(Cognitive Linguistics Research, 19.1) Susanne Niemeier, Martin Pütz - Theory and Language Acquisition-De Gruyter Mouton (2001)

Applied Cognitive Linguistics I: Theory and Language Acquisition
WDE
G
Cognitive Linguistics Research
19.1
Editors
René Dirven
Ronald W. Langacker
John R. Taylor
Mouton de Gruyter
Berlin · New York
Applied Cognitive Linguistics I:
Theory and Language Acquisition
Edited by
Martin Pütz
Susanne Niemeier
René Dirven
Mouton de Gruyter
Berlin · New York 2001
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter GmbH & Co. KG, Berlin
© Printed on acid-free paper

which falls within
the guidelines of the ANSI
to ensure permanence and durability.
Library of Congress Cataloging-in-Publication Data
Applied cognitive linguistics / edited by Martin Pütz, Susanne

Niemeier, René Dirven.
p. cm. - (Cognitive linguistics research ; 19)
Contents: 1. Theory and language acquisition - 2. Lan-
guage pedagogy.
ISBN 3110172216 (v. 1 : alk. paper) - ISBN 3110172224
(v. 2 : alk. paper)
1. Cognitive grammar. 2. Language acquisition. 3. Lan-
guage and languages-Study and teaching. I. Pütz, Martin,
1955- II. Niemeier, Susanne, 1960- III. Dirven, René. IV.
Series.
P165 .A66 2001
418-dc21
2001044895
Die Deutsche Bibliothek — Cataloging-in-Publication Data
Applied cognitive linguistics / ed. by Martin Pütz .... - Berlin ;

New York : Mouton de Gruyter
(Cognitive linguistics research ; 19)
1. Theory and language acquisition. - 2001
ISBN 3-11-017221-6
© Copyright 2001 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin
All rights reserved, including those of translation into foreign languages. No part of this book
may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopy, recording, or any information storage and retrieval system, without
permission in writing from the publisher.
Printing: WB-Druck, Rieden/Allgäu
Binding: Lüderitz & Bauer, Berlin
Printed in Germany
Acknowledgements
Most of the present contributions were originally presented at the 29th

International LAUD Symposium held in Landau, Germany, from
March 25-28, 2000. All were selected for inclusion only after a
lengthy process of refereeing and, in some cases, extensive revising.
Their work resulted in the publication of two volumes:
Applied Cognitive Linguistics. Vol. I. Theory and Language Acquisi-

tion
Applied Cognitive Linguistics. Vol. II. Language Pedagogy
While compiling the two volumes we have incurred a number of

debts that we wish to acknowledge. We are deeply indebted to two of
the editors of the book series Cognitive Linguistics Research: Ronald
W. Langacker and John R. Taylor, whose support was crucial to the
emergence of the two volumes.
We would like to thank the organizing staff of the symposium, in

particular Susanne Heid, Alexander Kraft, and Caria M. Sandy as
well as Heike Ramsauer for proofreading major parts of the manu-
script. Thanks are also due to Anke Beck and Birgit Sievert (Mouton
de Gruyter) for their kind assistance and cooperation with this ven-
ture.
Furthermore, our sincere thanks go out to the authors, who have re-
sponded with professionalism to all the requests that have been made
of them. In this regard, we would also like to express a great debt of
gratitude to the expertise of the many scholars who acted as our refe-
rees: Angeliki Athanasiadou, Frank Boers, Willis Edmondson, Carlos
Inchaurralde, Dirk Geeraerts, Stefan Gries, Peter Grundy, Juliane
House, Bernd Kortmann, Penny Lee, Lienhard Legenhausen, Bert
Peeters, Mechthild Reh, Sally Rice, Francisco Ruiz de Mendoza,
Doris Schönefeld, Rainer Schulze, Elzbieta Tabakowska, Jef Ver-
schueren, Marjolijn Verspoor, Helmut Vollmer, Michael Wendt,
Karin Wenz.
vi Acknowledgements
Above all, we want to thank Birgit Smieja, who did a marvelous job
in designing the layout of the book and in taking care of the laser
printout.
The Editors
Duisburg, Bremen, and Landau July 2001
List of Contributors
Jenny Cook-Gumperz
University of California, Santa Barbara, U.S. A.
Vyvyan Evans
Georgetown University, Washington, D.C., U.S.A.
CliffGoddard
University of New England, Annidale, Australia
Paul J. Hopper
Carnegie Mellon University, Pittsburgh, U.S.A.
Istvan Kecskes
State University of New York at Albany, U.S.A.
Amy Kyratzis
University of California, Santa Barbara, U.S.A.
Sydney M. Lamb Houston, Tx., U.S.A.

Rice University,
Ronald W. of
University Langacker
California, San Diego, U.S.A.
Katharina J. Rohlfing
University of Bielefeld, Germany
Sang Hwan Seong
University of Bonn, Germany
Andrea Tyler
Georgetown University, Washington, D.C., U.S.A.
Contents
Acknowledgements ν
List of Contributors vii
Introduction xiii
Martin Pütz, René Dirven and Susanne Niemeier
Section 1: Cognitive approaches to the English tense system
Cognitive linguistics, language pedagogy, and the

English present tense 3
Ronald W. Langacker
Pretend play: trial ground for the simple present 41

Jenny Cook-Gumperz and Amy Kyratzis
The relation between experience, conceptual structure

and meaning: non-temporal uses of tense and language
teaching 63
Andrea Tyler and Vyvyan Evans
Section 2: Facets of prototypes in grammatical constructions
Grammatical constructions and their discourse origins:

prototype or family resemblance? 109
Paul J. Hopper
Transitivity parameter and prominence typology:

a cross-linguistic study 131
Sang Hwan Seong
χ Contents
Section 3: Neurocognitive and cognitive issues of

language acquisition in general
Learning syntax - a neurocognitive approach

Sydney M. Lamb
Conceptual primes in early language development

Cliff Goddard
No preposition required. The role of prepositions for the

understanding of spatial relations in language acquisition
The 'Graded Salience Hypothesis' in second language

acquisition
Istvan Kecskes
Subject Index
Contents of volume II
Acknowledgements ν
List of Contributors vii
Introduction xiii
Martin Pütz, Susanne Niemeier and René Dirven
Section 1: Bottom-up approaches: Phrasal verbs and

phraseological expressions
English phrasal verbs: theory and didactic application 3

René Dirven
Teaching English phrasal verbs: a cognitive approach 29

Andrzej Kurtyka
A usage-based approach to modeling and teaching

the phrasal lexicon 55
Kurt Queller
Section 2: Top-down approaches:

Metaphor and idiom study
A cognitive linguistic view of learning idioms in

an FLT context 87
Zoltán Kövecses
On the systematic contrastive analysis of conceptual metaphors:

case studies and proposed methodology 117
Antonio Barcelona
xii Contents
Section 3: Systematical order instead of chaos

in morphology and lexis
A conceptual analysis of English -er nomináis 149

Klaus-Uwe Panther and Linda Thornburg
Basicness and conceptual hierarchies in foreign language

learning: a corpus-based study 201
Friedrich Ungerer
Section 4: Cultural models in education
The African cultural model of community in English language

instruction in Cameroon: the need for more systematicity 225
Hans-Georg Wolf and Augustin Simo Bobda
Subject Index 261

Introduction
Martin Pütz, René Dirven and Susanne Niemeier
The main title of the two volumes, i.e. Applied Cognitive Linguistics,
not only focuses on the theory of cognitive linguistics as it can be
applied to the teaching and learning of foreign languages; it also en-
compasses psycholinguistic models and theories with a focus on first
language acquisition. Thus, Volume I is concerned with the interac-
tion between language, cognition and acquisition in general (first and
second language acquisition), and Volume II addresses a series of
cognitive principles of linguistic, i.e. conceptual organization while
acquiring, learning and teaching second or foreign languages.
The two volumes address a number of important topics in the the-
ory, acquisition and pedagogy of languages seen from the perspective
of cognitive linguistics. We cannot discuss here in detail the main
principles, aims and findings of the cognitive enterprise, but would
nevertheless like to refer to the following assumptions and implica-
tions of the discipline of cognitive linguistics as set out by Rudzka-
Ostyn (1993: 1):
1. As one domain of human cognition, language is intimately linked

with other cognitive domains and as such mirrors the interplay of
psychological, cultural, social, ecological, and other factors.
2. Linguistic structure depends on (and itself influences) conceptu-
alization, the latter being conditioned by our experience of our-
selves, the external world and our relation to that world.
3. Grammar is motivated by semantic considerations.
4. Language units are subject to categorization which commonly
gives rise to prototype-based networks; much of it critically in-
volves metaphor and metonymy.
xiv Martin Pütz, René Dirven and Susanne Niemeier
5. Given the interaction among language subcomponents as well as

the interaction between language and other domains of cognition,
the various autonomy theses and dichotomies proposed in the lin-
guistic literature have to be abandoned; a strict separation of syn-
tax, morphology and lexicon is untenable; furthermore, it is im-
possible to separate linguistic knowledge from extra-linguistic
knowledge.
As a dynamic usage-based model (Langacker 2000: 9Iff.), cognitive

linguistics is predestined to have an impact on applied research in
such areas as language acquisition and language pedagogy, and, we
may add, on the various facets of ideology and society as such. Al-
though previous studies have already dealt with child language acqui-
sition from a cognitive linguistics perspective (cf. Slobin 1985, Sinha
et al. 1994, Niemeier and Achard 2000) as well as with the pedagogi-
cal implications of cognitive linguistics (Dirven 1989, Dirven and
Taylor 1994, Taylor 1987, 1993), the Landau Symposium was the
first to concentrate on "Applied Cognitive Linguistics" at a more
principled and encompassing level. Even if cognitive linguistics is a
rapidly expanding linguistic paradigm, the impact of this new and
revolutionary linguistic theory on various branches of applied re-
search and on their pedagogical implications is only now beginning
to be felt. The two volumes provide a systematic attempt to carve out
pathways from the links between language and cognition to the fields
of language acquisition and language pedagogy, and to deal with
them in one coherent framework called Applied Cognitive Linguis-
tics. Nevertheless, the editors are open to critical, alternative propos-
als and invite critical reflections from authors not agreeing with the
mainstream CL approach.
The papers in the present volume have been arranged according to
three major strings, namely (i) cognitive approaches to the English
tense system (Langacker, Cook-Gumperz and Kyratzis, Tyler and
Evans); (ii) facets of prototypes in grammatical constructions (Hop-
per, Seong); and (iii) neurocognitive and cognitive issues of language
acquisition in general (Lamb, Goddard, Rohlfing, Kecskes).
Introduction xv
The first three contributions deal with an account of the use of the
English tense system seen from various perspectives. In his paper
"Cognitive linguistics, language pedagogy, and the English present
tense" Ronald W. Langacker examines the English present tense, as
well as related phenomena like the progressive and the per-
fect/imperfect contrast, from the perspective of his well-known the-
ory of cognitive grammar. Although Langacker does not specifically
focus on issues of second language acquisition as such, he refers to
several pedagogical implications of cognitive linguistic theory and
possible directions for further research. Some of these directions
follow from the usage-based nature of cognitive grammar, in which
linguistic units are seen as being abstracted from usage events. Given
the fact that regular constructions of full generality constitute only a
small proportion of conventional patterns, Langacker assumes that
complete mastery of linguistic rules does not assure any degree of
actual fluency in a language. Furthermore, Langacker underlines the
importance of the communicative, social and cultural context of any
speech encounter and refers to the notion of 'construal', i.e. "our
ability to conceive and portray the same situation in alternate ways".
In conclusion, Langacker criticizes the traditional way of looking at
tense mainly on the grounds that these approaches have been objec-
tivist in nature. They lack an awareness of 'construal' and viewing
arrangements and the subjective basis of factors like homogeneity
and bounding. These are the two major viewing arrangements that we
impose on all phenomena. In the domain of things we can impose a
homogeneous view and conceptualize phenomena as homogeneous,
unbounded substances, or else as bounded objects with an internal
heterogeneous structure. This distinction corresponds in the language
to that between mass nouns and count nouns. Similarly in the domain
of processes, we impose a homogeneous arrangement and conceptu-
alize homogeneous states or activities, or else we impose a bounded
view and conceive of processes as bounded events. This distinction
corresponds in the verb system to imperfective tense forms (progres-
sive or habitual form) or to perfective, i.e. bounded tense forms. This
major insight may stimulate many language pedagogues and psycho-
linguists to rethink their entire language pedagogy or acquisition
xvi Martin Pütz, René Dirven and Susanne Niemeier
views. Although Langacker does not draw specific conclusions about

teaching, he argues that a clear understanding of its conceptual im-
port should at least allow pedagogical issues to be formulated in the
proper terms.
Some of the essential features of Langacker's analysis of the Eng-
lish present tense, e.g. the notions of 'construal' and 'viewing ar-
rangement' are taken up by Jenny Cook-Gumperz and Amy Ky-
ratzis in their article entitled "Pretend play: trial ground for the sim-
ple present", in which they illustrate Langacker's theoretical distinc-
tions
Their research pertains to the relationship between the conceptu-
alizes and the situation being viewed, and how it enables partici-
pants and readers/hearers to shift perspectives on the story events.
According to the authors, in storytelling a viewing arrangement is
sometimes imposed as a self-conscious yet suspenseful perspective.
Thus young children moving around with their toy cars or planes and
verbalizing the events in which they are involved use the simple pre-
sent to express the quick succession of events which they are bring-
ing about themselves. A major argument of the paper is that English
speaking children's early understanding of the tense system is a key
prerequisite for literacy. In order to attain this goal, children's en-
gagement in pretend play as well as in story telling and reading expe-
riences with adults seems to be indispensable. The authors cite vari-
ous other examples from children's books to illustrate the kinds of
narrative experiences young children participate in before they learn
to read for themselves. Their data are based on three studies, each
focusing on the speech of little children (aged 3-7) who are engaged
in collaborative character play with peers. In the center of the discus-
sion is the use of the present tense for temporally-sequenced action
and for habitual state, as well as present progressive for ongoing ac-
tion in in-role play with 3 and 4 year olds. The authors arrive at the
conclusion that through tense/aspect contrasts, children's stories are
able to provide more nuanced views of the characters and their situa-
tion, thereby shifting perspectives and viewpoints on events.
Uses of tense are also discussed by Andrea Tyler and Vyvyan
Evans in their paper "The relation between experience, conceptual
Introduction xvii
structure and meaning: Non temporal uses of tense and language

teaching". While Cook-Gumperz and Kyratzis put much emphasis on
the use of tense to shape different types of processes in terms of
viewing arrangements, Tyler and Evans concentrate on the temporal
function of tense, especially on the so-called exceptional, non-
temporal uses of the present tense, which do not signal time-
reference but serve various other conceptual functions such as the
expression of intimacy, salience, actuality, and attenuation. Tyler and
Evans reject the traditional view that such non-temporal usages are
arbitrary exceptions; rather they consider them as highly motivated
uses, deriving directly from how we conceptualize time. The authors
do not see these uses as metaphorical extensions of the temporal
functions of tense, either. Instead they share a view of 'experiential
correlations' that is also proposed by Joe Grady and Pinker. While
perceiving processes in time, we undergo many other experiences,
too. One set of these are spatial experiences. We associate these with
each other and conceive of them as being 'close' or 'distal'. This
higher principle can be invoked for both the temporal and non-
temporal uses of tenses. What is close is that which is present here or
can be intimate to us, or in the center of attention and actuality. Here
present tenses come in. What is distal is non-present or irrealis, or to
be softened or attenuated. The authors argue that the so-called 'ex-
ceptional' non-temporal uses of English tense is an area of grammar
often either ignored altogether or treated as arbitrary by pedagogical
grammars and language teachers. The main thesis in the paper is, as
just shown, that the non-temporal uses of tense are related to its time-
reference function in a fully motivated way. Central to the discussion
is the process of meaning extension or pragmatic strengthening, by
which a conventional non-temporal meaning can become associated
with a particular tense morpheme. Their rather elaborate theoretical
and practical underpinnings of their view illustrate convincingly that
by assuming the principles and methodology of cognitive linguistics
it is possible to relate the non-temporal senses associated with tense
and the time-reference function to each other in a plausible way.
They thus help to solve the difficulty of language teachers trying to
present insightfully the non-temporal uses associated with tense, i.e.
xviii Martin Pütz, René Dirven and Susanne Niemeier
to explain to the learner why the foreign language should be as it is

and to bring to the learner's consciousness the conceptualizations
conventionally associated with the structures of the foreign language.
The notion of 'exceptional uses' of grammar is also reflected in
Paul J. Hopper's paper "Grammatical constructions and their dis-
course origins: Prototype or family resemblance?" Whereas Tyler and
Evans emphasize the motivating force pertaining to so-called 'ex-
ceptional', peripheral uses of the present tense and drift away from a
view of a prototypical core and metaphorical extensions, Hopper also
takes up the implied notion of prototypicality but seriously calls into
doubt its validity as suggested by many, though not all, proponents of
cognitive linguistics, and also by many other scientists.1 The author
concentrates on one complex construction, the English pseudocleft
sentence, and discusses many examples taken from natural discourse
in which such a pseudocleft sentence appears in a fragmentary 'in-
complete' form. The author's view is that pseudocleft sentences do
not primarily serve to highlight any single identifiable sentence con-
stituent, but that this highlighting function is a mere side-effect. The
real function of pseudocleft sentences is rather to delay an assertion
for a number of pragmatic reasons. One of those may be the need to
create more processing time. This also explains why so many pseu-
doclefts are broken off in mid-plan: if the processing problem has
been solved, the construction can just be discarded. Thus the author
shows that there is a need to revise the prototype model in grammati-
cal construction and instead suggests adopting the family resem-
blance model. In such a view many different uses and meanings can
be vaguely related to one another without any of them being an ex-
tension from a core use. 'Canonical' constructions should be seen as
"highly stylized cultural artifacts ... put together and grammaticalized
through stylistic and normative conventions" and not as prototypes or
as the source of 'deviant' fragmentary instantiations in discourse.
Hopper thus also indirectly criticizes mainstream cognitive linguis-
tics in the sense that 'schemata' based on prototypical samples of
language (i.e. planned discourse) often do not reflect the observable
realities of language use in natural discourse. This observation, Hop-
per concludes, has important consequences for applied linguistics and
Introduction χίχ
foreign language teaching. In general, teachers should pay attention

to the role of "incomplete" non-prototypical utterances as the
"building blocks of fluent and appropriate oral discourse".
An interesting view of prototype theory is also proposed by Sang
Hwan Seong, who analyses the relationships of the "Transitivity
parameter and prominency typology" in the light of recent develop-
ments in cognitive grammar. Seong critically evaluates previous ap-
proaches dealing with particular linguistic schemata and prototypes
organized in and across languages such as English, German and Ko-
rean. It is traditionally argued that active clauses high in transitivity
typically undergo passivization in which agentive subjects, totally
affected objects and action predicates are correlated. Seong claims
that, on the contrary, the transitivity and prototype parameter should
be constrained such that the languages with high semantic transpar-
ency and a material case-marking system such as Korean proportion-
ally correlate with less grammaticization, e.g. they do not require
overt subjects. Seong also suggests that the grammatical properties of
a given language are closely linked to the presence/absence of the
built-in subjective reference point. The effects of this theoretical po-
sition are also reflected in L2 data based on an analysis of the writing
samples of German and Korean students. It was found, for example,
that German learners of Korean tend to produce, on the basis of their
German templates, sentences in Korean with overt subjects almost
without any exception early on and then gradually learn to drop these
subjects when they reach a higher level of proficiency. It was found,
furthermore, that typical Korean phenomena such as topic particles
and double nominative constructions are particularly difficult fea-
tures for Germans to acquire, even in the later stage of learning. It is,
in the editors' view, not exaggerated to conclude that prototypicality
is a highly language-specific phenomenon, which therefore does not
show up on a universal basis.
A neurocognitive perspective of learning syntax is advocated by
Sydney M. Lamb in his article "Learning syntax - a neurocognitive
approach". Lamb asserts that the two approaches to language, i.e. the
neurocognitive one and the analytical one, the latter including for him
cognitive linguistics, have different concerns from each other. The
XX Martin Pütz, René Dirven and Susanne Niemeier
distinction is based on the admittedly disputable claim that analytical

linguistics is concerned only with the analysis of linguistic data with-
out any reference to a cognitive and neurological basis, while neuro-
cognitive linguistics takes as the object of its cognitive study the neu-
rocognitive system of the individual.2 More particularly, Lamb criti-
cizes the traditional view that the child or adult learning a second
language must somehow acquire syntactic rules which the speaker
has internalized. Contrary to this assumption, Lamb proposes an al-
ternative view which holds that syntactic information is somehow
attached to individual lexical items. According to such a view, the
acquisition of syntactic knowledge is "part and parcel of the acquisi-
tion of lexical knowledge and therefore occurs little by little as indi-
vidual lexemes are learned". Learning consists mainly of building
and strengthening connections between synapses in the brain and
adjusting thresholds of nodes. In the case of the child learning lexical
items, it is commonly the case that conceptual connections are al-
ready present before they get connected to their linguistic expres-
sions. Lamb illustrates this point with words like crawl and eat. For
example, by the time a child learns these words, it has long since
been crawling without a crawler - and it is aware that eating has two
participants, since without eater and food there can be no eating. So
the child must learn a phonological expression plus the relative or-
dering of the already known participants.
Whereas Lamb's paper is an attempt to put forward a neurocogni-
tive approach to learning syntax which holds for both first and sec-
ond language acquisition, Cliff Goddard takes a stand on first lan-
guage acquisition with a clear focus on children's semantic and con-
ceptual development. The author is known to be one of the main pro-
ponents of the natural semantic metalanguage (NSM) approach origi-
nated by Anna Wierzbicka. To compare meanings expressed in dif-
ferent languages and different cultures, one needs to establish a se-
mantic metalanguage. This metalanguage is the set of semantic 'at-
oms' or primitives which cannot be further reduced or made simpler.
These primitives are also claimed to suffice to define or explain all
other terms in all possible languages. Since the semantic primes are
valid in all languages, they are by definition universal. These univer-
Introduction xxi
sal semantic primitives are carved out of natural language and also
have an inherent 'conceptual syntax'. Goddard's paper "Conceptual
Primes in Early Language Development" tackles the problem that
young children in their acquisition process do not start off with the
seventy odd semantic primitives, but with more complex concepts.
The author therefore explores various hypotheses about the nature
and identities of the innate concepts which may underpin language
acquisition. Since many of the child' s early words are more complex
and cannot be explained in terms of conceptual primes, a possible
explanation might be that these words have simpler meanings than
the corresponding words in an adult's vocabulary. Goddard con-
cludes from this that the child must have a certain 'conceptual vo-
cabulary' of prime concepts prior to their emergence in language
proper. In the remaining part of the paper Goddard engages in a de-
tailed semantic analysis of the child's production vocabulary (i.e. the
non-prime words and utterances) in order to investigate the concep-
tual primes which may be 'latent' in a child's early lexicon. An inter-
esting observation refers to the fact that the production vocabulary of
primes lags several months behind the conceptual vocabulary. As
Goddard's study is based on the language development of one single
child, i.e. Goddard's son Peter, it would be interesting to see to what
extent Peter's acquisition story is representative for other children's
production vocabulary and their acquisition of primes in the concep-
tual vocabulary.
Whereas Goddard in general investigates the development of the
linguistic and conceptual vocabulary of child language, Katharina J.
Rohlfing's paper "No preposition required. The role of prepositions
for the understanding of spatial relations in language acquisition" is a
report on experiments focusing on the role of one single syntactic
category, i.e. prepositions, for the understanding of spatial relations
also in first language acquisition. Rohlfing presents the results of an
empirical study of Polish-speaking children at the age of 2. The hy-
pothesis of her study was that, in prototype situations, children un-
derstand instructions even without prepositions, which however they
need in order to understand the instructions in abstract situations. The
experiments show that infants rely not only on the appropriate loca-
xxii Martin Pütz, René Dirven and Susanne Niemeier
tive prepositions, but also on non-linguistic strategies, such as an

object's relational character, its physical properties and the typicality
of a situation. Comparing her work with Goddard's, we can come to
an interesting hypothesis: the child may have a number of conceptual
categories for which (s)he does not yet have linguistic expressions
(Rohlfing) or preliminary labels from adult language (Goddard). The
prepositions themselves are acquired much later than their supposed
understanding indicates. Before the full acquisition of prepositions,
infants are able to master the basic relations "in" and "on" with the
help of non-linguistic strategies, and a linguistic instruction cannot be
understood context-independently at this age. Rohlfing concludes
that conceptual knowledge should be explored more task- and con-
text-dependently, "because the non-linguistic strategies are a part of
the infants' understanding and they establish the schematic meaning
of spatial relations".
Whereas Rohlfing's psycholinguistic study examines the role of
prepositions for understanding instructions with spatial relations in
first language acquisition, Istvan Kecskes considers both first and
second language acquisition, although with a clear focus on foreign
language learning. In his paper "The 'Graded Salience Hypothesis' in
second language acquisition" the author discusses the validity of the
'Graded Salience Hypothesis' (GSH), which was developed to ex-
plain LI processing. The GSH claims that figurative and literal lan-
guage use are governed by a general principle of salience in LI : Sali-
ent meanings (conventional, frequent, familiar) are processed first,
and parallel processing is induced when more than one meaning is
salient. Based on an empirical cross-sectional study, the author shows
that its application in second language acquisition can also be very
useful in the sense that salience is language- and culture-specific and
that it derives from certain knowledge structures that are essential
parts of the conceptual base of native speakers. When acquiring a
non-primary language, students have to learn not only the forms of
that particular language, but also the conceptual structures associated
with those forms. Salience is a very important part of the socio-
cultural heritage of native speakers, and it is something that reveals a
unique feature of the human mind.
Introduction xxiii
Notes
1. Problems of the prototype approach are also pointed out in CL, e.g. by Geera-
erts (1988). On the other hand, Hopper and Thompson (1985) themselves in-
voke prototype theory for typology. Without calling himself a cognitive linguist,
Kastovsky (1988) opposes prototype semantics as a superior alternative to
structural semantics. One example out of the hundreds of non-cognitive lin-
guists working with prototype ideas is Shaver et al. (1987).
2. Peeters (2001) devotes a whole paper to the - in his view - misnomer 'cognitive
linguistics' for the CL paradigm, which he believes should call itself 'Cognitive
Linguistics' with upper case, leaving the lower case appellation 'cognitive lin-
guistics' to its inventor, i.e. Sydney Lamb.
References
Dirven, René
1989 Cognitive linguistics and pedagogic grammar. In: Gerhard Leitner
and Gottfried Graustein (eds.), Linguistic Theorizing and Gram-
mar Writing, 56-75. Tübingen: Niemeyer.
Dirven, René and John R. Taylor
1994 English modality: A cognitive-didactic approach. In: Keith Car-
Ion, Kristin Davidse and Brygida Rudzka-Ostyn (eds.), Perspec-
tives on English. Studies in Honour of Professor Emma Vorlat,
542-556. Leuven: Peeters.
Geeraerts, Dirk
1988 Prototypicality as a prototypical notion. Communication & Cog-
nition 21: 343-355.
Hopper, Paul J. and Sandra A. Thompson
1985 The uses of the notion 'prototype' in typological studies. In: H.-J.
Seiler and G. Brettschneider (eds.), Language Invariants and
Mental Operations, 238-244. Tübingen: Gunter Narr.
Kastovsky, Dieter
1988 Structural semantics or prototype semantics? The evidence of
word-formation. In: Werner Hüllen and Rainer Schulze (eds.),
Understanding the Lexicon. Meaning, Sense and World Knowl-
edge in Lexical Semantics, 190-203. Tübingen: Niemeier.
Langacker, Ronald W.
2000 Grammar and Conceptualization. Berlin/New York: Mouton de
Gruyter.
xxiv Martin Pütz, René Dirven and Susanne Niemeier
Niemeier, Susanne and Michel Achard (eds.)

