns). It may in fact be a general principle that there is an average length
that dendrites achieve before bifurcation becomes a high probability (Lindsay and
Scheibel, 1981). Nevertheless, other factors besides dendritic length or number, such as
dendritic orientation, may show differential regional variation, as has been suggested by
the research of Elston and Rosa (1998).

Discussion Section 461
Editors: Another of the questions concerns the relationship between the size of dendritic
trees and spine number. Data presented in the chapter of Jacobs and Scheibel indicated
that, as the size of dendritic trees increases, the number of spines on a dendritic tree
increases more than dendritic length. We asked if this could be due, at least partly, to
thicker dendrites. Did the authors have the impression that dendritic thickness also
increases? It might be that the number of spines (and synapses) depends on size of
membrane area rather than on dendritic length10. Admittedly, we are aware of the fact that
the quantitative relationships in the neuropil do not leave much room for thicker dendrites:
Thus, Bok (1959) has shown in a comparison in different species that the density of
neurones decreases proportionally with increasing dendritic length, and may be explained
entirely by this increase. Data presented in the chapter of Jacobs and Scheibel, on areas 10
and 18 in the human cortex, are well in accordance with this. However, a slightly larger
proportional decrease in neuronal density than the increase in length does not completely
exclude an additional slight increase in thickness. We invited Jacobs to comment.
Jacobs: We cannot answer this directly. It is unclear at this point the extent to which
dendritic thickness affects overall dendritic complexity. Thicker dendrites certainly might
have more spines per unit length, but spine density itself is an independent variable, depend-
ing on a number of factors (e.g. synaptic density, distance from soma). Two additional
issues need to be emphasised in this regard: (i) Dendritic thickness may be more of a factor
in proximal (1st–3rd order) dendritic branches, which tend to be less spine dense, than in
more distal segments. Thus, segment thickness may contribute differentially to overall
dendritic complexity, depending on which portion of the arbor is being examined. (ii)
Since one cannot clearly visualise spines directly above or below dendrites with light
microscopy, most spine measures represent underestimates, especially for thicker dendrites
(Horner and Arbuthnott, 1991). We suspect that such underestimates might be greatest in
those regions exhibiting more complex (and perhaps thicker) dendritic systems. Thus, the
observed regional differences we have observed in both dendritic and spine measures may
actually be greater than typically reported. Further research is required to determine the
relative contributions of dendritic thickness to overall neuronal surface area.
Editors: We noted the suggestion of Jacobs and Scheibel that wider dendritic fields, and
increased spine counts, in higher cortical areas might mean greater polymodal integration
of information. We certainly agree that these features mean wider integration of inputs,
but suggest that it need not imply greater polymodal integration. For instance, in their
Figure 6.1, it is illustrated that in primary somatosensory cortex, the hand region has more
complex dendritic arbors than other parts of this area. Presumably, it is integrating
information in the same modalities as other parts of S1, though no doubt with a finer-
grained representation. Perhaps polymodal integration might be achieved in other ways,
without increase in dendritic arbor size, for instance with organized batteries of cells,
arranged hierarchically, which can achieve collectively a very wide integration. We invited
Jacobs to comment on these issues.
10
In contrast to what we assumed in a former discussion on allometric relationships (Schüz and Demianenko,
1995).

Jacobs: It is certainly possible that polymodal integration can be achieved by batteries of

cells rather than by converging polymodal input on a specific dendritic array. Given the
complexity of cortical neuropil, however, it seems highly probable that both types of
polymodal integration obtain. Future research into convergence patterns is required to
address this issue more directly and definitively.
Editors: In sensory cortex, not all regions have callosal connections (a fact mentioned, for
instance by Kaas), and we were interested whether this has implications for cytoarchitecture.
We asked Kaas, as well as Amunts/Zilles, if the regions which have callosal connections
have large and prominent pyramidal cells in lamina III, which could be a cytoarchitectonic
detail with exact cybernetic meaning. (This was an idea suggested by one of us in an older
paper [Miller, 1975]).
Kaas: I agree completely that large pyramidal cells likely mark the borders of at least V1
(which have callosal connections) and possibly other visual areas.
Amunts/Zilles: von Economo and Koskinas (1925) described a special subregion of OBg
at the very beginning of area OB (otherwise known as area 18, or V2) which is character-
ised by very large pyramidal cells in deep sublayer IIIc. Clarke (1993) has found that this
subregion may correspond to a callosal-rich region. She has identified degenerating myeli-
nated axons by the Nauta technique. We could identify consistently OBg by our observer-
independent approach in cytoarchitectonic preparations (Amunts et al., 2000).
Editors: We were interested in the cybernetic (or other) role of myelination of axons in
different cortical areas, as described by Hellwig. We took note of the role he suggested for
myelin in “consolidating” an early memory trace, so that parts of the cortex (especially
primary receiving areas) come to have dedicated functions for particular stereotyped
forms of information processing. We cannot refute this idea. However, it is possible that
myelination (and its consequence of rapid conduction) has an additional role in the cyber-
netics of nervous tissue. Hellwig’s argument depends on the assumption that conduction
time in local axon collaterals is always small compared to the neuronal integration time,
even when axons are unmyelinated, and therefore is unlikely to represent temporal
information for the recipient neurones. However, there is recent evidence that conduction
time may be longer (Bringuier et al., 1999; Gonzalez-Burgos et al., 2000), and neuronal
integration time can sometimes be shorter (Carandini et al., 1996) than were apparently
assumed. If so, it could be important from the cybernetic point of view whether axons are
myelinated or not. We asked Hellwig for his comments.
Hellwig: The answer is, of course, yes. In my chapter, I highlighted the idea that myelin
may be important in consolidating early learning processes. But this does not, ofcourse,
exclude the possibility that a broad spectrum of conduction times induced by myelination
may be very useful. In this context, I would like to mention Abeles’ “synfire chains” (Abeles,
1982) which work only if spikes arrive at a postsynaptic neurone in precise temporal coin-
cidence. It may be more likely to achieve this if neuronal activity at one cortical space is
projected to another piece of cortex with a broad spectrum of different conduction velocities.
Thus, the “good” conduction velocity which leads to temporal coincidence with other

spikes may be amongst those represented. In a narrow spectrum of conduction velocities,

this might not necessarily be the case.
Editors: Jain et al. (1998) describe, in area 3b of monkeys that the cortical representation
of the sensitive surface of each digit is highly myelinated, with myelin-light septa in
between the representations of each digit. This myeloarchitectonic feature is not changed
in the adult by deafferentation. Such a specialization may have both of the functions
mentioned above—a cybernetic function (for rapid communication between parts of the
surface of each digit), and a function of “consoldation” of a style of information process-
ing, in a rather immutable form in the adult.
We can also take another specific example—area MT—for discussion of the above
point. It is a heavily myelinated area, and there is evidence that the projection to this
area from V1 consists of large calibre axons (Rockland, 1995), which are myelinated
(Anderson et al., 1998). In this context, Kaas referred in his chapter to the very large
Meynert cells in V1 of primates. The large size of their cell bodies could be due not only
to their having long axons to support, but to the fact that their projections to MT are fast,
therefore requiring large-calibre axons. Valverde (1985, pp. 250–251) also discusses them,
and mentions that “the axons of Meynert cells could never be followed . . . beyond the
initial segment.” This suggests that the axons are myelinated.
Kaas: Yes, ofcourse the axons of Meynert cells need not only to be long, but thick and
myelinated in order to conduct rapidly.
Editors: Does myelination of incoming fibres from V1 contribute to the overall heavy
myelination in this area? Is the myelination in area MT confined to horizontal bands, or
more uniformly spread (as might be expected if it was partly associated with incoming
fibres)?
Hellwig: I assume that vertical fibres contribute to the heavy myelination of MT. Among
them are obviously incoming fibres from V1 considering the literature you quoted (Anderson
et al., 1998).
Kaas: The source of myelinations in MT appears to be intrinsic. Lesions of V1 remove V1

inputs to area MT, but do not alter the myelination in that region (J. Kaas et al., in prepara-
tion).
Miller: From the cybernetic point of view, this area is concerned with detecting coherent
motion. We ask whether a consistent story can be made of the fact that its inputs from
lower visual cortical areas are rapidly conducting, and probably also myelinated. The
unifying idea here is an informal model of motion processing in MT, dependent on rapid
conduction in inputs, which therefore exhibit little temporal dispersion. This could be
necessary for exact derivation of velocity in neurones which receive convergent input
from a number of primary motion detectors (e.g. in V1). With slower-conducting axons,
which have a wider spread of conduction time amongst a population, some temporal
dispersion would occur of the signals from “primary motion detectors” in V1. Thus, the
temporal accuracy of the information coming into MT from V1, by which coherent move-
ment over a field is integrated, would be degraded.

Shipp mentions that the V5/MT motion pathway is one of the most clearly segregated
of all. We asked him if the sharp segregation of MT is necessary because its cybernetic
requirements (fast, synchronized and very phasic processing) are very distinctive?
Shipp: It certainly seems reasonable to suppose that the higher temporal resolution of
signals arriving over fatter, faster-conducting axons might be of particular importance to
an area specialised for processing motion (MT/V5). The anatomical evidence is broadly
supportive here, as all three sources of ascending cortical input to V5 (layers 4B and 6 of
V1, thick stripes of area V2 and area V3) as well as V5 itself, are relatively rich in a class
of neurones picked out by the “CAT 301” antibody label (DeYoe et al., 1990; Levitt et al.,
1994; Tootell and Taylor, 1995). This same label distinguishes magnocellular from parvo-
cellular geniculate cells. It tags a structural element found in larger-than-average neur-
ones, and thus demonstrates that these large neurones are a ubiquitous feature of a wider,
cortical motion-processing circuitry. Thus, the CAT 301 data complements the other
anatomical features (input from V1, heavy myelination) that distinguish V5 from the cor-
tex immediately posterior, areas V4 and V4A. V5 is also notably different from these
areas in its functional activities. The distinction with areas anterior to V5, such as MST
and FST, which receive input from V5 and conduct higher forms of motion processing, is
much less marked.
On the other hand, slower forms of signal also make their way to V5/MT. Input from
the parvocellular/tonic channel can be detected there (Maunsell et al., 1990; Seidemann et al.,
1999) and, in human functional imaging studies, V5 can be shown to be activated by the
chromatic channels, which convey sensations of motion in pure colour-contrast graphics
(Wandell et al., 1999). Psychophysical approaches also identify fast and slow motion
systems. One formulation invokes 1°, 2° and 3° mechanisms of motion detection (Lu and
Sperling, 1995). The 3° mechanism is considerably slower than the other two (with a cut-
off temporal frequency sensitivity of 3 Hz versus 12 Hz), and it is the only mechanism
thought to be capable of signalling the motion of pure colour stimuli (Lu et al., 1999).
Hence, combining these observations, the 3° mechanism should also feed V5, in addition
to the 1° and 2° mechanisms. So far, it is unknown whether there is any segregation of
these mechanisms within V5.
Editors: In more general terms, there is recent evidence that latency of neuronal responses
to visual flashes does not increase much as one passes sequentially from region to region
in the “dorsal stream” of visual areas, but increases much more in the “ventral stream”
areas (Schmolesky et al., 1998; Bullier and Nowak, 1995). Given this, is there any indica-
tion of greater myelination in “dorsal stream” areas than in those of the “ventral stream”?
Hellwig: Higher myelination of dorsal regions is not strong enough to “leap to the eye”,
when looking through a series of photographs of myelin-stained sections from a human
brain. However, if one was to be sure, one would need to focus on this question in more
detail.
Editors: Another more general point, mentioned by Hellwig, is that primary sensory and
motor areas are more heavily myelinated than association areas. This may serve the cyber-
netic function of preserving temporal patterns in the incoming sensory areas as faithful as
possible to the actual stimulus. In the higher areas, slower conducting axons would allow

temporal dispersion, and therefore association between components of the signal which
arrived at slightly different times. Would this make sense?
Hellwig: Yes, this makes sense to me. Strong myelination leads to high conduction
velocities and short conduction times. This might be useful in the primary sensory areas in
order to preserve incoming temporal patterns as faithfully as possible. In the higher associ-
ation areas, it is less clear which temporal patterns are the useful ones. Therefore, there
may be less myelination. The broader the spectrum of conduction velocities (in particular
when slow conduction velocities are represented), the higher the temporal dispersion of
neuronal signals. For the reasons mentioned above this may be useful.
Amunts/Zilles: We would like to emphasise that the statement “heavily” or “less heavily
myelinated” is rather sloppy and misleading regarding the anatomical situation. The diag-
nosis “heavy myelination” of a cortical area can be caused by relatively few and wider
spaced axons with very thick myelin sheets (enabling fast conduction) or by relatively
more and closely spaced axons with relatively thinner myelin sheets (enabling slower
conduction). The diagnosis of the degree of myelination is normally performed at such a
low microscopical magnification that the two conditions cannot be discriminated. More-
over, the diagnosis “heavy myelination” is often caused by the fact that the myelinated
vertical fibre bundles ascend to the upper layer III in one cortical area, and thus, lead to the
impression of a more intense myelination of this area compared to a neighbouring areas,
where the fibre bundles terminate already in deeper layer III. The degree of myelination as
diagnosed in histological sections at low magnification does not allow an unequivocal
statement about conduction velocities. The present discussion of the criteria necessary for
a more differentiated evaluation of myelin stained sections apparently does not take into
consideration the sophisticated set of criteria already used by the Vogts (Vogt, 1911; Vogt
and Vogt, 1919). Consideration of these papers would clearly improve our discussion
about heavily and less heavily myelinated cortical areas.
2. PHYLOGENY IN RELATION TO LAMINAR AND AREAL

SPECIALIZATION OF THE NEOCORTEX
Editors: Valverde et al. provide some comparative data, with specific detail about the
hedgehog. In this species the neocortex has a thick lamina I, an accentuated lamina II and
absence of lamina IV. This finding might have a relation to the argument presented in the
chapter of Miller and Maitra, where it is suggested that in the typical mammalian neo-
cortex a slab of tissue is inserted into the more basic cortex, such as that found ventral to
the rhinal fissure. This slab appears to make up lamina III (and perhaps adjacent tissue) in
the adult, and thus splits the prominent superficial cell layer found ventral to the rhinal
fissure, leading to the formation of the definitive neocortical laminae II and IV. This
hypothesis might receive support from the similarity between cells in three locations: in
lamina II ventral to the rhinal fissure (pyramidal cells with short apical dendrites), in
lamina II of the neocortex (the same), and in lamina IV of the neocortex (spiny stellates or
star pyramids). In addition, in the chapter of Hellwig, it is mentioned that cell density in
laminae II and IV seems to covary across homotypical, agranular and granular cortices.
As far as the hedgehog goes, the implication seems to be that the splitting fails to occur.

We asked Valverde if he found this account convincing from a developmental or phylo-

genetic point of view.
Valverde: The six layered fundamental pattern of cortical lamination for the neocortex
was established by Brodmann. It was widely accepted as a tool for cortical parcellation,
implying that it was present in all neocortical areas and in all mammals. In contrast to the
six-layered neocortex, those parts extending ventral to the rhinal fissure (olfactory cortex)
or those parts on the medial side corresponding to the hippocampal formation constituted
the “older cortices”, i.e. paleocortex and archicortex. They do not show six layers and are
considered to be representatives of a primitive type of cortex from which the neocortex
evolved. Brodmann’s layering design was based on his belief that the six-layered schema
derived from a developmental (ontogenetic and phylogenetic) conception, thus arriving at
erroneous interpretations. Hence, if one is restricted to using only six numerals, one often
finds baffling layering truncations such as “layer II-III”, or expansions, like “sublayers
IVa, IVb and IVc”, not to mention the cases where even Greek characters are added to
these. As emphatically noticed by Cajal, the fundamental pattern of cortical organization
must be determined through the complete study of all cell varieties in specific cortical
areas, using methods providing a complete picture of cells and fibres. The Nissl method is
only the first step in this analysis (Lorente de Nó, 1949).
The limits separating neocortical from allo- and paleocortical formations are not sharp
and, several periallo- and peripaleocortical regions have been interpreted as transitional
areas between both structures. This led Sanides (1970) to put forward a new concept of
neocortical evolution, contemplating the entire neocortex as a set of concentric rings, or
different waves of progressive differentiation extending from the most primitive cortical
areas. This progressive differentiation takes place by thickening of the cortex, with the
appearance of new layers, and eventually the appearance of granular cells (granulariza-
tion). Pyramidal cells in primitive cortices (olfactory, hippocampal and fascia dentata)
have dendrites projecting predominantly towards the pial surface. They correspond to the
extraverted pyramidal cells of Sanides (1970) which make up most of Layer II, not only in
the paleo- and archicortex, but also in transitional cortices bordering these areas. The per-
sistence of these pyramidal cells in the neocortex of certain specimens of Erinaceinae,
Chiroptera and Cetaceans indicates that the final stage of neocortical evolution has not
been reached in these specimens. In addition, it also indicates that the major afferent input
to these particular areas is made in superficial layers (I and II), where these cells extend
their main dendrites.
Miller and Maitra (this Volume) raised the interesting question whether the laminar
pattern lateral to the rhinal fissure (i.e. towards the neocortex) and ventral to it (i.e.
towards the olfactory cortex) is essentially the same. The hypothesis is largely based on,
and adds to, the former study of Reiner (1991), who made a comparison of certain neuro-
transmitter- and neuropeptide-specific cell types in the dorsal cortex of the turtles with
those present in mammalian neocortex. For reasons that are outside the present context,
the dorsal cortex in modern reptiles has been considered a forerunner of the mammalian
neocortex. Reiner interpreted his results by suggesting that the dorsal cortex of turtles
lacks the neuronal populations characterising layers II-IV of the mammalian neocortex,
suggesting also that the evolution of the mammalian neocortex was characterised by the
addition of new neurones in these layers. Miller and Maitra went a step beyond, by show-
ing that layer II of the transitional cortices is in continuity, not only with the homologous

neocortical layer but also with the densely packed cells in Layers III and IV. Their images
do not leave any doubt in this respect and, what is most definitive, cell markers that were
found in specific neocortical layers are absent in equivalent regions in turtles, suggesting
that cells composing Layers II-IV were added in the mammal.
I do not want to question these observations, since the facts are there, clearly exposed
and convincing, but, I do ask “Where the devil do these new cells come from?”. Neither
Reiner, nor Miller and Maitra explain convincingly, why and how these new cells were
added. It has been assumed that in an hypothetical mammalian ancestor, the afferent fibres
ended predominantly in the most superficial cortical layers. Our data with Golgi studies in
the hedgehog—an insectivore—(Valverde et al., 1986) as well as the studies carried out in
Cetaceans (Glezer et al., 1988) have shown an extensive input of afferent fibres in the
exceedingly thick layer I, and in the extraverted Layer II pyramidal cells, and this type of
cortical input is reminiscent of the organization found in lower vertebrates. During some
advanced stages of cortical evolution the course and site of termination of afferent fibres
gradually changed to deeper cortical layers (layers III and IV), and this may have been
a driving force for the development of intrinsic cells in these internal layers which finally
became responsible for relaying afferent cortical fibres to pyramidal cells. Therefore, we
consider that during phylogenetic evolution, at least during the time span we may reason-
ably consider, there are no new types of neurones added to the cortex, but all of them
represent secondary modifications occurring in specific target cells. An example may be
given here: when we described for the first time the “chandelier cells” in the visual cortex
of the cat (Fairén and Valverde, 1980)—a type of cell making inhibitory contacts specific-
ally on the initial axon segments of pyramidal cells—(although Somogyi [1977] had
described a similar type of cell in the visual cortex of the rat), we, and many others,
thought that we were dealing with an entirely new type of neurone which had no equival-
ent in any other species or cortical area. We soon became aware that exactly the same type
of neurone was present not only in the visual cortex, but in other cortical areas, as well as
in all mammalian species thus far studied, from insectivores to man.
In the study of different forms of cells with spiny dendrites, one gets the impression that
all of them may share a common phylogenetic origin, and that a continuum can be traced
from cells in lower forms, to the most advanced mammal. In the neocortex of the hedgehog,
a complete series of intermediate forms between the most extraverted pyramidal cells and
typical pyramidal cells can always be found (Valverde, 1986). It is possible to think that,
during evolution, some pyramidal cells lose their apical dendrites in layer I but have retained
a thin apical branch tapering at some distance from the cell body. The cell becomes stellate
in form, and the dendrites appear concentrated in more restricted volumes. This seems to be
the case of stellate or “grain pyramids” in the barrel field of the somatosensory and visual
cortices of some rodents. In the final elimination of the remnant of their apical dendrites,
the cells turn into typical spiny stellate cells like those found in the visual cortex of the cat
and monkey (Valverde, 1971). These stages of pyramidal cell evolution also involve vari-
ations in the axonal patterns, which change from long projecting neurones (hedgehog, rat,
cat) to intrinsic cells with axons remaining inside the cortex (monkey). Following this
reasoning we may assume that, in the hypothetical mammalian ancestor, the primitive
pyramidal cells (sole inhabitants) accumulate genetic variations by analogy with the hypo-
thetical creation of new genes from a redundant duplicate of an old gene, as proposed by
Ohno (1970). The outcome is that evolution may have acted to select new connections that
conferred a selective advantage upon those animals possessing them (Valverde, 1988).

Editors: Another phylogenetic question we were also interested in concerned the

so-called “association cortex”. Kaas (1987) wrote, “Generalising from cats and monkeys
it appears that evolutionary advance in brain organization is marked by increases in the
numbers of unimodal sensory fields, not by increases in multimodal association cortex as
traditionally thought.” We asked if this still holds true, in the light of more recent findings.
How does the prefrontal cortex fit into this scheme? Does it mean that evolutionary
advance involves deeper analysis within major modalities, rather than more sophisticated
combination of information from several modalities?
Kaas: The statement can still be supported, but I’m not sure it is correct. Even V1 in some
sense is multimodal, and this complicates answering the question, and much of prefrontal
cortex may be multimodal. What is clearly true is that higher primates have more stations
in their sensory systems so that processing involves more steps.
Editors: We were also interested in the cases where long cortico-cortical connections
spread uniformly across an areal boundary. In such cases, is there any indication that the
two adjacent areas have an origin in a single area of lower species.
Kaas: If connections spread uniformly across a boundary, my first impulse would be to

question the boundary.
3. STATUS OF CORTICAL “MODULES”
Editors: The chapter of Levitt and Lund, as well as evidence reviewed in some of the
other chapters, speak in favour of some kind of columnar organization throughout the cortex.
However, Levitt and Lund also hint at the limitations of too strict a modular concept. The
question then is to strike the right balance between the concept of a structural parcellation
of the cortex into a series of discrete modules on the one hand, and that of structural homo-
geneity as it is suggested by Nissl- and fibre stains, as well as by the homogeneous distri-
bution of synapses (e.g. Figure 17 in Braitenberg and Schüz, 1998) on the other hand. We
were interested in this question.
In some cases, the discrete character of a column is clear (e.g. barrels, cytochrome
oxidase blobs and stripes, ocular dominance columns, superficial part of entorhinal
cortex), and there is also evidence for discreteness (as opposed to smooth shift) with
respect to intra-areal axonal patches (Amir et al., 1993). However, one may wonder
how many of the neuronal elements at a given place participate in columnar organiza-
tion: does this organization affect all elements and layers in parallel, or are there just
a few systems with modular structure superimposed onto an essentially homogeneous
network. The columnar structure seems mainly to concern the axonal side: inputs from the
thalamus (barrels, ocular dominance columns), cortico-cortical and callosal projections
(Gilbert and Wiesel, 1989; Goldman-Rakic and Schwartz, 1982) and intra-areal axonal
patches (as in the chapter of Levitt and Lund). The dendrites (with the exception of the
barrels) seem usually not to participate (Malach, 1994). However, not all the axonal
systems seem to participate in columnar organization, e.g. the diffuse feedback connec-
tions in the visual system, mentioned in the chapter by Shipp. Also, anterograde tracer
injections usually show a rather large and dense homogeneous field of collaterals

surrounding the injection site, suggesting that patchiness is mainly restricted to the far-
reaching axon collaterals, and—judging from the fibre density—each patch probably
contributes only a small proportion of the synapses at a given place (comparable
perhaps to the thalamic fibres).
In this context, the chapter of Dinse and Schreiner includes the following quotation:
“functional maps are overlaid so to ensure that many, if not all combinations of
the different parameters are represented in cortex. Recent theoretical studies
show that geometrical factors do not constrain the ability of the cortex to repres-
ent combinations of parameters in spatially superimposed maps of similar period-
icity. Considerations of uniform coverage suggest an upper limit of six or seven
maps. A higher limit, of about nine or ten, may be imposed by the numbers of
neurones or minicolumns available to represent each feature within a given
cortical microdomain (Swindale, 2000; Swindale et al., 2000). Thus, several
feature dimensions can be expected to be represented across VI, AI, and SI. . . .
From this point of view, each neurone and each location in VI, AI, and SI can be
understood as representing a specific set of many independent variables in the
sensory environment. Topographically, each location on the cortical surface
corresponds to a specific intersection of several systematic maps.”
This quotation could even be taken to imply that (outside special areas like barrels, CO
blobs, etc.) modules are an artefact of studying only one set of axonal projections at once.
We asked Levitt if he would agree to a “moderate modular concept” in which only
a moderate proportion of the neuronal elements participate. In this connection, we were
also interested to know how many different kinds of patches contribute to a column, and if
patches of different sources tend to be in register or rather to ignore each other.
Levitt: It has long been noted that the extent of visual cortex over which the preferred
orientation of recorded neurones shifts is smaller than the diameter of a single pyramidal
neurone’s dendritic tree. Moveover, data on the nonclassical receptive field surround
indicate that orientation selectivity of a given cortical neurone can be modified dynamic-
ally by the context in which it is embedded. Hence one asks: “What does it mean for there
to be fixed orientation columns when the tuning of individual cortical neurones can vary?”
It is difficult to think of a hard edged “column” in the cortex, at least physiologically. The
concept of a module is certainly more straightforward when defined anatomically. In add-
ition, while it is also the case that not all projection systems have a columnar or modular
organization (for example certain cortical feedback projections), we have some evidence
that some of these pathways have a very precise columnar termination pattern when
studied more carefully (for example the terminations of V3 projections to layer 4B in area
V1). Moreover, one should remember that the evidence for the “less modular” connections
comes from injections several hundred microns in diameter, labelling many neurones.
The connections of single neurones might well be more precise. We feel compelled to
agree with a more moderate concept of the module, if only because we remain unsure how
best to define one. It is even more difficult to define a module than to define art; like art
we have trouble defining a module, but unlike art we remain unsure if we would recognize
one even if we saw it.