2000 Cognitive Linguistics: Special issue on language acquisition.
Cognitive Linguistics 11-1/2, 1-151. Berlin/New York: Mouton
de Gruyter.
Peeters, Bert
2001 Does Cognitive Linguistics live up to its name? In: René Dirven,
Bruce Hawkins and Esra Sandikcioglu (eds.), Language and Ide-
ology. Volume 1 : Theoretical Cognitive Approaches. (Current Is-
sues in Linguistic Theory 204), 83-106. Amsterdam: John Ben-
jamins.
Rudzka-Ostyn, Brygida
1993 Introduction. In: Richard A. Geiger and Brygida Rudzka-Ostyn
(eds.), Conceptualizations and Mental Processing in Language.
A Selection of Papers from the First International Cognitive Lin-
guistics Conference in Duisburg, 1989, 1-20. Berlin: Mouton de
Gruyter.
Shaver, P., J. Schwartz, D. Kirson and C. O'Connor
1987 Emotion knowledge: Further exploration of a prototype approach.
Journal of Personality and Social Psychology 52: 1061-1086.
Sinha, Chris, Lis A. Thorseng et al.
1994 Comparative spatial semantics and language acquisition: Evi-
dence from Danish, English and Japanese. Journal of Semantics
11:253-287.
Slobin, Dan I. (ed.)
1985 The Crosslinguistic Study of Language Acquisition. Vol. 2: Theo-
retical Issues. Hillsdale, N.J.: Lawrence Erlbaum.
Taylor, John R.
1987 Metaphors of communication and the nature of listening and
reading comprehension interface: Journal of Applied Linguistics
1: 119-134
1993 Some pedagogical implications of cognitive linguistics. In: Rich-
ard A. Geiger and Brygida Rudzka-Ostyn (eds.), Conceptualiza-
tions and Mental Processing in Language. A Selection of Papers
from the First International Cognitive Linguistics Conference in
Duisburg, 1989,201-223. Berlin: Mouton de Gruyter.
Section 1
Cognitive approaches to the English

tense system
Cognitive linguistics, language pedagogy, and
the English present tense
Ronald W. Langacker
1. Introduction
Although I cannot claim any experience or substantial knowledge in

the area of language pedagogy, I do recognize how essential it is.
There seems little doubt that, in the broad field of language and lin-
guistics, the applied, pedagogical side outstrips the theoretical side in
terms of intrinsic importance, numbers of practitioners, and the se-
curing of institutional tolerance. The applied folk are, as it were, car-
rying linguistic theoreticians on their back. Not only that, but the
burden tends to be heavy, onerous, and quite unappreciative.
Perhaps cognitive linguistics will prove to be lighter, less onerous,
and more appreciative than certain previous theoretical burdens. I
hope it will even prove useful. Let me try to contribute by sketching
some pedagogical implications of cognitive linguistic theory, and
some possible directions of research. I see the effectiveness of peda-
gogical applications as an important empirical test for linguistic theo-
ries. My suspicion is that, in the long run, cognitive grammar will not
fare badly in this regard.
2. Implications of a usage-based model
Cognitive grammar is a usage-based model of linguistic structure, in

which linguistic units are seen as being abstracted from usage events
by the reinforcement of recurring commonalties (Langacker 1988,
2000). Such units run the gamut from the fully specific to the maxi-
mally schematic, with specific structures and local regularities being
4 Ronald W. Langacker
at least as important as high-level generalizations. Moreover, special

cases of general patterns can themselves have the status of linguistic
units, provided that they are learned by speakers and conventional in
a speech community.
Since linguistic units arise by abstraction from usage events, a us-
age-based approach is necessarily a construction-based approach.
That is, constructions are basic rather than epiphenomenal, and rules
are nothing more than schematic constructions (cf. Goldberg 1995).
A construction is merely an assembly of symbolic structures (form-
meaning pairings). Networks of constructions are deemed sufficient
for a full description of lexicon and grammar, which form a contin-
uum.
The usage-based and construction-based nature of cognitive
grammar (and other approaches to cognitive linguistics) has certain
implications for language pedagogy. One of them is the limited im-
portance of fully general rules. Since regular constructions of full
generality constitute only a small proportion of conventional patterns,
complete mastery of linguistic rules (as normally conceived) does not
assure any degree of actual fluency in a language. I can personally
attest that, by thoroughly learning all the rules and vocabulary found
in traditional textbooks and taught in traditional language classes,
one does not come even close to being fluent in a language. To
achieve fluency, one has to learn in addition a vast store of fixed ex-
pressions and normal ways of phrasing things in particular circum-
stances, out of all the ways the "rules" in principle permit. Only by
controlling this immense inventory of conventional expressions and
conventional modes of expression is it possible for speakers to put
together a continuous flow of complex expressions in real time
(Langacker 1987a: 35-36). If everything had to be computed from
scratch, from general grammatical rules and traditionally recognized
vocabulary, the exigencies of rapid, fluent speech would overwhelm
our processing capacities. This is the state to which I was brought by
traditional language training.
I have often said - and I think fairly accurately - that lexicon and
grammar as traditionally conceived and taught constitute only around
1% of the linguistic knowledge required for fluent speech, yet this
Cognitive linguistics, language pedagogy, and the English present tense 5
1% attracts around 99% of the effort and attention of linguists. The

usage-based and constructional perspectives encourage a reallocation
of effort that should bring these proportions more into balance. If it is
accurate, my assessment of the discrepancy between what linguists
focus on and what speakers need to know will not come as any great
surprise to those engaged in language pedagogy. It should be evident
to anyone confronted with real-life problems of language teaching.
Nor will I suggest any specific pedagogical techniques or strategies -
these are better left to the experts. But perhaps this new theoretical
perspective can support such initiatives by elevating the perceived
status of conventional expressions and facilitating their description.
Another facet of a usage-based model concerns the role of the
communicative, social, and cultural context. In particular, cognitive
grammar is contextually grounded because all linguistic units are
abstracted from usage events, comprising the full contextual under-
standings of socially engaged interlocutors with specific communica-
tive objectives in connected discourse. Any recurring facets of such
events can be incorporated in the conventional meanings of linguistic
units. It is only via progressive abstraction or decontextualization that
linguistic units sometimes approximate the situation of having con-
text-independent values. Yet I think they never fully achieve it. I be-
lieve that every expression, fixed or novel, is inevitably interpreted
with respect to some presupposed context, if only a vanilla context
derived from default assumptions. Hence the usage-based perspective
provides a theoretical underpinning for what we all know in practical
terms, namely the essential role of context and culture in language
understanding and language learning.
Let me illustrate with a banal example from my own experience.
The event took place on a train in France in 1963. I was trying to
hang up my jacket, but was having trouble due to the absence of the
loop of fabric normally provided for that purpose (this notion appears
to be a lexical gap in both English and French). One of the two
French women in the compartment summarized the situation with the
sentence in (1), which I remember precisely to this very day:
(1) Il n'y a pas de petit cordon.

Without my telling you the context, you could probably not deter-
mine the specific import of (1) or of the phrase petit cordon. Con-
versely, prior to the utterance - having been trained in French in a
classroom setting by traditional methods - I had no clue whatever
about how to describe the situation or the loop of fabric. Yet this one
utterance, immediately interpretable in context, was sufficient for me
to learn a natural way of expressing this situation in French. (Admit-
tedly, I have never since found myself in a situation where this
knowledge was useful, but if it ever happens again I am ready.)
Once more, I will not presume to make specific pedagogical rec-
ommendations. Let me simply state the obvious, namely that these
considerations argue for pedagogical approaches which emphasize
the interactive exposure to large quantities of natural speech in con-
text.
3. Construal
According to cognitive grammar (Langacker 1987a, 1990, 1991,

1999a), lexicon and grammar form a continuum, all elements of
which are symbolic in nature. It follows that all grammatical ele-
ments are meaningful. For everyone but certain linguistic theorists,
this is certainly a more attractive vision than the standard view of
grammar as pointless drudgery arbitrarily imposed, hence very hard
to learn. If this vision is correct - and by now I consider it essentially
proven - it offers numerous pedagogical opportunities.
The key to recognizing the meaningfulness of grammar lies in
adopting a conceptualist semantics that properly accommodates con-
strual, i.e. our ability to conceive and portray the same situation in
alternate ways (Langacker 1993). As an inherent aspect of its con-
ceptual-semantic value, every lexical and grammatical element in-
corporates a particular way of construing conceptual content - either
its own content or that evoked by other elements. It is by no means an
easy matter to correctly and convincingly describe this essential as-
pect of linguistic meaning, especially given the broad range of units
that need to be taught. In principle, however, an accurate appreciation
of the construal a unit imposes on a situation should allow one to

devise a more effective means of teaching it. Lexical and grammati-
cal elements are conceptual tools, and in learning to use a tool it is
helpful to know what it does.
Linguistic structures are conceptual tools for imposing particular
ways of viewing a situation. Hence their meanings are not directly
discernible or objectively present in the situations described - rather
they inhere in the cognitive process of apprehending those situations
and construing them for expressive purposes. Failure to appreciate
this fundamental point is the major reason why so many linguistic
theorists do not accept the meaningfulness of basic grammatical con-
structs. As a case in point, they continue to insist, quite erroneously,
that basic grammatical classes (noun, verb, adjective, etc.) are not
semantically definable. It is argued, for example, that since a word
like escape designates an event (and precisely the same event)
whether it functions as a verb or a noun, it must have the same
meaning as a member of either class, so verbs and nouns cannot be
distinguished semantically. The nominalization of a verb, deriving
nouns like performance or explosion, is therefore seen as having no
effect on its meaning. Thus statements like (2) still reflect standard
linguistic dogma, and (3) represents the only kind of argument ad-
vanced to refute the naive view that the parts of speech do have se-
mantic import.
(2) "[N]o constant semantic effect is associated with the function-

ing of a morpheme as a noun, as a verb, or as any other part of
speech." (Langacker 1968: 83)
(3) "[L]et's ask whether each part of speech really denotes a con-
sistent kind of meaning ... Now it is true that any word that
names an object will be a noun. But on the other hand, not
every noun names an object. 'Earthquake' names, if anything,
an action, as does 'concert'; 'redness' and 'size' name proper-
ties; 'place' and 'location' pretty obviously name locations. In
fact, for just about any kind of entity we can think of, there ex-
ist nouns that name that kind of entity. So the grammatical no-
tion of noun can't be given a definition in terms of what kind of

entity it names ... A particular kind of entity need not corre-
spond to a single part of speech either ... We conclude that
parts of speech ... are not definable in terms of meaning."
(Jackendoff 1994: 68-69)
What is missing from both traditional definitions and the standard

refutation is any realization that meaning is not objectively given but
reflects our apprehension of situations. Any degree of semantic sen-
sitivity (which I clearly lacked in 1968) should tell us that a change in
grammatical class involves a reconceptualization, an alternate con-
strual of the same content. Rather than being semantically vacuous, it
results in a subtly different meaning, in accordance with the abstract
semantic values of the classes. Going from a verb like perform to a
noun like performance involves a conceptual reification wherein a
process is reconstrued as a kind of thing. Such reification might be
characterized in terms of a process being construed metaphorically as
a physical object (in the case of count nouns) or a physical substance
(for mass nouns). But whatever the details, the standard argument
against a notional definition of grammatical classes hinges on the
tacit assumption that any characterization based on construal (rather
than the objective situation per se) is not even worth contemplating.
Numerous facets of construal are usefully thought of as being
analogous to visual phenomena. This leads Talmy (1996) to speak of
ception as the general process subsuming both perception and con-
ception. In the same spirit, I use the term viewing for both vision in
particular and conceptualization more generally, to the extent that
they seem parallel (Langacker 1995).
An example is our capacity to construe a situation either sche-
matically or at progressively greater levels of specificity. Conceptu-
ally, this dimension of construal is reflected in the meanings of a se-
ries of expressions like (4):
(4) thing > animal > dog > retriever > golden retriever
This increasing semantic specificity seems quite analogous to the

visual experience of seeing an object in progressively finer-grained
detail the closer we come to it. In any case, we are clearly able to
portray a situation at any desired level of specificity or schematicity,
and lexical items provide ranges of options for this purpose.
Some analogies are diagrammed in Figure 1. In vision, we can
distinguish between the maximal field of view (everything visible to
any degree at a given moment), within that the general locus of at-
tention (or "onstage region"), and within that the specific focus of
attention. The corresponding conceptual entities, adopted as semantic
constructs in cognitive grammar, are the maximal scope (i.e. the full
array of conceptual content an expression evokes), the immediate
scope (the portion directly relevant for a particular purpose), and the
profile (the entity the expression is construed as designating, or refer-
ring to).
Maximal field of view- Maximal scope (MS)
Focus of attention Profile
Locus of attention Immediate scope (IS)

("onstage region")
Figure 1. Viewing analogies
An expression's grammatical class is determined by the nature of its

profile (not its overall content). Basic grammatical classes have con-
ceptual characterizations. They invoke particular conceptual arche-
types for their prototypical values (e.g. 'physical object' in the case of
nouns), and particular cognitive abilities for their schematic descrip-
tions (e.g. conceptual reification). Class membership is therefore
neither arbitrary nor objectively determined. Rather, it inheres in a
particular way of viewing a situation and construing it for expressive
purposes.
At the schematic level, a noun profiles a thing, abstractly defined
as any product of conceptual reification. A verb profiles a process,
defined abstractly as a relationship scanned sequentially in its evolu-
tion through time. Various other classes (e.g. adjective, adverb, pre-
position, participle) profile relationships that are non-processual

(thus viewed holistically, rather than scanned sequentially, even
when they develop through time).
Consider a term like aunt. It is clearly relational, so why does it
function as a noun (and not, say, a preposition)? The reason, as
shown in Figure 2, is that it does not profile this relationship, but
rather the person who bears it to the reference individual (ego). Cru-
cial though it is to the meaning of aunt, the relationship is not its ref-
erent but is merely evoked to characterize its referent. (Observe that
there is no basis in this particular example for distinguishing between
maximal and immediate scope. This is analogous to a word consist-
ing of just a single morpheme, so that word and morpheme coincide.)
ego
MS/IS
aunt
Figure 2. A relational noun
Once we decide to use it, a form like aunt gives us no options: it is

always a noun by virtue of imposing the scope and profiling shown in
Figure 2. Despite polysemy and a certain flexibility of interpretation,
lexical items are fixed expressions, so basically they embody pack-
aged conceptualizations and their conventional symbolization. Ironi-
cally, it is grammar - so often thought of as an oppressor imposing
arbitrary limits on our expressive freedom - that gives us some relief
from these lexical strictures. Many grammatical elements have the
specific function of adjusting the construals that would otherwise be
inherited from lexical items.
For instance, the verb perform profiles a perfective process, i.e. a
process construed as being bounded. In Figure 3 a, the circle repre-
sents the focal participant; the vertical line and rectangle stand for the
activity this participant is engaged in; and the horizontal line (with
endpoints) indicates its progression through time (sequentially
viewed). Starting from this basic conception, the derivational mor-

phemes -er and -ance do not add any new content. Rather, their se-
mantic value resides in the construal they impose on that content.
Thus -er shifts the profile from the temporally evolving relationship
to its focal participant, as seen in Figure 3b. Performer is therefore a
noun, since it profiles a thing rather than a process, although the
original process provides its essential conceptual content. Perform-
ance is also a noun, but the thing it profiles is abstract, created by
conceptual reification. The ellipse in Figure 3c stands for this abstract
entity, a thing consisting of one instance of the process perform.
(a) perform (b) performer (c) performance
Figure 3. Profiling and grammatical class
Another example is the English progressive construction with be...

-ing. I claim that the progressive is only applicable to processes con-
strued as perfective, hence temporally bounded. It is an imperfectiv-
izing construction, so it does not apply to processes that are already
imperfective (not bounded), where its effect would be vacuous (Lang-
acker 1987b). Occurrence in the progressive is of course a standard
diagnostic for perfective (or "active") verbs, and non-occurrence for
imperfectives (or "statives"). The other standard diagnostic, with the
opposite distribution, is occurrence in the "true" present tense (with
an actual, "right now" interpretation). Here are some typical exam-
ples:
(5) a. * He builds a house. a'. He is building a house,

[perfective]
b. He knows the truth. b'. *He is knowing the truth,
[imperfective]
Intuitively, we can describe the progressive in terms of "zooming in"

and taking an "internal view" of a bounded event. Technically, I de-
scribe it as imposing an immediate temporal scope that excludes the
endpoints of the perfective process it applies to. This is shown in
Figure 4b, where - since only aspect is relevant here - 1 have reduced
the depiction of a process to a line representing its temporal exten-
sion. Thus, while the maximal scope is a span of time containing the
full, bounded process, the immediate scope subtends only an arbitrary
portion of its internal development. Only that portion is profiled,
since - as a matter of definition - the profile is the focal point within
the immediate scope (see Figure 1). The overall progressive expres-
sion is imperfective, because grammatical class is determined by the
profile and the profiled process is not bounded. (Also, as with any
imperfective, the profiled process is construed as being effectively
homogeneous.)
Perfective (b) Progressive Imperfective

MS
IS
Figure 4. Basic aspectual classes
Because it only applies to perfectives, the progressive construction -

though itself imperfective - signals that the original process is con-
strued as being bounded. The subtle contrast between a basic imper-
fective and one derived by using the progressive is seen in Figures
(4b-c). Both profile a process which is unbounded within the imme-
diate temporal scope (and construed as being effectively homogene-
ous). The difference is that a progressive expression creates this im-
perfective process by selectively attending to the interior of an overall
occurrence recognized as being bounded. Thus, in (6a), both the sim-
ple present and the progressive indicate a current residence in Chi-
cago, but the latter portrays this as part of an overall residential epi-
sode of limited duration. Likewise, with the past tense (6b) merely
describes the situation as stable and unbounded within the time span
in question {in that era), whereas the progressive construes it in rela-

tion to a longer period within which the rule of dinosaurs is seen as
being bounded. And in (6c), the progressive indicates that the statue's
location in the plaza is only temporary.
(6) a. I {live/am living} in Chicago.

b. In that era, dinosaurs {ruled/were ruling} the earth.
c. A statue of Martin Pütz {stands/is standing} in the plaza.
It should go without saying that whether a process counts as being

bounded is subject to construal. I say this anyway because I have of-
ten had linguists - even cognitive linguists - object to my characteri-
zation of perfectives by citing examples where bounding is not ap-
parent just by examining the objective circumstances. One such ex-
ample is (7). Since running around a pole is something that can go on
indefinitely, with no intrinsic endpoint, how can it be said that the
process is bounded?
(7) He is running around the pole.
The mistake here lies in assuming that the bounding of a process has
to be objectively given, with an inherent endpoint observable in the
situation itself. Ultimately, what counts for linguistic purposes is
whether a process is conceptualized as some kind of bounded epi-
sode, irrespective of whether a natural endpoint is discernible. There
is in general a strong tendency to conceptualize force-dynamic occur-
rences (those requiring the expenditure of energy - cf. Talmy 1988)
as being bounded in duration, even when the process is internally
homogeneous and nothing appears to be going on. Thus (8a) de-
scribes a stable situation, but since the stability results from a balance
of opposing forces, the basic process (i.e. the dam contains the surg-
ing floodwaters) is construed as an episode of bounded duration,
hence the progressive is possible. By contrast, in (8b) the basic proc-
ess (i.e. the barrel contains water) is non-force-dynamic. It merely
describes a spatial configuration, which as such can maintain itself
indefinitely. It thus receives an imperfective construal, and conse-

quently does not allow the progressive.
(8) a. The dam is containing the surging floodwaters.

b. The barrel contains water.
English has many perfective predicates (e.g. sleep, dream, run, walk,
sit, stand, lie, perspire, talk, chat, meditate, wear a tie) in which an
internally homogeneous activity is nonetheless construed as occur-
ring in bounded episodes. It is only their grammatical behavior - re-
sisting the simple present, occurrence in the progressive - which
alerts us to their perfectivity. This does not however imply that the
distinction is "purely grammatical", with no conceptual basis. The
grammatical classification hinges on a conceptual factor which is no
less real for being subject to construal.
Once the specific, subtly contrasting construals imposed by lexical
and grammatical elements are elucidated, innumerable structural de-
tails that otherwise seem quite arbitrary turn out instead to have an
intuitively graspable conceptual basis. The potential thus exists for
devising effective ways of teaching them. How much of this should
be explicitly taught to students? Should we burden the language
learner with technical constructs like profile and immediate scope? I
suspect not. It is not even evident that the classroom teacher should
be responsible for such technical details, which might better be
brought in at the level of overall planning and design of teaching
materials. I will leave this matter for those competent to assess it.
At the same time, these notions might find a natural place in a
language arts curriculum. Generative grammarians have often pro-
posed that linguistics ought to be more visible in school curricula at
all levels. In particular, they put it forth as a way of teaching science:
formulating hypotheses, testing them against the empirical evidence,
developing skills of argumentation, etc. While this may have some
merit, the very different vision of language embodied in cognitive
linguistics suggests another option: the concepts and descriptions of
cognitive linguistics might instead be used for inculcating an appre-
ciation of language as a means of evoking and symbolizing alterna-
tive conceptualizations of experience. If language arts training of this

sort were routinely available, it might then be possible to draw upon
such notions, even explicitly, in the language classroom.
4. Conceptual substrate
Traditional semantics focuses on the meanings and combinatory

properties of overt elements. Cognitive linguistics has clearly demon-
strated the limitations of this approach, showing that linguistic
meanings rest on a vast and multifaceted conceptual substrate. The
conceptions explicitly encoded by formal elements are merely the
"visible" portions of far more extensive conceptual structures that
support them and provide their coherence. Often left implicit - yet
critical for determining and interpreting what does appear overtly -
are factors such as the presupposed viewing arrangement, the nature
and force of the speaker-hearer interaction, and how expressions re-
late to the current discourse state. Linguistic understanding further
relies on elaborate processes of meaning construction involving
metaphor, metonymy, mental spaces, blending, idealized cognitive
models, and the evocation of myriad "fictive" entities.
Long ago, my own language training emphasized the translation or
transformation of individual sentences totally isolated from any con-
text. I thus find it interesting to contemplate how many levels of con-
ceptual organization support the interpretation of even a simple ex-
pression. The expression itself - overt linguistic elements and the
notions they directly encode - is of course merely the tip of the ice-
berg. The expression per se is part of a usage event, i.e. an actual
instance of language use, comprising the interlocutors' full contex-
tual understanding of the expression, including their apprehension of
its interactive force. The usage event is usually part of a longer dis-
course, and is one facet of the interlocutors' overall social interaction.
The interaction takes place in a particular situational context, which
in turn is embedded in a culture, which develops as a way of coping
with the world.
We rely on information from any or all of these strata even for the
interpretation of seemingly straightforward expressions. Let us once
again consider example (1 ), Il n'y a pas de petit cordon. This made
perfect sense given the situational context of wanting to hang up a
jacket, but without this contextual support it seems rather pointless
and hard to interpret. Apprehension of the situational context was
itself dependent on the knowledge of certain cultural practices, nota-
bly that of hanging up jackets and the usual provision of a loop of
fabric for this purpose. Moreover, certain basic properties of the
world we inhabit - such as temperature, gravity, and force dynamics
- are reflected in these cultural practices and implicitly invoked any
time we think of them.
A pivotal aspect of the conceptual substrate, quite clearly, is the
ground, comprising the speaker, the hearer, their interaction, and the
immediate circumstances. It is quite common - arguably even ca-
nonical - for the ground to remain offstage and not be mentioned. In
the unmarked situation, the ground functions as the tacit location
from which a scene is viewed and an expression's meaning is appre-
hended, as opposed to being onstage as the explicit focus of attention.
For instance, a tense marker locates a profiled process with respect to
the time of speaking (one facet of the ground), invoking it as a tem-
poral reference point, but does not directly mention it. Likewise, a
determiner - via its specification of (in)definiteness - invokes the
speaker and hearer as the individuals seeking to identify the nominal
referent, but leaves them offstage and unprofiled. The ground, then,
is the locus of conception. We do of course have ways of putting fac-
ets of the ground onstage and referring to them specifically, e.g. with
forms like I, you, here, and now. Even so, the ground's occasional
status as focused target of conception coexists with (and is subsidiary
to) its more fundamental role as the tacit locus of viewing.
Inherent in every usage event is a presupposed viewing arrange-
ment, pertaining to the relationship between the conceptualizes and
the situation being viewed. The default arrangement finds the speaker
and hearer together in a fixed location, from which they report on
actual occurrences in the world around them. There are however nu-
merous kinds of departures from this canonical circumstance. The
departures help make it evident that the default arrangement, so eas-

ily taken for granted, is nonetheless an essential part of the concep-
tual substrate supporting the interpretation of expressions. Whether
canonical or special, the viewing arrangement has a shaping influ-
ence on the conception entertained and consequently on the linguistic
structure used to code it.
For expressive purposes, we sometimes invoke a viewing ar-
rangement other than the actual one. On the most likely interpretation
of (9), for instance, the speaker is actually in motion and the tele-
phone poles are static (cf. Talmy 1996). Yet the situation is presented
as if it instantiated the default-case viewing arrangement, with a
static viewer observing the motion of other entities. Granted this fic-
tive viewing arrangement, the description is accurate: it represents
what the speaker actually sees, interpreted with respect to an ar-
rangement that, while canonical, is non-actual. A correct apprehen-
sion of the speaker's intent requires that the expression's overt con-
tent be properly related to both the fictive viewing arrangement
(which determines its form) and the actual one (where the viewer's
motion generates the visual experience coded by that form).
(9) The telephone poles are rushing past at ninety miles per hour.
Of course, language is not just used for describing what happens.

Most of us are unlike philosophers (or the idealized speakers some-
times imagined by philosophers) who spend their lives in detached
contemplation, producing only objectively verifiable assertions pur-
porting to truthfully describe the world. Beyond assertion and de-
scription, we use language for multifarious actions and purposes: for
asking questions, giving orders, making promises, performing official
acts; for attracting and directing attention; for eliciting approval,
agreement, acknowledgment, permission, and cooperation; for the
primal expression of pleasure, pain, fear, and disgust; and so on. The
apprehension of these actions and purposes provides a conceptual
substrate for the attachment of elements appearing overtly in expres-
sions. Whether or not they refer to the substrate specifically, these
elements - via their selection and their conformation - give witness

to its presence and its character.
For instance, modifiers sometimes pertain to the substrate rather
than the overt content. The since-clause in (10a) does not give a rea-
son for the bird being the kind of bird it is, but instead justifies the
speaker's action of requesting this information (Sweetser 1990).
Likewise, the honesty referred to in (10b) is that of the speaker in
tendering the advice, not that of the addressee in following it.
(10) a. Since you're an ornithologist, what kind of bird is that?

b. In all honesty, you should give up painting and get a job.
Part of the substrate resides in apprehending an expression's illocu-

tionary force, i.e. the type of speech act the interlocutors are engaged
in (Austin 1962). It is possible for the execution of that act to be put
onstage as the focused target of description. The result is a performa-
tive sentence, as in (11a). More commonly, though, it is left implicit
or signaled in a more peripheral manner. In the case of imperatives,
where the envisaged action is to be carried out by the addressee, the
agent can either be indicated in the usual way by the subject pronoun
you, as in (lib), or else left implicit, as in (11c). Since the hearer's
agentive role is an inherent specification of the speech act itself, it
can be omitted from the expression's overt content unless explicit
reference is felt necessary for emphasis. In particular cases, the
speaker's role in the envisaged action can also be left implicit, as in
(1 Id). To the extent that the speech event instantiates a standard type
of interaction, the content overtly expressed may be limited to those
facets of the situation not subsumed by the substrate.
(11) a. I order you to stay away from me.

b. You stay away from me!
c. Stay away from me!
d. Stay away!
Apprehension of the speaker-hearer interaction (including illocution-

ary force) is always part of the conceptual substrate, even for canoni-
cal assertions. It is inherent in the construal of every expression as

one facet of the viewing arrangement it presupposes. Expressions that
are non-descriptive or have non-assertive force constitute one large
class of departures from the standard viewing arrangement. There are
however many other kinds of departure from it, even with expres-
sions whose function is basically descriptive. Even when using lan-
guage to report on what happens in the world around us, there are
many ways in which we commonly deviate from the canonical situa-
tion of directly describing actual individuals and actual occurrences
involving them. Some - like metaphor, metonymy, and implicature -
are extremely prevalent if not utterly pervasive. In the case of meta-
phor, as in (12a), we are not simply viewing and describing an entity
in its own terms, but instead create a blend by selectively projecting
properties from the source domain onto the target domain; although
the target domain is the actual focus of our interest, it is the blend
that is directly encoded linguistically. In metonymy, the entity we
mention directly functions as a conceptual reference point, providing
mental access to an associated entity that we are actually referring to,
as in (12b). And of course, implicature allows us to convey the cru-
cial information without having to express it directly. Thus (12c) can
serve as an indirect means of telling the guests that it is time to leave
the dinner table.
(12) a. The brain tumor robbed him of the chance to finish his
novel. [metaphor]
b. The pen is mightier than the sword. [metonymy]
c. The living room is much more comfortable, [implicature]
I have already cited (9) as a case of invoking a fictive entity, namely

the viewing arrangement itself. Essentially any facet of the overall
conception an expression evokes can be fictive (or virtual) rather than
actual (Langacker 1999b). For instance, a rhetorical question like
(13a) can be analyzed in terms of a fictive speech act - the speaker
only pretends to ask a question, it being understood that the actual
interactive force is very different.
(13) a. Would my client - a faithful public servant, a devoted fam-

ily man, and a deacon in his church - even consider taking a
bribe? Of course not! [fictive speech act]
b. The trees got shorter as we approached the summit.
[fictive change]
In (13b), both the subject and the profiled event are fictive in nature.
The trees does not refer to any actual set of objects, but is rather a
role description, designating a feature of the landscape observable at
any altitude. Nor does any tree or set of trees actually change in
length. What the sentence describes is a virtual change generated by
viewing the trees instantiating the role at different altitudes as if they
were a single, changing entity.
Finally, let me mention the frequent but seldom noted phenome-
non of resorting to type specifications as a way of describing a set of
actual occurrences that are alike in some respect. Imagine a series of
actual events in each of which a single stranger - different each time
- reaches over a fence and picks a single apple - also different each
time. If there are three such events, the entire sequence can be sum-
marized by sentence (14a). It is not essential for our purposes that the
sentence is subject to alternate interpretations. What is essential is the
possibility of using (14b) for exactly the same event sequence. There
are three different strangers, and three different apples, yet these par-
ticipants are referred to in the singular: a stranger, an apple.
(14) a. Three times, strangers reached over the fence and picked
apples.
b. Three times, a stranger reached over the fence and picked an
apple.
I suggest that (14b) describes the events at the level of their common
type characterization. The three actual events each instantiate the
event type a stranger reach over the fence and pick an apple. At the
type level, representing what the three events have in common, there
is only one stranger and one apple. These are instances of the
stranger and apple categories, but they are not actual instances or
specific individuals. Rather, they are virtual (or arbitrary) instances

conjured up just to characterize the type of event in question. In
short, even though (14a) and (14b) are used to describe precisely the
same event sequence, they view it in different ways for linguistic
purposes. The latter involves the conceptual operation of extracting a
common event type, whose characterization invokes fictive entities
distinct from any actual individual. This special viewing arrangement
focuses on the type as the level that is directly coded linguistically,
even though the description pertains to actual events. The adverb
three times can be seen as an instruction for mapping between these
two levels.
The moral of all this, not at all surprising from the standpoint of
cognitive linguistics, is that the relation between linguistic expres-
sions and the world is non-transparent, being mediated by elaborate
mental constructions even when we seem to be merely reporting on
what happens (cf. Fauconnier 1997). For an accurate characterization
of linguistic elements, we need a detailed understanding of the many
subtleties of viewing, as well as the conceptual substrate that sup-
ports and shapes expressions. Though I will not suggest any specific
pedagogical implications, I have to believe that some appreciation of
these factors would be beneficial to both the language learner and the
language teacher. In what follows, I will try to show their relevance
to a proper analysis of a phenomenon noted for the pedagogical
problems it poses, namely the English present tense.
5. The English present tense
The one thing that is generally agreed upon concerning the English
present tense is that it is really not a present tense, i.e. its value can-
not be that of indicating that the process in question occurs at the
time of speaking. The arguments seem straightforward. On the one
hand, events that do occur at the time of speaking generally cannot be
expressed in the present tense. This is the case with perfectives, as
we saw earlier. As descriptions of actual, bounded events occurring
at the time of speaking, sentences like (15a) are consistently infelici-

tous:
(15) a. *Bill {sleeps/paints a fence/changes a tire/learns a poem}

right now.
b. Bill {is sleeping/is painting a fence/is changing a tire/is
learning a poem} right now.
To say these things, we must instead use the progressive, as in (15b).

On the other hand, many uses of the present tense do not refer to the
time of speaking. Standard uses of the "present" pertain to the future,
to the past, to "timeless" situations, or even to "eternal truths":
(16) a. Your driver's license expires on your next birthday.

b. I'm driving home last night and I hear a siren. I pull over
and stop. This cop comes up and starts writing me a ticket.
c. Hamlet moves to center stage. He pulls out his dagger. He
examines it.
d. Pi is irrational.
Despite these commonplace observations, I have long argued (e.g.