We agree completely with Dinse and Schreiner, and Swindale, that our uncertainties as
to the definition or rigidity of modules in the cortex may reflect simply the problems of
studying one parameter at a time. It might be that if one studied the organization of cells
coding several parameters at once, or defined by several anatomical constraints, one
would find a stricter modular organization. However, pondering the analysis of clustering
in a high-dimension space makes our brains spin. Thus, we are still left with the first prob-
lem we posed, that even for a single parameter the dimensions of a module can be smaller
than the extent of a single neurone’s dendritic field. These viewpoints can be reconciled
only by using newer data about the previously unappreciated biophysical complexities of
the individual cortical neurone, suggesting that each cell may consist of multiple inde-
pendent functional input domains throughout the dendritic tree. This makes for a somewhat
different notion of a module.
4. “HIERARCHY” OF AREAS
Editors: We raised with Young the general topic of “hierarchy”. It was clear from his
chapter that there is some kind of hierarchy, but it appears to be a very mild one if there
are thousands of possible solutions.
Young: Well, no! This is an interesting point. If you compare the number of violations for
optimised shuffled data (around 120), with the optimised real anatomical data, the number
of violations for real data is astonishingly low. The regularities are really very strong indeed.
Nature really cares about these relationships. So, the system is, surprisingly, strictly hier-
archical.
“Indeterminacy” (the result that millions of different solutions fit the anatomical data
equally well), is not a “real” problem. It is inherited from qualitative anatomical data
(since anatomists hate to count), via qualitative relational constraints (e.g. “higher-than”),
via an integer cost function (number of violations of a candidate hierarchy of the relational
constraints). It is abolished by quantitative data, as we have shown with Henry Kennedy’s
data (Barone et al., 2000). So, there is a single optimal structure that can be discovered. It
will be a bit complex, because there are subtle differences in the hierarchy between repre-
sentations of the visual periphery and those for central vision, and there are asymmetries
in the apparently reciprocal relationships between several stations—but I don’t believe
there is any meaningful functional interpretation of “indeterminacy”.
Editors: Obviously it is not very clear “who is above whom”. It seems to be a complex
hierarchy, like in a research institute in which the director can be uppermost or very low in
the hierarchy, depending on the problem to be solved (writing an article, repairing a broken
water tube, or chasing a computer virus). Thus, an occasional visitor to this institute can
come up with many different hierarchies if the criteria is the flow of information between
people.
Young: Perversely, notwithstanding my remark above, you’re right about that! The rela-
tionship between hierarchical connectivity and information flow (dynamics) is a complex
one. Both complex dynamics (including simultaneous onset latencies) and hierarchical
connectivity can be properties of the same system (Scannell et al., 1999). “Information

flow and dynamics” is related to hierarchy, but in a complex manner. Hierarchy is only
one property of the visual network. It also has topological and hodological properties,
among several others, and these other aspects of organisation are also important in deter-
mining the system’s dynamics. The misapprehension in the recent suggestions that the
visual system can’t be hierarchical because some dorsal stations respond simultaneously
to flashes is that “hierarchical” equates to “serial”.
So, at least three non-equivalent aspects of the anatomical organization of the visual
system are known, each of which would be expected to be reflected in rather different
aspects of the system’s function.
First, laminar hierarchy relates to strong anatomical regularities in the laminar origins
and terminations of projections, and might be expected to reflect itself in the elabora-
tion of stimulus preferences in stations at different levels of the hierarchy. Laminar
hierarchy does not specify the number of stations through which signals must pass to
reach a particular area: there are very many connections that pass between areas on
non-adjacent levels (Felleman and Van Essen, 1991). Nor is it related directly to
differences in the area-to-area pattern of inputs to different structures. Consequently,
it would not be expected that this aspect of organization should reflect itself directly
in timing of activity onsets in different stations, nor directly in the similarity in function
of different stations.
Second, hodology relates to the number of stations, or processing steps, through
which signals must pass to reach one structure from another. Hodology specifies the
sets of relationships between stations, which are different from those specified by pat-
terns of their laminar connectivity; and hodology is not related directly to differences
in the area-to-area pattern of inputs to different structures. Consequently, it would not
be expected that this aspect of organization should reflect itself directly in the elab-
oration of receptive field properties in different stations, nor directly in their similar-
ity and dissimilarity in function.
Third, the topology of the system arises from the patterns of area-to-area connectivity
within the system. This aspect of organization need not be related simply to either
laminar hierarchy (Young, 1992) or to hodology (Young et al., 1994). However, an
area’s information processing functions are constrained by the inputs they receive,
and the destinations of their outputs (e.g. Young et al., 2000b). It would hence be
expected that the more similar is the set of areas giving rise to inputs to two struc-
tures, and the more similar is the set of stations to which they send outputs, the more
similar should be their functional roles (Young et al., 2000). Hence, topology should
be related more directly to the functional similarity of different stations, and less
directly to the elaboration of receptive field size and selectivity, or to the onsets of
activity in different stations.
The term “hierarchical”, carries a special and precise meaning in the context of the
organization of the visual system (Felleman and Van Essen, 1991; Hilgetag et al., 2000a).
It refers to aspects of organization specified by regularities in the patterns of laminar
origin and termination of projections. By contrast, the term “serial”, also carries a specific
meaning, relating to a property of the area-to-area connection patterns that is apparent
both in the system’s topology and hodology. “Hierarchical” does not imply “serial”, nor

vice versa. Indeed, quite different architectures can be specified by the combinations of
these terms. It is possible to envisage an architecture that is both hierarchical and serial, as
for example in sensory cortex; an architecture that is serial but not hierarchical, as perhaps
in prefrontal cortex; one that is hierarchical but not serial (perhaps the dorsal pathway in
the visual system); and an architecture that is neither hierarchical nor serial (as perhaps in
classical auto-associative nets, or the CA3 field of the hippocampus).
5. STATUS OF AREA 4
Editors: Within the overall topic of hierarchy of cortical areas, we were particularly
interested in area 4 and its position in the level of cortical organization (loosely, in the
“hierarchy”; but see the distinctions in the use of this term from Young [above]). It is
interesting to compare the various anatomical criteria which might be used. Most of them
seem to place area 4 at a low position in the hierarchy, but the situation is a bit equivocal:
the large amount of myelin, as well as (according to the measures by Jacobs and Scheibel)
the fact that area 4 does not have particularly large dendritic arbors in supragranular layers
(though larger than in primary sensory areas) place it at a low position in cortical
hierarchy. On the other hand, the low cell density (see chapter by Jacobs and Scheibel)
gives area 4 a place at a very high position. Jacobs and Scheibel point out that the
increased size of dendritic trees correlates with a decrease in the density of neurones.
However, the decrease in cell density seems to be far more dramatic than can be explained
by the size of dendritic arbors: area 4 belongs to the areas with very lowest cell density.
Since a decrease in the density of neurones cannot be due to anything other than an
increase in neuropil, this suggests that in area 4—in contrast to other areas—there is
a more than proportionate increase in the axonal contribution to the neuropil, or perhaps to
the abundance of large calibre myelinated axons in the neuropil. However, as far as we
know, not enough data are available to decide if there is an overproportionate increase in
axonal volume in area 4 which could explain the low cell density.
Interestingly, the hierarchical position of area 4 is also somewhat equivocal from the
axonal side, as mentioned in the chapter by Levitt and Lund. In the visual system it has
been shown that, away from the primary sensory area, axonal patches tend to become
larger, and this is even more so for the overall size of the patch field produced by a single
tracer injection (Amir et al., 1993; Lund et al., 1993: see chapter of Levitt and Lund, this
Volume). In area 4, however, the patch size is larger than in the primary somatosensory
areas (as is the size of the dendritic trees) but the size of the total patch field is not (Levitt
and Lund, this volume; and personal communication). Also, from the overall arrangement
of the cortico-cortical connections between areas, Young suggests a low position in
hierarchy. We invited several contributors to comment on this, and/or on the “location” of
area 4 in cortical hierarchy in general.
Jacobs: It would not be surprising if area 4 had a larger axonal input per unit area than
areas 3, 1 and 2, especially since the motor cortex constitutes a kind of final common path-
way out of the cortex. In that sense, it might well be thought of as a “higher hierarchical
centre,” but in a somewhat different sense from those in the sensory-related realms. Most
high hierarchical cortical centres have a good deal of re-entrance, that is, mutual feedback
connections to other contributing systems. Motor cortex might first be thought of as not

having this much feedback output (except to brain stem and spinal cord), but when one
considers all the “afference copy” lines it must send to sensory areas to alert them of
“what is about to happen”, it can be considered a higher cortical region as well.
Levitt: We acknowledge the conflicting data concerning the classification of area 4 as

a primary or higher-order cortical area. We also recognize that cell density in area 4 is
among the lowest in the cortex, and note the question whether an overproportionate
increase on the axonal side could explain the low cell density. However, our measure-
ments did seem to indicate that the spacing of terminal patches from intrinsic connections
was scaled to the basal dendritic field of superficial layer pyramidal neurones, as in all
other areas examined, i.e. they are not more closely spaced than would be predicted by
dendritic field size. However, it is conceivable that these patches might contain a higher
density of boutons or ramifying fibres. To our knowledge, there are no counts of terminal
fibres or synaptic boutons that would allow such a comparison of area 4 with other cortical
areas, so we cannot comment on whether the lower cell density in that area is offset by
other factors. The problem is that purely anatomical criteria for classification of areas,
either singly or into a hierarchy, ultimately depend on some form of functional measure-
ments for confirmation. So we turn the question back to students of motor function to ask
if there is anything distinctive about motor cortex that might shed light on the anatomical
anomalies. One representative study illustrating that we do not completely understand
how to classify area 4 is that of Schieber and Hibbard (1993). These authors showed that
single M1 neurones were active during movements of several fingers, and that regions of
M1 active during movement of different fingers overlapped extensively. We tend to think
of primary cortical areas as having a precise topographic representation, with higher areas
having less precise topography (due to bigger receptive fields or genuinely degraded
maps). If this were so in M1, we might expect single M1 neurones to fire only during
movements of a particular finger, or for given regions devoted to a single digit. One inter-
pretation of the data from this study is that M1, nominally a primary area, contains a map
more like that of a higher order area.
Young: The neuroinformatic analyses of connectivity definitely say that area 4 is low in
the hierarchy, as you describe. That’s just the way the results are, so there’s not a lot
I can say in mitigation. The internal histology might more readily be expected to be the
anomaly, since area 4 is certainly a bit unlike sensory areas in a functional sense.
Miller: One resolution to this issue is that area 4 is an exception to any general scheme,
and is best considered together with the primary somatosensory cortex, as part of an inte-
grated cortical “organ” in this region. One can suggest this because area 4 lacks lamina IV
(a principle input lamina) while area 3 (part of the primary somatosensory cortex) has very
few pyramidal cells in lamina V (the main output lamina). Thus, in this combined cortical
organ, inputs would be dealt with in area 3, followed by direct transfer of information by
short cortico-cortical links to area 4, which then elaborates the output. In this context,
I note that Young writes: “One feature of the (somatosensory) system, however, that it is
rather unlike either the visual or auditory systems, in that parts of the system comprise the
cortical motor system, viz: the medial supplementary motor area (SMA), the premotor
cortex (area 6), and the primary motor cortex (area 4). These motor areas are arranged in
what appears to be a hierarchy, in which the primary motor area is associated with primary

somatosensory areas, the premotor area with higher somatosensory areas, and SMA with
still higher ones”.
We also noted that Rademacher had referred to the “extremely low neuronal density in
layer V” in area 3. As part of the evaluation of the above idea (of sensorimotor cortex as
an “integrated cortical organ”), we asked him whether the low pyramidal cell density in
lamina V of area 3 is typical of primary sensory cortex generally, or just of area 3 in the
somatosensory cortex?
Rademacher: It is difficult to set up rules, since individual variations are quite striking.
However, for primary auditory cortex, cell density in layer V is approximately 10% lower
than in layers II to IV (mean values from 10 brains).
Jacobs: Given the patterns of interconnectivity that characterise primary somatosensory

and motor cortex, your suggestion that these two regions together constitute a cortical
module of complementary cytoarchitectonics is an intriguing one. Certainly, from a
developmental perspective, these regions mature hand-in-hand with each other
(Chugani et al., 1987). Functionally, area 4 works very closely with areas 3, 1 and 2 by
synthesising various sources of input, including processed proprioceptive and tactile
information from layer II and III pyramidal neurones in area 3, 1 and 2 (Porter, 1997),
prior to initiating smooth voluntary movements. Thus, from cytoarchitectonic, develop-
mental, and functional perspectives, these two regions may indeed seem to constitute
a cortical module.
Amunts/Zilles: The concept of cortical hierarchy is related to connectivity, and the

amount of connectivity is reflected by the amount of neuropil. The cortex is roughly sub-
divided by layer IV into a supragranular and an infragranular part. Layer IV is the major
input layer of thalamo-cortical afferents. Layers I-III are targets of intracortical connec-
tions. Thus, an increase in neuropil in layers I-III may indicate an increased complexity in
intracortical connectivity and a relatively high position in the hierarchy of cortical areas.
Layers V and VI are mainly output layers and, therefore, cannot be used as separate indi-
cators for cortical hierarchy.
Based on our previous quantitative cytoarchitectonic analyses, area 4 has the lowest cell
packing density in the cortex, both in supra- and in infragranular layers i.e., the amount of
neuropil of area 4 is high. Using this criterion, it follows that area 4 reaches a high hier-
archical order. In addition, whereas areas 3, 17, 41 (etc.) receive strong sensory input, area
4 is a motor area beside all the converging input from cortical areas. As such, it handles
information which is most complex, that is area 4 integrates information of different
modalities and cortical regions (in other words, it is high in the hierarchy). Area 3 has
a much higher packing density in supragranular layers and therefore, a low amount of
neuropil, or connectivity (and so is low in the hierarchy). The two areas are thus (concep-
tually) at “opposite ends” of the cortex. There are only a few areas in the cortex which are
so different in architecture as are these two areas. Finally, behaviour as defined by motor
activity is controlled by area 4. Thus, area 4 is contributing as the last cortical instance to
the ultimate goal of the brain, i.e. behaviour.
Schüz: I would like to add here some personal communication by Braitenberg. The low
density of neurones in area 4 may be due to the particularly high degree of myelination

submerging the pattern of horizontal striation characteristic for other areas. In this sense,
there may indeed by an overproportionate increase in axonal volume in this area.
6. INTERPLAY BETWEEN LAMINAE, AND BETWEEN CORTEX

AND THALAMUS
Editors: In the chapter of Miller (“Wheels within wheels”) a hypothesis was presented
that pyramidal cells in lamina V could act to support or “prime” the activations of connec-
tions within laminae II/III of the cortex, this enabling cell assemblies to form which could
be ignited, without this implying uncontrolled explosions of activity. Any cell in lamina V
might serve its “priming function” for a variety of connections within laminae II and III,
and thus could be part of a number of independent cell assemblies. In the chapter of Shipp,
one part refers to integration by “OR” logic, as follows: “A successful stimulus need pos-
sess just one of the requisite features, and the form selectivities need not be concordant.
This may seem a bizarre conception, but it may be closer to the experimental picture: the
paper illustrates a superficial V2 neurone driven by either (a) rightward motion of an
achromatic slit, or (b) flashed presentation of a relatively large, blue spot; the effect of
amalgamating these features in a single stimulus is not mentioned (Tamura et al., 1996).
It is perplexing to think how the ambiguity inherent in the signal from this cell might be
interpreted by higher visual centres.”
The chapter of Miller (Wheels within wheels) proposes that interplay between
laminae II/III and V is involved in the functioning of cell assemblies. The role pro-
posed for lamina V cells, as a primer for many assemblies, themselves located mainly
in superficial laminae may correspond to what Shipp refers to as “OR” logic. From the
cell assembly viewpoint it is not at all puzzling, though it may be if one imagines that
single neurones are, in themselves, the vehicle for information representation. How-
ever, it is noted that Shipp places the “OR” logic cells in superficial laminae which is
different from the hypothesis in “Wheels within wheels”. Shipp was invited for his
comments.
Shipp: My interpretation of the dual colour and motion selective cells (DCMS cell for
short) was influenced by the ideas of Wolf Singer and colleagues involving binding-
by-synchrony (Singer and Gray, 1995; Singer, 1998). However, having read your account
I can see how the data also fits your perspective.
The “AND” logic seems to be required for a DCMS cell that might be signalling a fea-
ture conjunction to a higher area. However our data suggests that colour cells in thick
stripes of V2 are rarest in lamina III, which houses most of the cells projecting to higher
areas, and where directionally-tuned cells are most common. Hence some other explana-
tion for a DCMS cell is required. “OR” logic seems to fit the bill for a cell whose role is to
link up a synchronised assembly of direction-specific cells and of colour specific cells
(located external to thick stripes) that are all responding to the same object. The effect of
synchrony is to let this assembly win a competition with other potential assemblies,
responding to other coloured moving objects (perhaps corresponding to selective attention
to the object in question).
If I’m reading you correctly, your proposal is more to do with how any assembly gets
ignited and keeps going, rather than with which one gets ignited. However, the DCMS cell

is like the proposed lamina V cell with regard to receiving more indiscriminate input,
resulting in its dual selectivity. Moreover, although there is, as far as I know, little know-
ledge about such a cell’s resting threshold, the “OR” property means that either colour or
motion inputs acting alone can tip it over threshold. Thus, I can agree that a DCMS cell in
some senses fits your idea about recruiting additional neurones into an active assembly,
allowing it to achieve a critical mass.
Although I didn’t emphasise it in my chapter, our V2 data does suggest that lamina V is
less selective than lamina III. Our actual conclusion is that the distinctiveness of stripes
(i.e. thick vs thin vs interstripes, measured in terms of % selectivity for direction, orienta-
tion, colour or length) is maximal in lamina III, and declines both in deeper layers, and
more superficially. In our data, lamina VI is where stripes are least distinct; lamina II and
V are intermediate, and lamina IIIB is where they are most distinct.
Thinking liberally, I like to picture the less selective neurones as part of a non-commit-
ted population that the more committed neurones compete to recruit. (Apart from the
outer layers of V2 [i.e. layers 2, 5 and 6], the non-committed neurones could be in the
pulvinar, and/or the superior colliculus, or caudate nucleus or other structures.) The
committed neurones need support from the non-committed neurones in order to sustain
long-lasting activity of the kind that results in conscious awareness. Thus, the non-
committed neurones act as a limited central resource, equating to the restricted capacity
of the focus of attention.
Editors: That seems reminiscent of the relationship proposed between neurones in lamina
II/III (“committed”) and lamina V (“uncommitted”).
Shipp: This brings me to the reference in the article “Wheels within wheels” to cortico-
thalamic connections. I have just completed a study of connections from cortical areas V4
and V5 to the pulvinar. The two areas maintain almost separate pulvinar fields. However,
in our material, their mutual cortico-cortical connections are also meagre to non-existent.
By contrast, both areas have strong connections with V3, and the pulvinar fields of both
V4 and V5 overlap with that of V3. Hence, further evidence is provided that the pulvinar
interlinks cortical zones that are themselves interconnected, as your article assumes.
Furthermore, analysis of past literature makes it clear that the same is true for the whole
string of areas constituting the ventral pathway (V1, V2, V4, TEO, TE).
7. DEVELOPMENTAL FORCES DETERMINING GYRIFICATION,

AND CYTOARCHITECTONIC DIFFERENCES
Editors: Rademacher introduced questions about the process of gyrification, including the
following: “ . . . one may postulate that an ontogenetic process exists that can produce pro-
found morphological shifts as determined by random environmental factors without much
genetic change.” We asked if it could be that the nongenetic part of the determination of
gyral and sulcal patterns (such as determines the difference between monozygotic twins)
is actually just random (a purely statistical difference, emerging as part of a complex
developmental process), rather than a random environmental effect?
Rademacher: I agree with that suggestion.

Miller: Rademacher also mentioned the “tension model of gyrus formation”, based on
minimizing axonal length. This has much to recommend it. However, there is a possibility
that conduction time in axons has an important significance in the cybernetic operations
carried out by cortical tissue (e.g. Miller, 1991, 1996). Conduction time is determined by
both axon calibre and axonal length. Thus a general rule that the gyri form to minimise
total axon length may limit the available range of axonal conduction times. Indeed there
are some axonal bundles (such as the fornix, or the arcuate fasciculus) which appear to
take deliberately circuitous routes, as if to increase total axonal length and therefore con-
duction time. If these arguments are correct, the third alternative proposed by Rademacher
for gyrus formation (a complex process of “gyrification”, the resultant of many subsidiary
developmental processes) may sometimes apply. We invited his comments.
Rademacher: As discussed in the main text, according to the “tension model”, minimiz-
ing axonal length is a fundamental process in gyrus formation. However, this may not be
the only force which drives the final shaping of the convoluted brain. Since axon calibre,
axonal length, and degree of myelination put structural constraints onto functional
measures such as the computational speed of network processes, application of the
tension model might limit the range of possible axonal conduction times (Miller, 1991,
1996). The latter may represent an important parameter conveying relevant information
in the “fourth dimension”. Information theory surely provides a theoretical framework
for addressing this question concerning the detailed nature of neural codes. Anatomical
information also suggests that there is indeed a possibility that conduction time in axons
plays a role. The fundamental question is whether such a mechanism limits a general
principle supposed to regulate the range of fibre tract length; or, alternatively the relation
between structure and “time” may be hardwired exclusively in distinct “centres” or
networks, which analyse this modality specifically, by analogy with other cognitive oper-
ations such as language processing.
It has been shown recently that there may be something like an “internal clock” in a dis-
tributed network, composed of the dorsolateral prefrontal cortex, the posterior part of the
inferior parietal cortex, the posterior cingulate cortex and the basal ganglia (Onoe et al.,
2001). How this system accomplishes a temporal monitoring process in time perception
is not known to date. However, I would not expect that a flexible neural system could be
hardwired to the degree that time is translated in a linear fashion into axonal length.
Neuronal firing rate and interspike interval distribution provide patterned information
which can be represented on much more variable time scales than the distribution of
axonal length could ever permit. It is also known that neurones can modify their sensitiv-
ity towards incoming (afferent) signals which suggests that control strategies may change
for different tasks. Thus, I would rather imagine some kind of gating mechanism, as has
been described for the thalamus, as a relay centre controlling transformation from tem-
poral to rate coding (Ahissar et al., 2000), with the hippocampus as a phase and interval
comparator for oscillating signals (Glassman, 2000). The climbing fibre system of the
cerebellum appears also to be involved with timing. Some insight into precisely how
such timing processes may work is available for the primate visual system (Victor,
2000). In the human auditory system, the temporal “envelope” of sound is processed by
distinct and hierarchically organized series of neural substrates (i.e. superior olivary
complex, inferior colliculus, medial geniculate body and Heschl’s gyrus) (Giraud et al.,
2000).

Schüz: I agree that long conduction times play an important role for cortical functions and
that they do exist (Miller, 1975). In spite of this, the tension model (van Essen, 1997)
seems plausible to me. Gyrencephaly is a feature of large brains, so their sheer size may
give enough room for long conduction times without any detours. Large brains may even
run into trouble with respect to short conduction times: in order to make the human brain
as fast as a small brain (by using thicker axons), one would have to blow it up to an
impossible size (Ringo et al., 1994). Indeed, in large brains, only some axons seem to
compensate for the longer conduction times by a correspondingly larger thickness (Jerison,
1991; Schüz and Preißl, 1996). Also, the system of long-range connections is arranged as
if the cortex would strive to save axonal length (Scannell et al., 1995; Young, this Volume;
Schüz and Braitenberg, this volume).
Editors: In the cross-modal primary auditory cortex A1 (studied by Pallas), which has
acquired some of the properties of the visual cortex, we were interested whether there is
any evidence for the existence of cytochrome oxidase “blobs” typical of visual cortex, or
for radial arrangement of columns of orientation selectivity? We ask this since a simple
intracortical mechanism has been postulated which explains many of the features typical
for the primary visual cortex, in particular orientation selectivity, and the radial arrange-
ment of orientation columns (Braitenberg and Braitenberg, 1979; Braitenberg, 1985, 1992):
inhibitory centres which—in primates—should be located in the cytochrome oxidase
blobs.
Pallas: Cytochrome oxidase is an activity marker, and is found preferentially in cells

whose metabolic activity is high. The study of its distribution is thus a very indirect way
of inferring the organization of functional pathways. I am not aware of any study report-
ing on cytochrome oxidase blobs or stripes in ferrets, rewired or otherwise. Cytochrome
oxidase blobs have been well-described in primates, but until 1995 had not been found in
other mammals. In 1995, Murphy et al. (1995) demonstrated blobs thought to be related
to ocular dominance columns in cat primary visual cortex. Boyd and Matsubara (1996)
showed that the blobs in cats were co-extensive with patches of inputs from the C-layers
of the dLGN, indicating an origin from W and especially Y pathways. The work of
Shoham et al. (1997) supports this interpretation. Although blobs have not been reported,
eye-specific fields have been reported in cross-modal ferret MGN by Angelic et al.
(1997). Our unpublished data suggest a related pattern in A1. Charm et al. (2000) have
recently demonstrated “pinwheel” type radial arrangements of neurones with differing
orientation tuning characteristics in cross-modal ferrets. They also show physiologically
that similarly tuned neurones are interconnected, supporting our anatomical data on
horizontal connections. These findings together show that modular organization, typical
of visual cortex, can apparently be induced by early visual input in a foreign cortical
region.
Editors: Pallas also mentions the work of Dehay et al. This was also referred to in another
chapter (Rademacher), as showing that early bilateral enucleation leads to reduction in the
size of area 17 and corresponding increase in the size of area 18. We were interested
whether the distinctive features of primary sensory cortex (especially the distinctive lam-
ina IV) are dictated by patterned sensory input, with the neighbourhood relations possible
in a one- or two-dimensional sensory surface.

Pallas: The laboratories of Rakic and Kennedy argued some time ago about whether or
not bilateral enucleation in foetal macaques could result in the creation of a new occipital
cortical area, or an expanded area 18. (They agreed however that enucleation results in
a smaller area 17) (Rakic, 1988; Dehay et al., 1991, 1996; Rakic et al., 1991). It is my
opinion that this question remains unresolved, because the enucleation itself makes it dif-
ficult to define cortical areal borders. Bilateral enucleation is a severe deafferentation
that causes massive cell death in the lateral geniculate nucleus when done early in devel-
opment. Death of LGN cells leads to substantial excess cell death in primary visual
cortex, particularly in layer 4 (Windrem and Finlay, 1991), in part because thalamic
axons from other parts of thalamus tend not to sprout and innervate denervated cortex
(Miller et al., 1991). Finlay and Slattery (1983) reported that the number of layer 4 cells
in different cortical areas is inversely related to the amount of thalamo-cortical input, the
area receives during normal development. This does not necessarily imply that patterned
sensory input is necessary for creation of a distinctive layer 4, only that physical input of
some sort is required. It is not surprising that bilateral enucleation of foetal macaques
would lead to striking changes in visual cortical cytoarchitecture (such as the loss of
layer 4, the primary identifying characteristic of area 17), and a concomitant shrinkage in
its area, leading to changes in those features of extrastriate cortex dependent on input
from 17. Ablation of one cortical region is also well-known to cause changes in the
typical convolutions of neighbouring regions, as they expand to fill the empty space.
Without layer 4, it is difficult to define the border between areas 17 and 18 precisely.
Although it is possible that enucleation creates a new cortical area, it does not seem to be
the most parsimonious explanation. This could be clarified by injecting tracers in, or
recording from, the “area X” of Rakic (or the expanded “area 18”), but to my knowledge
this has not been done.
Editors: We were interested in whether the transmutability of primary sensory areas,

shown well by Pallas, could also involve non-sensory areas. Could a non-sensory area be
converted into a sensory area (with its distinctive lamina IV) or vice versa? In terms of an
actual experiment, would it (for instance) be possible to persuade the mammilothalamic tract
to innervate a sensory thalamic nucleus, and observe the effects on the cortical projection
region?
Pallas: As mentioned above, sensory cortex is typified by a thick layer 4. If non-sensory

cortex received heavy thalamic input, it is conceivable, from what we know now, that
a thick layer 4 would appear. Whether this would convert non-sensory to sensory cortex or
vice versa is unknown but seems possible. The actual reason that we can “persuade” the
retina to invade the medial geniculate nucleus is quite fortuitous: the optic tract passes by
MGN on its way to its normal targets and presumably can read signals emanating from the
deafferented MGN in cross-modal animals. Only if the mammilothalamic tract normally
passed close to sensory thalamus would innervation be likely.
Editors: Given that the thickness of lamina IV appears not to be a result of differential
cell loss during development (Finlay and Slattery, 1983), and since rearing in total dark-
ness does not affect neurone numbers in lamina IV (Borges and Berry, 1978; Gabbott et al.,
1986), it is possible that the presence of a thick lamina IV in sensory regions reflects
a pre-determined developmental program.