1987b) that the English present tense does in fact locate the desig-
nated process at the time of speaking (coincident with the ground).
More precisely, the present tense indicates that a full instantiation of
the profiled process occurs and precisely coincides with the time of
speaking. The proposed account will serve as a case study illustrating
many of the points made earlier. I do not know how the present tense
should be taught, but an understanding of how it really works must
surely be relevant to the problem.
I analyze a tense marker as imposing an immediate temporal scope
for the focused viewing of the process it grounds. For English, there
are just two, as shown in Figure 5, where a box with squiggly lines
represents the speech event. The past tense morpheme imposes an
immediate scope located prior to the speech event, while the present
tense morpheme (at least in English) imposes one that coincides with
it.
(a) Past Perfective Present Perfective

MS
IS
t ΚΛΛΛΛΛ
(c) Past Imperfective (d) Present Imperfective

MS MS
IS IS
t wvWw ^ w w v
Figure 5. Tense and aspect
Observe that in the past tense there is no inherent limit on the length
of the immediate scope, so a perfective process of any length can
always be made to fit inside it. By contrast, in the present tense the
immediate temporal scope must be the same in duration as the speech
event. If an imperfective process endures for a span of time that in-
cludes the immediate scope (in either the past or the present), only
that portion of it subtended by the immediate scope is profiled. Re-
call that an expression's profile is necessarily confined to its immedi-
ate scope (the general locus of attention). Moreover, since an imper-
fective process is internally homogeneous and not characterized in
terms of bounding, any subpart singled out for profiling will itself
constitute a valid instance of the process type in question. (Imperfec-
tives are quite analogous to mass nouns in this respect.)
Let me start by pointing out how much this analysis accounts for,
straightforwardly and even rather elegantly. First, it accounts for im-
perfectives being able to occur in the present tense, as sketched in
Figure 5d. Since any part of an imperfective process itself counts as a
full instantiation of the process type, this will also be true for the
portion that coincides with the time of speaking. Observe that the
analysis does not imply that the stable situation described is valid
only for the brief duration of the immediate scope. For instance, an
utterance of (5b), He knows the truth, does not entail that his knowl-
edge of the truth is limited to the time of the utterance. What is being
claimed, instead, is that the speech event defines a "window" for fo-
cused viewing. In using a present tense imperfective, the speaker is
taking a temporally coincident sample of the overall situation and
observing that - for the portion sampled - the situation is stable and
unbounded. It is possible, without contradiction, for the same overall
situation to be sampled at different times, as in (17):
(17) He knew the truth then, and he still knows it now.
The same overall situation endures without interruption for a span of

time that includes both immediate scopes, those imposed by the past
tense in the first conjunct and by the present tense in the second
conjunct. There is one overall situation, but from it each clause se-
lects a different sample for focused viewing, resulting in two distinct
profiled processes, each characterized by local stability and the ab-
sence of bounding. It is in the nature of imperfectives that a local
sample is sufficient to reveal a stable situation of indefinite duration.
Because progressives are derived imperfectives, the analysis pre-
dicts their occurrence in both the past and present tense, as in (18).
This is diagrammed in Figure 6. Observe that two immediate scopes
are indicated. The progressive itself imposes an immediate temporal
scope, labeled ISi, whose boundaries are internal to those of the per-
fective process it applies to (see Figure 4b). The past or present tense
marker then applies to the imperfective process profiled by the pro-
gressive. As shown in Figure 5, it imposes its own immediate tempo-
ral scope, given as IS2, which either precedes or coincides with the
time of speaking. The profile of the complex expression is confined
to the sample of the imperfectivized process that falls within the
scope imposed by the tense. Recall that the ground is the locus of
conception, and a tense marker specifies the locus of attention (or
focused viewing) with respect to it.
(18) She {was/is} {working/writing a letter/wearing a sweater}.

(a) Past Progressive (b) Present Progressive

MS
IS,
IS2
vvvvv — ^ -
l
Figure 6. Tense and the progressive
The analysis also accounts for the usual infelicity of present tense
perfectives, as in (15a). The difficulty is not however a matter of
conceptual incoherence. Indeed, the conceptual configuration de-
picted in Figure 5b is perfectly coherent and non-anomalous. There is
nothing inherently contradictory about a bounded event temporally
coinciding with the speech event. And indeed, true present tense per-
fectives are sometimes permissible, as we will see. Instead, the
problem with present tense perfectives is that certain factors make the
configuration in Figure 5b hard to achieve in practice.
A perfective process is bounded, so a full instantiation of such a
process includes its boundaries. Thus, if a perfective process is to
coincide with the time of speaking, its beginning point has to coin-
cide with the initiation of the speech event, and its endpoint with its
termination. This poses both a durational problem and an epistemic
one. The durational problem is that there is no inherent connection
between the length of the event described and the length of the
speech event describing it. It takes longer to paint a fence, for exam-
ple, than it does to utter the clause He paints a fence. The epistemic
problem resides in having to observe an event in order to identify it
as a prerequisite to describing it. By the time an event is observed
and identified, it is already too late to initiate a speech event that pre-
cisely coincides with it. These problems do not arise with imperfec-
tives, given their mass-like character and the property that any por-
tion of the overall process counts as a full instantiation of the process
type. Hence an imperfective has no specific duration, and a portion
which follows a period of observation and identification can still

count as a valid instance.
If a perfective event has a long enough duration, therefore, and
extends through the time of speaking, we can describe its occurrence
"right now" by means of a present tense progressive. This was dia-
grammed in Figure 6b. The progressive derives an imperfective proc-
ess delimited by ISi, and the present tense selects for focused view-
ing the portion of it that coincides with the time of speaking (IS2). Of
course, this solution is not automatically available if the perfective
event is punctual, effectively consisting of just an onset and an offset,
with no interior phase of any significant duration. A sentence like
(19a) is consequently infelicitous if intended as the description of
something actually occurring right now (not as merely indicating that
the popping is imminent). The reason is that the event is too short for
imposition of an immediate scope that excludes its endpoints - there
is no interior phase of sufficient duration for focused viewing.
(19) a. * A balloon is popping (at this very moment),

b. *She is blinking, [single blink]
Likewise, (19b) is infelicitous if construed as referring to a single

blink in progress. Of course, it can always be construed as repetitive,
and since repetitives are perfective, they allow the progressive and
require it in the present tense. We can also rescue (19b) by imagining
a special viewing arrangement. It would be felicitous, for instance, if
we were watching a slow motion film in which a single blink were
viewed as occurring over a span of several seconds. Such examples
show that well-formedness judgments depend on default assumptions
about the world as well as a presupposed viewing arrangement.
Also accounted for by the proposed analysis is a striking system-
atic exception to the usual non-occurrence of present tense perfec-
tives, namely performatives, as in (20). Performatives clearly profile
bounded events and not only tolerate but actually require the present
tense. The reason is that a performative represents a special viewing
arrangement in which the process put onstage and profiled is the
speech event itself. Since the profiled process and the speech event
are the same, they have to be temporally coincident, as shown in Fig-

ure 7. Because of this property, the durational problem does not arise
with performatives. Nor does the epistemic problem. The latter arises
when it is necessary to observe an event and identify it prior to initi-
ating its description. But with performatives, the speaker is responsi-
ble for carrying out the profiled event and acts with prior intent. As
the intentional agent of the process, the speaker has no need to ob-
serve it in order to know its identity.
(20) a. I promise to cooperate.

b. I beg you to give me another chance.
c. I hereby sentence you to 30 years in prison.
Performative
MS
is]
, ΙΛΛΛΛΙ ^
Figure 7. Performatives
Performatives indicate that the durational and epistemic problems are

not problems with present tense perfectives per se. Rather, they stem
from particular circumstances of viewing, namely the default viewing
arrangement. In the default arrangement, the event to be described is
independent of the speech event and beyond the control of the
speaker, who merely observes the occurrence and then reports it. In
this case the speaker can hardly begin its description coincident with
its initiation, nor is its duration likely to match the time needed to
utter a finite clause. But performatives, being intentional actions
which implement the very events described, avoid these problems by
their intrinsic nature.
Are there other kinds of viewing arrangements which, by their
nature, avoid the durational and epistemic problems? We simply
need to imagine a situation where the speaker controls both the oc-
currence and the duration of the event described and can therefore
make the description coincide with the occurrence. In one situation I
can imagine, a child is playing with toy cars and a play village and
accompanies each action she takes with a descriptive sentence. The
successive utterances in (21) then coincide with the successive acts of
pushing a toy car from one place to another. This use of the present
tense for perfective events seems perfectly natural and unproblem-
atic.
(21) Now I drive to work. Now I go to the store. Now I drive home.
More generally, the present tense is naturally used for the narration
of demonstrations. Imagine an origami class, where each clause in
(22a) accompanies the action it describes:
(22) a. I pick up a sheet of paper and I fold it in two. I fold it again.

Now I take the scissors, and I make an incision from one
corner to the center,
b. I put a tablespoon of butter in the pan. It melts quickly. Now
I put the fillet in. I cook it at a low temperature for five min-
utes.
Or a cooking program on television, as in (22b). Note that the subject

does not have to be the speaker (e.g. It melts again). It is only neces-
sary that the speaker have sufficient control over events to avoid the
durational and epistemic problems.
The last sentence in (22b) is included to raise certain issues con-
cerning duration and coincidence. Obviously, the chef does not take
five minutes to utter this sentence. Nor is the butter likely to melt in
exactly the span of time required to say It melts quickly. There are
two approaches we might consider for dealing with this type of
problem. I suspect that both of them are valid and play some role in
overlapping ranges of data. One approach is to recognize a certain
amount of tolerance in applying the notion "precise coincidence" to
actual circumstances. Except perhaps with performatives, exact tem-
poral coincidence - in the strictest sense - is not a realistic expecta-
tion (for perfectives). It has to be understood as coincidence apparent
on a certain time scale, or some approximation close enough for the
purpose at hand. This of course is characteristic of language use in

general. A person who says that All politicians are dishonest is not
considered to be lying if he believes that somewhere in the world
there might be one who is honest.
Tolerance in what counts as temporal coincidence clearly figures
in the play-by-play mode of speech used by sportscasters. In a way,
their role exemplifies the default viewing arrangement, since they
occupy a fixed position and do in fact report on actual occurrences.
How, then, do they overcome the durational and epistemic problems
in their frequent use of present tense perfectives, as in (23)?
(23) Stockton dribbles along the baseline. He passes out to Malone.

Carl makes a nice grab. He puts up a three-point shot. He
scores!
Note that the events described in this way have approximately the
right duration for temporally coincident description. In the context of
a sporting event, they are also quite stereotyped, so the announcer has
a good idea of what is likely to transpire at any instant. One is there-
fore able to shadow the events fairly closely, sometimes even to an-
ticipate and describe them simultaneously with their occurrence. The
goal at least is to come as close as possible to coincident description,
and the conventions of play-by-play reporting rest on either the fic-
tion that this is feasible or else the tolerance of a certain time-lag.
Undoubtedly we have to recognize flexibility and degrees of ap-
proximation in what counts as "precise coincidence". These are, after
all, matters of construal rather than objective scientific measurement.
Yet this hardly seems adequate for examples like the final sentence in
(22b). Moreover, I can easily imagine an alternative mode of narra-
tion for a demonstration, where - instead of coinciding with it - each
statement precedes the action it describes, thereby telling the listener
what to watch for. At least for cases like these, we appear to need
some other approach. The approach I suggest is to posit a distinct
viewing arrangement, one that does not specifically involve the si-
multaneous narration of actions. Although, in practice, the events in
question are correlated with actions, they are conceptualized more
abstractly as entries in a list, collectively constituting a kind of script

or scenario that is being followed. In this respect they would be
roughly analogous to stage directions, as in (16c). My proposal for
such cases is that the present tense verbs are not in fact being used for
the direct description of actual events - instead they indicate the
reading off of entries on some kind of list or scenario. In other words,
what is being coded linguistically is not the actual occurrence of
events, but rather their virtual occurrence as part of a non-canonical
viewing arrangement. The viewing arrangement is such that the vir-
tual occurrence does coincide with the time of speaking.
If this proposal should seem far-fetched, it is only because we are
so accustomed to thinking in terms of the default viewing arrange-
ment and regarding linguistic expressions as direct descriptions of the
world. I have tried to show, on independent grounds, that there are
many other kinds of viewing arrangements, and that we often resort
to the direct linguistic coding of virtual entities even when our real
concern is with actual ones. Recall (14b), Three times, a stranger
reached over the fence and picked an apple. To describe an actual
situation involving three strangers, three apples, and three events of
picking, we resort to an expression that directly codes a virtual event
involving fictive instances of the stranger and apple categories. We
"conjure up" these fictive entities as a way of capturing the abstract
commonality of the actual events. Once we recognize the true lin-
guistic prevalence of virtual entities and non-canonical viewing ar-
rangements, the proposal to posit a virtual event occurrence is not at
all far-fetched, but rather the sort of thing we ought to anticipate.
Let us begin with the scheduled future use of the present tense, as
in (24a). My proposal is that such expressions relate only indirectly
to the actual event in question. What a sentence like this directly de-
scribes is not the actual event per se, but rather a representation of
that event on some kind of virtual schedule, some kind of plan or
projection concerning the anticipated occurrence and timing of events
in the future. Several considerations support the notion that some-
thing like a schedule is involved. For one thing, the scheduled future
strongly favors a time expression, as seen by the infelicity of (24b).
Moreover, it does not work well for events that cannot be scheduled
or anticipated. Thus (24c) is awkward unless it is uttered by God, or

perhaps by a scientist with supreme confidence in a method of quake
prediction. Finally, there is sometimes an actual schedule that is be-
ing read, as in (24d).
(24) a. Our new furniture comes tomorrow.

b. ??Our new furniture comes.
c. ??An earthquake strikes next week.
d. There it is on the monitor - our bus leaves at noon.
A virtual schedule pertains to the future, but its own status and loca-
tion are another matter. When a plan is in effect, the schedule itself is
stable and mentally accessible through a span of time that includes
the present. The schedule consists of virtual events, which are repre-
sentations of anticipated actual events. Moreover, the time interval
through which each virtual event is conceived as unfolding is identi-
fied with a particular time in the future, as shown by the dotted corre-
spondence lines in Figure 8. However, the events constituting the
schedule are only virtual.
/ Eventj Event2 Event3 /

I I 1 1 / Virtual
/ ~ 1 1 ^ /Schedule
Actual /
hvents I t
/
Figure 8. The scheduled future
Metaphorically, we can think of a virtual schedule as a "document"

available to be "read" at any time. In producing a sentence like (24a),
the speaker is essentially reading off one of its entries. Reading an
entry amounts to the virtual occurrence of the event it comprises, and
since that event is profiled by the sentence produced, a (virtual) oc-

currence of the profiled process precisely coincides with the time of
speaking. Use of the present tense thus conforms to the proposed
characterization, taking into account the special viewing arrangement
in which the speaker is "reading" aloud from a virtual schedule. In
that context, where all the events are virtual, they occur in the sense
of being read, and the reading is necessarily coincident with the
speech event.
I take this as being typical of the so-called "non-present" uses of
the present tense in English. Though details vary, a number of them
are plausibly described metaphorically as the reading of a virtual
document; the differences reside in the kind of document envisaged.
In the case of (22b), the document would be an imagined script of
how the cooking demonstration is supposed to proceed, step by step.
In the case of stage directions, as in (16c) [Hamlet moves to center
stage. He pulls out his dagger. He examines it.], the script may well
be physically embodied. But even conceived as a virtual document, it
comprises a series of inscribed events available to be read at any
time. Reading them, and thereby apprehending the successive event
descriptions, induces their virtual occurrence in the form of mentally
constructing the prescribed event sequence. What about the historical
present, as in (16b) [I'm driving home last night and I hear a siren. I
pull over and stop. This cop comes up and starts writing me a
ticket.]? Here the virtual document consists of a series of recalled
events that the speaker can mentally "replay" at leisure, at the pace
required for linguistic encoding. Another kind of recall figures in a
photo caption, as in (25). Here the statement is physically instantiated
and read quite literally. The photo captures one moment of the event
described by the caption, and helps to evoke its virtual occurrence in
the guise of apprehending the statement's import.
(25) Nixon says farewell from the steps of his helicopter,

[photo caption]
More generally, the key to understanding "non-present" uses of the

present tense lies in recognizing the special viewing arrangements
they presuppose. They all diverge from the default arrangement by

invoking some kind of mental construction - such as a schedule,
script, or mental replay - consisting of event representations. Even
when these correspond in some fashion to actual events, the repre-
sented events are the ones directly coded linguistically and profiled
by the present tense verb. What counts as the occurrence of such a
process is therefore not an actual occurrence, but rather a virtual one
consisting in its apprehension in the manner indicated by the special
viewing arrangement (e.g. reading the virtual schedule, or running
the mental replay). Indeed, the profiling of actual events in accor-
dance with the default viewing arrangement can be seen as a special
case of this more general scheme, the case of identity between the
mental construction of represented events and the observation of ac-
tual ones.
One broad class of mental constructions comprises generalizations
extracted to represent the world's basic structure, as opposed to spe-
cific, contingent occurrences that arise within this stable framework
(Goldsmith and Woisetschlaeger 1982; Langacker 1996, 1997). Ex-
pressions describing the world's basic structure - which I refer to as
structural statements - include generics, habituais, and "timeless
truths". Even when the verbs employed are perfective, they do not
refer to specific, actual events. For example, (26a) does not designate
any actual instance of a kitten being born. The profiled event is a
virtual instance of the process type in question, conjured up to ex-
press a generalization about one facet of the world's basic nature.
The events coded by structural statements belong to mental construc-
tions purporting to represent the world's basic structure rather than
any actual occurrences. Each of these event representations corre-
sponds to an open-ended set of actual instantiations, distributed
throughout the time span during which the generalization holds.
Tense marking on a structural statement specifies the time at which
the event representation can be consulted as a way of apprehending
this facet of the world's structure. Present tense indicates its viewing,
and thus its virtual occurrence, coincident with the time of speaking.
(26) a. A kitten is born with blue eyes. [generic]

b. My cousin goes to a singles bar on Friday night, [habitual]
c. Water decomposes into hydrogen and oxygen.
["timeless truth"]
Another special use of the present tense in English involves causal

relationships between events and situations. This curious construc-
tion, illustrated by (27a), is often heard from sports announcers. I
suspect it is rather common in casual speech. Despite the present
tense, the sentence does not describe anything happening right now,
nor is it a direct description of actuality. It is, moreover, ambiguous.
Sentence (27a) may be construed as comparable to (27b), pertaining
to a facet of future potentiality. Alternatively, it has an interpretation
similar to (27c), pertaining to something that failed to materialize in
the past. But unlike (27b-c), which make these interpretations ex-
plicit, in (27a) they are left implicit: how the events described are
connected to actuality is not directly specified by any overt element.
(27) a. He makes the freethrow and the game is tied.

b. If he makes the freethrow, the game will be tied.
c. If he had made the freethrow, the game would have been
tied.
An expression like (27a) profiles a sequence of two virtual processes,

the first representing an event, and the second a situation resulting
from that event. This mental construction manifests a special viewing
arrangement allowing the simplified presentation of a contingency
inherent in the past or future evolution of reality. Let's face it, (27b)
and especially (27c) are quite complex both conceptually and gram-
matically, involving modals, perfect aspect, mental spaces, and shifts
of viewpoint within those spaces. The viewing arrangement of (27a)
does away with all this at the level of explicit linguistic coding. It
simply abstracts the causally related processes and presents them as
virtual occurrences available for direct viewing at the moment of
speech. The complexity of their epistemic status is still all there, but
is incorporated in the presupposed viewing arrangement instead of

being overtly expressed.
The overall situation is sketched in Figure 9. The two virtual proc-
esses are connected by a double arrow to represent the causal rela-
tionship between them. It is only this virtual process sequence that is
put onstage and profiled by the successive coordinate clauses in a
sentence like (27a). Each clause is in the present tense, hence their
profiles coincide with the respective speech events associated with
the two finite clauses. In producing and understanding the expres-
sion, the speaker and hearer view and apprehend the virtual process
sequence, and their apprehension of it constitutes its virtual occur-
rence coincident with the time(s) of speaking.
/
/ Γ-
1 Processai—^ Process2| /
^ /Virtual
Α ,F
/ ,Λ w
w ΛΛΑ,
w
Figure 9. A virtual process sequence
The presupposed viewing arrangement incorporates a conception of

how the profiled process sequence relates to actuality. There are two
options in regard to time. As indicated by the dotted correspondence
lines, the time span during which the virtual process sequence is en-
visaged as occurring is identified with a temporal interval either prior
to the speech events or subsequent to them. Under either option, the
virtual processes are taken as representing a valid contingency, such
that the actual occurrence of Processi results in the actual occurrence
of Process2. However, the viewing arrangement also specifies that
Processi is not (or not yet) actual. When applied to the past, where
the course of reality has already been determined, the non-actuality of
Processi entails that it did not in fact occur, and thus implicates that
the situation coded by Process2 does not obtain. When applied to the
future, where the course of reality has not yet been determined so that
Processi is necessarily non-actual, presenting it for consideration has
the effect of suggesting that its actual occurrence (and the subsequent
occurrence of Process2) may be quite imminent.
Finally, I should mention the use of present tense in certain types
of subordinate clauses, as in (28):
(28) {If / when / until / before / after / while} you make a decision,
you should consider all your options.
Despite its present tense form, in each case the predicate {make a
decision) refers to a process envisaged as occurring in the future.
Here I basically follow the analysis proposed by Fauconnier (1997) in
terms of mental spaces. The subordinators introducing these clauses
are space builders: //establishes a hypothetical space, and the others
set up spaces defined by their temporal location. They further shift
the viewpoint to the space they establish. In other words, they incor-
porate the instruction to adopt a special viewing arrangement in
which the clausal content is apprehended from a temporal vantage
point other than the actual time of speaking. Its apprehension does of
course occur at the time of speaking, but when this is Actively identi-
fied with the time span internal to the mental space, a process con-
ceived as occurring in this space is fictively viewed as coincident
with the speech event.
6. Conclusion
I have sketched an analysis of the English present tense that is quite

non-standard and undoubtedly controversial. It is still preliminary
(even after all these years), and certainly incomplete. For instance,
following Brisard (1999), I have not sufficiently emphasized the
epistemic immediacy that constitutes the flip side of temporal coinci-
dence. Still, I have little personal doubt that the account is basically
valid, or at least suggestive of what a valid account might look like.
For sake of discussion, let us suppose that you agree. What then fol-
lows? What conclusions can we draw for language pedagogy?
I am of course eschewing any specific pedagogical proposals. I
will however suggest that the traditional way of looking at tense,
even in linguistics, engenders confusion by obscuring its basic nature.
Standard discussions are objectivist in spirit. They ignore construal
and the subjective basis of factors like homogeneity and bounding.
They have no conception of the myriad viewing arrangements that
mediate between objective circumstances and the formulation of lin-
guistic expressions. Thus they attempt to account for tense directly in
terms of the temporal relation between the actual time of speaking
and the full duration of an envisaged actual occurrence. They do this
even when - according to the analysis presented here - the process
being viewed and temporally located is only a portion of the actual
occurrence (notably with imperfectives), or else a virtual process
connected to it in a manner specified by the viewing arrangement. It
is no wonder, then, that a cogent description remains elusive, and that
the present tense is claimed to be anything but a present tense.
The consideration of pedagogical issues can only be aided by an
accurate understanding of what is being taught. In the case of lan-
guage, unfortunately, traditional and modern understandings are usu-
ally far from adequate, even for things as fundamental as the present
tense. It is premature to suggest that cognitive linguistics is coming to
the rescue. I do however see it as a positive development, providing
new and revealing perspectives on specific problems as well as our
overall conception of language and how it relates to culture, cogni-
tion, and social interaction. In short, I think we are starting to get a
real grip on how things work. If so, it should eventually give rise to
successful pedagogical applications, which will lend it empirical
support.
References
Austin, J. L.
1962 How to Do Things with Words. Cambridge, MA: Harvard Uni-
versity Press.
Brisard, Frank
1999 A critique of localism in and about tense theory. Ph.D. disserta-
tion, University of Antwerp.
Fauconnier, Gilles
1997 Mappings in Thought and Language. Cambridge: Cambridge
University Press.
Goldberg, Adele E.
1995 Constructions: A Construction Grammar Approach to Argument
Structure. Chicago and London: University of Chicago Press.
Goldsmith, John and Erich Woisetschlaeger
1982 The logic of the English progressive. Linguistic Inquiry 13: 79-
89.
Jackendoff, Ray
1994 Patterns in the Mind: Language and Human Nature. New York:
Basic Books.
1968 Language and its Structure: Some Fundamental Linguistic Con-
cepts. New York: Harcourt, Brace & World.
1987a Foundations of Cognitive Grammar, Vol.1, Theoretical Prereq-
uisites. Stanford: Stanford University Press.
1987b Nouns and verbs. Language 63: 53-94.
1988 A usage-based model. In: Brygida Rudzka-Ostyn (ed.), Topics in
Cognitive Linguistics, 127-161. (Current Issues in Linguistic
Theory 50.) Amsterdam/Philadelphia: John Benjamins.
1990 Concept, Image, and Symbol: The Cognitive Basis of Grammar.
(Cognitive Linguistics Research 1.) Berlin/New York: Mouton de
Gruyter.
1991 Foundations of Cognitive Grammar, Vol.2: Descriptive Applica-
tion. Stanford: Stanford University Press.
1993 Universals of construal. Proceedings of the Annual Meeting of
the Berkeley Linguistics Society 19: 447-463.
1995 Viewing in cognition and grammar. In: Philip W. Davis (ed.),
Alternative Linguistics: Descriptive and Theoretical Modes, 153-
212. (Current Issues in Linguistic Theory 102.) Amsterdam/
Philadelphia: John Benjamins.
1996 A constraint on progressive generics. In: Adele E. Goldberg (ed.),

Conceptual Structure, Discourse and Language, 289-302. Stan-
ford: CSLI Publications.
1997 Generics and habituais. In: Angeliki Athanasiadou and René Dir-
ven (eds.), On Conditionals Again, 191-222. (Current Issues in
Linguistic Theory 143.) Amsterdam/Philadelphia: John Ben-
jamins.
1999a Grammar and Conceptualization. (Cognitive Linguistics Re-
search 14.) Berlin/New York: Mouton de Gruyter.
1999b Virtual reality. Studies in the Linguistic Sciences 29(2): 77-103.
2000 A dynamic usage-based model. In: Michael Barlow and Suzanne
Kemmer (eds.), Usage Based Models of Language, 1-63. Stan-
ford: CSLI Publications.
Sweetser, Eve E.
1990 From Etymology to Pragmatics: Metaphorical and Cultural As-
pects of Semantic Structure. (Cambridge Studies in Linguistics
54.) Cambridge: Cambridge University Press.
Talmy, Leonard
1988 Force dynamics in language and cognition. Cognitive Science 12:
49-100.
1996 Fictive motion in language and "ception". In: Paul Bloom, Mary
A. Peterson, Lynn Nadel and Merrill F. Garrett (eds.), Language
and Space, 211-276. Cambridge, MA/London: MIT Press/Brad-
ford.
Pretend play:
trial ground for the simple present
Jenny Cook-Gumperz and Amy Kyratzis1
1. Introduction
Here is Edward Bear, coming downstairs now, bump, bump,

bump, on the back of his head, behind Christopher Robin. It
is, as far as he knows, the only way of coming downstairs, but
sometimes he feels that there really is another way, if only he
could stop bumping for a moment and think of it. And then he
feels that perhaps there isn't. Anyhow, here he is at the bottom
of the stairs, and ready to be introduced to you. Winnie-the-
Pooh [Milne 926: 3].
And so begins one of the most self consciously literary, yet most fre-
quently reissued children's books of the last 70 years. In this the
original beginning to the story, since adapted by Disney and others
into a universal tale of a "cute little bear", the literary devices of
tense/aspect provide a specific set of what Langacker (this volume)
calls the viewing arrangements that enable participants and read-
ers/hearers to shift perspectives on the story events. The following
text adds a further dimension to the tellers', participants', and
reader/hearers' perspectives, when A. A. Milne inteqects his own
character/voice: "When I first heard his name, I said, just as you are
going to say, 'But I thought he was a boy?'"
The author here enters the story briefly as a protagonist, that is an
authorial "I" telling how the tale came to be told. After this interlude
the story proper begins in the traditional story telling past tense, the
third person preterite (Benveniste 1971). "Once upon a time, a very
long time ago now, about last Friday, Winnie-the-Pooh lived in a
forest all by himself under the name of Sanders." We cite these, and
42 Jenny Cook-Gumperz and Amy Kyratzis
later examples from the Dr. Seuss books, to illustrate the kinds of
narrative experiences children participate in before they learn to read
for themselves. In attending to the often told/read tales, they are ex-
posed to linguistic constructions that can have complex semantic
implications such as signaling shifts in perspective. Through tense/
aspect contrasts, children's stories are able to provide succinct infor-
mation about the characters, the tale's context and the relationship of
the teller to the protagonists that make simply told tales dependent on
complex conceptualizations for their interpretation. As Langacker
argues: "Inherent in every usage event is a presupposed viewing ar-
rangement, pertaining to the relationship between the conceptualizers
and the situation being viewed" (16).
The progressive, as described by Langacker, for example, is dis-
tinct, not for its focus on the present time but for the viewing ar-
rangement it imposes: allowing the "zooming in" and "taking an in-
ternal view" of a bounded event (12). In story telling, and in literary
contexts, such a viewing arrangement is not appropriate; the events
must be told as though the narrator can view their endpoint. In narra-
tivizing discourse, then, to quote White's seminal paper on Narrativ-
ity: "we can say, with Benveniste, 'Truly there is no longer a 'narra-
tor'. The events are chronologically recorded as they appear on the
horizon of the story. Here no one speaks. The events seem to tell
themselves" (White, quoting Benveniste 1981). Here the events are
told as though they are unfolding at the moment of reading, and the
narrator takes a distal, uninvolved viewpoint; s/he is not in the event.
Yet sometimes, the narrator seeks to manipulate viewpoint to in-
volve the reader more, by imposing a viewing arrangement that com-
bines the affordances of the narrator's distal viewpoint which allows
story events to be viewed unfolding, with a closer-seeming viewing
arrangement that creates suspense. The simple present is a literary
form which does just this. It imposes just such a self-conscious yet
suspenseful perspective. The distal aspect of its viewpoint comes
from reporting perfective events as though the teller could view their
endpoint. As Langacker claims this imposes a seemingly "impossi-
ble" viewing arrangement for the conceptualizer unless she/he recon-
ceptualizes the event as virtual rather than actual (19). Langacker
Pretend play 43
goes on to point out that the simple present is difficult to use in de-
scribing perfective events, occurring at the moment of speaking be-
cause of what he termed the durational problem. As he puts it "be-
cause a perfective process is bounded, a full instantiation of it in-
cludes its boundaries, so if a perfective process is to coincide with the
time of speaking, its beginning point has to coincide with the initia-
tion of the event of speaking, and its endpoint with its termination"
(25). In other words, the speaker has to coordinate the event time
with the beginning of speech and wait for the termination of the re-
ported events to coincide with the time of speaking. The main use of
the present would be in event casts, which describe actions taking
place in the present time, such as sports commentaries, cooking
shows, direction giving, and even more rarely, advanced driving
tests. In event casts, such a viewpoint is possible because events are
happening quickly and each event is completed before the other be-
gins. Since such events rarely occur in everyday discourse, we tend to
assume a virtual, that is a hypothetical event is at issue when we
hear/see a simple present, as in the above beginning of the story.
It is more usual in descriptions of present actions to use progres-
sive constructions rather than the simple present. However with ha-
bituais, where there is no endpoint, the present is also used. When
progressives ("imperfectives") are used, the speaker is taking a tem-
porally coincident sample of the overall situation and observing that,
for the portion sampled, the situation is stable and unbounded. For
statements such as "I am driving to work", because a termination
viewpoint is not necessitated, and similarly for generic statements or
statements of habitual events (e.g. "I drive to work"), the endpoint
does not have to be encompassed in the time of speaking. Therefore,
to summarize Langacker's argument, there are two issues at stake in
the choice of tense-aspect. One is the focus on the event time as si-
multaneous with or prior to the speech event, that is the choice as to
whether one takes an involved or distal viewpoint; and the other is
the viewing arrangement or the choice of perspective, that is whether
the endpoint is revealed at the time of the telling.
If we accept that the simple present is unusual in ordinary, every-
day discourse, occurring only in special contexts, then children rarely
get practice with it. This argument is further substantiated in looking

at child language studies. Already in Roger Brown's early "A First
Language" (1973), he showed that the progressive present occurs
much earlier than the simple present. Brown found a range of acqui-
sition ages of 1;10-2;10, for the progressive to between 2;4 and 3;8
for the simple present in the three children he studied. Additionally,
there is one area of children's early experience where the simple pre-
sent is used: in picture-book storytelling between adults and children.
This pre-literacy experience bears out Langacker's argument about
the perspectival shift between continuative (or imperfective) action
that you are a part of and the perfective aspect. In acquiring literacy,
children must learn to negotiate this shift and adopt a perspectival
role very different from the demands of everyday spoken discourse.
They need to be able to take a viewpoint that looks in on events as
potentially complete and finite, as well as viewing ongoing events
from a distanced or continuative perspective. Children gain this expe-
rience in preliteracy through examining the pictures in a storybook as
the story read to them unfolds. The events taking place in the pictures
of a book are not realized until the reader/hearer looks in on them
from their appropriate perspective which is further reinforced by the
story text.
As we saw in the first examples, the perspectival shift heightens
involvement, since the present aligns the viewpoint of the reader/
hearer with that of the narrator, while simultaneously allowing the
"viewing" of the unfolding of events (and their endpoints) that may
in actuality take/have taken a long time to play out. From a writer's
viewpoint, because viewing arrangements are virtual, using the sim-
ple present makes an increased involvement possible and so is a de-
sirable literary move to take.
The excerpt below from Dr. Seuss' "One Fish, Two Fish, Red
Fish, Blue Fish", conveys this manipulation of viewpoint, although to
fully appreciate what is happening both pictures and text need to be
read together, as is common in children's books.
Pretend play 45
Here are some

who like to run.
they run for fun
in the hot, hot sun.
Oh me! Oh my!
Oh me! Oh my!
What a lot
of funny things go by.
We see them come.