Valverde: I entirely agree with the proposal that the specification and number of Layer IV
cells in various sensory areas reflects (and depends on) a pre-established developmental
program. This leads us directly to consider the paradigm of areal and cellular specification
at relatively early stages of embryonic development, which do not depend on the inter-
action with specific thalamic, or other sensory, inputs. An instructive example is found in
our recent studies in the “Pax-6” mutant mouse (Jiménez et al., 2000). For many years, it
was considered that the development of the olfactory bulb depends on the arrival of olfact-
ory fibres from the nasal placode. Absence of olfactory receptor cells results in animals
that lack olfactory bulbs (as is the case in the homozygous mutant Pax-6 mice). However,
we were able to demonstrate the presence in these mutant mice of an olfactory-bulb-like
structure inside the brain, containing cells hodologically and phenotypically similar to
the mitral cells of the olfactory bulb in wild type animals. This argues in favour of the
occurrence of intrinsic subsets of functional domains that exist and develop—albeit
abnormally—even when the establishment of the proper afferent connections was lacking.
This links beautifully with the suggested existence of a “protomap” in the embryonic
cerebral cortex, which may be responsible not only for areal specification but also for
defining the identity and number of particular cell groups (Rakic, 1988). In fact, recent
gene expression studies have demonstrated that, during embryonic development, cells appear
to be committed to become specific phenotypes and layers well before they reach their
final position in the cortical mantle (see Donoghue and Rakic, 1999 and references therein).
Pallas: Rockel et al. (1980) proposed that the decision for determining between pyramidal
or stellate morphology occurs after arrival of thalamic input. If no or little thalamic input
arrives, cells which might have been destined to end up in lamina IV, as stellate cells, end
up instead as pyramids in lamina II/III, and then they or other cells die off in excess, to
produce the thinner cortex seen in non-primary areas. This idea was supported by Windrem
and Finlay (1991). They showed that thalamic ablations on P1 produced an increased
incidence of cell death in lamina II/III, an absence of lamina IV and its stellate cells, and
a reduced thickness of the affected cortex. In relation to Valverde’s comment, the amount
of thalamic input can be pre-determined in a sense and can thus determine the cytoarchi-
tecture, but by information extrinsic to cortex (thalamo-cortical axons). There is also data
which is relevant here, that lamina IV is twice as thick in binocular cortex as in monocular
cortex (Beaulieu and Colonnier, 1983). Also supportive are the multiple studies using the
DNA-alkylating agent methylazoxymethanol (MAM) to ablate a subpopulation of cortical
precursors. If one kills those that would have given rise to deep layers, the system read-
justs and nonetheless produces a cortex with all of its layers (Gillies and Price, 1993),
again suggesting that laminar identity and possibly morphology can be regulated.
Jacobs: To state the obvious, those of us who work in the cerebral cortex need to
remember that the cortex does not exist apart from subcortical structures such as the thal-
amus, anymore than the brain itself functions independently of the body. As such, ultimate
understanding of cortical cytoarchitectonics, hierarchies, and interconnections cannot be
obtained without a complete understanding of subcortical contributions to information
processing.
Editors: We asked Pallas one further question about the processes involved in areal
differentiation: is it possible to tell whether the sorts of activity-dependent plasticity

which Pallas has shown rely entirely on local influences between cells (e.g. Hebbian
processes) or might they involve also action at a distance (e.g. diffusion of chemical
markers).
Pallas: Our manipulation involves only a change in the spatiotemporal patterning of activ-
ity reaching MGN. We did not change the identity of the thalamocortical fibres carrying
the patterned activity. If changes in the activity pattern lead to alterations in diffusible
markers then yes, action at a distance is possible. At present we do not have data to
address this point although we are currently performing experiments that would address it.
Recent experiments by the Chapman and Katz labs (Chapman et al., 1999; Crowley and
Katz, 1999; Chapman, 2000; Crowley and Katz, 2000) show that much of the initial
modular organization of visual cortex in terms of ocular dominance and orientation col-
umns is independent of activity, although activity is required for maintenance and plasti-
city of modularity.
8. SYNOPSIS OF CYTOARCHITECTONIC DIFFERENCES
Editors: From reading several chapters submitted for this book (for example that of Kaas)
we obtained the following tentative picture of the cybernetic implications of cytoarchi-
tecture:
(i) The basic cytoarchitecture of the cortex is rather uniform, as befits a network for per-
forming association on information which has to be described multidimensionally.
There are, admittedly, differences between areas in the width of the dendritic arbor of
pyramidal cells as well as in the extent of local axonal ramifications (as documented
in the chapters by Jacobs and Scheibel, and Levitt and Lund). These differences are
associated with differences in packing density and size of cells, but overall the impact
on Nissl cytoarchitecture is not very great.
(ii) On the other hand, for primary sensory areas, the information is different, described
by two spatial dimensions (or in the auditory cortex by the single dimension of pitch).
In these cases much more specialized cybernetic operations may be needed, perhaps
for relatively stereotyped pre-processing of sensory information. The cytoarchitectonic
features that arise in the sensory parts of the cortex as a result may also be more
highly specialized, perhaps with critical involvement of a distinctive lamina IV (packed
with spiny stellate cells), and for the visual system several more complex arrange-
ments, including those to process motion.
(iii) Apart from these considerations, throughout the cortex, cell size may be determined
largely by the length and calibre of axons to be supported.
This scheme is perhaps a simple-minded summary of cytoarchitectonics. How much will

it explain? Not all, no doubt. There may be developmental (or other) reasons why (for
instance) area 4 has a much reduced lamina IV, and much of the prefrontal cortex is the
same; why sensory cortex has a granular lamina IV; and why area 3 has very few cells in
lamina V, perhaps again for developmental reasons. We invited Kaas to comment.
Kaas: I agree with your brief overview.

9. NEOCORTEX: UNIFORM OR BIPARTITE PROCESSING ALGORITHMS?
Editors: Underlying these last questions (especially the second point, above) is another,
more fundamental one: to what extent is the cerebral neocortex an organ for analysing the
three main sensory inputs (visual, auditory and somatic sense), with the neighbourhood
relations possible in a two- or one-dimensional sensory surface? To what extent is it
a more generalised information processing organ, which can analyse information even
when it does not have this inherent structure? Perhaps the cortex is a mixture of both of
these. If so, we ask two questions: How do the special features of the primary sensory
areas equip them for analysis of information already organized dimensionally? Are other
areas of cortex equipped for a different style of information processing—perhaps an “all-
purpose” associative processor, rather than one specialized for one- or two-dimensional
arrays? The alternative is that the cortex has a more uniform style, similar across all areas,
for its cybernetic operations. In this case, one could envisage that the distinctive lamina IV
of primary sensory organs is more of an “amplifier” for specially-important input from the
outside world, rather than a lamina with a categorically different cybernetic function.
Dinse/Schreiner: In our view, the principles of processing discussed for primary sensory
areas hold equally well for all other cortical areas. The fundamental processing step is
assumed to consist of a local operation modified (contextually) by long-range connections.
The only way to generate a local operation of this type is by combinations of excitatory
and inhibitory mechanisms, quite as outlined in our chapter for sensory cortical areas. In
this view, this would provide the basis for an uniform manner of processing. In addition,
this processing is performed within a 2-dimensional sheet allowing the combination of
local operations with a continuous representation of parameters (see also next question),
maintaining (locally) neighbourhood relationships. While local processing obeys per se
rules of proximity, the 2-dimensional sheet defines proximities of various kinds.
For sensory areas, the “representation” of the outside world implicates both 2-dimen-
sionality and proximities within the various types of physical worlds, i.e. dimensions
derived by further processing, rather than implicit in spatial two-dimensionality. In this
sense this scheme is highly intuitive for early cortical representation, where it is reason-
ably clear what is represented. However, this scheme has been shown to hold also for
intermediate states: perfect examples are the highly specialized and detailed maps as
described by Suga et al. (1984) in the bat. Because of the highly specialized ultrasound
environment of bats it was possible to deduce, and then to find higher-order maps that
contain ordered representations of echo frequency and echo delay.
The main problem in higher cortical areas arises from the fact that the relevant parameter
spaces are unknown and, in principle, are difficult to deduce. The nature and properties of
such parameter spaces can (and need to) be determined to understand cortical processing
outside the primary sensory or motor domains. For example, it is likely that there are highly
abstract parameter spaces representing a profile of a face in terms of emotional expression.
In any case, the basic principle is to compute and assemble behaviourally relevant aspects
of proximity, or similarity and dissimilarity in the projected parameter space.
This view can easily be extended to incorporate temporal aspects as well. It is
conceivable that temporal proximity defines entire feature spaces (For instance, a move-
ment is a complicated space-time pattern; even more complicated space-time or spectral-
temporal patterns are possible in the auditory system). As already shown for primary

visual cortex, several of such highly abstract parametric maps can be superimposed. With-
out a priori knowledge, this renders an analysis of what is being represented quite difficult.
Editors: In the chapter of Dinse and Schreiner, there are hints of another formulation of
this, where they contrast “parametric mapping” with “distributed representation”. Could it
be that the former are limited to sensory areas, while the latter are more widely useful
throughout the neocortex?
Dinse/Schreiner: In our view, the same arguments are valid for the distinction between
local and distributed representation of sensory, motor or even conceptional aspects: in
principle, parametric mapping is identical to distributed processing. As outlined in our
chapter, the difference is mainly methodological, and arises from different reconstruction
algorithms (e.g. the optical imaging-based feature maps are just due to an “iceberg effect”,
by ignoring 90% of neural activation). We agree, however, that “distributed representation”
sounds a bit more general, or to put it the other way round, “parametric-mapping” is more
intuitive by directly relating to sensory representations of known physical features. How-
ever, as outlined above, in principle we don’t see any systematic and general differences
between these concepts. Accordingly, we believe that the concept of “distributed repre-
sentation” better encapsulates what is going on in the brain, and this concept holds equally
in lower (or early) and higher cortical areas.
Miller: Closely related to this broad issue, is a much more detailed question. We were
interested on the distinctive lamina IV of primary sensory areas of cortex. Usually it is
crowded with spiny stellate cells or other pyramid-like cells which have either no apical
dendrite, or a much reduced one. This could mean that this lamina is, to a degree, shielded
from the full impact of long cortico-cortical input to a region of cortex. An extension of
this theme is to suggest that the lamina IV of primary sensory areas is specialized as a pre-
processor of sensory thalamic input, to analyse sensory input “as itself”, before that input
mingles very much with the rest of the information coursing through the cortical network.
Against this idea, Young pointed out the evidence that, even in sensory lamina IV, only
about 5% of excitatory synapses are derived from the specific sensory thalamic input,
implying that input from other regions, including that from other cortical areas is far more
important than thalamic input in this layer.
Kaas: Young’s statement is correct, but one has to be careful about what conclusions to
draw. The retina provides only a small portion of the synapses in the LGN, but all of the
neurones in the LGN depend on this input for activation. Other inputs are modulatory, and
many connections are intrinsic. The functional roles as well as the numbers of connections
need to be determined.
Miller: It seems that several further factors have to be taken into account before one can
evaluate the relative importance of specific thalamo-cortical and cortico-cortical inputs to
lamina IV of sensory cortex:
(i) There is evidence that thalamo-cortical inputs have greater efficacy than cortico-
cortical ones. This conclusion has been reached both from studies in electrophy-
siological studies in slices (Stratford et al., 1996; Gil et al., 1999) and from

cross-correlation studies in vivo (Toyama et al., 1981; Tanaka, 1983; Reid and
Alonso, 1996) as well as from the anatomical observation that thalamo-cortical affer-
ents tend to be more densely ramified than the axons of pyramidal cells (Braitenberg,
1978).
(ii) According to Lund (1979), much of the axonal arborization of spiny stellate neurones
in lamina IV of cat visual cortex is distributed horizontally in the same lamina. If this
is the case, it might be that many of the excitatory synapses on lamina IV spiny stel-
lates neurones are ones from other similar spiny stellate neurones, rather than from
distant cortical areas, or other structures. If this were the case, lamina IV in primary
sensory cortex could be specialized as an amplifier of thalamic input.
(iii) This suggestion can also be derived from evidence that synapses between neighbour-
ing spiny stellate cells are of specially high efficacy and reliability (Feldmeyer et al.,
1999; Tarczy-Hornoch et al., 1999).
(iv) Beyond the number and efficacy of synapses, is the actual firing frequency in afferent
axons. It is probable that thalamic projection neurones have a substantially higher rate
of firing in waking animals than do most cortico-cortical neurones. The data I draw
on for thalamic projection neurones are: Steriade and Hobson, (1976); Mukhametov
and Rizzolatti (1970); Sakakura (1968); Lamarre et al. (1971); Glenn and Steriade
(1982). For cortico-cortical neurones I draw on several papers by Harvey Swadlow in
Journal of Neurophysiology, in rabbit. In larger animals there is not much useful data,
and what there is from striate cortex, which may not be typical. However, if the
thalamic cells really do fire at substantially faster rates than the cortico-cortical ones, the
chances of impulses in the several thalamo-cortical inputs to a spiny stellate neurone
arriving sufficiently close together in time to summate and produce suprathreshold
excitation may be much higher than that for the cortico-cortical inputs.
(v) Even beyond primary sensory areas (e.g. in MT/V5 of primate) there is evidence that
the “ascending” input from primary sensory areas has an advantage over that from the
rest of the cortex. In area MT electronmicroscopy shows that synapses of projections
from V1 are exceptionally large, (often completely surrounding the recipient dendritic
spines) and are therefore probably of high efficacy (Anderson et al., 1998). There is
also evidence that blockade of the magnocellular layers of LGN essentially eliminates
driven activity in V5, apart from a few sites where P-driven responses can be detected
(Maunsell et al., 1990). Background activity (presumed to be derived from non-sensory
inputs) was less than a third of the peak rates of firing in response to a stimulus.
Given these factors (for which, I admit, empirical evidence is not always completely com-
pelling), could not the idea that lamina IV of sensory cortex (and perhaps of other regions)
is a preprocessor and amplifier of sensory input, be more attractive than might otherwise
be implied? We asked several contributors about this issue. Initially we asked Lund for
up-to-date views on the extent of mutual connectivity between spiny stellate cells in lamina
IV of sensory areas.
Lund: It is clear that the spiny stellate neurones in layer 4 make a large contribution to
their spiny stellate cell neighbours in the same layer (30% of their excitatory inputs
according to Ahmed et al., 1994). However, numerical weight of synapses may be less
important than their individual efficacy. Physiologically these lateral connections are less
reliable than the thalamic inputs, even though far more numerous, as a drive to the cells.