We see them go.
Some are fast.
And some are slow.
Some are high.

And some are low.
Not one of them is like another.

Don't ask us why.
Go ask your mother.
Simple present descriptions such as "we see them go" by implication

put the reader/hearer into the frame of the action, and this involve-
ment is quite literally represented by pictures showing a boy and girl
watching the events as they are unfolding in the virtual picture space,
as if the "real" little boy or girl hearer/listener were inside this space.
However, the present tense conceptually also allows the events to
appear limited with potential endpoints. This viewpoint is reinforced
figuratively by the illustrations on each page which actually appear to
delimit the action, because they end with the page. These contrasts
allow the other constraint of "objectivity" in narrative, as suggested
by narrative theorists such as Benveniste (1971) and White (White
1980), also to be realized.
Children must begin to understand such complex and seemingly,
"impossible" viewing arrangements in order to participate in early
book-reading routines and understand literacy events. In Heath's

(1986) description of a two-year-old Roadville child's participation
in a bookreading routine, the child was able to utilize the simple pre-
sent "Puppy sees the ant"; "ant bite puppy" in responding to her
aunt's questions ("what does the puppy do?") (Heath 1986: 107).
Later, children will need to accomplish even greater milestones in
using tense and aspect shifts for literary effect; they will have to learn
to utilize the progressive not only in the sense of its affording of a
"zooming in" effect on a bounded event, but for backgrounding in
narrative. As noted by Hopper (1979), the contrast between ing-ed is
not a sentence-level, semantic phenomenon but an essentially dis-
course-level one between events and states presented as back-
grounded and events presented as individualized and sequenced.
Our argument in this paper is that English speaking children's
early understanding of the tense system is a key prerequisite for liter-
acy, and that a major source of this preparation for literacy comes
from children's engagement in pretend play with peers as well as
with story telling and reading experiences with adults. A viewing
arrangement for use of the simple present which avoids Langacker's
durational problem is planning during pretend play, particularly plan-
ning talk for the manipulation of character figures. Planning action
sequences for third-party characters in fantasy-play allows the
speaker to achieve a distal viewpoint on the activity. Because the
action sequence that is planned is hypothetical, or irrealis, a full in-
stantiation of it and its boundaries is not necessitated. Since the ac-
tions themselves are imaginary, the boundaries of the actions can be
hypothesized. In this way, planning action sequences for character
figures in manipulative pretend play can be a context which scaffolds
and provides a training ground for the use of the simple present.
Our argument is grounded in a perspective that views grammar as
discourse-embedded. More specifically, this view (Budwig 1990,
1995, Ervin-Tripp 1993, Gerhardt and Savasir 1986, Slobin 1982,
1985) maintains that children construct and appropriate grammatical
contrasts to index particular clusters of semantic, pragmatic, and dis-
cursive notions. As stated by Ervin-Tripp (1987, 1993), "my advice
to students of syntax is to get a video camera and look at what is go-
Pretend play 47
ing on" in the discourse. In other words, patternings in grammar are

appropriated by children to construct or index patternings in discur-
sive phenomena. For example, Budwig (1995) argued that some chil-
dren use the contrast between "I-my" self-reference forms in first
person to index a cluster of semantic/pragmatic features denoting the
contrast between low and high agency/intentionality. Along similar
lines, Gerhardt and Savasir (1986) argued that preschool aged chil-
dren utilize the form contrast between simple present and progressive
to index the contrast between normativity and negotiation. In this
paper, we combine the arguments by Langacker (this volume) and
Gerhardt and Savasir (1986) and argue that children use the simple
present, not to index normativity alone, but to signal normative, tem-
porally-sequenced action from a privileged viewpoint which allows
the endpoints of perfective events not to have to occur during the
time frame of the linguistic description of the events. This privileged
viewpoint is the distal perspective afforded by the genre of hypotheti-
cal planning talk, particularly as it occurs during collaborative char-
acter play with peers.
2. Examples
The data: excerpts in this paper come from three studies: one a study
of two 3.6 year old girls who play regularly in each others' home
(Cook-Gumperz 1985, 1992). The second set come from a study of
character play with four- and seven-year-old dyads (Kyratzis 1992).
The third set of data is taken from spontaneous play situations
videotaped in a nursery school in a combined three- and four-year-old
classroom (Kyratzis and Guo 2001, Kyratzis forthcoming).
2.1. Present tense for temporally-sequenced action: 3-year olds
Two three and half year old girls, Lucie and Susie are playing a game
of mummies and babies, indoors in small play space, surrounded by a
collection of dolls and stuffed animals. The girls make up a loose
narrative plot for a game of Mummies and Babies where mummies

feed, give drinks to their babies, scold them, put them to bed, go into
the garden with their babies, go on a walk to the park etc. The game
is fast-paced in that the girls are continuously talking or vocalizing
sounds, such as drinking or soothing noises. Through their talk they
create three distinct voices, a narrative planning voice in which they
negotiate and jointly construct the plot, a voice of mothers to each
other, or a voice of mothers speaking to their babies, each marked by
a distinct tone of voice. For example, when the mothers address their
babies their voice has short clipped phrasing with many rising/falling
and lengthened pitch contours especially at the end of phrases. The
narrative voice has an evenly paced tempo, close to their ordinary
voices. The mother to mother voice is louder, with higher but more
even pitch register. The narrative voice is shown with underline, the
mummies to babies in italics, and the mummies to mummies in bold
typeface. Ordinary voices are shown as unmarked text in the same
type face as the rest of the paper.
The first example shows the girls using the present tense as part of
a narrative describing the play action that they are about to perform.
Example 1.1
Susie: We've got to go up here and cause it's a sunny day we've
got to go outside and sit down in our garfden]
Lucie: Oh... oh yes that., that can be there..that can be there.,
yes
Susie: And we sit down and have a glass of orange iuicev. (whis-
pered mummy to baby) there sit down.
(play drinking and pouring sounds)
Shh..shh..now some more., of., gumpy-shek.. it's mac-
cambe... it's maccamba
Susie: The babies don't like it
Lucie: No the babies don't like macacamba
[Two turns later]
Susie: And he sits.... And the babies sit on our laps with us
Pretend play 49
In this example the two little girls used sounds, (Shh..Shh..) and
made up words (maccacamba, gumpy shekt) to set up a rhyth-
mic/metric beat for the game. Tense contrasts, along with voice-tone
differences, mark the different perspectives represented and enacted
by the "game voices". The narrative plot is predominantly repre-
sented in the simple present tense conveying a distal perspective as if
the interactants are looking in on the stage where they are about to
perform their actions. The initial statement by Susie relies on the
model phrase "got to", but as the aspectual sense of these phrases has
something of a habitual character to them, they could be interpreted
as 'when it's sunny we always go and sit in the garden'. In Susie's
second and third utterance the simple present describes the actions as
they are about to happen, yet with the sense that these activities once
performed can be visualized as computable actions. Their potential
completeness contributes to their immediacy.
In the next example the girls get into a dispute about the conse-
quences of the action's narrative which is resolved in the last line by
Susie's addition to the plot that signals her acceptance of the story
line.
Example 1.2
Lucie: Anyway.... and you say.. "Sandra., have you got pins"., and
I'll sav 'Ves"
Susie: Have you got some pins?
Lucie: Yes
Susie: I want to hold her.. I want., it's not fair
Lucie: She's having your pin
Susie: She's mine
Lucie: ..Now you sav "Sandra., have you got pins., and I'll sav
"yes"
Susie: ..NO
Lucie: No you..And and you sav "Sandra..have you got pins to
stick in your baby and and you and I say "yes"
Susie: And um., then you come and give me two pins.
Susie's acceptance of Lucie's idea for the plot development is recog-

nized by her conjoined contribution "and then you come and give me
two pins". Initially the use of the discourse marker "anyway" is part
of the narrative plot outlining speech, which along with use of direct
speech markers 'you say' and 'I'll say' give the narrative voice
something of the quality of a story in a children's book. The acting
out of the plot takes place in a "mummy to mummy" voice with
raised pitch, enacting the dialogue described by the narrative. These
dialogue stage directions and enactments are illustrative of Lang-
acker's durational problem. The talk describing the narrative plot can
project a completable event or set of actions because it is hypothe-
sizing the whole short sequence of dialogue, action and exchange
from a distal perspective. These directions are then carried out but
only in so far as they voice the suggested dialogue. If the girls' dia-
logue itself became the source for further actions arising either from
mummy to mummy or mummy to baby talk, it is likely that they
would use a tense shift to a present progressive to signal a new focus
on continuing action, so 'zooming in' on the action as it takes place
over a longer time span. This use of the present progressive in
'mummy-to-mummy talk' can be seen in the next Example 2.1.
However while the simple present tense implies a temporal/dura-
tional aspect, as Langacker explains such a durational issue can be
used to implicate habitual usage. Young children rarely use this form.
In the first example (1.1) we have commented on the use of the ha-
bitual in a modal construction "we've got to go and sit down in our
garden cause it's sunny" meaning 'when it's sunny we always go out
in the garden'. The following example has something of a similar
habitual generic quality to it, implying that a journey to Belgium is
always similar to one being described.
Example 1.3
Diane: How do you get to Belgium Susie?
Susie: First you go on a boat and then you get to Belgium and go
on a car the...then ..you get to Belgium
Lucie: and then you go on a train and...and sea and ships
Pretend play 51
Susie: ahh...and then you go in your car and you.your daddy

put.gives.takes you to Belgium with your Daddy and
Mummy and then you get to Belgium
Diane: oh umm
Susie: and it's a long way and it's inside Dominique's house and
you'll see Dominique won't you?..
2.2. The present progressive for ongoing action in in-role play

with 3 year olds.
Another set of examples where the two girls alternate their "mother-
to- mother" talk with mother-to-baby talk. Apart from the distinctive
voice/intonation patterning the former is more likely to be character-
ized by use of the progressive tense usually the present progressive as
the girls (as mummies) report their activities to each other, and the
mother to baby talk contains the greatest range of tense shifting and
syntactic complexities, using conditionals, imperatives with justifi-
cations, indirect imperatives e.g. "I don't want you to shout at San-
dra" reprimands in the form of questions e.g. "why did you spit water
at Sally Manga?"
Example 2.1
Lucie: Your baby is crying... Samolina...Santolina...allright...I'll
give you to your mummy...I'll give you to your mummy
now...Y our baby wants you
S: [inaudible] [out of range]
L: allright..here..she comes ..And then we are going to walk
around..should we go for a little walk?...s7za// we go for a
little walk...with Samolina..And try not to hurt her Samo-
lina...Samolina look..don't hit her ...oh let's go for a walk
you two...let's go for a walk with Sandra when she fin-
ishes There's Sandra...Sandra ..we are going for a little
walk
Susie: I know it.. where's my cup
Lucie: allright baby...have a drink on that one now...on mummy's

knee...you know where it is... and you 7/ be happily ever after
Susie: I hold my baby 'cause she was crying for me
Lucie: N0..N0 you hold my baby and I hold your baby
In this example Lucie is organizing the activity of her two 'pretend'

babies and calling out to the other "mummy" (Susie), who is tempo-
rally given the name Sandra by Lucie at this stage in the game, that
she is taking her babies for a walk. This statement serves as a direc-
tive to tell Susie about the new direction in the new plot. As Lucie's
talk goes seamlessly back and forth between her babies and the other
mummy, tense choices is one of the main indicators of her shifting
game roles, the progressive describes actions being described by the
mummies to each other such as "your baby is crying" or "we are go-
ing for a walk", the simple present either depicts mummies' talk to
babies, particularly as directives or imperatives "let's go for a walk"
"Samolina look., don't hit her".
2.3. Present tense for temporally-sequenced action: 4-year olds
We now shift to examples of speech from character play involving

four-year-olds. The narrative excerpts are from two four-year-old
girls, Abby and Fran. With the circus props, these girls are emplot-
ting a narrative about a Mom and twin children putting on an ice-
skating show. This excerpt is rich in tense shifting. Background, set-
ting-like information uses the progressive, both past ("pretend I was
moving this up and down" and "pretend somebody was ice-skating
on the rink") and present ("and you're sitting where the man- lady
zooms by") and copula verbs, both past ("there was a mom"; "pre-
tend they were boyfriend and girlfriend, okay?") and present ("this is
you, the teen-ager"; "now somebody's on the ice skate rink"; "but
*he is the audience"; "he's my boyfriend"; "that's my mom"). In
contrast, temporally-sequenced action is presented in past subjunctive
("I didn't let the seals fall off' "you got off the rink to watch for
awhile"; "I got off') and present tense ("she slows down when she
Pretend play 53
gets to you"; "they say who's that"; "I say mama"; "you take a good
look and say"; "I say that's my mom"; "and he says that's my mom").
In the narrative voice, when the girls say "I", they are referring not to
themselves {in realis) but to the character of the Playmobile figure
they have designated as acting out their role. In this voice, the girls
seem to be reading off a script representation that is already con-
structed. As Langacker (this volume) states, "Tense marking on a
structural statement specifies the time at which the (already con-
structed) event representation can be consulted as a way of appre-
hending the world's structure" (Langacker this volume). In contrast,
when the girls are talking as themselves, in stage-manager voice us-
ing the first person, they include modal marking on the verb (e.g.
"I'm supposed to act myself'; "pretend I could sit the seals on"), just
as they also use commands and questions in this voice ("do it like
this", "make it like this so the seals can sit"; "wait, I'll be right
back").
The excerpts below exemplify how this dyad shifts tense and as-
pect to mark different phases of the play, using statives in the set-up,
beginning phase, then shifting to progressive and simple present to
mark background and temporally sequenced action, respectively.
[up to now, they've been doing an animal tricks show. Now they shift
to ice skating show scenario]
Example 3.1
Fran: and pretend somebody was ice-skating on the rink, there was
a mom.
wait, pretend you got off the rink to watch for a while.
[F grabs A's figure and places it in front A]
Abby: okay, I got off it. [picks up the figure]
Fran: now, now somebody's on the iceskate rink, 'cuz when
she...[grabs A's figure again]
I'll set her up to be standing.
Abby: okay, but *he is the audience, [grabs a male figure and
places it in front of F] wait. I'll be right back. [A goes to F's
side] pretend = t h e y were boyfriend and girlfriend, okay?

[A returns to her seat]
In these lines, the girls are setting up the scenario, and they use a lot
of copulas "somebody is on the ice-skating rink" and statives "there
was a mom", "they were boyfriend and girlfriend", including one in
present tense "but *he is the audience". Statives are appropriate for
the setting-up phase of the pretend. However, after this is accom-
plished, they shift to progressive and simple present.
Example 3.2
Fran: and you're sitting where the man - where the lady zooms by,
but she slows down when she gets to you. okay? [F runs her
figure along the "ice" until it reaches F's side of the circus
ring] they say 'who's that'?
Abby: and I say- and I say- and I say [gets up and walks towards F]
Fran: 'mama!'
Abby: 'oh, he's my boyfriend', we're talking.
Fran: no, no, you ask 'who's that?', and then you take a good look
and say 'that's your mom!'
Abby: um,'who's that'?
Fran: and I say, 'that's my mom'.
Abby: and he, and he says, 'that's my mom'.
Fran: yeah, they both say that because they're twins.
In the first line, "and you're sitting where the man - where the lady
zooms by, but she slows down when she gets to you", F presents an
event sequence (lady zooms by, slows down), and uses the simple
present to denote the temporally sequenced action. Simultaneously,
she presents the background event (the audience sitting), using the
progressive, since this is background material and its endpoint is not
a focus of the story. Lower, three turns later, A uses progressive as-
pect to again denote background material - the explanation of her
character's speech ("we're talking"). Throughout the rest of the ex-
cerpt, through a long chain of story action, the two girls use simple
present to denote the temporally sequenced acts ("you ask", "and
Pretend play 55
then you take a good look" "and say", "and I say" and "he says". In
the last line, they use third person in what Langacker called the sta-
tive sense, because the endpoint of the act "say" is not a focus and the
information is merely explanatory, hence background. In example
3.2, as in the Dr. Seuss excerpt above, the simple present is used
when there is an "observer" to the event (in this case, the Mom).
Other times in their narrative, when there isn't such involved view-
point, the girls use the preterite to tell and plan their story.
Example 3.3
Fran: hello sweetie! Pretend you said, "hey that's my mama".
Abby: 'hey, that's mama/and then she...no, pretend that, I'm then,
the um, the trumpeters,
well, like,
Fran: ==the seals.
Abby: the seals come on
Fran: and I do the, um, the
Abby: and they go 'shooo--(x)
Fran: put that one on this!
Abby: no, I'm gonna do it my way.
Example 3.3 shows A and F advancing their collaborative story to-

gether, using simple present ("seals come on", "I do", "they go
'shoo'"), though they then break off into a stage manager argument
(Wolf and Hicks 1989). They shift to modals ("gonna") and com-
mands to do their arguing to one another in the stage manager voice.
In contrast, in the voice that went with simple present, the "I's" and
"you's" referred to characters, not to the girls themselves. The girls
shift tense and aspect to index shifts in story levels (background-
foreground) and story phases (stage-manager/director) in ways that
are very impressive for this age.
In sum, of the 10 four-year-old dyads in this study with character
figures, only two used simple present in the sophisticated way shown
here, for "consulting" or reading off temporally sequenced action
from a script in the present time, and differentiating this planning
phase both from the stage-manager voice and from background de-
tail. Abby and Fran displayed quite sophisticated tense-aspect shift-

ing for children in the preschool period.
2.4. Present tense for habitual state, progressive for ongoing action
in role-play with four-year-olds
In Kyratzis' other data base with four-year-olds, which was role-play

in the classroom, children seldom used character figures. They acted
in roles themselves. Consequently, use of the simple present did not
occur other than for habitual state (e.g. "I'm a Chinese sister, and I
look pretty"; "I come from Korea".) For temporally sequenced action
in their dramatic narratives, children used progressive tense (e.g.
"I'm making rice"). We argue that this is because they are taking an
involved, in contrast to a distal, viewpoint on the action, which they
did in the ice-skating show excerpt featured above. In role-play, they
are themselves engaged in the action. In contrast, in the character
play excerpted above, they were manipulating character figures and
talking about and reflecting on the action.
In the excerpt below, three four-year-old girls (Viv, Aly, and Joan)
are enacting a scenario where they are pretending to be Chinese sis-
ters and make rice for their brothers.
Example 4
Aly I'm making soup// (kalatoya)// (kalatoya)// (kulakepa)//
Viv (malakapu)// pretend I'm making food// =(xx) pretend= I'm
making Chinese stuff//
Aly =I'm making food for my brother=// I'm
making food for my brother//
Viv pretend I'm making (rice)
Aly == I'm making Chinese ri- soup for my brother//
Viv yeah// and I'm making soup//
In this scenario, the children are conducting the action themselves, as

when the mummies in the Linde/Sophie example were describing
their own actions while doing them. They use progressive to do this
Pretend play 57
("I'm making Chinese stuff"; "I'm making food for my brother";

"I'm making soup"), since their perspective on the action is quite
involved.
In addition to describing temporally sequenced action, these chil-
dren use simple present to talk about habitual or generic action. For
example, they describe habitual states ("I look pretty", "we always
need it", "pretend I come from- from America"; "I come from
China").
Example 5
Aly pretend I'm Chine::se// I'm a Chinese sister/ and I look
pretty//
Viv pretend this (xx)// pretend (xx) are called (rices)//
Aly and those are called chaniza// because they're- they're spe-
cial kind of tortilla// and we always need itII
Joan we make (lots of fresh rice) 'cause ("because") we're Chi-

nese people// pretend/ I'm a Chinese//
Aly pretend/ this is cabbage// this is cabbage right?
Viv yeah// pretend I'm- pretend I'm American// pretend I come
from- from America//
Aly = { I come from China [f]}//
Viv yeah// 'cause ("because") you're a China person// pretend I
come from- from- from Korea 'cause ("because") I'm a Ko-
rean person
Aly =I'm a=
Viv =that's= where my mom comes from/ Korea//
These uses of habitual simple present, particularly the link to "al-

ways", tie in with the previous examples from the three year old girls.
While the four year old girls in this example have lexicalized the
habitual quality of their action with "always", it is clear that both the
three and four-year-old girls understand the habitual perspective
taken on the events described.
As our six examples above show, between the ages of early three
and late four, children are learning a great deal about temporality as
marked by tense. They show understanding of what we've referred to

as the durational problem, and perhaps more importantly, the need to
shift perspective on events for different discourse purposes. The
ability to shift among the simple present for temporally sequenced
action and for giving a generic description and the progressive for
achieving a zooming-in perspective is quite advanced. As Langacker
(this volume) states, "tense marking on a structural statement speci-
fies the time at which the (already constructed) event representation
can be consulted as a way of apprehending the world's structure".
3. Conclusion
Children gain a special experience from the textual quality of stories

as they are read aloud. An aspect of this textual quality is shifting in
perspectival view, which is constituted by tense and aspect shifting.
These shifts, in turn, are reinforced by the visual experience (i.e.
looking at story pictures) as part of the textual quality in children's
books. Perspectival shifts and visual experience taken together pro-
vide a textual experience beyond understanding of narrative plot.
Through tense/aspect contrasts, children's stories are able to provide
more nuanced views of the characters and their situation. These make
the oral reading of simply told tales dependent on complex conceptu-
alizations for their interpretation (as afforded by the syntax), and ren-
der the story reading interesting to the adult reader as well as to the
child listener. From these oral readings, the child gleans a sense of
the rich textual experience that stories provide.
Although preschoolers are already fairly adept at perspectival
shifting in their pretend play, as the data shown here suggests, their
ability to recognize and make use of tense shifts is one of the things
that changes most between three and four years. The three year olds
shown here sometimes seemed to have an unstable use of tense,
shown here in the interjection of future modals within present tense
(contractions of the future auxiliary, as in "I'll", as though it were
part of the pronoun). Moreover, the four year olds had a quite so-
phisticated shifting between progressive and simple present to mark
Pretend play 59
the contrast between foreground and background events. More prac-

tice with these oral narratives, and greater exposure to bookreading,
no doubt help negotiate this developmental transition.
Another preparation from literacy gleaned from character play is
the use of simple present in the habitual sense. This may be an im-
portant precursor for literacy events, such as giving definitions and
explanations (Kurland and Snow 1990, Aukrust and Snow 1998). In
these genres of talk, habituais are common. For example, explana-
tions featured in Aukrust and Snow's (1990) American data con-
tained simple present habituais (e.g. "I like pancakes; They make me
strong" (1998: 235) and "like when you talk, it goes on" (1998: 240).
Definitions taken from Kurland and Snow (1997) use habituais as
well (knife defined as "thing that cuts things" (1997: 609) and um-
brella as "something that you hold over your head in case it rains"
(1997: 610). The ability to give definitions is an important type of
decontextualized language skill and is "related to being part of an
academic culture" (Kurland and Snow 1997: 603), hence is an im-
portant aspect of emergent literacy. Children who have practice in
using habituais in pretend play may have a leg up and a natural segue
into literacy-related talk genres such as definitions and explanations,
which utilize similar tense-aspect markers, thereby taking similar
perspectives or viewpoints on events.
However, it is from oral experiences outside of books as well, the
character play with peers and parents, that children gain the tools to
accomplish tense and aspect shifting, and thereby perspective. In
character play, children rework the scripts of everyday experience.
Onto these scripts, they import the language of perspective shifting,
importing habituais, progressives, and simple presents from their
preliterate experience of storybooks to rework and reflect upon their
scripts of everyday experience.
Note
1. Names are listed in alphabetical order.

References
Aukrust, Vibeke Grover and Catherine E. Snow

1998 Narratives and explanations during mealtime conversations in
Norway and the U.S. Language in Society, 27: 221-246.
Benveniste, Emile
1971 Problems in General Linguistics. Coral Gables, Florida.
Brown, Roger
1973 A First Language: The Early Stages. Boston: Harvard University
Press.
Budwig, Nancy
1995 A Developmental-Functionalist Approach to Child Language.
Mahwah, NJ: Lawrence Erlbaum Associates.
Cook-Gumperz, Jenny
1985 Keeping it together: text and context in children's language. In:
James Alatis and Deborah Tannen (eds.), Languages and Lin-
guistics: The Interdependence of Theory, Data, and Application.
Washington, D.C.: Georgetown University Press.
1992 Gendered contexts. In: Peter Auer and Aldo Diluzio (eds.), The
Contetxualization of Language. Amsterdam: John Bejamins
Ervin-Tripp, Susan M.
1987 Speech acts and syntactic development: Linked or independent?
Keynote address to the Boston Child Language Conference, Oc-
tober 1987.
1993 Constructing syntax from discourse. In: Eve V. Clark (ed.), Pro-
ceedings of the twenty-fifth annual Child Language Research Fo-
rum, 333-341. Stanford, CA: Center for the Study of Language
and Information.
Geisel, Theodor Seuss
1960 One Fish, Two Fish, Red Fish, Blue Fish. New York: Random
House.
Gerhardt, Julie and Iskander Savasir
1986 The use of the simple present in the speech of two 3-year-olds:
Normativity not subjectivity. Language in Society 15: 501-536.
Heath, Shirley Brice
1986 What no bedtime story means: Narrative skills at home and
school. In: Bambi B. Schieffelin and Elinor Ochs (eds.), Lan-
guage Socialization Across Cultures. New York: Cambridge
University Press.
Pretend play 61
Hopper, Paul
1979 Aspect and foregrounding in discourse. In: Talmy Givón (ed.),
Discourse and Syntax. Syntax and Semantics, vol. 12, 213-241.
New York: Academic Press.
Kurland, Brenda F. and Catherine E. Snow
1997 Longitudinal measurement of growth in definitional skill. Journal
of Child Language 24: 603-625.
Kyratzis, Amy
1992 Gender differences in the use of persuasive justifications in chil-
dren's pretend play. In: Kira Hall, Mary Buchholtz and Birch
Moonwomon (eds.), Locating power : Proceedings of the Second
Berkeley Women and Language Conference 2, 326-337. Ber-
keley: Berkeley Women and Language Group.
Forthcoming Language socialization across time and context: Emotion talk in
preschool same-sex friendship groups. To appear in: Nancy Bud-
wig (ed.), Language Socialization during the Preschool Years
[Special issue]. Early Education and Development.
Kyratzis, Amy and Jiansheng Guo
2001 Preschool girls' and boys' verbal conflict strategies in the U.S.
and China: Cross-Cultural and contextual considerations. In: Amy
Kyratzis (ed.), Gender Construction in Children's Interactions: A
Cultural Perspective [Special Issue]. Research on Language and
Social Interaction.
This volume Cognitive linguistics, language pedagogy, and the English present
tense.
Milne, Alan Alexander
1926 Winnie-the-Pooh. New York: E. P. Dutton
Slobin, Dan I.
1982 Universal and particular in the study of child language. In: E.
Wanner and L.R. Gleitman (eds.), Language Acquisition: The
State of the Art, 128-170. Cambridge: Cambridge University
Press.
1986 Crosslinguistic evidence for the language-making capacity. In:
Dan I. Slobin (ed.), The Cross-Linguistic Study of Language Ac-
quisition, Vol. 2: Theoretical Issues, 1157-1256. Hillsdale, NJ:
Lawrence Erlbaum Associates.
White, Hayden
1980 The value of narrativity in the representation of reality. In: W. J.
T. Mitchell (ed.), On narrative, 1-23. Chicago: The University of
Chicago Press.
Wolf, Dennie and Deborah Hicks

1989 The voices within narratives: The development of intertextuality
in young children's stories. Discourse Processes 12(3): 329-351.
The relation between experience, conceptual
structure and meaning: non-temporal uses of
tense and language teaching
Andrea Tyler and Vyvyan Evans
1. Introduction
1.1. The problem
It is widely accepted that the primary function of English tense is to

signal time-reference (Binnick 1991, Comrie 1985, Jespersen 1924;
Langacker 1991, this volume, Quirk et al. 1973, Reichenbach 1947).
While this paper is concerned with both tense and language teaching,
we will not be primarily addressing the time-reference function asso-
ciated with tense, nor will we be offering an approach for teaching
tense. Nonetheless, the problem we address is directly motivated by
an issue faced by language teachers. While tense morphology canoni-
cally signals time-reference, it is often associated with other non-
temporal functions (Celce-Murcia and Larsen-Freeman 1998, Comrie
1985, Cutrer 1994, Fleischman 1989, Langacker 1991, Quirk et al.
1973, Riddle 1986, Swan and Smith 1987, Ter Meulen 1995, West-
ney 1994).
There are four kinds of non-temporal uses of tense that we will
consider. The first function relates to a designation of intimacy be-
tween the speaker/s and others. The second relates to what we will
term salience (commonly referred to as foregrounding and back-
grounding in the discourse literature). The third concerns what we
will term actuality, in which tense is used to signal the extent to
which the experiencer (or speaker) believes the event described cor-
responds to the actual world-state and conditions holding (this has
64 Andrea Tyler and Vyvyan Evans
been variously termed epistemic stance, cf. Fillmore 1990, or as a

distinction between realis and irrealis). The fourth function concerns
what we will term attenuation, in which certain speech acts are "sof-
tened" or mitigated in terms of their threat to face (cf. Brown and
Levinson 1987). This function is commonly referred to in terms of
linguistic politeness phenomena. These functions will be described in
greater detail in Section 2.
In previous approaches to tense it has been common to distinguish
the time-reference function from such non-temporal functions on the
basis of "meaning" versus "use" (e.g. Comrie 1985). That is, while
the literal meaning of tense relates to time-reference, non-temporal
uses are simply uses derived from contextual interpretation. The ar-
gument is that in a sentence such as: I just wanted to ask you if you
could lend me a pound (ibid.: 19), the addressee can determine from
context that the use of the past tense does not relate to a past desire,
but rather to a current situation. This common approach has tended to
reinforce the view that non-temporal uses should be treated as ex-
ceptions.
The difficulty for language teachers, and one we have faced our-
selves in classroom settings, is how to insightfully present the non-
temporal uses associated with tense. The approach offered by re-
ceived wisdom, as reflected in course books and pedagogical gram-
mars, is to treat them as exceptions, or worse to ignore them alto-
gether. For instance, Westney (1994) has observed that in pedagogi-
cal grammars: "[T]ime reference is treated as dominant and other
uses are simply appended" (ibid.: 79). Riddle (1986) notes that most
pedagogical texts ignore the uses of tense to signal intimacy, sali-
ence, and attenuation. If any of them are addressed, the general posi-
tion is that these uses are arbitrary; presentation of non-temporal
usages are often scattered throughout a grammar with no attempt to
tie the non-temporal use back to the basic temporal sense. Such
analyses have resulted in second language learners being instructed to
simply learn formulaic phrases to express polite requests, indirect
commands, conditionality, etc. with little or no explanation for why
the tense marking in the phrases they are asked to memorize does not
correspond to temporal uses of tense. The ultimate result, as Riddle
The relation between experience, conceptual structure and meaning 65
(1986) documents, is that second language teachers are at a loss for a

satisfactory explanation of the phenomena and even relatively ad-
vanced second-language learners often experience difficulty acquir-
ing these non-temporal uses of English tense.1
Our central thesis in this paper is that the non-temporal uses of
tense are related to its time-reference function in a motivated way.
Rather than treating tense as being analyzable into a literal time-
reference meaning and an unrelated range of non-literal exceptional
uses, we suggest that two important aspects of the cognitive linguistic
perspective allow us to view the non-temporal meanings as being
related to the temporal meaning in a systematic way. First, cognitive
linguists treat grammatical markers, such as tense morphemes, as
being meaning-bearing elements in the same way, in principle, as full
lexical items. This entails that tense markers can be treated as form-
meaning pairings. The second is that symbolic elements (i.e. form-
meaning pairings) are subject to usage-based meaning extensions. As
such, through use, additional meanings can become associated with a
particular form, resulting in the lexical form becoming related to a
semantic network of distinct, although ultimately related, senses. We
follow Elizabeth Traugott (e.g. 1989) in referring to this process of
meaning extension as pragmatic strengthening.
In other words, we argue that the non-temporal meanings associ-
ated with tense are conventional meanings or senses associated with a
particular lexical form, in the same way as the time-reference mean-
ing. While the time-reference sense might be the diachronically pri-
mary meaning element, our approach suggests that distinguishing
meanings based on their "literalness" or whether or not a particular
meaning counts as an exception may be misplaced. Moreover, we
will argue in detail that due to the way in which we actually experi-
ence the notions of intimacy, salience, actuality, and politeness,
namely in terms of proximal-distal spatial relations, and the fact that
time-reference is experienced in terms of analogous spatial relations,
in certain situations tense morphemes which canonically signal time-
reference can implicate a non-temporal relation. Through usage-
based conventionalization, i.e. pragmatic strengthening, a conven-
tional non-temporal meaning can become associated with a particular

tense morpheme.
The main purpose of the present paper, as we noted above, is not
to offer a new way of teaching tense, whether we are concerned with
temporal or indeed non-temporal meanings. Rather, the aim of pre-
senting the rather elaborate study outlined below, is to illustrate that
by assuming the perspective and methodology of cognitive linguistics
it is possible to relate the non-temporal senses associated with tense
with the time-reference meaning in a plausible way. This approach,
we suggest, should facilitate language teaching by providing a sys-
tematic model of the links among the semantics of tense markers,
which is to say between temporal and non-temporal meanings. Lang-
acker (this volume) suggests that cognitive linguistics may "prove to
be lighter, less onerous, and more appreciative than certain previous
theoretical burdens. I hope it will even prove useful". We offer the
present analysis in the same spirit, and suggest that by relating the
various meanings associated with tense morphology, language teach-
ers have a model by which to understand and so relate the diverse
functions performed by tense markers. It is our hope that the pro-
posed theoretical perspective may assist in the language classroom.
1.2. Previous analyses and overview of the proposed solution
Our goal is to present a motivated approach to the multiple uses of

English tense. We argue that there is an experientially-based com-
monality reflected at the conceptual level that holds between the con-
cepts of time-reference (externalized linguistically via tense), and
intimacy, salience, attenuation, and actuality; this commonality moti-
vates the use of the tense morphemes to implicate the multiple, at-
tested meanings associated with tense. As such, the tense mor-
phemes, which primarily encode time-reference, have come to be
associated with non-temporal meanings.
In our account, we depart in certain ways from previous explana-
tions of tense within cognitive linguistics (e.g. Dirven and Radden
2000, Langacker 1991). We note two heretofore unaddressed prob-
lems with the analyses of the usage of English tense morphology. The
first involves a contradiction between the representation of the basic
sense of the past tense morpheme and several of its non-temporal
uses. Langacker (1991), whose work forms the basis of many subse-
quent analyses, represents past time as being equated with "non-
immediate reality" (ibid.: 242), and present time with "immediate
reality" (ibid.: 242). The situations or states of affairs being refer-
enced by the English tense morphemes are represented as known
reality, i.e. they "are accepted by a conceptualizer as being real"
(Langacker 1991: 242). Reality status is argued to be signaled lin-
guistically by the presence or absence of modal marking. Past time,
which largely equates with non-immediate reality, is signaled by the
past tense morpheme; present time, which largely equates with im-
mediate reality, is signaled by present tense morphology. In contrast,
future and modal forms are represented as signaling irrealis. How-
ever, English speakers regularly use the past tense morpheme to sig-
nal relatively less commitment to the reality of an event or state of
affairs; past tense is also used in certain politeness phenomena, which
are clearly not interpreted as states of affairs accepted as real. These
uses are at odds with the "known reality" representation.
The second problem concerns the explanatory power of certain
claims about metaphorical distance and tense morphology. Recog-
nizing that English tense morphemes are not solely interpreted in
terms of temporal relations, Langacker (1991) argues that there is an
epistemic opposition between "immediate and non-immediate real-
ity" (ibid.: 245-246) which is marked by the absence or presence of
past tense morphology;2 the present and past tenses contrast with
"immediate and non-immediate irreality" (ibid.: 245-246) which are
marked by various modal forms. On this basis, he labels the past
tense morpheme as a "distal marker". Significantly, his discussion in
relation to the past tense morpheme centers on temporal distance.
Expanding on Langacker's analysis, Dirven and Radden (2000:
Chapter 9) note that a sentence such as: I wanted to ask you a favor
"illustrates a metaphorical shift of the past tense...The use of past
tense in [I wanted to ask you a favor] achieves an effect of politeness;
it distances the situation in time and, as a result makes the request
seem less face-threatening." What is left unspecified in such an ac-

count is why placing a speech event in the past, i.e. in non-immediate
reality, where it would normally be assigned the interpretation of
known (or established) reality, should be understood as less face-
threatening. For instance, the normal interpretation of: I wanted to
ask you a favor still involves the understanding that the speaker is
about to ask a favor. Why should establishing that the speaker had a
favor to ask before the speech event, a state which continues at the
time of the speech event, result in less face-threat? Simply saying this
is an example of metaphoric shift and "temporal distance" (which
itself is a metaphor that brings together two distinct domains - the
temporal and spatial) does not actually offer an account of the phe-
nomena. Crucially missing from this metaphorical analysis is an ex-
planation of the relationship between 'known reality', time, and cog-
nitive distance.
Intuitively it seems that tense-related politeness phenomena do in-
volve a notion of distance, but the distance seems to be interpreted in
a non-temporal way. The nagging question remains as to why a mor-
pheme which primarily signals non-immediate reality or, metaphori-
cally speaking, "temporal distance", should come to be associated
with "non-temporal distance". The analysis we put forward in this
paper seeks to offer an explanation for these apparent dichotomies.
In particular, we expand on two fundamental insights initially dis-
cussed in Grady (1997, see also Evans 2000). First, many linguistic
phenomena which have previously been treated homogeneously as
metaphor are more accurately understood as arising from different
sources; the most important for our purposes involves experiential
correlation, i.e. independently motivated and recurring correspon-
dences in experience. Second, previous accounts of conceptual
structure have tended to represent concepts pertaining to internal
states as abstract and indirect while representing concepts pertaining
to experiences with the external world as concrete and direct. Grady
(1997) persuasively argues that this representation is questionable
and that a more appropriate distinction can be made in terms of im-
age and response concepts. Image concepts pertain to sensorimotor
information, derived from external experience. In contrast, response

concepts pertain to information arising from internal states.
Specifically, we suggest that the concepts of time-reference, inti-
macy, salience, actuality, and attenuation all represent response con-
cepts that are structured or elaborated in terms of the same experien-
tially-based image content. In certain contexts, tense (which denotes
time-reference) may implicate one or another of these concepts.
Through continued use, these implicatures can come to be conven-
tionally associated with the tense morphemes. This process of mean-
ing extension we term, following Traugott (e.g. 1989) and Hopper
and Traugott (1993), pragmatic strengthening. In essence then, we
will argue that morphologically bound English tense forms (past and
present tense markers) are meaningful elements which typically
prompt for time-reference meaning. However, these forms have been
extended to convey non-temporal meanings in systematic and moti-
vated ways.
In view of the foregoing it is worth briefly contrasting the nature
of the present analysis of tense with that presented by Langacker (this
volume). In his paper, Langacker argues that canonically "the present
tense indicates that a full instantiation of the profiled process occurs
and precisely coincides with the time of speaking...[while]...[t]he
past tense morpheme imposes an immediate scope located prior to
the speech event" (ibid.: 22). Put another way, for Langacker, coding
time (the time of speaking), and reference time exactly co-occur in
canonical uses of the present tense. One of the issues that his paper
addresses is the way in which present tense is employed in situations
in which the present tense, for instance, does not refer to a situation
in which coding time and reference time coincide. By way of exam-
ple, a speaker who lives in London but works in New York might
say: "I work in New York", even though it is the weekend, and cod-
ing time coincides with not working and being in London as opposed
to New York. Langacker is at pains to address such 'non-present'
uses of the present tense in terms of the possibility of multiple con-
struals or viewing arrangements that change the interpretation of the
particular linguistic element under examination. In some instances
"what is being coded linguistically is not the actual occurrence of
events, but their virtual occurrence as part of a non-canonical view-

ing arrangement" (ibid.: 30). While we in large measure subscribe to
Langacker's stance, it is worth pointing out that in the present paper
we are not dealing with 'non-present' uses of the present tense, in the
sense of Langacker. Such 'non-present' uses are still, after all, tempo-
ral in nature. Rather, we seek to examine why tense morphology can
be employed to code meaning elements which are non-temporal in
nature. After all, politeness, for example, is not in any obvious way a
temporal notion. Yet, in English, and indeed, in a range of other lan-
guages, notions such as politeness can be designated by utilizing
tense morphology (cf. Fleischman 1989). Our approach, as intimated
above, and as will become evident, is to claim that there are distinct
meaning elements or senses, relating to concepts such as politeness,
which are conventionally associated with the past and present tense
morphemes. While construal is an important part of understanding
non-canonical temporal uses of tense, some non-temporal meanings
associated with tense are, we argue, due to the polysemy exhibited by
tense morphology. This parallels the polysemy exhibited by free
morphemes such as the much studied English preposition over, for
instance (cf. Tyler and Evans forthcoming a, forthcoming b).
The remainder of the paper is organized in the follow manner. In
Section 2, we briefly present the linguistic phenomena under consid-
eration. In section 3, we lay out the theoretical foundations for our
analysis. First, we examine the evidence in favor of adopting the dis-
tinctions of image concepts versus response concepts rather than
concrete concepts versus abstract concepts (contra Lakoff and John-
son 1980, 1999). We further argue that all the meaning elements as-
sociated with English tense that we discuss are best understood as
response concepts. Next, we turn to Grady's insights regarding expe-
riential correlation. We argue that experiential correlation provides
a powerful mechanism for explaining how distinct events come to be
associated at the conceptual level. In Section 4, we examine the non-
tense linguistic patterns associated with the domains of temporal ref-
erence, intimacy, salience, actuality, and attenuation and show that all
are elaborated through distal-proximal image content. In Section 5,
we argue that the fact that all these concepts are elaborated by the
same image content links them at the conceptual level. This concep-
tual linking, then, licenses the use of tense morphology to signal
these various uses. Finally, in Section 6, we sketch some of the im-
plications of the analysis for language pedagogy.
2. The phenomena
In order to give a sense of the various meanings of tense to be inves-

tigated, we present some examples below. We also emphasize that
we will only be considering the synthetic tense forms in English,
namely the present and past tense bound morphemes. The present
tense morpheme we will represent by [zero/s] or [0/s], to capture the
fact that this tense morpheme constitutes the default verb marking.
The past tense morpheme we will represent by [Id], which represents
a generalization over regular and irregular past tense verb marking in
English. These two tense morphemes can be employed to signal the
following meanings:
Time-reference
(1) a. I work in advertizing (present time-reference)
b. Yesterday I went to the cinema (past time-reference)
Intimacy
(2) a. A: Jane just bought a Volvo.
B: Maureen has one.
A: John, you've got to quit talking about Maureen as if
you're still going together. You broke up three months
ago. (Riddle 1986)
The context in which the utterance occurred is that Speaker B, John,

and Maureen were previously in an intimate relationship but the rela-
tionship ended and the two have not seen each other for some time.
Speaker A interprets Speaker B's use of the present tense as an un-
justified claim to intimacy.
(2) b. My daughter's father was Brazilian. He stays in contact with

Suzanna, but I haven't seen him in years. (Tyler, personal
data)
The past tense appears to be employed by the speaker to signal emo-

tional distance from Suzanna's father. Given the context, we cannot
interpret this use to mean that the father is dead or no longer Brazil-
ian. Nor can we assign the interpretation that the man from Brazil is
no longer in the role of father.
Salience
(3) a. In November 1859, Charles Darwin's The Origin of Species,
one of the greatest and most controversial works in the lit-
erature of science, was published in London. The central
idea in this book is the principle of natural selection. In the
sixth edition.. .Darwin wrote: "This principle of preservation
of the survival of the fittest, I have called Natural Selection."
(Eigen and Winkler 1983: 53)
Tense is employed to signal the relative status of the information or

salience, i.e. past tense signals background and supporting status and
present tense signals foreground status.
(3) b. Bateson introduced the notion of frame in 1955 to explain

how individuals exchange signals that allow them to agree
upon the level of abstraction at which any message is in-
tended. Even animals can be seen to use frames to interpret
each other's behavior, by signaling, for example, "This is
play." Bateson (1972) insists that "frame" is a psychological
concept, but to characterize it he uses "the physical analogy
of the picture frame and the more abstract...analogy of the
mathematical set", (p. 186) (Tannen 1993: 18)
Again, past tense is used to establish the background information and

present tense is used to foreground the information the author deems
most salient.
Actuality
(4) a. I wish I knew what he'll say next. (Westney 1994)
b. I wish the students liked phonetics. (Fleischman 1989)
Past tense is used in conjunction with specific lexical items, such as

wish, to signal a state of affairs, which the speaker believes to differ
substantially from actuality.
c. Suppose your house burned down. Do you have enough in-

surance? (Frank 1983)
Past tense is used to signal that a situation is hypothetical, i.e. a

situation which given the current world-state does not hold, and
hence is imaginary or contingent in some way.
d. If he studied harder now, he would get better grades. (Frank

1983)
Past tense in the if clause (the protasis) signals a negative stance vis-
à-vis the conditional situation and reality, i.e. the situation described
is non-actual. The sentence roughly paraphrases as "He is not study-
ing hard at the moment. If he changes the present circumstances, and
he studies harder, it would be possible for him to get better grades."
Attenuation
(5) a. I was thinking about asking you to dinner. (Fleischman
1989)
b. I was hoping we could get together next week. (Fleischman
1989)
Past tense is used to attenuate invitations, and hence to decrease the

threat to face for the hearer (e.g. the imposition of an invitation which
must be responded to immediately), and to the speaker (e.g. possible
immediate rejection of the invitation), (cf. Brown and Levinson
1987), and is thus conventionally interpreted as polite.
c. Receptionist (answering telephone): Good afternoon, Dr.

Keller's office.
Caller: Yes, I wanted to ask you a question. (Davies per-
sonal communication)
Past tense is used to attenuate requests.
d. Did you want to take a look at this? (Fasold personal com-

munication)
e. I thought you might like to try this (Westney 1994)
Past tense can be used to attenuate suggestions.
f. It's high time we left (Westney 1994)

Past tense attenuates commands and reprimands, rendering them
conventionally more polite.
In the foregoing examples we have seen that tense is crucially em-
ployed in English to signal a number of distinct and fundamentally
non-temporal meanings. In (2), tense is used to signal intimacy. In (3)
tense signals the relative salience or status of the information being
conveyed. In (4), tense is employed to signal the stance towards the
actuality of a particular scenario, i.e. speaker's degree of commitment
to the reality of the scenario, or alternatively how likely it is to come
about. Finally, in (5), tense can have an attenuating function in re-
quests, commands and invitations, mitigating the amount of imposi-
tion on the addressee or mitigating potential threats to our public per-
sona or face.
In the remainder of this paper, we explore the hypothesis that the
various meanings associated with tense are highly motivated, being
related in a systematic, principled way. Thus, we attempt to move to
a deeper level of explanation of the phenomena than afforded by pre-
vious metaphor-based accounts of tense.
3. The nature of conceptual structure
3.1. Image concepts versus response concepts
One of the most important contributions of cognitive linguistics to

our understanding of language has been the insight that much of eve-
ryday language involves discussing/understanding one concept in
terms of another. Within cognitive linguistics it has traditionally been
assumed that there is an a priori distinction to be made between ab-
stract concepts, which are not directly perceived and with which we
have less experience, e.g. time, and emotion concepts such as anger,
etc., and concrete concepts, which are directly perceived and with
which we have more extensive experience, e.g. motion, heat, etc.
(Lakoff and Johnson 1980, 1999). It has been further argued that it is
the distinction between direct and indirect perceptual experience that
is responsible for the organization at the conceptual level of abstract
concepts in terms of concrete concepts. Hence, an utterance such as
Time flies is held to demonstrate that English speakers conceptualize
and accordingly lexicalize the abstract, less directly experienced no-
tion of time in terms of the more directly experienced notion of a
physical entity in motion.
However, it is not at all clear that so-called abstract concepts are
less directly or less fully experienced than so-called concrete con-
cepts. For instance, Ortony (1988) observes that putative abstract
emotion concepts such as love are experienced by children much ear-
lier and more extensively than putative concrete concepts such as
journeys, which Lakoff and Johnson (1980, 1999; Lakoff 1987,
1993) and others have claimed serve to structure the abstract con-
cepts. More recently, Grady (1997) has pointed out that while so-
called abstract concepts may not result from external sensorimotor
input (i.e. perceptual experience from the external world), this does
not entail that they do not derive from equally direct experience.
Grady's argument is that so-called abstract concepts pertain to inter-
nal, subjective states, but it does not follow that such internal states
are not directly experienced. As Grady observes: "[T]he awareness
that we are conscious (as opposed to unconscious) is perhaps the

quintessential subjective experience, and yet we do not consider this
awareness to be merely an intellectual construct or abstraction - it is
a real and direct experience." (Grady in preparation: Chapter 5, 22).
In a detailed study of the conceptual system for time, Evans
(2000) argues that the concept 'time', which has previously been as-
sumed to be a parade example of an 'abstract' concept, in the sense
of an intellectual construct, can be traced to physiological mecha-
nisms and processes. This work offers support for Grady's suggestion
that 'abstract' concepts derive from frequently experienced internal
states which humans are aware of at the level of phenomenological
experience. As such, these experiences would seem to be just as basic
as those arising from external sensorimotor input. Thus, there appears
to be little justification for distinguishing many concepts based on
their level of abstractness (Evans 2000, Grady 1997, in preparation).
In addition, it is also worth reminding ourselves that even senso-
rimotor experience of the world 'out there' is mediated by our par-
ticular physiology and neurological architecture. As the develop-
mental psychologist Jean Mandler (1992) points out, information
arising from observation and experience of the real world does not
directly enter into consciousness in an unmediated form. Even the
most basic concepts are represented in the human conceptual system
as redescriptions and only then are accessible to conceptual structure.
That is, sensorimotor experience itself cannot be assumed to be 'di-
rect'.
Accordingly, following Grady, we suggest that the bifurcation in
the nature of concepts has little to do with whether a concept is ab-
stract or concrete, but rather is determined by what kind of informa-
tion the concept represents a redescription of. In essence, a more ac-
curate distinction may be that while image concepts represent rede-
scriptions of 'external' i.e. sensorimotor experience, response con-
cepts represent redescriptions of internal states (cf. Evans 2000).
Perceptual information derived from external sensorimotor infor-
mation and redescribed into a format accessible to the conceptual
system we term image content. This represents substrate available at
the conceptual level which serves to derive image concepts, such as
motion, heat, etc. Perceptual information derived from internal states

(also redescribed into a format accessible to the conceptual system)
gives rise to what we term, following Grady (1997), response con-
tent. Response content gives rise to response concepts. Response
concepts derive from internal body states. In some cases, these inter-
nal states themselves may ultimately be responses to external infor-
mation, hence the term response. For instance, love refers to an inter-
nal state, yet at the same time constitutes an emotional response to
another being in the external world.
What is clear from the foregoing is that while response concepts
lack the 'objectivity' of image concepts, such as those pertaining to
physical proximity or physical distance (deriving from sensorimotor
information), they are no less basic (Grady 1997). As concepts such
as time-reference (in the sense of temporal deixis rather than clock-
time), intimacy, salience, actuality and attenuation are internal in
nature and hence subjective, rather than being external and inter-
subjective, it follows that they constitute response concepts.
The next question to consider is why it should be that response
concepts tend to be elaborated in terms of image content. One possi-
ble explanation comes from the variability across individuals and the
non-verifiability inherent in internal states.3 Perhaps because we do
not have access to each other's minds, when we attempt to external-
ize information about our internal states through language, humans
have resorted to talking about such internal states in terms of experi-
ences which are verifiable, inter-subjective and so consistent across
individuals. The issue may not be that the internal experience is less
direct or less fully experienced than the sensorimotor experience for
the experiencer/speaker but that the speaker's internal experience is
less direct for the listener.
A second explanation for why response concepts tend to be elabo-
rated in terms of image concepts, may be because the parts of the
brain which process response information may be less well connected
to the conceptual system (the repository of concepts which language
externalizes) than the parts of the brain which process image infor-
mation (Evans 2000, Grady 1997, cf. Jackendoff 1987, 1992). In
evolutionary terms it makes sense that the conceptual system should
have better access (in terms of better neurological connections) to

sensorimotor information than to response information. In terms of
ecological viability, it is more pressing that external information be
more readily available to the conceptual system (and hence con-
sciousness) than response information (cf. Edelman's 1992 sugges-
tions regarding the evolution of consciousness). Since response con-
tent may be only partially accessible to the conceptual system, re-
sponse concepts may opportunistically appropriate image content in
order to develop structure accessible (i.e. "visible") to the conceptual
system, and hence available for encoding by language (Evans 2000,
Grady 1997).
On this view, it is due to the elaboration of a response concept in
terms of image content that a concept such as time-reference (exter-
nalized by the tense morphemes [0/s] and [Id]) is elaborated in terms
of locational content pertaining to proximal-distal relations, rather
than in its own terms (whatever such terms might be). However, this
does not deny that at least some of the content which elaborates a
response concept is response content, as attested by the fact that we
intuitively "know" what it is that tense symbolizes, without being
able to adequately express this without the assistance of image con-
tent such as time-lines etc. (cf. Comrie's 1985 explanation of tense
for instance, which appeals to notions such as a time-line in order to
explain time-reference phenomena). Hence, the hypothesis that this
response content is only partially accessible to the conceptual system
explains 1) why response concepts such as time-reference are diffi-
cult to define, and 2) why response concepts tend to be expressed
linguistically in terms of image concepts such as 'distance'.
We now turn to a consideration of the mechanism which serves to
elaborate response concepts in terms of image content.
3.2. Experiential correlation
One of the remarkable insights to have emerged from cognitive lin-

guistics has been the realization that conceptual structure is largely
organized in terms of substrate deriving from external sensorimotor
experience, which we are identifying as image content. However, it

remains to be explained what mechanism motivates response con-
cepts to appropriate certain kinds of image content and not others.
A promising candidate mechanism is experiential correlation,
which has been studied in detail by Grady (1997, 1999a, 1999b, in
preparation). Grady notes that a consequence of the nature of the in-
teraction between humans and their environment is that certain kinds
of experiences are frequently correlated. For instance, a common,
recurring experience in the world is the correlation between the verti-
cal elevation of a physical entity and an increase in the quantity of the
entity, which is to say when there is an increase in vertical elevation,
there is frequently a correlative increase in the original amount of the
entity. By way of illustration, if there are two boxes stacked one on
top of the other and a delivery person adds another two boxes to the
stack, the height of the stack increases. Hence, height (vertical eleva-
tion) and number of boxes (quantity) are correlated in our experience.
Similarly, if there is a certain amount of liquid in a container, and
more liquid is added, the level of the liquid rises. So, humans fre-
quently experience greater quantity in terms of an increase in vertical
elevation.
Grady argues that tight and recurring correlations in experience
motivate associations at the conceptual level. Hence, as quantity and
vertical elevation are correlated in experience, they come to be linked
at the conceptual level, as attested by the example in (6):
(6) Prices have gone up recently.
This sentence has a conventional interpretation in which prices have

increased. Yet, this reading is achieved by utilizing the linguistic
prompt gone up. The point is that language systematically utilizes
expressions, which conventionally denote vertical elevation to pro-
vide an interpretation of greater quantity.
We suggest that correlations in experience serve to relate certain
kinds of image content with specific response concepts. Kurath
(1921, cited in Sweetser 1990) studied Indo-European words for
emotions and noted that they are often derived from the physical ac-
tions or sensations accompanying relevant emotions. Kurath attrib-

uted this historical development "to the inseparability of physical
sensation from emotional reaction, or of emotional state from con-
comitant physical changes" (Sweetser 1990: 29). Putting this in pres-
ent terms, the response concept 'anger' is elaborated in terms of im-
age content pertaining to heat, precisely because anger and a physical
sense of feeling hot (caused by increased adrenaline and increased
heart rate) form a tight correlation in experience.4
4. Concepts elaborated by proximal-distal content
During the course of this paper, we will argue that time-reference, the
primary meaning associated with tense forms, comes to implicate
non-temporal meanings, which in turn through entrenchment (a pro-
cess we term pragmatic strengthening) comes to be conventionally
associated with the tense morphemes. However, in order to able to
offer a motivated account for the association of non-temporal mean-
ings with the two tense morphemes, we need to establish why it
should be that time-reference should implicate intimacy, salience,
actuality and attenuation in the first place. We will argue that all five
of these response concepts are elaborated in terms of locational con-
tent pertaining to the relative physical proximity of the experiencer
because we conceptualize each of the concepts of time-reference,
intimacy, salience, actuality and attenuation in terms of proximal-
distal relations with respect to the experiencer. Hence, these distinct
concepts, while not literally being spatial concepts are all elaborated
in terms of spatial deixis, which, as we will demonstrate, is motivated
in each instance by a distinct, tight correlation in experience. It is by
virtue of being elaborated in terms of similar image content, we will
suggest, that the tense morphemes, which denote time-reference, can
come to implicate and ultimately denote non-temporal meanings.
In the remainder of this section, we examine the evidence for ex-
periential correlations which give rise to each of these distinct con-
cepts being associated with the proximal-distal dimension. We will
also illustrate how these experiential correlations are reflected in non-

temporal language patterns.
4.1. Time-reference
As noted by a number of scholars (e.g. Comrie 1985, Rauh 1983,

Traugott 1975, 1978) tense is a deictic phenomenon which signals
time reference with respect to a reference point or deictic center. The
reference point typically assumed is coding time, or time of speaking.
Coding time is a temporal concept, which in principle is distinct from
spatial image-based content. However, it quickly becomes clear, at
least for English speakers, that it is virtually impossible to conceptu-
alize temporal deixis without appealing to spatial image content. This
is implicit even in formal definitions of tense. For instance, Comrie
(1985) characterizes tense as "the grammaticalisation of location in
time" (ibid.: 1). Given that location is a spatial notion and hence
pertains to image content, by virtue of offering a definition of time-
reference in such terms, it is evident how deep-seated the conceptu-
alization of time-reference is in terms of the locational information.
Grady (1997) has noted that in experiential terms there is a tight
correlation between the temporal concept of 'now' and the particular
physical location, which is proximal to the human experiencer, i.e.
'here'. In other words, we cannot help but experience the present
moment in terms of our immediate physical surroundings and our
sensory perceptions of them. This tight correlation in experience
between the present moment and the particular location we happen to
occupy, motivates the elaboration of the concept identified by the
lexeme now in terms of our experience of our physical location and
the vicinity proximal to us. Extending this insight, we note that 'not
now' is similarly experienced (through memory or imagination) as a
location other than the immediate 'here'. In other words, 'not now' is
experienced as 'not here.'
A second experiential source linking time and space involves the
human experience of getting from point A to point B. Traversing a
certain distance inevitably correlates with the elapse of a certain
amount of time. Thus, elements of the spatial domain, such as

movement from one location to another and distance, have become
strongly associated with the elapse of time. This is reflected in many
aspects of language. For instance, it is not uncommon to hear ex-
changes along the lines of the following:
(7) A. How far is the restaurant from here?