They may depend on timing of other inputs, or withdrawal of inhibition to show their full
potential as an efficient drive or modulatory input. These axon terminals have type 1 mor-
phology, and so are presumed to be excitatory. In work by Anderson et al. (1993) filling
single neurones, spiny stellate cells in the monkey visual cortex had either extensive
lateral intralaminar connections but no rising connections to other layers, or projections to
other layers and only local axon arbors in the same layers. However, they had only a few
filled examples. Yabuta and Callaway (1998) have fuller observations on the detail of
these local versus interlaminar projections.
Young: I emphasise the relatively small proportion of synapses in geniculo-recipent layers

that are “owned” by geniculate afferents—while Miller emphasises the specialised and
effective nature of these inputs. We are both right!
These inputs are special, but there are still very many other inputs, arising from internal
brain sources, on these cells and their local network neighbours. Our original discussion
was in relation to the consequences of these arrangements for “amplification”. I don’t dis-
pute that there is amplification, and indeed I suppose that this is the principal means by
which the visual system is oriented computationally to make representations about the
outside world. But amplification, in the sense implied by the traditional view, is a multi-
plication of the input by a gain factor. Maybe the high gain of these special synapses is
part of that multiplication. I only note that—probably—the other sources of information
washing over these cells from internal sources also play a part in the multiplication, and so
in what is represented. In fact, the traditional view (that it’s all “bottom-up”) makes a very
much stronger claim than mine: that the 95% of other inputs on these cells don’t have any
effect. Seems unlikely! So, I’m presently comfortable with my position on this.
Miller: I accept this accommodation of the two viewpoints, especially since there is
evidence that inputs to lamina IV from lamina VI (which itself receives many inputs from
distant cortical regions) tend to show considerable facilitation with stimuli repeated in
close succession (Stratford et al., 1996; Tarczy-Hornoch et al., 1999), apparently an
“amplifier” for inputs other than thalamo-cortical ones. To evaluate properly the relative
importance of “internal” inputs and thalamo-cortical (“external”) inputs, one needs evid-
ence about whether the prior probability of stimuli (represented by the input from distant
regions of cortex) has a decisive role in the firing of lamina IV cells in visual cortex.
Shipp: You ask whether the comparative isolation of the magnocellular (M) pathway
from LGN to V5 via layers 4Ca and 4B of V1 could be characterised as serving “to analyse
sensory input as itself”. There are a number of relevant factors here. Each has its uncer-
tainties, but they combine to weigh against this viewpoint.
Firstly, only layer 4C is equivalent to “layer 4” as cited above; layer 4B is probably not
layer 4! The terminology is that of Brodmann, and it has stuck, but many who have delib-
erated over layer terminology in primate V1 prefer to refer to 4B as lower layer 3, i.e.
layer IIIc (e.g. Casagrande and Kaas, 1994). Why does this matter? Well, layer 4B is
unusual in receiving direct feedback from areas V5 (and from V3 and V2). In other areas
(e.g. V2, etc.) such feedback normally avoids layer 4 and is concentrated in layers 1 and 6;
it can spread from layer 1 into layers 2 and 3, but it is not focused on layer 3. Layer 4B is
thus unusual in receiving highly focused feedback from V5 (that also targets layer 6 but
avoids layers 2, 3 and 5 and only involves layer 1 in the peripheral field). The above

debate concerns the relative efficacy in layer 4 of ascending and intrinsic inputs to
neurones. In layer 4B there is the added component of descending input (and note that the
ascending input is not thalamic, but is from layer 4Ca).
Secondly, the “isolation” of this relay to V5 refers to its domination by magnocellular
(M) input, and a relative lack of contamination by other retino-geniculate channels. There
are no obvious intrinsic axonal relays of parvocellular (P) or koniocellular (K) inputs
into 4B. Some V5-efferent neurones in layer 4B are stellate cells, so they could transmit
untainted M signals. However, other V5-efferents are obviously pyramidal, and these
demonstrably relay P signals through influences upon their apical dendrite. So here is at
least one route of P signals to V5 (see my earlier comments; also see my chapter in this
volume “Fundamentals . . . ” for references). P and K signals are mixed together in layers 2
and 3, where M inputs are also substantial, so the (mainly) M output from layer 4B to V5
is purer by comparison. But, obviously, isolation from rival geniculate input channels is
not the same thing as isolation from other ongoing cortical processing.
Finally, for academic purity, I have to note that the whole notion of magno-based direc-
tion processing in the 4B-V5 relay has recently been called into question. Physiological
data on the temporal response properties of V1 neurones, allied to motion energy models
of directional tuning, imply that direction selectivity depends on an obligatory combina-
tion of fast M and slower P inputs, in order to achieve the modelled arrangement of RF
subunits in spatial and temporal quadrature (De Valois et al., 2000). This evidence is
compelling in isolation, but awkward to reconcile with everything else.
Miller: More generally, in my paragraphs above, I raised two possibilities about the dis-
tinctive lamina IV for primary sensory cortex, either that it might confer special cybernetic
capabilities on sensory cortex, or that it serves as an “amplifier” for input to these regions.
In the light of the discussion presented above, I tend to favour the second (less specific) of
these hypotheses, as more clearly formulated, and supported by substantial evidence,
while the idea of a special cybernetic role, generalized across sensory cortical areas, is
difficult to formulate in a testable manner. However, the “amplification” need not apply
just to sensory input. This in turn persuades me to think (in agreement with Dinse) of the
neocortex as having a rather uniform type of cybernetic operation (though with local
quantitative variations), rather than being a mixture of areas with two basically different
types of information processing.
10. THE HISTORICAL DEBATE: LOCALIZATION VERSUS DISTRIBUTION

OF FUNCTION, AND THE NATURE OF ORGANIC UNITY OF THE
CORTEX
Editors: The last issue, on which we invited the comments of Shipp and Seitz, was per-
haps the underlying focal point for the whole of this book. Shipp wrote, in his introductory
part: “localisation can be thought of as segregation, the idea that any structurally (or
neurochemically, or genetically) distinct module of tissue is likely to have a corresponding
functional identity, a specialisation that distinguishes it from neighbouring structures. To
anyone willing to accept the label ‘neuroanatomist’, this notion is probably axiomatic.”
Later he wrote: “Clearly, any global theory of ‘how the brain works’ has to match the
functions to the anatomical structures”. Nevertheless he softened this view with the

following remark: “Segregation and integration are the twin fundamentals of cortical
organisation. In an elementary sense, the latter cannot exist without the former. It is easy
to cast them as opponent processes—but this, surely, is to miss the evident reality that they
must cooperate to subserve the goals of cortical processing.” In similar vein Malcolm
Young writes, in his chapter: “V4 and MT exchange quite robust projections, but each is
the home of neurones with very different stimulus selectivities.”
The balance between segregation and integration of function in the cortex is a difficult,
fundamental and profound issue. Obviously there must be some aspects of function which
can, in some experimental designs, be correlated with areal differences in architecture,
while other functions can only be derived from coordination of several or many areas.
Which sorts of function can be matched with morphological areas, and which require
integration of many areas?
A number of neurologists in the early and middle years of the twentieth century, in
several countries (Britain, Germany, France, Ruissia) opposed the implications of localiza-
tionist neurologists, and neuroanatomists such as Brodmann. To give a flavour of their way
of thinking we quote below from Alexander Luria (The Working Brain, Penguin Books).
Luria was impressed by the way in which animals and humans had recourse to many
ways of solving a problem, and tended to solve it in different ways at different stages of
maturation of their cogntive apparatus. He also emphasised that, in its actual function, the
brain was in continual interaction with the external environment (including its social
dimension), so that any functional system was not defined simply in the brain, but by
continual interplay between the brain and the outer environment. As a result, he writes as
follows:
“ . . . all mental processes such as perception and memorizing, gnosis and praxis,
speech and thinking, writing, reading and arithmetic, cannot be regarded as iso-
lated or even indivisible ‘faculties’, which can be presumed to be the direct
‘function’ of limited cell groups, or to be ‘localized’ in particular areas of the
brain . . . . the fundamental forms of conscious activity must be considered as
complex functional systems; consequently the basic approach to their ‘localiza-
tion’ in the cerebral cortex must be radically altered.”
“ . . . mental functions . . . cannot be localized in narrow zones of the cortex, or in
isolated cell groups, but must be organized in systems of concertedly working
zones, each of which performs its role in the complex functional system, and which
may be located in completely different and often far distant areas of the brain.”
“It is accordingly our fundamental task not to ‘localize’ higher human psycholo-
gical processes in limited areas of the cortex, but to ascertain by careful analysis
which groups of concertedly working zones of the brain are responsible for the
performance of complex mental activity; what contribution is made by each of
these zones to the complex functional system; and how the relationship between
these concertedly working parts of the brain in the performance of complex men-
tal activity changes in the various stages of development.” [Luria’s italics, in the
above quotations]
We suspected that Shipp would not disagree with any of these remarks, and some of the
modern proponents of functional imaging are certainly looking for the “contribution . . .

made by each of these zones to the complex functional system.” We invited Shipp to com-
ment on these quotations.
Shipp: Quite so! As you predict, there is nothing in Luria’s comments to reject. They
seem both accurately and expressively worded.
Let me return to the introductory remark in my chapter that a global brain theory needs
to match functions to structures (the “matching problem”). The list of structures (at a cor-
tical level) is complete at the level of lobes, approaching completion at the level of areas
(less so in humans than monkeys), but far from complete in details of sub-areal modules.
The list of functions (“faculties” is a preferable term) tends to be one of everyday abilities,
which we all understand intuitively, but which cognitive scientists (including psycho-
physicists, clinicians, etc.) have the task of systematising and “anatomising”, to give the
everyday terms a workable scientific definition.
The problem here is an inescapable one: Any function that is defined on a cognitive,
or perceptual basis is almost certainly going to derive from a distributed brain system (as
pointed out by Luria). This is as true for say, “colour vision”, as it is for reading or memory.
So the function of individual lobes, areas modules (etc.) will never be perfectly sum-
marised in these terms.
At this point, it is vital (if also dull and pedantic) to avoid purely semantic confusions.
So, for example, a statement like “V1 blobs are specialised for colour vision” has to be
understood not to mean: (a) that other parts of the brain (e.g. thin stripes in V2), or other
parts of V1 (e.g. layer 4Cb), are not also involved in colour vision; (b) that blobs are not
also involved in a function other than colour vision (e.g. perhaps the signals they process
also contribute, eventually, to something like “shape from shading”). Such qualifications
(i.e. clarifying the usage of the term “specialisation”) apply even more strongly to the
equally true statement “area V4 is specialised for colour vision”.
In other words, the term “specialisation” is consistent with (does not necessarily
exclude) the fact that the neural basis of cognitive functions may be distributed, although
clearly the operative nub of the term is that the distribution is not pluripotent (i.e. some
areas are more equal than others).
Now let us view the matching problem from the opposite perspective. It is incontrovert-
ible that different lobes (or areas or modules) have different physiology, different response
properties and different neural wiring. Hence they cannot have identical functions. This is
such a trite truism that it barely merits printing, except insofar as it leads to the following:
if a function is to be satisfactorily localised to an individual structure, it must be described
or defined in neural terms. Evidently this is largely impracticable at present, because
neural processing is so incompletely understood.
However, I try to give an example. Following David Marr, it is useful to think of neural
processing at three different levels: a computational goal, an algorithm, and a hardware
(i.e. neural) implementation of the algorithm. So, “colour vision” involves a number of
abilities, one of which is the recognition of the same surface hue under variant illumina-
tion (“colour constancy”). The computational goal for colour constancy is to retrieve the
spectral reflectance function of a surface. Algorithms for colour constancy have been
devised, initially by Edwin Land, in which the first stage is a local centre-surround “differ-
encing” operation, where centre and surround operate on the same part of the spectrum.
Physiologically, this equates to the classical definition of a “double-opponent cell” found
in goldfish retina and, to some extent in V1 blobs. Hence, this might be mooted as a function

localised to V1 blobs—except that double opponent cells are also found elsewhere in
primate brains (e.g. V2), so the ultimate definition of blob function has to be yet further
specified. Further, and importantly, the end-product of this function may be not only
colour constancy; once established, evolution may find another use for it.
Seitz: I agree with the contention that conscious activity must be considered as a unitary
functional system of the human brain. Conscious activity represents a state of complex
human brain functions which cannot be subdivided into a diversity of different funda-
mental subfunctions. Nonetheless, attention provides a window limiting the amount of
information that can be processed consciously in the human brain, at a certain point in
time, this being operative within a predefined capacity. Thus, conscious activity accom-
modates graded levels of awareness for executive, perceptive and cognitive processes.
Over time, these processes are channelled into foci of attention, depending on the situ-
ational circumstances.
I agree with the idea that mental functions are organized in systems of concertedly work-
ing zones, each of which performs its role in complex functional subsystems. These
systems include the different sensory modalities providing the brain with external
information, the motor system allowing for physical activity, and the so-called higher cor-
tical areas. Both unity and diversity of mental functions can be appreciated, for example,
in object exploration. In this activity, the subject performing exploratory movements
perceives the object’s characteristics, and constructs a mental image of the manipulated
object, but does not notice the exploratory movements themselves; nor can he describe the
finger movements performed. The diversity is conveyed by the involvement of the various
subsystems described, while the unity is reflected by the top-down combination, in con-
cert, of the exploratory movements that sample the somatosensory information optimally
in terms of completeness and speed.
The question of how the different areas of the cortex can have diverse functions, and yet
are part of a united and unified cortical structure seems to me to reflect the level of observ-
ation rather than to be a contradiction. One obvious reason is that what appears unified in
terms of histology proves to be quite heterogenous in terms of quantitative cytology, neu-
roreceptor distribution, interareal connectivity, and neuroelectric properties, as reflected,
for example, by the receptive or executive field size or cortical columns. The question to
be resolved, however, is how the different modalities of highly differentiated information
processing are tied together to a conscious percept, and to the intimately linked self-
experience of the subject. It is the aim of human systems physiology to disentangle the
contribution which is made by each of these zones to the complex functional system, and
to come up with refined methodology to address the question of conscious behavior.
Historically, the first approach to be adopted was the study of neurological patients by cor-
relating brain lesions with clinical deficits, as well as by describing electrical stimulation
effects during open brain surgery. The advent of computer technology provided means to
design non-invasive neuroimaging and electrophysiological methods which allow investig-
ators to unravel the temporal evolution and to identify the distributed topographic repre-
sentations of intact human brain function. These tools bear the promise of also elucidating
the processes mediating conscious information processing.
We have learnt from neurological “syndromology” that circumscribed brain lesions
induce partial deficits usually involving some specific functions, while other functions are
preserved. Examples are hemineglect and anosognosia, conditions in which perception of