B. Oh, about a five minute walk.
Given that tense morphemes typically obtain their time reference

with respect to the temporal 'now' (the coding time), which is elabo-
rated at the conceptual level in terms of locational image content, and
the experiential correlation between traversing a particular distance
and the elapse of time which co-occurs with the traversing, it is
hardly surprising that we should find that time-reference is conceptu-
alized and lexicalized in terms of physical proximal-distal relations
with respect to the experiencer. It is this elaboration, we suggest,
which may be largely responsible for our ability to conceptualize
temporal events in terms of physical location and distance, as attested
by expressions such as: Christmas is getting close; The present mo-
ment has arrived, etc. (cf. Evans 2000, Grady 1997, Moore 2000).
Moreover, expressions such as the near past, the distant past, etc.,
which employ the spatial language near and distant in order to signal
time-reference with respect to coding time, illustrate that time-
reference is elaborated in terms of physical proximal-distal relations
vis-à-vis the experiencer.
In many ways, this account of the cognitive relationship between
time and space may seem very like that for tense articulated by Dir-
ven and Radden (2000), and Langacker (1991), and indeed the more
general account of the concept of time offered by Lakoff and Johnson
(1999). We readily acknowledge that the present account builds on
those earlier ones. However, we believe the explanatory value pro-
vided by the notion of experiential correlation adds substantially to
our understanding of how humans come to understand one concept
from a distinct domain of experience, in terms of another, and begins
to provide the illusive explanation for the apparent contradiction
between the primary sense associated with the tense morphemes and
their non-temporal uses.
4.2. Intimacy
Due to the nature of human interaction, there is a tight and recurring

correlation in experience between intimacy and physical proximal-
distal relations. In physical terms, two people cannot be intimate, e.g.
touch, kiss, have face-to-face conversations, etc., unless they are
physically proximal. In terms of familial relations, which are typi-
cally considered to be more intimate than the relationship between
acquaintances, family members tend to spend a greater proportion of
their time in physical proximity with each other than with acquain-
tances or casual friends. Consequently, there is a tight correlation
between intimacy and physical proximity. We suggest that due to this
recurring experiential correlation between intimacy and physical
proximity, intimacy (which is a response concept), is elaborated at
the conceptual level in terms of physical proximal-distal relations.
This conceptual elaboration is attested by expressions such as the
following:
(8) Peggy and I have been close for many years, but lately she has
been acting a little distant.
In this sentence, the notion of physical proximity, as denoted by the

word close, has a conventional reading of intimacy and that of physi-
cal distance, as denoted by the word distant, has a reading of lack of
intimacy. In fact, the reading is so highly conventionalized that on
first inspection we might fail to notice that the literal meanings of
close and distant are not of intimacy and non-intimacy but rather of
physical proximity and distance. The point is then, that the nature of
experience, and particularly recurring correlations, gives rise to re-
sponse concepts such as intimacy becoming elaborated at the con-
ceptual level in terms of proximal-distal content. Hence a close rela-
tionship is one which involves intimacy, while a distant relationship

is lacking in intimacy.
4.3. Salience
Salience is another concept elaborated in terms of image content

pertaining to the physical experience of being proximal or distal. Due
to the nature of our sense organs, particularly our eyes, that which is
physically closer to us is more salient, that which is at a distance less
salient. That which is closer tends to be that which is in foveal vision
and more clearly observable, while that which is physically distant
tends to be in peripheral vision and less clearly observable. That
which is closer appears to be relatively larger, that which is distant
appears to be relatively smaller. Entities which are located physically
closer to humans, or events which take place physically closer are
more likely to demand immediate attention than those which are
physically distant. Thus, there is a tight correlation between the sali-
ence of an entity and how close it is to the experiencer.
By way of example, in a situation in which a human can see two
tigers, the tiger which is ten feet away is likely to be more salient
than the one a quarter of a mile away. This tight experiential correla-
tion between salience and physical proximity in human experience is
illustrated by the following:
(9) We have to keep focused on the pressing issues of the day,

those which are close at hand, not some distant threat or pe-
ripheral controversy.
In this sentence, the lexical items focused, pressing and close at

hand, which represent image content pertaining to physical proximity
and distance, are used to indicate which issues the speaker holds to
be most salient; the items distant and peripheral are used to indicate
those which the speaker holds less salient or important.
4.4. Actuality
The next concept we examine is that which we are terming actuality.

By actuality we mean that which a person believes to be objectively
true and reliably known, particularly about the current world-state.
Generally our cognitive commitment to (and hence belief in) the ac-
tuality of something which can be verified perceptually is much
stronger than to something which we cannot verify. Given the
physiological constraints on human sense-perception, that is, given
that the ability of our sense organs allows us to see, hear, smell, etc.
most acutely that which is within an area which is physically proxi-
mal to us, our beliefs about what currently holds in the world, corre-
late tightly with what we have personally experienced, by virtue of
our sense-perceptory apparatus. Given that our sense-perceptory ap-
paratus gathers information about our physical proximity, our actual-
ity correlates with that which is physically proximal. This correlation
is attested by expressions such as:
(10) a. I saw it with my own eyes.

b. I know someone was in the house. I heard the sound of foot-
steps with my own ears.
Moreover, it follows that while we can be sure of that which is veri-

fiable by our own senses, we are less sure of that which is not avail-
able to our own senses. Thus, we are cognitively committed to what
is proximal and physically verifiable and we conceptualize these en-
tities and events as constituting our actuality; we are much less com-
mitted to the actuality of that which is distant and not physically veri-
fiable. In view of the foregoing we suggest that actuality is elaborated
at the conceptual level in terms of content pertaining to that which is
physically proximal to the experiencer.
4.5. Attenuation
The relation which concerns us here is that between attenuation of

invitations, requests, suggestions, etc., on the one hand, and proxi-
mal-distal relations between the speaker and the addressee on the
other. Attenuation represents a conventionalized linguistic means of
mitigating the imposition placed on the addressee by a request, ques-
tion, etc., and a means of mitigating the potential threat to face on the
speaker, should the hearer fail or refuse to comply (cf. Brown and
Levinson 1987). Like the other concepts considered in this paper,
attenuation can be elaborated at the conceptual level in terms of
physical proximity-distance.
Talmy (1988) noted that real world force dynamics, in terms of
barriers and forces needed to overcome those barriers, provides a
powerful model for our understanding of much of the rest of our ex-
perience, including social interactions. Behind our understanding that
certain linguistic acts (such as requests, commands, etc.) place impo-
sitions on the addressee, is a complex set of notions involving
authority, status, desire to keep social relations in balance, and desire
to be seen as cooperative, which are themselves largely seen in terms
of force dynamics. Sweetser (1990) argues that language allows us to
affect one another's actions without having to rely on physical force.
"[Language] is ... our major means of intellectual and emotional in-
fluence on each other. As linguistically capable human beings, we
have no need to constantly resort to physical pushes and pulls to in-
fluence other speakers of our language; we can do so in a far more
sophisticated and effective manner via the vocal organs and the
auditory sense-channel" (ibid.: 41). She adds, "In the real world, we
don't usually use force unless we need to overcome reluctance on the
part of the person we are forcing. ... In the real world, force is usu-
ally resented by the victim because freedom is valued" (ibid.: 61). In
sum, Sweetser argues that consideration of socio-physical force lies
behind many speech acts and much of politeness phenomena.
In terms of politeness specifically, Sweetser notes that it is less
face-threatening to enable your interlocutor to be cooperative rather
than to evoke your restrictive powers of authority. Drawing on Prin-
ciple 3 from R. Lakoff s (1973) discussion of politeness: "Make the

interlocutor feel good. Be friendly" - Sweetser notes that "[this] goal
[is] best achieved by minimal exercise of overt authority." (ibid.:
153) One important way in which speakers give the appearance of
lessening their overt power and simultaneously giving their addressee
the option to be cooperative (R. Lakoff s 2nd principle "Give op-
tions"), is to linguistically implicate that the speaker is not in the po-
sition to force compliance. English has developed a number of con-
ventional ways of enacting such mitigation. We believe that one of
the most pervasive has to do with implicating that the speaker is
physically distant from the addressee.
In humans' everyday experience, there is a tight correlation be-
tween physical proximity and one's ability to affect an entity. For
instance, if the experiencer is proximal to an object he or she can
pick the object up, scrutinize it and manipulate it. If the experiencer
is located away from the object he or she is unable to affect the object
to the same degree. This experience is reflected in linguistic exam-
ples such as the following:
(11) She kept a tight grip on the budget.
In this sentence, the degree of control over and hence ability to affect
the budget is articulated by the phrase tight grip, which literally de-
notes very close physical contact. Examples such as these are li-
censed by virtue of the experiential correlation between the ability to
affect something and physical proximity. Experience also tells us that
physical distance results in a lessening of the ability to affect a par-
ticular entity. This is reflected in sayings such as:
(12) When the cat's away the mice will play.
Invitations, requests, etc. create a situation in which the addressee's

actions are potentially affected by the speaker. Attenuation represents
an attempt on the part of the speaker to imply a lessening of his/her
ability to affect the addressee. Given that being located away from an
object correlates with the experiencer being unable to affect the ob-
ject, we suggest that attenuation is elaborated in terms of locational

content in which the speaker is physically distal with respect to the
entity being affected, i.e. the addressee.
5. Accounting for non-temporal uses of tense
To this point, we have argued that the concepts time-reference, inti-

macy, salience, actuality and attenuation represent response concepts,
in part elaborated in terms of response content. As response content
derives from internal states, it is less accessible to the conceptual
system than externally derived image-based content. Hence, response
concepts appropriate image based-content in order to elaborate them-
selves. Following Grady (1997) we suggest they do so by exploiting
naturally occurring and ubiquitous correlations in experience. Each of
the response concepts dealt with correlates with proximal-distal con-
tent with respect to the experiencer. Hence, each of these concepts
shares similar image content, as depicted in Figure 1.
Response Content Image Content
Response Concepts Image Concepts
Legend: 1: time-reference; 2: intimacy; 3: salience; 4: actuality; 5: attenuation;

6: proximity; 7: distance
Figure 1. The elaboration of response concepts in terms of image content
The unshaded spheres on the left correspond to the response con-

cepts: time-reference, intimacy, salience, actuality and attenuation
respectively. The two shaded spheres on the right correspond to

proximity and distance, which represent redescriptions of sensori-
motor information (image content) pertaining to physical proximal-
distal relations. The response content is shaded in grey, indicating it
is only partially accessible to the conceptual system, as further indi-
cated by the broken arrows between the response content and the
response concepts. Accordingly, image content (shaded in black)
which is wholly accessible to the conceptual system is appropriated
so as to make the response concepts accessible to the conceptual
system. As noted earlier, in this paper we are dealing with the two
bound tense morphemes in English namely [0/s] and [Id], which cor-
respond to present and past tense respectively. Such morphemes con-
stitute form-meaning pairings. Given the experiential correlation
between the temporal reference point, i.e. the experiencer's aware-
ness of now, and the location occupied by and proximal to the expe-
riencer, time-reference is elaborated in terms of proximal-distal con-
tent, such that the present tense lexicalized by [0/s] is elaborated in
terms of locational content proximal to the experience, and the past
tense, lexicalized by [Id] is elaborated in terms of locational content
distal with respect to the experiencer.
We have also argued that other concepts such as intimacy, sali-
ence, actuality, and attenuation are elaborated in terms of similar lo-
cational content. Accordingly, we hypothesize that due to the parallel
conceptual elaborations, in certain contexts the use of tense to signal
time-reference can implicate a distinct concept, such as intimacy for
instance. Through continued use, we suggest that tense morphology
has become associated with the additional meaning, a process we
term pragmatic strengthening, such that a new meaning component
becomes associated with the tense morphemes.
5.1. Tense and intimacy
Now we return to the examples of non-temporal uses of tense in (2)

through (5) with which we began our discussion. Turning first to
intimacy, we have argued that intimacy is elaborated at the concep-
tual level in terms of locational content proximal to the experiencer.

This is due, we suggested, to the tight correlation in experience be-
tween intimacy and physical proximity. Now let us examine the ex-
ample in (2) reproduced below:
(2) a. A: Jane just bought a Volvo.

B: Maureen has one.
A: John, you've got to quit talking about Maureen as if
you're still going together. You broke up three months
ago. (Riddle 1986)
In this example the use of the present tense form has by B, is inter-
preted by A as a claim of (unwarranted) intimacy. John does not have
direct knowledge as to whether Maureen still owns a Volvo. He is
speaking as if they are in an on-going relationship which would give
him that knowledge. As such, in examples such as this the use of
tense provides an intimacy reading. The point is that the tight corre-
lation between intimacy and physical proximity is a corollary of the
correlation between time-reference and physical proximity. Being
temporally "located" provides immediate and verifiable experience.
As such, knowing that Maureen has a Volvo in the present represents
a stronger claim to intimacy than knowing that at some point in the
past she had one.
Due to an intimacy reading having become conventionally associ-
ated with tense markers, we suggest that A is able to interpret the
tense usage not in terms of time-reference, but rather in terms of in-
timacy, due to tense being conventionally employed in certain situa-
tions to signal relative intimacy.
A second example illustrates the use of past tense to signal lack of
emotional intimacy:
(2) b. My daughter's father was Brazilian. He stays in contact with

Suzanna, but I haven't seen him in years. (Tyler, personal
data)
In this case, the speaker presents information about her daughter's

father in the past tense. Out of context, this sentence is ambiguous
and could suggest that the father is no longer alive. However, the
additional information, that the father and the daughter continue to
see each other, rules out the interpretation that the father is dead.
Given our knowledge of the world, we can also rule out the interpre-
tation that he is no longer Brazilian, as one's country of origin typi-
cally does not change. Finally, we can reject the interpretation that
the man from Brazil is no longer Suzanna's father since biological
fatherhood is unchangeable and he seems to continue, at least in
some aspects, in the social role of father. We conclude that the
speaker used the past tense to signal her own psychologi-
cal/emotional attitude of non-intimacy towards her daughter's father.
We hypothesize that intimacy has become associated with tense
for the following reason. In certain situations, such as when talking
about human relationships, as in (2a-b), relative intimacy is impli-
cated. As time-reference and intimacy share similar image content,
tense can become reanalyzed as the linguistic component in the sen-
tence signaling the intimacy relation. Through continued usage of
tense in situations in which intimacy is implicated tense develops
intimacy as an additional meaning component, which can become
instantiated in semantic memory along with the time-reference func-
tion. This process of conventionalization we term pragmatic
strengthening.
5.2. Tense and salience
We find an analogous recruitment of English tense to indicate sali-

ence (i.e. whether information is being foregrounded or background-
ed). We illustrated this phenomenon with the example in (3a-c) re-
produced below:
(3) a. In November 1859, Charles Darwin's The Origin of Species,

one of the greatest and most controversial works in the lit-
erature of science, was published in London.
b. The central idea in this book is the principle of natural se-

lection
c. In the sixth edition... Darwin wrote: "This principle of pres-
ervation of the survival of the fittest, I have called Natural
Selection." (Eigen and Winkler 1983: 53)
The main point of these sentences is not the precise date of publica-
tion of this book, but rather the central topic of the book. The infor-
mation presented in the first sentence establishes the frame for the
focal information which occurs in the second sentence. The informa-
tion in (3 c) provides supporting evidence for the key point, and as
such is less prominent in terms of information status. Discourse ana-
lysts have often referred to the relative status of information such as
that in sentence (3 a) and (3 c) as background and information such as
that in sentence (3b) as foreground. In this example, tense is em-
ployed to signal the relative status of the information, i.e. past tense
signals background status and present tense signals foreground status.
As before, we suggest that tense comes to signal salience for the
following reason. In certain contexts, when, for instance evaluating
the relative importance of information, some pieces of information
are implicitly more or less important than others. As time-reference
correlates with physical proximity, so too degree of salience corre-
lates with that which, as noted earlier, tends to be in foveal vision and
thus physically proximal. Accordingly, as time-reference shares
similar image content with salience, tense can in some contexts im-
plicate relative salience. We suggest that through continued use of
tense in contexts in which salience is implicated, namely pragmatic
strengthening, tense morphology has developed a conventionalized
meaning component, in which past tense denotes relatively less sali-
ence while present tense denotes greater salience.
This use of past and present tense in order to signal relative sali-
ence relates to Langacker's (this volume) arguments concerning non-
canonical uses of the present tense. He discusses vivid narrative,
play-by-play sports casting, historical present, and other discourse
uses of present tense which clearly do not conform to the typical rep-
resentation of present tense usage. It is not entirely clear how his ex-
planation of non-canonical uses of the present tense as coding special

viewing arrangements involving "virtual" events and texts might
apply to the particular uses of present tense morphology we are con-
sidering. However, the use of the present tense to make a particular
event more 'real' would seem to be related to the use of the present
tense to denote greater salience and hence importance in terms of
information structure.
5.3. Tense and actuality
As we have previously observed, tense can be employed to signal

actuality (the degree to which the experiencer believes the event de-
scribed matches the current or actual world state). Let us reconsider
example (4c), reproduced below:
(4) c. Suppose your house burned down. Do you have enough

insurance?
Normal interpretation of this sentence is that the speaker is hypothe-

sizing about an event which the speaker does not believe to have
taken place, i.e. the speaker is not referring to an actual event in
which the addressee's house burned down at some time before the
moment of speaking. The fact that the speaker does not believe that
the house concerned has actually burned down is signaled by the lex-
eme suppose.5 In addition, in this example past tense seems to under-
score that the speaker does not believe that your house burned down
references a situation holding in the world. That is, the use of past
tense cannot be interpreted as referring to a past event, but rather
signals a lack of commitment to the actuality of the situation.
While English speakers regularly use the past tense in hypothetical
constructions, they also have the option of using the present tense.
So, Suppose your house burns down is also perfectly acceptable.
Moreover, the choice of present tense would not entail that the
speaker believes that the house is in fact burning down at the moment
of speaking. Rather the choice between past and present tense seems
to signal a somewhat subtle shift in the speaker's commitment to the

probability that the hypothetical event will happen. In cases such as
these, the choice of the present tense seems to signal a stronger
commitment on the speaker's part to the possibility of the hypotheti-
cal state of affairs occurring.
In terms of example (4c), without further context or access to the
speaker, it is difficult to say with confidence why the speaker chose
the past tense rather than the present. Given the language, this sounds
like a sales pitch for insurance. It may be that the speaker was con-
cerned about sounding overly aggressive and, not wanting to have his
or her arguments dismissed out of hand as mere 'scare tactics', chose
the past tense in order to soften the message.
The point is that in these constructions, neither past nor present
tense is interpreted in its canonical temporal meaning as coding in-
formation about events, which the conceptualizer believes to be real.
Again, we suggest that as time-reference (signaled by tense) and ac-
tuality are both elaborated in terms of similar image content, i.e. the
present correlates with the physically proximal, as does the notion of
actuality, in situations in which a stronger or weaker commitment to
the actuality of the event described is implicated, tense is reanalyzed
as a lexical form contributing to this implicature. Through continued
use of tense in such contexts, this meaning comes to be convention-
ally associated with the tense morphemes, such that the present tense
form [0/s] signals a stronger commitment to the potential actuality of
the situation, while the past tense form [Id] signals a weaker com-
mitment.
5.4. Tense and attenuation
We now turn to the final non-temporal meaning of tense, which we

will address in this paper. Consider the following example, which is
adapted from (5 c) above.
(13) I wanted to ask you a question.

In sentences such as (13) we conventionally understand that the use

of past tense does not place the desire to ask the question in the past,
but rather that it attenuates and so makes such requests less face-
threatening and hence more polite.
We noted earlier that attenuation - which constitutes a linguistic
lessening of the speaker's assertion of authority and an offering of
options to one's addressee to cooperatively be affected by the
speaker's request, hence lessening the face-threat involved - corre-
lates with not being physically proximal to the addressee. That is,
attenuation can be elaborated at the conceptual level in terms of
proximal-distal locational content, the same image content used to
elaborate time-reference. Put another way, while being in past time-
reference correlates with not being physically proximal with the cur-
rent location, so too being polite correlates with not being physically
proximal. After all, very close proximity can be perceived as being
overly assertive and/or aggressive. This may be related to physical
proximity being a pre-requisite for physical control. As the image
content which serves to elaborate the concepts time-reference and
politeness is closely related, we suggest situations in which politeness
(and hence attenuation) are implicated, tense has come to be reana-
lyzed as the lexical means of signaling attenuation.
5.5. Further issues
In view of the foregoing it is now apposite to ask why it should be

tense, rather than some other linguistic form, which has become as-
sociated in English with the non-temporal meanings described in the
foregoing. Indeed, tense appears to have developed some of the
functions described above in a cross-linguistically robust way (cf.
Fleischman 1989). This situation is even more intriguing as there is
evidence that in some languages some of the meanings described in
the foregoing are lexicalized not by tense but by other closed-class
elements. For instance, in Japanese the demonstrative are 'that',
which locates an object distant from both the speaker and listener,
can serve to express attenuation:
(14) A (male): doo shita no?

how did interrogative marker
'What's wrong?'
Β (female): kon 'na koto iu no are
this such thing say complement that
nan-desu kedo anata kimochi waruin-desu.
be but you creepy be
'It might be that to say this to you, but you're
creepy.' [Mitsuyo Sato]
In this attested example, a female Japanese speaker employs are in

order to attenuate a face-threatening remark. What is interesting is
that the form used is a demonstrative, which locates an object distant
from both the speaker and hearer. Given that Japanese employs a
form with a primary meaning of a distal location in order to lexicalize
attenuation, this pattern supports our suggestion that attenuation is
elaborated at the conceptual level in terms of locational content distal
to the speaker and hearer. Moreover, this pattern coheres with our
explanation for the situation in English. We argued that as past time-
reference is elaborated in terms of locational content which is distant
from the experiencer, the parallel elaboration between attenuation
and past time-reference at the conceptual level means that time-
reference can come to develop an attenuation meaning.
In English (and many other languages) the appropriation of tense
as the relevant marker may simply be opportunistic based on fre-
quency of appearance, i.e. due to usage. As English tense is so fre-
quent, being marked on most verb forms, it is natural, given the par-
allel elaboration in terms of image content between time-reference
and intimacy, salience, actuality and attenuation, that tense should be
re-analyzed as the form which marks these meanings, rather than
another form being developed for this purpose. However, the fre-
quency of tense does not preclude another lexical item being devel-
oped to express these meanings, as we have just seen to be the case in
Japanese.
6. Consequences for language teaching
A number of researchers have noted the difficulty second language

learners regularly encounter with non-temporal uses of English tense.
For instance, Riddle (1986) documented that advanced learners of
English often experience difficulty appropriately interpreting and
producing non-temporal uses of tense in the areas of intimacy and
salience. While advanced learners tend to learn to mitigate requests
and suggestions through the use of the so-called past tense modals
could and would, they experience a good deal of confusion over the
type of attenuated invitations, requests, and suggestions exemplified
in (5) above (Celce-Murcia and Larsen-Freeman 1998). Lock (1996)
notes that uses of tense to indicate actuality presents problems for
learners.
We believe that at least part of this difficulty stems from the way
English tense is represented in contemporary reference grammars and
English as a Second Language (ESL) texts. As Riddle (1986) pointed
out, ESL texts tend to present past tense solely as meaning 'com-
pleted before the time of speaking'; 'exceptional' uses, especially in
the areas of intimacy, saliency, and attenuation, are generally ignored
all together. For instance, while The Collins Cobuild English Gram-
mar lists "vivid narrative" and "firm plans for the future" as the
"other uses of tenses" (ibid.: 257), this information is placed in a us-
age note, several pages removed from the discussion of the primary
uses of past and present tense. Moreover, no attempt is made to ex-
plain the connection between the temporal and extended uses. Their
sections on politeness make no mention of the use of past tense to
convey attenuation. Lock (1996) presents uses of tense in a variety of
extended discourse contexts, but does not mention tense to convey
foreground and background (i.e. salience); neither does he mention
use of past tense to convey attenuation.
When non-temporal uses are presented, for instance in hypotheti-
cal constructions, they tend to be discussed in terms of arbitrary con-
structions to be memorized. For example, from The Collins Cobuild
English Grammar, "When you are talking about an unlikely situa-
tion, you use simple past tense in the conditional clause." (ibid.: 350).
No explanation of why past tense is used in this situation is given.

Similarly, Lock (1996) offers the following rule for unlikely condi-
tions: "The general rule is that a condition is marked as unreal by the
tense of the finite verb group being one step back in the past relative
to its tense in the expression of a real condition..." (ibid.: 199). No
further explanation follows. Finally, Westney (1994) reported that
English teachers and texts often have difficulty in clearly articulating
generalizations which cover many everyday uses of English tense. He
notes that uses involving actuality and attenuated suggestions are
simply unexplained exceptions which render rule-making very diffi-
cult.
In sum, the typical ESL student appears to be presented with a
partial list of uses of English tense, rather than a unified model, along
with the advice that memorization of the arbitrary patterns is the best
policy as there is no systematicity in the various non-temporal uses.
We believe that insights from cognitive linguistics have real merit
in offering more systematic, motivated accounts of how English
works. However, we also believe that simply stating that tense can
have additional usages due to "metaphoric extension" is in itself not
very helpful. As we noted in Section 1.2., simply labeling something
as metaphoric does not necessarily provide a revealing explanation,
unless we actually explain how the information from the two do-
mains come to be associated. We have attempted to show how the
"exceptional" meanings associated with tense are grounded in expe-
rience, by virtue of experiential correlations, and derived as semantic
elements associated with lexical forms, through language use, i.e.
pragmatic strengthening. The nature of meaning extension is itself a
function of treating grammatical elements as meaning-bearing units.
This way of viewing language, we suggest, will be particularly help-
ful for teachers and language learners alike.
Notwithstanding the utility and, we suggest, the plausibility of the
present analysis, it is obvious that the details and concomitant com-
plexity of the discussion presented here would be largely inappropri-
ate for language learners or even many language teachers.
Clearly further research is required in order to develop materials,
based upon the foregoing proposals, which teachers could present in
a classroom setting. Such must be the aim of cognitively oriented

applied linguists.
7. Conclusion
In this paper we have argued that the primary meaning of tense is

temporal reference, which is elaborated at the conceptual level in
terms of spatial proximal-distal relations. The elaboration of this par-
ticular image content is motivated, we suggested, by experiential
correlation. Due to the nature of tight correlations in experience,
other response concepts such as intimacy, salience, actuality and at-
tenuation are also elaborated in terms of similar image content. In
certain contexts where tense is employed and these meanings are
implicated, tense markers can come to be reanalyzed as the markers
of these non-temporal meanings. This is the result, we have argued,
of the parallel elaboration of time-reference on the one hand and in-
timacy, salience, actuality and attenuation on the other, in terms of
similar image content. Through continued use, these meanings can
come to be conventionally associated with the tense markers [0] and
[Id], a process we have termed pragmatic strengthening.
In terms of language teaching, this account has great utility as it
provides a unified account of tense phenomena. Hence, it would be
more teachable and coherent than accounts that assume that non-
temporal meanings are arbitrarily related to the temporal reference
meaning of tense.
Notes
1. Riddle's point is dramatically underscored by the presentation of tense in

Collins Cobuild English Grammar (1990), which specifically purports to be a
functional grammar aimed at second language learners with the intent of "con-
centrating on the real patterns of use in today's English." (back cover). No
mention is made of the use of tense in relation to politeness in invitations, re-
quests, or suggestions (ibid.: 204-206; 228-232). The exception to this silence
is in a section 'Expressions used instead of Modals' where we find the state-
ment, "You can use 'want' instead of 'would like' to give an instruction or
make a request... 'Wanted' is also sometimes used. It is more polite than
'want.' (ibid.: 240) No further explanation is provided. Similarly, in the section
on expressing importance, there is no discussion of tense as a signal of the rela-
tive importance of information within the discourse (ibid.: 236-237; 257); the
only remotely possible mention is in a usage note associated with the present
tense under the rubric of 'Vivid narrative' (ibid.: 257). This deals only with per-
sonal narratives coded exclusively with the present tense; present tense is repre-
sented as a device to increase audience involvement. The only example involv-
ing tense given in relation to actuality is the use of the present tense in the fol-
lowing, "Suppose we don't say a word and somebody else finds out about it."
Although native speakers also use the past tense in such structures, this use is
not mentioned This grammar also clearly exemplifies the tendency to scatter
non-temporal uses of tense, essentially representing them as arbitrary, and of-
fering no attempt at a unified, systematic account that would tie them to the
central temporal sense.
2. Specifically, Langacker argues that immediate reality is marked by the absence
of a modal and the absence of the distal morpheme; non-immediate reality is
marked by the absence of a modal and the presence of the distal morpheme. The
immediate/non-immediate contrast is argued to be "a proximal/distal contrast in
the epistemic sphere." (ibid.: 245) "Immediate reality coincides temporally with
the time of speaking, so to the extent that the notion of time is specifically in-
voked, present time is conceived as one facet of immediate reality... In precisely
analogous fashion, the predication of non-immediate reality is equivalent to one
of past time...These notions are basically epistemic, i.e. they do not refer to
time, yet they have an obvious interpretation with reference to the time-line
model: since reality subsumes the past and present (but not the future), and im-
mediate reality constitutes the present, the temporal projection of non-
immediate reality can only be the past. Presumably, then, the distal morpheme
has a prototypical value that invokes the time-line model and is reasonably con-
sidered a past-tense predication. That, however, is only one manifestation of its
basic epistemic import." (ibid.: 246).
3.Sweetser (1990) notes that experiences which have been labeled concrete tend
to be those which allow continuity across individuals. For instance, "vision
is...identical for different people - that is to say, two people who stand in the
same place are generally understood to see the same thing ... Identity across
people is a highly objective characteristic..." (ibid.: 39) In contrast, "abstract"
concepts tend to be identified with internal states. Since human beings cannot
communicate through mental telepathy, one human being cannot directly ob-
serve or know another person's mental or emotional state; we can only guess.
Internal states simply do not allow for verifiable identity across people in the
same way that entities and events in the external world do. Moreover, internal
states and subjective reactions are proverbially variable across individuals. We

hypothesize that in order to communicate more effectively with their interlocu-
tors about internal states, speakers have come to use concepts from the more
observable, objectively agreed upon (intersubjective) world to discuss the unob-
servable (subjective) (cf. Grady 1997).
4. It is interesting to note that evidence is beginning to emerge that correlation
may be a fundamental operation at all levels of cognitive processing. For in-
stance, at the neurological level, integration or "binding" of perceptual infor-
mation which is spatially distributed in the brain in order to form a coherent
percept, may result from the correlated firing of the relevant neurons (Crick
1994, Crick and Koch 1990, 1998, Pöppel 1994, Stryker 1991). That is, the
particular neurons associated with the sensory qualities constituting the percep-
tion of an object fire in correlated fashion. This synchronous firing serves to
integrate the various spatially-distributed sensory qualities into a coherent per-
cept, without requiring that the information be transmitted to and hence inte-
grated at a single site in the brain.
5. In Fauconnier's (1994, 1997) terms, suppose is a space builder, which signals a
hypothetical space.
References
Binnick, Robert
1991 Time and the Verb. New York: Oxford University Press.
Brown, Penelope and Stephen Levinson
1987 Some Universals in Language Use. New York: Cambridge Uni-
versity Press.
Celce-Murcia, Marianne and Diane Larsen-Freeman
1998 The Grammar Book. Rawley, MA: Newbury House.
Collins Cobuild English Grammar
1994 London and Glasgow: Collins Publishers.
Comrie, Bernard
1985 Tense. Cambridge: Cambridge University Press.
Crick, Francis
1994 The Astonishing Hypothesis. New York: Simon Schuster.
Crick, Francis and Christof Koch
1990 Towards a neurobiological theory of consciousness. Seminars in
the Neurosciences 2:263-275.
1998 Consciousness and neuroscience. Cerebral Cortex 8: 97-107.
Cutrer, Michele
1994 Time and Tense in Narrative and in Everyday Language. USCD
Dept. of Cognitive Science: Technical Report 9501.
Dirven, René and Günter Radden

2000 A cognitive linguistics grammar of English. Manuscript.
Edelman, Gerald
1992 Bright Air, Brilliant Fire. New York: Basic books.
Eigen, M. and R. Winkler
1983 Laws of the Game: How the Principles of Nature Govern
Chance. New York: Alfred A. Krapf.
Evans, Vyvyan
2000 The Structure of Time: Language, meaning and temporal cogni-
tion. Doctoral thesis. Linguistics dept., Georgetown University.
Fauconnier, Gilles
1994 Mental spaces. Cambridge: Cambridge University Press.
1997 Mappings in Thought and Language. Cambridge: Cambridge
University Press.
Fillmore, Charles
1990 Epistemic stance and grammatical form in English conditional
sentences. Papers from the Twenty-Sixth Regional Meeting of the
Chicago Linguistics Society, 137-162. Chicago: University of
Chicago.
Fleischman, Suzanne
1989 Temporal distance: A basic linguistic metaphor. Studies in Lan-
guage 13(1): 1-50.
Frank, Marcella
1983 Modern English: A Practical Reference Guide. Englewood
Cliffs, NJ: Prentice-Hall.
Grady, Joseph
1997 Foundations of meaning: Primary metaphors and primary scenes.
UC Berkeley: Doctoral Thesis
1999a A typology of motivation for conceptual metaphor: Correlation
versus resemblance. In: G. Steen and R. Gibbs (eds.), Metaphor
in Cognitive Linguistics. Philadelphia: John Benjamins.
1999b Crosslinguistic regularities in metaphorical extension. Paper pre-
sented at the Annual Meeting of the Linguistic Society of Amer-
ica, Los Angeles, January, 1999.
In prep. Foundations of meaning. (Manuscript) Krasnow Institute for Ad-
vanced Study, George Mason University.
Hopper, Paul and Elizabeth Closs Traugott
1993 Grammaticalization. Cambridge: Cambridge University Press.
Jackendoff, Ray
1987 Consciousness and the Computational Mind. Cambridge, MA.:
MIT Press.
1992 Languages of the Mind. Cambridge, MA.: MIT Press.
Jespersen, Otto
1924 The Philosophy of Grammar. London: Allen and Unwin.
Lakoff, George
1987 Women, Fire and Dangerous Things. Chicago: Chicago Univer-
sity Press.
1993 The contemporary theory of metaphor. In: Andrew Ortony (ed.),
Metaphor and Thought, 202-251. Cambridge: Cambridge Uni-
versity Press.
Lakoff, George and Mark Johnson
1980 Metaphors We Live by. Chicago: Chicago University Press.
1999 Philosophy in the Flesh. New York: Basic Books.
Lakoff, Robin
1973 The logic of politeness; or minding your p's and q's. CLS 9, 292-
305.
Langacker, Ronald
1987 Foundations of Cognitive Grammar, vol. 1. Stanford: Stanford
University Press.
1991 Foundations of Cognitive Grammar, vol. 2. Stanford: Stanford
University Press
This volume Cognitive linguistics, language pedagogy and the English present
tense.
Lock, Graham
1996 Functional English Grammar: An Introduction For Second Lan-
guage Learners. Cambridge: Cambridge University Press.
Mandler, Jean
1992 How to build a baby: II. Conceptual primitives. Psychological
Review 99(4): 587-604.
Moore, Kevin Ezra
2000 Spatial experience and temporal metaphors in Wolof. Ph.D thesis.
Linguistics dept., University of California at Berkeley.
Ortony, Andrew
1988 Are emotion metaphors conceptual or lexical? Cognition and
Emotion 2: 95-103.
Pöpel, Ernst
1994 Temporal mechanisms in perception. In: O. Sporns and G. To-
noni (eds.), International Review of Neurobiology 37, 185-202.
San Diego, CA.: Academic Press.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik
1973 A Grammar of Contemporary English. New York: Harcourt
Brace Jovanovich Inc.
Rauh, Gisa
1983 Tenses as deictic categories: An analysis of English and German
tenses. In: G. Rauh (ed.), Essays on Deixis, 229-275. Tübingen:
Gunter Narr.
Reichenbach, Hans
1947 Elements of Symbolic Logic. New York: The Free Press.
Riddle, Elizabeth
1986 The meaning and discourse function of the Past Tense in English.
TESOL Quarterly 20: 267-286.
Stryker, Michael
1991 Seeing the whole picture. Current Biology 1 (4): 252-253.
Swan, Michael and Bernard Smith
1987 Learner English. Cambridge: Cambridge University Press.
Sweetser, Eve
1990 From Etymology to Pragmatics: Metaphorical and Cultural As-
pects of Semantic Structure. Cambridge: Cambridge University
Press.
Talmy, Leonard
1988 Force dynamics in language and cognition. Cognitive Science 2:
49-100.
Tannen, Deborah
1993 What's in a frame? Surface evidence for underlying expectation.
In: Deborah Tannen (ed.), Framing in Discourse, 14-56. Oxford:
Oxford University Press.
Ter Meulen, Alice
1995 Representing Time in Natural Language: The Dynamic Interpre-
tation of Tense and Aspect. Cambridge, MA: MIT Press.
Traugott, Elizabeth Closs
1975 Spatial expressions of tense and temporal sequencing: A contri-
bution to the study of semantic fields. Semiotica 15(3): 207-230.
1978 On the expression of spatio-temporal relations in language. In: J.
Greenberg (ed.), Universals of Human Language. Vol. 3: Word
Structure, 369-400. Stanford: Stanford University Press.
1989 On the rise of epistemic meanings in English: An example of
subjectification in semantic change. Language 65: 31-55.
Tyler, Andrea and Vyvyan Evans
Forthc. a Reconsidering prepositional polysemy networks. Language.
Forthc. b Spatial Scenes. New York and Cambridge: Cambridge University
Press.
Westney, Paul
1994 Rules and pedagogical grammar. In: T. Odlin (ed.), Perspectives
on Pedagogical Grammar, 72-96. Cambridge: Cambridge Uni-
versity Press.
Section 2
Facets of prototypes in
grammatical constructions
Grammatical constructions and their discourse
origins: prototype or family resemblance?
Paul J. Hopper
"Language is a form of organized stutter." - Marshall McLuhan
1. Introduction1
Future historians of linguistics will, one might speculate, find the

most significant turning point of the 1990's to be not the develop-
ment of more subtle theories of grammatical structure, whether based
on autonomous linguistic structure or on cognitive schémas, but the
revolution in the nature of linguistic data brought about by massive
storage and high-speed search software. In this respect, linguistics
will be seen to have moved closer to other scientific fields in allow-
ing the nature of its enterprise to be radically affected by changing
technologies. The availability of large corpora as a source of data
coincides with the rejection by many linguists of the exclusive use of
"sentences" devised from intuition (introspection) and with a wide-
spread movement to find motivations for grammatical constructions
in cognitive dispositions and usage. In the present paper I will present
a study of an English construction, and will suggest how observations
about such constructions based on corpus data might lead us to dif-
ferent conclusions about their nature from those based on introspec-
tive data.
For an example of this from syntax, we will consider the case of
pseudocleft sentences. The orthodox construction of pseudocleft as
laid out in the standard reference grammar of English (Quirk, Green-
baum, Leech, and Svartvik 1985: 1387-9) involves "an SVC [=
Subject Verb Complement] sentence with a w/z-nominal clause as
110 Paul J. Hopper
subject or complement". This work illustrates some of the types us-

ing the following sentences:
What you need most is a good rest

What he's done is (to) spoil the whole thing
What John did to his suit was (to) ruin it
What I'm going to do to him is (to) teach him a lesson
What I'm doing is teaching him a lesson
Some of the uncertainty, the tension between what people say and
write and what is "canonical", comes to the surface when the wh-
clause is in the perfect and the resumptive verb is the -en participle
rather than the base form or infinitive, as in ?What he has done is
spoilt the whole thing. Revealingly, the authors suggest that this us-
age is a reduced form of two discrete utterances: "of doubtful accept-
ability...may be an ellipted form of: What he's done is this: he's
spoilt the whole thing".
In the by now considerable literature on pseudoclefts, two themes
stand out. The first is that the function of the pseudocleft sentence is
to background the material in the wh- clause so as to throw certain
constituents of the predicate into contrastive relief. Thus in What my
car needs is a new battery, only the direct object a new battery is
new; the subject my car and the verb needs are both old or "given".
The constituent-focus theory runs through both sentence-level and
discourse studies of pseudoclefts. This distribution is claimed, for
example by Prince (1978), to correspond to contexts in which (in this
hypothetical instance) my car and needs are in some way recoverable
("given", "presupposed" in her terminology) from the previous dis-
course, either directly or inferentially. Similarly, in What John did to
his new suit was ruin it, the agent and the grammatical object are
already known and only the verb is asserted. Such treatments moti-
vate pseudoclefts by reference to the ambient and preceding context,
rather than by reference to upcoming segments. They derive from a
linguistic theory that holds that linguistic forms code current mean-
ings. I will here argue that in spoken discourse pseudoclefts are an-
Grammatical constructions and their discourse origins 111
ticipatory, and serve among other things to alert listeners to an immi-

nent significant segment or to permit speakers a brief period of down
time to prepare the formulation of an utterance.
The second theme that permeates treatments of clefting is that "in-
complete" pseudoclefts are exceptional and to be explained as re-
duced and deviant versions of the full construction. Thus Collins
(1994) gives the following example of an incomplete pseudocleft
from the London-Lund corpus:
(1) now what I'm fascinated is to know that COMMUNISTS (coughs)

and FASCISTS are such bad SHOTS that when they have a battle
between THEMSELVES ALL their MISSILES hit the UNFORTUNATE
Mr. BROOKE, who as I understand is NEITHER (cited by Collins
1994:45. Some transcriptional detail has been omitted; caps as
in original, italics are supplied.)
Fragments such as what I'm fascinated is are said (Collins 1994: 44)
to be "system-deviant". System-deviance, in Collins' view, arises
from processing factors: "There are, broadly speaking, two categories
of incomplete pseudo-clefts and clefts: those whose completion is
prevented by a variety of contextual or processing factors, and ellipti-
cal constructions in which omitted material is textually recoverable."
In either case, the assumption is that the forms are exceptions to the
normal realization of the pseudocleft. Speakers lose their drift or do
not bother to state material that is obvious from the context. The lan-
guage system (the speaker's competence) provides for canonical
pseudoclefts, but speakers mutilate them in attempting to produce
(i.e. perform) them.
In this paper I will suggest that when natural spoken discourse is
considered, it appears that pseudoclefts do not function primarily, or
perhaps at all, to highlight any single identifiable sentence constituent
such as a verb or a noun phrase. If such highlighting occurs, it is a
by-product - an epiphenomenon - of a more rhetorical use of the
pseudocleft: pseudoclefts function in natural discourse to delay an
assertion for any of a number of pragmatic reasons. Furthermore, the
"system-deviant" fragments are just as capable of carrying out this
112 Paul J. Hopper
function as are the more canonical pseudoclefts. Such functional

equivalence suggests that we should hesitate to privilege "complete"
constructions over "incomplete" or "fragmentary" ones, and should
reject such notions as deviance and other terms implying a defi-
ciency.
2. Pseudocleft fragments
In what follows, many examples are considered in which a pseudo-

cleft appears in a fragmentary, "incomplete" form. Of course, as we
have seen, attributing incompleteness to these forms presupposes a
certain view of the relationship between the ideal forms of grammar
and the reality of spoken discourse such that spoken discourse is a
degenerate or deviant partner to a grammatically perfect ideal con-
struction. In order to avoid this presupposition, I will adopt the more
neutral term piece. A pseudocleft piece is minimally a segment of
discourse that begins with what and continues with a verb. The
verb's subject is either the what or some other noun phrase. After
these, there may be other elements (objects and complements). Often,
but again not always, the pseudocleft piece terminates with the cop-
ula is/was. Pseudocleft pieces as described here are not a unitary
schema, but link up in family-resemblance fashion with functionally
similar pieces such as the thing, the one, the only, and several others.
Following the pseudocleft piece there is usually, though not al-
Λ
ways, a continuation that may consist of a full clause:
(2) Anna: But have you discussed the fact that you don't believe
[...] that this person can actually help you with this
person
Mike: Yeah I have Anna. And what they're s what my psy-
chologist has told me is that they will make arrange-
ments for me to see a psychiatrist if they f th if I still
feel this
or a complete predicate:
(3) Yeah. But at the moment I I think it sh should be enhanced

because what it actually does is looks through the whole of the
die dictionary.
But while the continuation of a pseudocleft piece is frequently a

clause or a predicate in a conventional sense, it should not be imag-
ined that we can set up a pseudocleft "construction" such that the
pseudocleft piece and the continuation together form a unit such as a
"sentence". Quite often the pseudocleft piece is not followed by any
continuation at all:
(4) the kind of newspapers that they want erm are are are really in a
sense not er any longer er those down-market tabloids. And so
it's it's b it's a struggle for in what I suppose er [= a?] popula-
tion of about ten million I'm not quite certain myself exactly
what it is but it's a huge population the two er main daily news
er popular daily newspapers can barely sell eight hundred.
And at the other extreme, the continuation may be pursued over a gap
of several subsequent clauses, as in this example:
(5) I mean maybe this is because of the school. I only have them on
a Monday and then I have them next week Monday. Now what
happens is we tried in the past different groups. You know dif-
ferent forms. So you ask somebody from each of these about
three or four different forms Collect the homework for me on
Tuesday...
The event presaged by what happens is is not in the continuation, nor

even in the following clause. In fact it goes considerably beyond the
scope of the given segment, during which time the interlocutor must
patiently wait and listen. Similarly:
(6) ...wood stain almost black. It's really beautiful and it's got all
the cornices and the ceiling rose and high skirtings all original.
Anyway then you go in on this it's like a parquet floor and then
114 Paul J. Hopper
what they've done is the cellar it's got a cellar and you go down
the steps to the cellar but there's like a proper two proper
rooms so on your left you've got a sort of cellar with a quarry
tile.
The follow-up to what they 've done is a general domain of architec-

ture or interior decorating rather than a grammatical complement of
the verb. The pseudocleft here marks an up-coming new segment
introduced by the theme noun cellar. Many other examples of this
kind of thing could be cited.
Observations such as these should caution us against any explana-
tion of pseudoclefts that involves some structural arrangement be-
tween the pseudocleft piece and the continuation, such as "focusing"
on a constituent in the continuation, at least in less formal registers.
There may be such a grammatical relationship in some cases, but
there may not be. If there is a grammatical relationship, it is more
likely to be a coincidental one, or else an interpolation from a more
formal register. For whatever its role in formal and written registers
and in the decontextualized sentences used by some linguists and
philosophers, direct transcriptions of recorded speech point to prag-
matic or rhetorical motivations rather than structural or strictly se-
mantic ones as the functional basis of the English pseudocleft.
While it is possible to identify several such functions, they all de-
rive from a single fact: The pseudocleft works to delay the delivery
of a significant segment of talk. It accomplishes this by adumbrating
(foreshadowing) the continuation in general terms without giving
away the main point:
(7) House to house collections are be going to begin this month in

Hallgreen. What we 're going to do is we're going to target one
area per month.
Here, what we 're going to do is points to an activity as the theme of

the up-coming continuation, with we as the subject and are going to
as the tense-aspect. Nothing else is supplied. This delaying function
permits other useful strategies to be deployed. In addition to the gen-
eral "delaying" function, we may discern from the corpus at least the
following five:
i. To alert the listener that an upcoming utterance is noteworthy

ii. To make an attitudinal comment on an upcoming utterance
iii.To state a general theme for the upcoming utterance
iv. To buy time while alternative wordings are considered
v. To hold the floor pending the upcoming utterance
Usually it is possible to see more than one of these at work, so that

any single example might be used to illustrate several or all of them.
The headings of the following examples are not therefore intended to
suggest that the example is illustrative of only one of them.
i. To alert the listener that an upcoming utterance is noteworthy

A pseudocleft almost always suggests that an upcoming utterance or
segment of an utterance is important, and that the listener should pay
attention:
(8) ...it need to kill something like a thousand times its own size.
There's η there don't seem to be a r- real need. And in defence
I mean what snakes or what animals try <pause> like what
most animals try to do is if they tha have got a poisonous prop-
erty is another animal attacks them they give them er a dose of
venom which will not kill them it will just deter them next
time.
The important point that the speaker wishes to focus attention on is

the fact that most poisonous animals deliver less venom than would
be needed to kill their victim. The build-up to this point includes also
the conditional clause if they have got a poisonous property, revealed
by the inclusion of a second occurrence of is to be part of the pseudo-
cleft piece itself, and the further conditional another animal attacks
them. The listeners are being prepared for the focal segment not only
by the pseudocleft piece but by its extension with a first conditional
116 Paul J. Hopper
and by a second supplementary conditional. By this time they will

have their pencils ready.
ii. To make an attitudinal comment about an up-coming utterance

A considerable number of instances of pseudoclefts include an ex-
pression of the speaker's attitude toward the up-coming statement.
Typical examples include what worries me, what makes me smile,
what amazes us, what saddens me greatly, what makes them exciting:
(9) Joe: Er I mean if if people want to plant bombs there's al-

ways ways of doing it whatever the security.
Rachel: Yeah.
Joe: I mean what is worrying is that it's happening er peo-
ple could get hurt and we don't know who's doing it
really it might be the IRA it might not.
Interjecting a personal stance toward the upcoming segment naturally

increases its prominence and provokes the listener's curiosity. While
(as in this example) there may not be a first-person pronoun in the
pseudocleft piece, attitudinal verbs like worry, amaze, etc. add a
subjective element to the upcoming assertions, implicitly introducing
a first-person reference where there otherwise was none.3 In this
same vein, a number of examples show an interposed I suppose
within the pseudocleft piece that again serves to subjectivize the dis-
course:
(10) Robin: .. .you feel <pause> I mean <pause> the campaign has
been going on in Scotland for longer than in in this
country do you think it's erm a a victory for what I
suppose you ΊI see it as a victory for the people over
government I suppose would you?
Walter: Yeah I do. I mean actually I mean it didn't work in
Scotland from the start and that was like
(11) to to clean up Mexico City air. And obviously this then became
a pretty major objective for himself in terms of his own ambi-
tions and his own presidential er ambitions for er what I sup-

pose must be the ninety-four ninety-four the next election
ninety-four. Erm anyway the erm an emergency program was
started er late I think in nineteen-eighty-nine. Erm I'm certain
two or three
At times, in fact, what I suppose functions as a virtually empty for-

mula that delays the continuation while adding nothing in the way of
independent meaning:
(12) What what I wondering about er dr dr dr drifting away really

from looking at individual ...claims and individual <pause>
erm personal problems to what I suppose were policy is-
sues... and er to do with the area of the er the type of serv-
ice. . .that's offered to claimants
The subjective nature of the pseudoclefi in spoken registers distin-

guishes it from both written registers and from the construction with
which it is often paired, the cleft sentence (see Collins 1994: 103).
iii. To state a general theme for the up-coming utterance

A striking fact about pseudoclefts in natural discourse is that the
verbs in the pseudoclefi are drawn from a very restricted lexical set.
Among the most frequent are do, happen, need, make, use, say (usu-
ally in some form such as what I'm saying), tell, and a few others.
These verbs have in common that they are semantically simple, basic,
and general. They serve to adumbrate the continuation by supplying a
general domain within which the more specific verb of the continua-
tion will fit. Actions are adumbrated with do:
(13) Derek: Thing is though you've got to preserve a a certain sort

of scruffiness t to add to your programmer authentic-
ity.
Deirdre: No j just get an anorak.
Jerry: Well what I thought I might do is erm <laughs>
Derek: <laughs>
118 PaulJ. Hopper
Jerry: when I go into the office I'll nut the first person I see
(14) I mean I think Peggy and Sheila will be interested to hear what
the software actually does here because this is relevant obvi-
ously to them. Erm (...) I mean basically what it does is it di-
vides the the word are defined into various groups.
Events are adumbrated with happen:
(15) Frank: ... she can ring Charles and speak to him.
Lisa: Yes. Erm <pause> try yes try and get her to ring you
know <pause> t over the next day or so. Well I sup-
pose what'll happen is she'll ring and you'll answer
probably and it'll be good if I answer it won't I ooh-er
Happen here, unlike do in the earlier example, covers several single

actions (ring, answer, answer). Its scope therefore extends over more
than one clause, a fact that makes a functional explanation in terms of
a single constituent in a single clause out of the question.
Modalities of various kinds are expressed by want, need, and rao-
dals:
(16) Yeah but you see y y <sighs> <pause> you've got as much of a
problem in expecting one of them to be [handy] when some-
thing happens as you have with a police officer. What we need
<pause> actually is more money spent on the police isn't it?
(17) John: ...what kind you know? I've got a few brochures
today er d is there any kind of do you could you tell
me what it must need to have you know like er in
the way of disc drives and things?
Imogene: Well erm what you need is a erm there's a small
disc drive you don't need a five and a half inch disc
drive.
(18) If we're saving then what we want to know is how much more
could we spe buy how much more goods and services could we
acquire at the end of the year er by saving now
(19) 'Cos like I wanted really what I wanted was I'd never seen
anyone all day so what I suppose I wanted was someone to talk
to
(20) This is a kind of a special approach because we are operating

from a standpoint that we don't have the necessary science in
order to - for us to be exact in terms of our adjudication of
these cases. So what we have to do is to make some conclu-
sions based on what we refer to as the doctrine of reasonable
doubt. (CSPAE)
(21) Actually, what you might do is leave this part and say, these
were the parameters originally to the committee. (CSPAE)
The deployment of modality markers here is further significant be-

cause of the subjective dimension that such forms always contribute
to the discourse. Almost always the subject of the modal verb is a
speech act participant (first or second person pronoun). More gener-
ally, the speaker's strategy for choosing verbs in the pseudocleft
piece is to anticipate some lexical material from the upcoming dis-
course such that an indefinite amount of projected utterance is pre-
framed as "action", "event", "need", "want", "possibility", etc. This
pre-framing is of course the speaker's choice, and is not determined
by what is to come. It is therefore hardly surprising if, as in this next
example (cited earlier), the pre-frame that is proclaimed in the pseu-
docleft does not appear, in retrospect, to match the actually imple-
mented subsequent discourse:
(22) and then what they 've done is the cellar it's got a cellar and you
go down the steps to the cellar but there's like a proper two
proper rooms so on your left you've got a sort of cellar with a
quarry tile
120 PaulJ. Hopper
The speaker has here pre-framed the continuation as a set of actions

of the builders/decorators rather than as a description of the interior;
but her enthusiasm for the design overrides her admiration for the
designers, and the continuation moves from being a narrative ("what
they've done") to being a portrayal of the interior ("what it is now
like").
In anticipating part of the lexical content of the up-coming utter-
ance^), the speaker is furthermore distributing information over a
wider stretch of discourse. This dispersal of information can be ex-
ploited rhetorically to introduce new information that would other-
wise be more condensed in a more measured way, so as to be distrib-
uted over a longer time period and thus to be more easily digested by
listeners. This feature no doubt explains the high frequency of pseu-
doclefts in academic discourse and other formal registers where
complex issues are being articulated.
/v. To buy time while alternative wordings are considered

There is often reason to believe that the use of a pseudoclefit is related
less to the focus demands of the discourse than to the need for time
while formulating an utterance. It thus functions as a pleonastic seg-
ment that fills what would otherwise be heard as a hesitation (Ong
1982: 40-41). The formulas of epic verse work in a similar way.
Homer's rhododaktylos ëôs "rosy-fingered dawn" comes as a ready-
made piece of two and a half feet of verse that can be "plugged in"
whenever the sense and the prosody permit it. The singer, while pro-
ducing it, is not thinking about the current line but planning the next
one under the pressure of real-time performance (see Lord 1960: 22
and 30-67). The analogy is especially striking when we consider
fixed and semantically empty pseudocleft pieces such as what I sup-
pose is, what I suggest is, what happens is:
(23) a rather modern play and it seems to avoid the kind of language
<pause> traditionally expected from drama pointed well-
formed eloquent witty erm and it's what I suppose is one of the
reasons why we see this play as a sort of er modern or experi-
mental piece of drama 'cos it seems to deny itself a lot of the

erm traditional resources of drama
(24) Alan: also that other facets of the curriculum had to be er

brought on so people took on technology which will
be coming on board after that
Harriet: Mm.
Alan: er history and geography. And what's happened is is
as they've come on board and that we feel that <snif£>
we have got to grips with them as well as we can at
this moment in time we've abandoned them and then
gone on to
The speaker "Alan"4 has a faltering delivery and is clearly fishing for
words. The two hesitation markers in a short stretch, and even the
sniff, and the resort to ready-made expressions all betray this. What's
happened is is not the only cliché in Alan's utterances (facets of the
curriculum, coming on board [twice], got to grips with, as well as we
can, at this moment in time). To the reader of the transcript he is
practically unintelligible, but of course the listener, who had access to
the entire conversation to that point, may have had no such difficulty.
The frequent association of pseudocleft pieces with disfluencies
such as hesitation markers and clichés points to a role in processing
spoken discourse on line. Pseudocleft pieces seem to occur with
some regularity after a pause and also at points in the conversation
where one speaker has not taken an expected turn and the other in-
terlocutor has to step into the breach:
(25) Fred: ...engineering at Aston and so we've always had a fair

amount of autonomy
Tim: Yeah.
Fred: and so I don't think it's made a great deal of difference.
Tim: Yes.
<pause>
Fred: But I think what has made the difference is er civil engi-
neering degrees are now awarded by about forty institu-
122 Paul J. Hopper
tions and some of the polys take students of very low

calibre indeed and so er in a way we could do
(26) Joe: Er I mean if if people want to plant bombs there's al-

ways ways of doing it whatever the security.
Rachel: Yeah.
Joe: I mean what is worrying is that it's happening er peo-
ple could get hurt and we don't know who's doing it
really it might be the IRA it might not.
The often noticed repetition of is in examples such as the next one,

which is representative of a considerable number of such examples,
suggests that the speaker has processed the pseudocleft as an unana-
lyzed unit to the extent that she is unaware of having said is, and
feels bound to produce a further is to mark the boundary of the pseu-
docleft:
(27) Adrian: I think that the difference is probably in interpretation.

Whereas I would use the word power I can understand
why you'd gib at the use of that word.
Hannah: But surely surely what what this suggests is is that you
have you have to have class erm community and cul-
turally-specific notions of what masculinity
Adrian: Mm.
Hannah: and femininity are
Hannah's struggle to produce these more complex ideas is seen not

only in the repetitions {surely, what, you have) but in Adrian's back-
ground mm, probably to be interpreted as encouraging the speaker to
continue in spite of the difficulty of the ideas.
v. To hold the floor pending the upcoming utterance

Several decades of work5 have confirmed the centrality of turn-taking
in the organization of conversation, and a growing body of work sug-
gests that grammar is heavily invested in strategies for delivering
well-formed turns ("turn-construction units", see Schegloff 1996,
"transition-relevant places", see Ford and Thompson 1996). An im-

portant motive for deploying a pseudocleft in conversational dis-
course is that of shielding an utterance from possible interruption by
signaling that the speaker's turn, contrary to expectations deriving
from form, length and content, is not yet over.
The speaker who is embarked on a discourse longer than a normal
"turn" must always risk losing the floor through interruption by a
listener. This risk - the loss of a rhetorical investment - is amplified
if the speaker delays, for whatever reason, the delivery of an effective
or otherwise prominent statement. Yet such delay is often very desir-
able. The pseudocleft piece is ideally structured to deal with this
situation, for it keeps the airwaves moving (where silence would in-
vite an interruption), holds the listener's attention by releasing small
amounts of anticipatory material (what we would like to know is,
what is especially worrying is, what these facts suggest is, etc.), and
keeps the floor by avoiding the intonational fall that would mark the
end of a turn.
Haj Ross, in a recent paper (Ross forthcoming), has also drawn
attention to the indeterminate nature of this construction, going so far
as to label it a "constructoid". In an informal communication on the
same subject, he remarks: "It is like this: there is a construction - NP
be NP - and pseudos are attracted to it (maybe 'strange attractors' is
not a bad metaphor) - but they don't get all the way there." Ross'
observation seems to rest on the idea of a prototype: there is an ideal
pseudocleft that is based on a simple copular sentence (as in What we
use is a sonic aid [example from Cobuild]), and pseudoclefts ap-
proach it to a greater or to a lesser degree. This thought, however,
leaves out the entire question of why pseudoclefts exist in the first
place and what the functions of the various "imperfect" examples
might be that are found in live natural discourse. It seems to me that
by inverting this picture we obtain a clearer idea of both the con-
struction and its discourse functions. In the case of the pseudocleft,
the function appears to be not so much the focusing function that is
generally attributed to the pseudocleft. To be sure, focus is part of the
story, for the pseudocleft serves to alert the hearer to the imminence
of a significant segment of talk, and no doubt the constituent focus-
124 Paul J. Hopper
ing function that has been noted in written or edited spoken texts
derives from its less specific discourse focusing function. But more
important, as has been argued here, seems to be the delaying func-
tion, a rhetorical effect of impressing the listener with the "social"
significance of something about to be said, and making the listener
aware that what follows is part of a considered argument worthy of
attention and not a casual comment. From this delaying function, in
one way or another, derive the various other functions that make the
pseudocleft such a widely used and prominent construction in Eng-
lish.
I would surmise that this general picture is true not only of pseu-
doclefts but of all of the more complex grammatical constructions.
Grammatical constructions are normativized rationalizations of
families of smaller and more fragmented quasi-lexical parts. They
result from the grammaticalization of these fragments.
A further observation is that the discourse motivations for the use
of pseudocleft fragments are essentially temporal rather than struc-
tural. These motivations reside not in the static distribution of pre-
suppositions across a completed construction, but in the timing and
delivery of utterances: anticipating upcoming material, delaying an
utterance for assessed rhetorical and cognitive effectiveness, finding
strategies for warding off interruption. The presuppositional structure
that is such a striking feature of written pseudoclefts is, one would
surmise, a by-product of the transfer of the pseudocleft to a
monologic atemporal medium and its expansion there as an extended
grammatical construction.
3. Conclusions
A common assumption of cognitive linguistics is that speakers frame

utterances in terms of templates known as "schémas" that are based
on prototypical instantiations of constructions.6 However, the leading
researchers in cognitive linguistics have not usually seen any neces-
sity to calibrate schémas beside actual utterances, since in principle
utterances are themselves seen as mere reflections of immanent cog-
nitive structures.7 Yet isolated samples of language often bear only a

partial similarity to the forms found in transcribed texts taken from
natural discourse (Hopper 1996, Shirai 1990). Joan Bybee in a num-
ber of recent publications (Bybee 1998, forthcoming, see also Bybee
and Hopper forthcoming) has shown the importance of considering
structure in language as a function of the discourse frequency of
forms, and stresses the need to study forms through their uses as ex-
emplified in transcribed texts (usage-based phonology and grammar).
The study of complex constructions like the pseudocleft from the
perspective of their distribution in discourse points to a need to revise
the prototype picture of constructions or schémas so as to bring
grammar in line with what Mühlhäusler (1983, see Davis 1992) has
called "the observable realities of language use."
I suggest that the demands of consistency, the requirement that
grammatical constructions be measured against a single ideal proto-
type, lead us to privilege in linguistic analysis sentence types that
have achieved this consistency only in the context of reflective,
planned discourse, and that in natural spoken discourse the visible
parts of these canonical construction types appear as fragments. But
these fragments should not be seen as incomplete partíais. Instead,
the reverse is the case: the fuller "canonical" constructions should
rather be seen as highly stylized cultural artifacts, amalgamations of
fragments put together and grammaticalized through stylistic and
normative conventions. The proper model for understanding this dis-
tribution, I would suggest, is not the central vs. marginal instance
model (prototype model), but the family resemblance model, in
which the fragments that end up as canonical instances are interre-
lated through partial formal-functional similarities. The grammatici-
zation of apparent prototype instances such as:
(28) What we need <pause> actually is more money spent on the