the extrapersonal space, or of the personal “state” is impaired in the presence of an other-
wise intact perception. However, often there is no specificity of the clinical deficits relat-
ing to a unique lesion topography. Thus, similar clinical symptoms may originate from
a variety of lesion locations. For example, there is the “alien hand syndrome”. This con-
dition renders a limb—and preferentially the hand—alien to the patient, such that he
perceives a touch as foreign, and he does not know how to use the limb as his “agent”. The
underlying lesions may be located in the fronto-mesial region but also in posterior parietal
cortex. Conversely, patients may suffer from hallucinations or illusions, such that they
perceive non-realistic events in the absence of objective phenomena, with these experi-
ences often limited to one modality. Furthermore, in amputees, phantom limb sensations
may fool the patients such that they have the feeling of a still-existing limb. Thus, damage
or abnormalities of the nervous system may not only induce drop-outs from the entire
repertoire of possible brain functions, but may also stimulate over-activity resulting in
misperception.
A further interesting aspect of brain lesions is that they may be “silent” when they are
small and limited to the cortex. However, they become clinically apparent as soon as they
extend over a certain cortical area, and into the underlying white matter. In these situ-
ations, damage is not restricted to the resident cortical neurones, but also affects, in addi-
tion, the descending efferent projections to the different subcortical relay nuclei including
the cerebellum, the corticopetal afferents, and the cortico-cortical connections. In general,
clinically apparent brain lesions damage a portion of the entire neuronal apparatus and its
connections to other cortical areas and subcortical structures. Conversely, the size of the
neuroreceptive or neuroexecutive fields, as well as the locally-inherent connectivity pat-
terns, determine whether a lesion of a given volume becomes clinically apparent. Further,
recent evidence suggests that the wealth of parallel projection systems in the human brain
plays an important role in the cerebral reorganization mediating recovery from acute
neurological deficits. In consequence, brain diseases exemplify the diversity of brain
function.
Activation studies appear to supplement lesion studies, in that single areas, or a few
brain areas are activated in brain activation tasks. However, these methods of measuring
brain activity have been operationalized such that they are based on mathematical oper-
ations including psychophysical task subtractions, signal averaging and multidimensional
data processing. Topographic information is inherent in these measuring devices, it being
given by the number of recording units in EEG or MEG, or by the pixel matrix in tomo-
graphic imaging representing the elements of maximal resolution. Results show that the
more specific the hypothesized difference between the task and control conditions, the
fewer and the more circumscribed the number of haemodynamic or metabolic activation
areas. Similarly, dipole calculations pinpoint activity-locked, bioelectric changes to cor-
tical foci at a certain time point following processing onset. At first glance, these data
support the focussed interest in the diversity of brain function. It should be emphasized,
however, that the unsubtracted tomographic images and the raw bioelectric recordings
show that each experimental state shows some activity changes throughout the entire
brain, reflecting the complexity and unity of the working human brain.
Editors: How can the different areas of the cortex have diverse functions, and yet be part
of a united and unified cortical structure? There are similar debates in other areas of brain
research. For example, for Hebb’s cells assemblies, one can ask how the component

neurones of an assembly can individually have different response properties, while at the
same time (when they are activated together as part of a functional whole) the assembly
can have its own unique and more sophisticated representation of information. Likewise,
one can ask how left and right cerebral hemispheres can, at the same time, have very dif-
ferent capabilities, and be part of a unified forebrain. These are perhaps examples of
a recurring problem in biology, concerning the nature of the organic unity of living things.
A major source of confusion when trying to analyse these questions is that in the same
sentence, in the same breath, it is very easy to juxtapose assumptions referring to two quite
different levels of organization: the level in which one seeks exact correlations between
structure and function within an organism, and the level of the function of a whole organ-
ism. The very word “function” is itself highly ambiguous, since it is so easily used without
specifying whether it refers to cybernetic function, or to the functional role for an intact
organism. Probably it is impossible to construct arguments by which the latter can be
derived from the former. The insoluble “three-body problem”, forerunner of chaos theory
(Peterson, 1993), seems elementary by comparison! Even so, both are legitimate and
important objectives for study. Despite this, it may be that cerebral localization of function
can be exactly defined only in terms of cybernetic function; or as Shipp writes: “if a func-
tion is to be satisfactorily localised to an individual structure, it must be described or
defined in neural terms”.
REFERENCES
Abeles, M. (1982) Local cortical circuits. (Studies in Brain Function, No. 6). Springer, Heidelberg.
Ahissar, E., Sosnik, R. and Haidarliu, S. (2000) Transformation from temporal to rate coding in a somatosensory
thalamocortical pathway. Nature (London), 406, 302–326.
Ahmed, B., Anderson, J.C., Martin, K.A.C. and Nelson, J.C. (1994) Polyneuronal innervation of spiny stellate
neurons in cat visual cortex. Journal of Comprative Neurology, 341, 39–49.
revisited: Cytoarchitecture and intersubject variability. Journal of Comparative Neurology, 412, 319–341.
brought into stereotaxic space—where and how variable? Neuroimage, 11, 66–84.
Anderson, J.C., Martin, K.A.C. and Whitteridge, D. (1993) Form, function, and intracortical projections of
neurons in the striate cortex of the monkey Macacus nemestrinus. Cerebral Cortex, 3, 412–420.
Anderson, J.C., Binzegger, T., Martin, K.A.C. and Rockland, K.S. (1998) The connection from cortical area of
V1 to V5: a light and electron microscopic study. Journal of Neuroscience, 18, 10525–10540.
Angelucci, A., Clascá, F., Bricolo, E., Cramer, K.S. and Sur, M. (1997). Experimentally induced retinal projec-
tions to the ferret auditory thalamus: Development of clustered eye-specific patterns in a novel target.
Barone, P., Hilgetag, C.-C., Kennedy, H. and Young, M.P. (2000) A determinate hierarchy for macaque visual
cortex derived from quantitative laminar data. Society for Neuroscience Abstracts, 26, 449.17.
Beaulieu, C. and Colonnier, M. (1983) The number of neurons in the different laminae of the binocular and
monocular regions of area 17 in the cat. Journal of Comparative Neurology, 217, 337–344.
Bok, S.T. (1959) Histonomy of the cerebral cortex. Princeton, New Jersey, Van Nostrand-Reinhold.
Borges, S. and Berry, M. (1978) The effects of dark rearing on the development of the visual cortex of the rat.
Boyd, J.D. and Matsubara, J.A. (1996). Laminar and columnar patterns of geniculocortical projections in the cat:
Relationship to cytochrome oxidase. Journal of Comparative Neurology, 365, 659–682.
Braitenberg (1978) Cortical architectonics: general and areal. In: M.A.B. Brazier and H. Petsche (eds), Architec-
tonics of the cerebral cortex. New York: Raven Press, pp. 443–465.
Braitenberg, V. and Braitenberg, C. (1979) Geometry of orientation columns in the visual cortex. Biological
Cybernetics, 33, 179–186.

Braitenberg, V. (1985) Charting the visual cortex. In: A. Peters and E.G. Jones (eds), Cerebral Cortex, Vol. 3,
(Visual Cortex). New York, London: Plenum Publishing Corp, pp. 374–414.
Braitenberg, V. (1992) How ideas survive evidence to the contrary: a comment on data display and modelling.
In: A. Aertsen and V. Braitenberg (eds), Information processing in the cortex, experiments and theory.
Berlin, Heidelberg, New York: Springer, pp. 447–450.
Braitenberg, V. and Schüz, A. (1998) Cortex: Statistics and Geometry of Neuronal Connectivity. (Revised
edition of Anatomy of the Cortex. Statistics and Geometry, 1991). Berlin, Heidelberg, New York:
Springer.
Bringuier, V., Chavane, F., Glaeser, L. and Fregnac, Y. (1999) Horizontal propagation of visual activity in the
synaptic integration field of area 17 neurons. Science (New York), 283, 695–699.
Bullier, J. and Nowak, L.G. (1995) Parallel versus serial processing: new vistas on the distributed organization of
the visual system. Current Opinion in Neurobiology, 5, 497–503.
Carandini, M., Mechler, F., Leonard, C.S. and Movshon, J.A. (1996) Spike train encoding by regular-spiking
cells of the visual cortex. Journal of Neurophysiology, 76, 3425–3441.
Casagrande, V.A. and Kaas, J.H. (1994) The afferent, intrinsic and efferent connections of primary visual cortex
in primates. In: A. Peters and K.S. Rockland (eds), Primary Visual Cortex in Primates. New York: Plenum
Press, pp. 210–259.
Chapman, B. (2000) Necessity for afferent activity to maintain eye-specific segregation in ferret lateral genicu-
late nucleus. Science (New York), 287, 2479–2482.
Chapman, B., Godecke, I. and Bonhoeffer, T. (1999) Development of Orientation Preference in the Mammalian
Visual Cortex. Journal of Neurobiology, 41, 18–24.
Chugani, H.T., Phelps, M.E. and Mazziotta, J.C. (1987) Positron emission tomography study of human brain
functional development. Annals of Neurology, 22, 487–497.
Clarke, S. (1993) Callosal connections and functional subdivision of the human occipital lobe. In: B. Gulyas,
D. Ottoson and P. Roland (eds), Functional organization of the human visual cortex. Oxford: Pergamon
Press, pp. 137–149.
Crowley, J.C. and Katz, L.C. (1999) Development of ocular dominance columns in the absence of retinal input.
Nature, Neuroscience, 2, 1125–1130.
Crowley, J.C. and Katz, L.C. (2000) Early development of ocular dominance columns. Science (New York), 290,
1321–1324.
Dehay, C., Giroud, P., Berland, M., Killackey, H. and Kennedy, H. (1996) Contribution of thalamic input to the
specification of cytoarchitectonic cortical fields in the primate: Effects of bilateral enucleation in the fetal
monkey on the boundaries, dimensions, and gyrification of striate and extrastriate cortex. Journal of
Dehay, C., Horsburgh, G., Berland, M., Killackey, H. and Kennedy, H. (1991) The effects of bilateral
enucleation in the primate fetus on the parcellation of visual cortex. Developmental Brain Research, 62,
137–141.
De Valois, R.L., Cottaris, N.P., Mahon, L.E., Elfar, S.D. and Wilson, J.A. (2000) Spatial and temporal receptive
fields of geniculate and cortical cells and directional selectivity. Vision Research, 40, 3685–3702.
DeYoe, E.A., Hockfield, S., Garren, H. and Van Essen, D.C. (1990) Antibody labelling of functional subdivi-
sions in visual cortex: cat-301 immunoreactivity in striate and extrastriate cortex of the macaque monkey.
Visual Neuroscience, 5, 67–81.
Elston, G.N. and Rosa, M.G.P. (1998) Morphological variation of layer III pyramidal neurones in the occipito-
temporal pathway of the macaque monkey visual cortex. Cerebral Cortex, 8, 278–294.
Fairén, A. and Valverde, F. (1980) A specialized type of neuron in the visual cortex of cat: A Golgi and electron
microscope study of chandelier cells. Journal of Comparative Neurology, 194, 761–779.
Feldmeyer, D., Egger, V., Lubke, J. and Sakmann, B. (1999) Reliable synaptic connections between pairs of
excitatory layer 4 neurones within a singel “barrel” of developing rat somatosensory cortex. Journal of
Physiology, 521, 169–190.
Finlay, B.L and Slattery, M. (1983) Local differences in the amount of early cell death in neocortex predict adult
local specializations. Science (New York), 219, 1349–1351.
Gabbott, P.L.A., Stewart, M.G. and Rose, S.P.R. (1986) The quantitative effects of dark-rearing and light expos-
ure on the laminar composition and depth distribution of neurons and glia in the visual cortex (area 17) of
the rat. Experimental Brain Research, 64, 225–232.
Gil, Z., Connors, B.W. and Amitai, Y. (1999) Efficacy of thalamocortical and intracortical synaptic connections:
quanta, innervation and reliability. Neuron, 23, 385–397.
Gilbert, C.D. and Wiesel, T.N. (1989) Columnar specificity of intrinsic horizontal and corticocortical connec-
tions in cat visual cortex. Journal of Neuroscience, 9, 2432–2442.
Gillies, K. and Price, D.J. (1993) The fates of cells in the developing cerebral cortex of normal and
methylazoxymethanol acetate-lesioned mice. European Journal of Neuroscience, 5, 73–84.

Glassman, R.B. (2000) A “theory of relativity” for cognitive elasticity of time and modality dimensions support-
ing constant working memory capacity: involvement of harmonics among ultradian clocks? Progress in
Neuropsychopharmacology and Biological Psychiatry, 24, 163–182.
Giraud, A.L., Lorenzi, C., Ashburner, J., Wable, J., Johnsrude, I., Frackowiak, R. and Kleinschmidt, A. (2000)
Representation of the temporal envelope of sounds in the human brain. Journal of Neurophysiology, 84,
1588–1598.
Glenn, L.L. and Steriade, M. (1982) Discharge rate and excitability of cortically projecting intralaminar thalamic
neurons during waking and sleep states. Journal of Neuroscience, 2, 1387–1404.
Glezer, I.I., Jacobs, M.S. and Morgane, P.J. (1988) Implications of the “initial brain” concept for brain evolution
in Cetacea. Behavioral and Brain Sciences, 11, 75–116.
Goldman-Rakic, P.S. and Schwartz, M.L. (1982) Interdigitation of contralateral and ipsilateral columnar projec-
tions to frontal association cortex in primates. Science (Washington), 216, 755–757.
Gonzalez-Burgos, G., Barrionuevo, G. and Lewis, D.A. (2000) Horizontal synaptic connections in monkey pre-
frontal cortex: an in vitro electrophysiological study. Cerebral Cortex, 10, 82–92.
Horner, C.H. and Arbuthnott, E. (1991) Methods of estimation of spine density: are spines evenly distributed
throughout the dendritic field? Journal of Anatomy, 177, 179–184.
Jain, N., Catania, K.C. and Kaas, J.H. (1998) A histologically visibled representation of the fingers and palm in
primate area 3b and its immutability following long-term deafferentations. Cerebral Cortex, 8, 227–236.
Jerison, H. (1991) Brain size and the evolution of mind. 59th James Arthur Lecture. New York: American
Museum of Natural History.
Jiménez, D., García, C., De Castro, F., Chédotal, A., Sotelo, C., De Carlos, J.A., Valverde, F. and López-
Mascaraque, L. (2000) Evidence for intrinsic development of olfactory structures in Pax-6 mutant mice.
Kaas, J.H. (1987) The organization of neocortex in mammals: implications for theories of brain function. Annual
Review of Psychology, 38, 129–152.
Lamarre, Y., Filion, M. and Cordeau, J.P. (1971) Neuronal discharges of the ventrolateral nucleus of the
thalamus during sleep and wakefulness in the cat. 1: Spontaneous activity. Experimental Brain Research,
12, 480–498.
Levitt, J.B., Kiper, D.C. and Movshon J.A. (1994) Receptive fields and functional architecture of macaque V2.
Lindsay, R.D. and Scheibel, A.B. (1981) Quantitative analysis of the dendritic branching pattern of granule cells
from adult rat dentate gyrus. Experimental Neurology, 73, 286–297.
Lorente de Nó, R. (1949) Cerebral cortex: Architecture, intracortical connections, motor projections. In Fulton’s
Physiology of the Nervous System. London: Oxford University Press, pp. 288–313.
Lu, Z.L., Lesmes, L.A. and Sperling, G. (1999) The mechanism of isoluminant chromatic motion perception.
Proceedings of the National Academy of Sciences, U.S.A., 96, 8289–8294.
Lu, Z.L. and Sperling, G. (1995) The functional architecture of human visual motion perception. Vision
Research, 35, 2697–2722.
Lund, J.S., Henry, G.H., Macqueeen, C.L. and Harvey, A.R. (1979) Anatomical organization of the primary
visual cortex (area 17) of the cat. A comparison with area 17 of the macaque monkey. Journal of Comparat-
ive Neurology, 184, 599–616.
Luria, A. (1973) The working brain. London: Penguin Books.
101–104.
Maunsell, J.H.R., Nealey, T.A. and DePriest, D.D. (1990) Magnocellular and parvocellular contributions to
3323–3334.
Miller, B.M., Windrem, M.S. and Finlay, B.L. (1991) Thalamic ablations and neocortical development:
Alterations in thalamic and callosal connectivity. Cerebral Cortex, 1, 241–261.
Miller, R. (1975) Distribution and properties of commissural and other neurons in the cat sensorimotor cortex.
Miller, R. (1991) Cortico-hippocampal interplay. Berlin, Heidelberg: Springer Verlag.
Miller, R. (1996) Axonal conduction time and human cerebral laterality. Amsterdam: Harwood Academic.
Mukhametov, L.M. and Rizzolatti, G. (1970) The responses of lateral geniculate neurons to flashes of light
during the sleep-waking cycle. Archives Italiennes de Biologie, 108, 348–368.
Murphy, K.M., Jones, D.G. and Van Sluyters, R.C. (1995) Cytochrome-oxidase blobs in cat primary visual
Ohno, S. (1970) Evolution by gene duplication. Berlin, Heidelberg: Springer Verlag.
Onoe, H., Komori, M., Onoe, K., Takechi, H., Tsukada, H. and Watanabe, Y. (2001) Cortical networks recruited
for time perception: a monkey positron emission tomography (PET) study. Neuroimage, 13, 37–45.

Peterson, I. (1993) Newton’s Clock: Chaos in the Solar System. New York: W.H. Freeman and Co.
Porter, L.L. (1997) Morphological characterization of a cortico-cortical relay in the cat sensorimotor cortex.
Rakic, P. (1988) Specification of cerebral cortical areas. Science (New York), 241, 170–176.
Rakic, P., Suner, I. and Williams, R.W. (1991) A novel cytoarchitectonic area induced experimentally within the
primate visual cortex. Proceedings of the National Academy of Sciences, U.S.A., 88, 2083–2087.
Reid, R.C. and Alonso, J.-M. (1996) The processing and encoding of information in the visual cortex. Current
Opinion in Neurobiology, 6, 475–480.
Reiner, A. (1991) A comparison of neurotransmitter-specific and neuropeptide-specific neuronal cell types
present in the Dorsal Cortex in turtles with those present in the isocortex in mammals: Implications for the
evolution of Isocortex. Brain Behavior and Evolution, 38, 53–91.
Ringo, J.L., Doty, R.W., Demeter, S. and Simard, P.Y. (1994) Time is of the essence: a conjecture that
hemispheric specialization arises from interhemispheric conduction delay. Cerebral Cortex, 4, 331–243.
Rockland, K.S. (1995) Morphology of individual axons projecting from area V2 to MT in the macaque. Journal
Rockel, A.J., Hiorns, R.W. and Powell, T.P. (1980) The basic uniformity in structure of the neocortex. Brain,
103, 221–244.
Sakakura, H. (1968) Spontaneous and evoked unitary activities of cat lateral geniculate neurons in sleep and
wakefulness. Japanese Journal of Physiology, 18, 23–42.
of neocortex evolution. In: C.R. Noback and W. Montagna (eds), The Primate Brain. New York: Appleton
Century Crofts, pp. 137–208.
Scannell, J.W., Blakemore, C. and Young M.P. (1995) Analysis of connectivity in the cat cerebral cortex.
Scannell, J.W., Hilgetag, C.C., Panzeri, S. and Young, M.P. (1999) Hierarchical organization and latency in the
primate visual system. Society for Neuroscience Abstracts, 25, 821–14.
Schieber, M.H. and Hibbard, L.S. (1993) How somatotopic is the motor cortex hand area? Science (Washington),
261, 489–492.
Schmolesky, M.T., Wang, Y., Hanes, D.P., Thompson, K.G., Leutgeb, S., Schall, J.D. and Leventhal, A.G.
(1998) Signal timing across the macaque visual system. Journal of Neurophysiology, 79, 3272–3278.
Schüz, A. and Demianenko, G.P. (1995) Constancy and variability in cortical structure. A study on synapses and
dendritic spines in hedgehog and monkey. Journal für Hirnforschung, 36, 113–122.
Schüz, A. and Preißl, H. (1996) Basic connectivity of the cerebral cortex and some considerations on the corpus
callosum. Neuroscience and Biobehavioral Reviews, 20, 567–570.
Seidemann, E., Poirson, A.B., Wandell, B.A. and Newsome, W.T. (1999) Color signals in area MT of the
macaque monkey. Neuron, 24, 911–917.
Sharma, J., Angelucci, A. and Sur, M. (2000) Induction of visual orientation modules in auditory cortex. Nature
(London), 404, 841–847.
Shoham, D., Hübener, M., Schulze, S., Grinvald, A. and Bonhoeffer, T. (1997) Spatio-temporal frequency domains
and their relation to cytochrome oxidase staining in cat visual cortex. Nature (London), 385, 529–533.
Singer, W. and Gray, C.M. (1995) Visual feature integration and the temporal correlation hypothesis. Annual
Singer, W. (1998) Consciousness and the structure of neuronal representations. Philosophical Transactions of
the Royal Society of London, B, 353, 1829–1840.
Somogyi, P. (1977) A specific ‘axo-axonal’ interneuron in the visual cortex of the rat. Brain Research, 136,
345–350.
Steriade, M. and Hobson, J.A. (1976) Neuronal activity during the sleep-waking cycle. Progress in Neuro-
biology, 6, 155–376.
Stratford, K.J., Tarczy-Hornoch, K., Martin, K.A.C., Bannister, N.J. and Jack, J.J.B. (1996) Excitatory synaptic
inputs to spiny stellate cells in cat visual cortex. Nature (London), 382, 258–261.
Suga, N. (1984) The extent to which biosonar information is represented in the bat auditory cortex. In: G.M.
Edelman, W.E. Gall and W.M. Cowan (eds), Dynamic Aspects of Neocortical Function. New York: Wiley
and Sons, pp. 315–373.
Swindale, N.V. (2000) How many maps are there in visual cortex? Cerebral Cortex, 10, 633–643.
Swindale, N.V., Shoham, D., Grinvald, A., Bonhoeffer, T. and Hübener, M. (2000) Visual cortex maps are
optimized for uniform coverage. Nature Neuroscience, 3, 822–826.
Tanaka, K. (1983) Cross-correlation analysis of geniculostriate neuronal relationships in cats. Journal of
Neurophsyiology, 49, 1303–1318.
Tarczy-Hornoch, K., Martin, K.A.C., Stratford, K.J. and Jack, J.J.B. (1999) Intracortical excitation of spiny
neurons in layer 4 of cat striate cortex in vitro. Cerebral Cortex, 9, 833–843.
Tootell, R.B.H. and Taylor, J.B. (1995) Anatomical evidence for MT and additional cortical visual areas in
humans. Cerebral Cortex, 5, 39–55.

Toyama, K., Kimura, M. and Tanaka, K. (1981) Cross-correlation analysis of interneuronal connectivity in cat
visual cortex. Journal of Neurophysiology, 46, 191–201.
Valverde, F. (1971) Short axon neuronal subsystems in the visual cortex of the monkey. International Journal of
Valverde, F. (1985) The organizing principles of the primary visual cortex in the monkey. In: A. Peters and
E.G. Jones (eds), Cerebral Cortex, Vol. 3 (Visual Cortex). New York and London: Plenum Press, pp. 207–257.
Valverde, F. (1986) Intrinsic neocortical organization: some comparative aspects. Neuroscience, 18, 1–23.
Valverde, F. (1988) Competition for the sake of diversity. Behavioral and Brain Sciences, 11, 102–103.
van Essen, D.C. (1997) A tension-based theory of morphogenesis and compact wiring in the central nervous
system. Nature (London), 385, 313–318.
Victor, J.D. (2000) How the brain uses time to represent and process visual information (1). Brain Research,
886, 33–46.
Vogt, O. (1911) Die Myeloarchitektonik des Isocortex parietalis. Journal für Psychologiie und Neurologie, 18,
107–118.
Vogt, C. and Vogt, O. (1919) Allgemeinere Ergebnisse unserer Hirnforschung. Journal für Psychologiie und
Neurologie, 25, :292–398.
Berlin: Springer.
Wandell, B.A., Poirson, A.B., Newsome, W.T., Baseler, H.A., Boynton, G.M., Huk, A., Gandhi, S. and Sharpe,
L.T. (1999) Color signals in human motion-selective cortex. Neuron, 24, 901–909.
Windrem, M.S. and Finlay, B.L. (1991) Thalamic ablations and neocortical development: alterations of cortical
cytoarchitecture and cell number. Cerebral Cortex, 1, 230–240.
Yabuta, N.H. and Callaway, E.M. (1998) Functional streams and local connections of layer 4C neurons in the
primary visual cortex of the macaque monkey. Journal of Neuroscience, 18, 9489–9499.
Young, M.P. (1992) Objective analysis of the topological organization of the primate cortical visual system.
Young, M.P., Scannell, J.W. and Burns, G. (1994) The Analysis of Cortical Connectivity. New York: R.G.
Landes; Heidelberg: Springer.
Young, M.P., Barone, P., Hilgetag, C.-C., Kennedy, H. and Falchier, A. (2000a) A determinate hierarchy for
macaque visual cortex derived from quantitative laminar data. Society for Neuroscience, Abstracts, 26,
449.7.
Young, M.P., Hilgetag, C.C. and Scannell, J.W. (2000b) On imputing function to structure from the behavioural
effects of brain lesions. Philosophical Transactions of the Royal Society: Biological Sciences, 355, 147–161.

Almut Schuez, Robert Miller - Cortical Areas - Unity and Diversity (Conceptual Advances in Brain Research) - CRC Press (2002)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Almut Schuez, Robert Miller - Cortical Areas - Unity and Diversity (Conceptual Advances in Brain Research) - CRC Press (2002)

Uploaded by

Copyright:

Available Formats

Cortical Areas:

Unity and Diversity

© 2002 Taylor & Francis

© 2002 Taylor & Francis

London and New York

© 2002 Taylor & Francis

Taylor & Francis is an imprint of the Taylor & Francis Group

© 2002 Taylor & Francis

British Library Cataloguing in Publication Data

© 2002 Taylor & Francis

1 Introduction: Homogeneity and Heterogeneity of Cortical Structure:

2 Cyto- and Myeloarchitectonics: Their Relationship and Possible

6 Regional Dendritic Variation in Primate Cortical Pyramidal Cells 111

10 The Cerebral Cortex of Mammals: Diversity within Unity 195

12 Cross-Modal Plasticity as a Tool for Understanding the Ontogeny and

15 Connectional Organisation and Function in the Macaque Cerebral Cortex 351

19 Discussion Section 459

© 2002 Taylor & Francis

© 2002 Taylor & Francis

Katrin Amunts Hubert R. Dinse

Laura López-Mascaraque Jörg Rademacher

© 2002 Taylor & Francis

Stewart Shipp Malcolm P. Young

Facundo Valverde Karl Zilles

© 2002 Taylor & Francis

THE EMPIRICAL STATUS

© 2002 Taylor & Francis

KEYWORDS: areas, connectivity, cytoarchitectonics, human cerebral cortex, myelin, myelo-

© 2002 Taylor & Francis

© 2002 Taylor & Francis

1.3. Cyto- and Myeloarchitectonics: Different Aspects of the Same Underlying

© 2002 Taylor & Francis

© 2002 Taylor & Francis

2. SIMPLE RULES RELATE THE CYTO- AND

2.1. Cytoarchitectural Data

2.2. Two Basic Assumptions

2.2.1. First assumption

2.2.2. Second assumption

© 2002 Taylor & Francis

2.3. Procedure and Results

© 2002 Taylor & Francis

© 2002 Taylor & Francis

3. CAN CYTO- AND MYELOARCHITECTONICS BE INTERPRETED

3.1. Connectivity and Function

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

(i) Borders between cortical areas can be identified by observer-independent statistical

© 2002 Taylor & Francis

© 2002 Taylor & Francis

© 2002 Taylor & Francis

3. HOW DIFFERENT ARE TWO CORTICAL AREAS IN THEIR

Whether a cortical region is homogeneous in architecture and thus constitutes a single

© 2002 Taylor & Francis

© 2002 Taylor & Francis

(Dis-)similarities between these areas were quantified by calculating a multivariate