police isn't it?
should thus be seen as a normativized derivative of several frag-

mented possibilities, one in which written conventions and conven-
tionalized and reflective ways of presenting complex information
126 Paul J. Hopper
have no doubt contributed to a complex "schema" that in formal and

written registers controls the assembly of the formula.
I would further suggest that these are not isolated examples, but
are in fact typical of recognized canonical grammatical constructions
in general. By the time a grammatical construction comes to the at-
tention of linguists it is already highly normativized. Worse still,
Westernized normative standards may be smuggled into descriptions
of unwritten languages when linguists base their elicitations on Eng-
lish equivalents. To view the canonical constructions as prototypes
and as the source of "deviant" fragmentary instantiations in discourse
is to put the cart before the horse.
It is sometimes said (e.g. Chomsky 1972, 1965) that children are
exposed to "degenerate" data that cannot account for the fiali forms of
language. Chomsky in fact explicitly denies that the fragments that
occur in discourse could form the basis of a coherent systematic
grammar:
The native speaker has acquired a grammar on the basis of very restricted
and degenerate evidence. (Chomsky 1972:23)
Many children acquire first and second languages quite successfully even
though no special care is taken to teach it to them and no special attention is
given to their progress. It also seems apparent that much of the actual speech
observed consists of fragments and deviant expressions of a variety of sorts.
(Chomsky 1965:201).
Corpus studies suggest instead that these "degenerate" data are the
true substance of natural spoken language, and that what our descrip-
tive and prescriptive grammars give us are normativized assemblies
of these fragments that tend to impress themselves on us as mental
prototypes because of their greater social prestige - their associations
with schooling, with literacy, and with complex discourse character-
ized by long periods and uninterrupted turns. This observation has
significant consequences for both linguistic theory and applied lin-
guistics. One moral to be drawn is that foreign language teachers and
applied linguists in general should attend to the role of shorter, "in-
complete" utterances, even ones that violate rules of canonical
grammar, as the building blocks of fluent and appropriate oral dis-

course.
Notes
1.1 am grateful to Rosalind Moon and Jeremy Clear for their help in introducing me
to the Collins-Birmingham University International Linguistic Database (Co-
build, <http://www.cobuild.collins.co.uk/>) during a visiting fellowship at the
Center for Advanced Research in English at Birmingham University in the spring
of 1999; to Malcolm Coulthard for securing this appointment for me and for his
hospitality; to the Faculty Development Fund at Carnegie Mellon University for
financial support; and to David Kaufer, Head of the English Department at
Carnegie Mellon, for arranging academic leave for me during the period of the
fellowship.
2. Citations are from two sources. Those taken from the UK Spoken subcorpus of
Cobuild are unattributed. A small number of additional examples have been
taken from the Corpus of Spoken American Professional English, by Michael
Barlow, available from Athelstan <http://www.athel.com/cspa.html>. These are
tagged with the initials CSPAE. As is customary in corpus studies, citations are
lifted from the original corpus without modifications of spelling or punctuation.
3.For a recent account of the importance of subjectivity in language, see Scheib-
man 2000.
4. Speakers in Cobuild are identified only as "male voice" or "female voice". Ficti-
tious names have been supplied in the present article.
5. For a summary and important analysis of the mutual relevance of turn-taking to
cognition, see the Introduction to Ochs, Schegloff, and Thompson (eds.), 1996.
6.Constructional schémas are roughly equivalent to prototypes. See Taylor 1998:
177-179 and footnote 1 (p. 199) for discussion.
7. On the Cognitive Linguistics view of schémas as abstractions that are immanent
to linguistic expressions, cf. Langacker (1998): "It is reasonably supposed that
schémas are immanent in their instantiating expressions, and emerge as cognitive
entities by reinforcement of the structural properties they share at a certain level
of abstraction" (page 13; italics as in the original). In Langacker's view, grammar
resides in "schematized representations of sound-meaning pairings, abstracted
from (and immanent in) the specific symbolic configurations observable in com-
plex expressions" (page 2).
128 Paul J. Hopper
References
Bybee, Joan L.
1998 The emergent lexicon. Papers of the 34th Annual Meeting. Chi-
cago Linguistic Society.
Forthcoming Phonology and Language Use. Cambridge: Cambridge Univer-
sity Press.
Bybee, Joan L. and Paul Hopper
Forthcoming Introduction. In: Joan L. Bybee and Paul Hopper (eds.), Fre-
quency and the Emergence of Linguistic Structure. Amsterdam:
John Benjamins.
Chomsky, Noam
1965 Aspects of the Theory of Syntax. Cambridge: The MIT Press.
1972 Language and Mind. New York: Harcourt Brace Jovanovich.
Collins, Peter C.
1994 Cleft and Pseudocleft Constructions in English. London: Rout-
ledge.
Davis, Hayley
1992 Drawing the morphological line. In: George Wolf (ed.), New De-
partures in Linguistics, 90-115. New York: Garland.
Ford, Cecilia E., and Sandra A. Thompson
1996 Interactional units in conversation: syntactic, intonational, and
pragmatic resources for the management of turns. In: E. Ochs,
E. A. Schegloff, and S. A. Thompson (eds.), 134-85.
Hopper, Paul J.
1996 When grammar and discourse clash. Essays on Language Func-
tion and Language Type dedicated to T. Givon, 231-246. Edited
by Joan Bybee, J. Haiman, and S. A. Thompson. Amsterdam:
John Benjamins.
Langacker, Ronald
1998 Conceptualization, symbolization, and grammar. In: M. Toma-
sello (ed.), 1-40.
Lord, Albert Β.
1960 The Singer of Tales. (Harvard Studies in Comparative Literature,
24) Cambridge: Harvard University Press.
Mühlhäusler, Peter
1983 Stinkiepoos, cuddles, and related matters. Australian Journal of
Linguistics 3: 75-91.
Ochs, Elinor, Emanuel A. Schegloff, and Sandra A. Thompson (eds.)
1996 Interaction and Grammar. (Studies in Interactional Sociolinguis-
tics, 13) Cambridge: Cambridge University Press.
Ong, Walter
1982 Orality and Literacy: The Technologizing of the Word. London:
Methuen & Co. Ltd.
Prince, Ellen
1978 A comparison of WH- and IT clefts in discourse. Language 54:
883-906.
Quirk, Randolph, S. Greenbaum, G. Leech, and J. Svartvik
1985 A Comprehensive Grammar of the English Language. London:
Longman.
Ross, Haj
To appear The frozenness of pseudoclefts: towards an inequality-based
syntax. Papers of the Annual Meeting of the Chicago Linguistics
Society.
Schegloff, Emanuel A.
1996 Turn organization: one intersection of grammar and interaction.
In: E. Ochs, E. A. Schegloff, and S. A. Thompson (eds.), 52-133.
Scheibman, Joanne
2000 Structural Patterns of Subjectivity in American English Conver-
sation. Dissertation (Ph.D.): Department of Linguistics, Univer-
sity of New Mexico.
Shirai, Yasuhiro
1990 Putting PUT to use: Prototype and metaphorical extension. Issues
in Applied Linguistics 1: 78-97.
Taylor, John R.
1998 Syntactic constructions as prototype categories. In: M. Tomasello
(ed.), 177-202.
Tomasello, Michael (ed.)
1998 The New Psychology of Language. Mahwah, NJ: Lawrence Erl-
baum.
Transitivity parameter and prominence
typology: a cross-linguistic study*
Sang Hwan Seong
This paper investigates the relationship of the transitivity parameter

and prominence typology in light of recent developments in cognitive
grammar. After critically evaluating previous approaches (Hopper
and Thompson 1980, Shibatani 1985, Talmy 1988 and Voorst 1996)
dealing with particular linguistic schémas and prototypes organized
in and across languages, we will provide a parameterized cognitive
functional perspective to account for language learnability and lin-
guistic variation in English, German, and Korean. This research is
based on the assumption that the grammatical structure of native lan-
guage guides or influences the output of adult second language learn-
ers (L2). The grammatical structure of each individual language is
shaped by the performance of the language speaker. Thus, this line of
research is different from formalist acquisition theory which suggests
that the way in which argument structure acquisition takes place can
be reduced to merely linking a universal conceptual structure to syn-
tactic components (cf. Goodluck 1991). By incorporating diachronic
perspectives as well, the present topic will contribute to the study of
cognitive principles of linguistic variations and organizations deriv-
ing from English, German and Korean. The present author advances
a new cognitively motivated claim that the transitivity parameter
should be constrained such that pragmatically determined word order
languages with high semantic transparency proportionally correlate
with less syntacticization.
132 Sang Hwan Seong
1. Prototype theory and transitivity
Cognitive grammar (CG) views a linguistic construction as the pair-

ing of a specification of form with a specification of meaning (Taylor
1995: 198). Langacker (1988: 18) also maintains that "grammar is
inherently symbolic, and therefore distinct from neither lexicon nor
semantics." It is therefore assumed that in linguistic constructions (or
grammatical constructions) the syntactic component cannot be set
apart from the semantic representation. Grammar has no independent
existence apart from semantic and phonological structure (Langacker
1988: 5). In this line of argument, the interpretations of prototypical
transitive constructions are subject to the more general theory of cog-
nitive grammar. According to Hopper and Thompson (1980: 251),
transitivity is traditionally interpreted "as a global property of an en-
tire clause such that an activity is 'carried-over' or 'transferred' from
an agent to a patient." Under this assumption, it is observed that the
kinds of constructions that deviate from prototypical transitive con-
structions are also treated as marginal examples instantiated from the
construction's prototype specification. Thus, we find the following
unagent-like entities as grammatical subjects in English.
(1) a. This tent sleeps six.

b. The room seats 500.
c. The fifth day saw our departure.
The fact that these unagent-like noun phrases can appear as syntactic
subjects in English does not pose serious problems in cognitive
grammar since they are considered mere deviations from the central
semantic specification of prototypical transitivity (Taylor 1995:
214).1 Compared to English sentences like Mary killed a spider or
Tom hit a ball, the semantic relation between subject and each event
involved in (1) is less intense. The subjects in (1) do not execute the
processes denoted by the verbs. On the basis of Dutch, French and
English data, Voorst (1996) demonstrates that there is cross-linguistic
variation in the typical transitive constructions even with verbs such
as break and buy as far as the level of intensity is concerned. Thus,
Transitivity parameter and prominence typology 133
transitive constructions in Dutch show a high level of intensity be-

tween the subject and the event denoted by the verb and between the
event and the object, whereas in English both relations may be indi-
rect or less intense (The rock broke the windshield / D *Het steentje
heeft de voorruit gebroken). However, Voorst does not provide a
functional account as to why this kind of cross-linguistic regularity
exists. More recently, Talmy (1988) uses "energy transfer" as the
metaphor for the transitive process. In Talmy's terms, energy transfer
does not take place in (la-lc), because these subjects actually do
nothing to execute the unfolding of the events denoted by the verbs.
The same kind of idea is also very much visible in the semantic
analysis of transitive constructions by Voorst (1996). Under this
context it seems natural that active clauses high in transitivity or en-
ergy flow typically undergo passivization in which agentive subjects,
totally affected objects and action predicates are correlated (Hopper
and Thompson 1980: 292-294).
(2) a. *Six are slept by this tent.

b. *500 people are seated by this room.
c. G. ?? Unsere Ankunft wurde von dem fünften Tag gesehen.
'Our arrival was seen by the fifth day.'
d. *Our departure was seen by the first day.
Thus, the passive constructions in (2) are all judged ungrammatical

because the active counterparts of these sentences all exhibit low
transitivity. The subjects of the active clauses are not agents. The
arguments occupying the subject and object positions in the active
clauses in (1) are not well differentiated semantically. Prototypical
transitive constructions are cases in which the arguments filling the
subject and object slots are diametrically opposed to each other in-
volving "agent" and "patient" in the semantic make-up (e.g. kill and
hit in contrast with resemble). Since the cases involved in (2) deviate
significantly from the prototypical transitive constructions, the reason
that the constructions in question in (1) cannot be passivized readily
obtains. The following German examples can also be predicted not to
undergo wm/ew-passivization due to the inherently non-actional
134 Sang Hwan Seong
lexical properties of the verbs involved. Note that the corresponding

distinction in English is lost to a large extent.
(3) a. *Das Buch wurde nicht gekannt. (Helbig/Buscha 1994: 171)

'The book was not known.'
b. *Der Pinto wurde von ihm besessen.
'The Pinto was owned by him.'
c. *80 Pfennig werden von dem Heft gekostet. (Helbig/Buscha
1994:171)
'80 pennies are cost by the notebook.'
Even in cases where the nature of the verb is prototypically actional,

passives often fail to apply if the subject is not an agent.
(4) a. Die Mutter schneidet das Brot.

'The mother cuts the bread.'
b. Das Brot wird von der Mutter geschnitten, (passive counter-
part of 4a)
'The bread is cut by the mother.'
c. Das Messer schneidet das Brot.
'The knife cuts the bread.'
d. *Das Brot wird von dem Messer geschnitten, (passive
counterpart of 4c)
'The bread is cut by the knife.'
(Helbig/Buscha 1994: 165)
The subject in German sentence (4c) clearly denotes an instrumental

case which plays an essential part in the cutting process. However, it
does not control the energy flow per se involved in the event.
Shibatani (1985: 831) makes a valid observation when he argues
that "passives center around agents, and their fundamental function
has to do with the defocusing of agents". If there is no agent to defo-
cus in the active clause, passivization is usually impossible. The
prototype approach significantly undermines the attempt to charac-
terize the structure and/or function of the passive as a counterpart of
the active voice (see also Shannon 1987). However, Shibatani does
not explain why languages exist in which prototypical transitive con-
structions are not passivizable, when he states (1985: 831) that "even
in transitive sentences, passives often fail to apply if the subject is not
an agent." That is, he does not account for why not all prototypical
transitive clauses with agentive subjects find passive counterparts in
a language such as Korean. Our approach is based upon the func-
tional motivation that the richer a morphological case-marking sys-
tem a language has, the more degenerate the system of passivization
becomes.
2. Transitivity and passives
If the assumption that clauses high in transitivity are strongly corre-

lated with prototypical passivization is right, we should find positive
evidence in a language like Korean where the grammatical relations
of transitive constructions are typically coded by nominative and ac-
cusative markers as illustrated in (5).
(5) Kyengchal-i totwuk-ul cap-ass-ta

policeman-NOM thief-ACC arrest-past-Decl
'The policeman arrested the thief.'
This construction can be passivized as follows (cf. O'Grady 1991:

47):
(6) Totwuk-i kyengchal-eykey cap-hi-ess-ta

thief-NOM policeman-DAT arrest-pass-past-Decl
'The thief was arrested by the policeman.'
In the passive construction (6) we notice that the accusative argument

totwuk-ul in (5) is changed into the nominative argument in (6),
whereas the nominative argument "policeman" is construed with an
oblique case marker in (6). Usually the oblique marked noun phrase
in (6) can be dropped, as is also the case with English and German
13 6 Sang Hwan Seong
because the passive construction involves "agent defocusing". We

also note that in the construction (6) a passive suffix hi is inserted
between the verbal stem cap and the rest of the verbal compound. To
account for the structural properties of the Korean passive type in (6),
O'Grady (1991: 47-52) takes a Categorial Grammar approach based
on the formal system of morphological marking conventions in Ko-
rean. Sohn (1994: 301) also claims that there are three types of pas-
sives in Korean: (a) suffixal, (b) lexical, and (c) phrasal. Suffixal
passives are the most basic and typical, while the other two types are
a kind of pseudo passive. The choice among the three passives is
dependent on what types of predicates are involved. In this paper,
however, the present author does not treat lexical passives and
phrasal passives as Korean passive constructions because they do not
employ the grammatical passivization device (i.e. they are not mor-
phologically derived passives). In lexical passives, despite the pas-
sive meanings they trigger, the forms of the stems are entirely differ-
ent from active verb stems (e.g. macìa 'be hit' in contrast with chita
'hit'). Korean lexical passives are also composed of constructions in
which the inchoative verb stem toy 'become' is attached to Sino-
Korean verbal nouns (e.g. wiim-ha 'entrust' vs. wiim-toy 'be en-
trusted'). Phrasal passives are made up of a verb stem and the in-
choative verb cita 'get to be, become' (e.g. cwu 'give' vs. cwu-e cita
'be given'). The evidence for not treating lexical and phrasal passives
as Korean passives comes from the grammaticization of Germanic
passives.
Valentin (1999: 145) argues that old Germanic passive construc-
tions with werdhan (meaning 'happen', 'begin to exist', or 'enter into
being') are perfective constructions, and that the ones with bim/was
(first person singular present/preterite indicative form 'to be') are
non-perfective. Valentin demonstrates that old Germanic passive
constructions with werdhan do not coincide with the periphrases of
New High German or Dutch, where the aspectual oppositions are
exactly contrary to the old ones.2 The following New High German
examples deserve our attention:
(7) a. Das Haus wird verkauft.

'The house becomes sold'
b. Das Haus ist verkauft worden, (preterite version of 7a)
'The house is sold become' (i.e. the house has been sold)
c. Das Haus ist verkauft.
'The house is sold.'
Unlike English, there are two different passive strategies available in

Modern German: namely werden-passive and sein (statal)-passive
respectively. In werden-passives (7a-b) expressing an action or oc-
currence it goes without saying that the change of perspective from
the active counterpart takes place without affecting the lexical se-
mantics of the verb verkaufen and that the original meaning of the
verb werden (to become) is lost. The statai passive in (7c) which used
to be a usage more prevailing in earlier periods should be interpreted
as a derived construction of (7b), since the construction (7c) denotes
a state that has resulted from previous action (see also Drosdowski
1995: 180). The Modern German statai passive with sein is a perfec-
tive construction. According to Valentin (1999), however, it is diffi-
cult to posit a passive as a grammatical device in the old Germanic
dialects since wirdhu (first person singular present indicative form of
werdhan) and bim/was exhibit the same "full" meaning; they are not
auxiliaries. The present author assumes that this is also the case with
the putative Korean passive constructions. We note that productive
constructions such as the Korean lexical and phrasal passives with
the inchoative verb stems toy and ci are fully compositional in their
morphological make-up. The original meanings of these inchoative
verb stems are retained. Thus the validity of the dominant view to
classify the aforementioned constructions under the rubric of Korean
passivization based on the translation equivalents (cf. O'Grady 1991:
50) does not simply obtain. With respect to prototype theory, the suf-
fixal passives in Korean also pose problems. Despite the nominative
and accusative marked transitive construction in (5), it cannot be
maintained that all Korean prototypical active transitive constructions
find corresponding passive constructions. For many passive sen-
tences there are no active counterparts either. It is also the case that
138 Sang Hwan Seong
many passives and their corresponding active constructions are not

semantically equivalent. Nam and Koh (1993: 297) state that "Ko-
rean passives are not always derived from transitive counterparts and
that there are more transitive verbs that cannot trigger derived suf-
fixal passives than that can".
(8) a. Saram-dul-i mul-lul mani massi-nta.

man-pl.-NOM water-ACC a lot drink-pres.
'People drink milk a lot'
b. *Mul-i marti massi-i-nta.
water-NOM a lot drink-pass-pres.
'Milk is drunk a lot [by people]'
In (8) we find a prototypical transitive clause with the verb drink in

which no passive counterpart is available. Thus, as far as the Korean
passive is concerned, a morphology-based structural description of
the rule does not simply obtain.
What interests us in the description of Korean grammar is not the
transitivity phenomenon per se, but the animacy constraint imposed
upon the voice system. Let us observe the examples in (9) and (10).
(9) a. ?mos-i os-ul ccic-ess-ta

nail-NOM clothes-ACC tear-past-Decl
Lit. Ά nail tore the clothes.'
b. os-i mos-ey ccic-ki-ess-ta
clothes-NOM nail-by tear-pass-past-decl
'The clothes were torn by a nail'
(Sohn 1994: 307)
(10) a. *sikan-i na-lul ccoch-ko isseyo

time-NOM I-ACC chase-ing is
'Time is chasing me' (Time is regarded as personified)
b. nan-nun sikan-e ccoch-ki-ko isseyo
I-TOP time-by chase-A/'-ing am
Ί am being chased by time' (i.e. Things are hectic for me)
(Klaiman 1988: 56-57; hi = passive suffix)
The two transitive constructions in (9) and (10) tell us that inanimate
subjects are avoided in controlling the state of affairs in Korean dis-
course, regardless of the argument's thematic status. This animacy
constraint in the Korean diathesis might explain why the middle
voice coupled with reflexivization is not available in the Korean and
Japanese systems.3
(11) a. K. */ chaek-un casin-ul cai pania

this book-TOP itself-ACC well sell
b. i chaek-un cal pal-i-nta
this book-TOP well sell-pass.-Decl.
'This book sells well.'
c. G. Dieses Buch verkauft sich gut. (passive reading)
this book sells (reflexive) well
d. D. Deze stoel zit heel aangenaam. (Abraham 1995: 20)
this chair sits very pleasantly
'This chair sits very pleasantly'
One might argue that this kind of reasoning is not quite valid, given
the fact that compared to English alone, Dutch and German transitive
constructions show a relatively transparent semantic encoding system
of subject selection analogous to Korean, while German is nonethe-
less identified as a language with reflexive construction. This seem-
ingly plausible counter-argument is untenable, since we subscribe to
the view that the reflexivization as an important feature of subject-
prominence applies even to the German transitive construction in
(11) with passive reading. That is, in German, even inanimate patient
subjects trigger reflexivization in the middle reflexive construction.
The corresponding Korean construction (lib) employs a pseudo-
passive with a passive morpheme. Reflexivization in Korean is re-
stricted only to animate entities (cf. also Seong 1999b).
When it comes to passivization with respect to transitivity, there is
also a misunderstanding over the identification of the transitive ob-
ject. Hopper and Thompson (1980: 259) suggest that "the special
markings on definite objects, found in many languages, are better
interpreted functionally as signals of the high transitivity of the
140 Sang Hwan Seong
clause as a whole." They also argue that "the arguments known to

grammar as INDIRECT OBJECTS should in fact be transitive ob-
jects rather than what might be called 'accusative' objects, since they
tend to be definite and animate". And this observation further sup-
ports the findings of Givón (1979: 54) that, in one pair of English
texts, out of 115 indirect objects, 112 (or 97%) are definite and
overwhelmingly animate. This is argued to be also true of some
Bantu languages where topic-related syntactic processes including
passivization are correlated with the fact that the 'dative' argument
takes precedence over the patient noun phrase. Despite the cross-
linguistic insights by Hopper and Thompson, the present author has
some reservations based upon German and Korean data. Let us look
at the following dative and accusative noun phrase serialization pat-
tern of typical ditransitive constructions in German and Korean.
(12) a. Hans hat dem Anwalt das Geld gegeben. (German)

'Hans has to the lawyer the money given'
b. Hans-nun pyunhosa-eykey ton-ul
TOP lawyer-DAT/BEN money-ACC
cwu-ess-ta. (Korean)
give-pret.-Decl.
'Hans gave the lawyer the money.'
In both constructions in (12a) and (12b) we note that the dative noun
phrase precedes the accusative patient noun phrase as the unmarked
order.4 In German, this is only possible if the two combinations are in
free variation. This tells us that the more referential animate object
appears before the inanimate object, as far as the precedence rule is
concerned. However, this cannot be taken to mean that the dative
animate object is indeed the true transitive object of the verb, since
the dative marked argument cannot be promoted to the nominative
case in German or Korean passives (e.g. *Der Anwalt wurde das
Geld gegeben; Das Geld wurde dem Anwalt gegeben). Thus, the
identification of true transitive objects cannot be established unitarily
by means of syntactic tests such as passivization (cf. *Er wurde ein
Preis verliehen 'He was awarded a prize.'). This is in turn in conflict
with another syntactic strategy called 'English dative movement' in

which beneficiary/recipient argument precedes patient argument in
the ditransitive construction. In English, 'dative movement' is also
taken to support the preference selection of the dative recipi-
ent/beneficiary argument for identifying the true transitive object in
English (cf. Hopper and Thompson 1980: 260).
(13) a. Clara wrote a letter to Santa Claus.

b. Clara wrote Santa Claus a letter.
The two sentences in (13) have the same meaning (D. Shaffer, p.c.).
The only difference may be a slight difference in focus. In the first,
the focus is on letter: I wrote a letter (not something else) to Santa
Claus. In the second, the focus is on Santa Claus: I wrote Santa Claus
(not someone else) a letter. Since in (13b) the human recipient argu-
ment appears in the "object position" adjacent to the verb, this dative
object would be interpreted as the true object according to the logic
of Hopper and Thompson. Since the comparable nominal as an in-
herent case is not subject to promotion to the nominative argument in
German and Korean passivization, the attempt to establish the true
object in terms of animacy in ditransitive constructions does not suc-
ceed cross-linguistically. Thus, definiteness and animacy criteria as-
sociated with a high degree of transitivity cannot be a true diagnostic
for selecting a true transitive object, at least with respect to passivi-
zation in German and Korean (cf. G. *Er wurde geholfen 'He was
helped.'). In view of the criticism of the traditional transitivity theory,
the transitivity parameter should be constrained such that languages
with high semantic transparency and an overt case-marking system
proportionally correlate with less syntacticization.
The traditional analyses of passives centered around the formal
aspects of the constructions per se or the lexical properties of the
verbs involved (Haider 1985 for German and O'Grady 1991,
Klaiman 1988 for Korean). Contrary to the previous studies, the pre-
sent author attempts to account for the motivation behind the passive
system by setting the discourse category "topic" apart from the
grammatical category "subject".
142 Sang Hwan Seong
3. Subject and topic revisited
Korean is well known for its subject and topic-prominence (Li and
Thompson 1976). For the discussion of "subject and topic" the pres-
ent author adopts Chafe's (1976: 50) functional definition of "topic"
as something which "limit[s] the application of the main predication
to a certain restricted domain," or as something which "sets a spatial,
temporal, or individual framework within which the main predication
holds." In previous research (cf. Hawkins 1986, Müller-Gotama
1994), it was pointed out that a grammatically prescribed fixed word
order language like English shows more ambiguous surface syntax
with respect to the pragmatic functions compared to German and the
Slavic languages. In these approaches, however, the traditional no-
tions of subject and object still play a central role in interpreting the
core cases of grammatical organization. The present author maintains
that for a more adequate description of the typological regularities,
we need to set the notion of topic apart from the category of subject
(Seong 1999a, 1999c). Gundel (1988) in her elegant study also illus-
trates what kinds of strategies are available to mark "topic-comment
structures" across languages, following the tradition of Li and
Thompson (1976). However, she does not show how the relative de-
gree of topic-comment structure can vary among the languages in-
vestigated. In this section, we want to further investigate how the
word orders of English, German, and Korean are pragmatically
regulated in controlled dialogue situations. Let us compare the fol-
lowing question and answer pairs.
(14) English (Chafe 1976: 48)

Q: What happened to the lamp?
Al : The dog knocked it over.
A2: ?It was knocked over by the dog.
(15) German
Q: Was ist mit der Lampe passiert?
A1 : Der Hund hat sie umgeworfen.
Lit. 'The dog has it knocked over.'
A2. Die hat der Hund umgeworfen.

it-ACC. has the dog-NOM knocked over
A3 : ?Sie wurde von dem Hund umgeworfen, (passive)
A4: ?? Sie hat der Hund umgeworfen.
(16) Korean
Q: chundung-i etteke doinke-ya?
lamp-NOM how became-Q
[definite] [wh-rheme:focus]
Lit. 'What has become of the lamp?'
Al: (guguss-un) gae-ka nume-tturyu-ss-tta.
[it -TOP] dog-NOM fall-cause-pret-Decl.
[TOPIC] [FOCUS] [VERB]
A2: ??guguss-un gae-eyeuyhae nume-ci-ess-ta.
[it - TOP] dog-by fall-inchoative-pret-Decl
Lit. 'It was knocked over by the dog.'
In the sentences in (14-16) the present author provided a controlled

dialogue situation whereby various answers (A) to the question (Q)
'what happened to the lamp?' can be formulated for the three lan-
guages in the order of preference. As we know from the functionalist
research tradition associated with Talmy Givón, Charles Li, Sandra
Thompson and others (cf. Traugott and Heine 1991), the fixing of
discourse strategies influences the grammatical packaging of the dis-
course and evolution of the syntactic and morphological structure.
We also hold the view that the grammatical structure of each given
language is constantly shaped by the performance of the language
speaker. Hence, the differences of discourse strategies identified in
the dialogue exchanges (14-16) provide important implications for
the grammaticization processes in each language investigated.
In the English Q & A pair (14), we observe that while it is the
topic, the dog is chosen here as subject. The preferred response (Al)
treats the dog as the subject, although the disfavored response (A2)
which promotes the topic it is entirely grammatical. In the German Q
& A pair (15) as well, we find that der Hund is selected as the subject
over the topical pronouns die or sie. Between English and German,
144 Sang Hwan Seong
we also note another difference - that passivization as a topicaliza-

tion strategy can be a more feasible alternative answer in English
than in German. In the Korean Q & A pair, the preferred unmarked
answer (16A1) to the question (16Q) shows that the topical pronomi-
nal constituent with nun appearing in the sentence initial position
precedes the pre-verbally focused rhematic element dog. This topical
item can be freely dropped in the most natural discourse setting. It
seems in Korean that even though the constituent with subject marker
ka appears as the sentential subject in (16A1), its position is not syn-
tactico-semantically but pragmatically determined (see also Shannon
1999 for Early New High German [ENHG] data on this issue).5 This
topic-focus constraint also applies to the precedence rule ordering of
dative, accusative and oblique arguments in Korean as the present
author crucially demonstrated in his previous article on word order in
German and Korean (cf. Seong 1999a). Based on the different dis-
course strategies attested in (14-16), the present author demonstrated
that the linear arrangement of Korean as a topic-prominent [TP] lan-
guage follows the scale of "communicative dynamism" (cf. Firbas
1992) far more closely than it does in a subject-prominent [SP] lan-
guage.
This finding suggests that the passivization strategy is highly re-
stricted in TP languages, since they exhibit a "built-in topic-com-
ment structure". In these languages the use of syntactic promotion to
achieve topicalization of a non-subject participant is not necessary;
such a strategy can be attained directly. In English, however, we do
not find this strategy. Van Oosten (1984) already observes that
agency or, more generally, "primaryhood" often wins out over topic-
hood in English subject-selection in basic, active sentences. Even
though German also seems to conform to this principle in basic sen-
tences, the passivization in which topichood prevails over agency in
English subject-selection does not strictly apply to German, due to
the existence of verb-second rule (cf. examples in [15]). In English,
the passive construction is employed in cases where a non-agentive
topic beats the agent or the primaryhood for subject-selection, and
English also has other various syntactic devices such as zY-cleft (or
extraposition), there-existential, and towg/?-construction (e.g. This
book is tough to read) to satisfy the secondary need to identify the

topic of the sentence. In this respect, the topic or pragmatic pivot
prevails over agency in English subject-selection, as claimed by
Foley and Van Valin (1984: 115). Korean, on the other hand, strictly
conforms to the topic- and focus (rheme) positional constraint in
which two separate categories of topic and subject are available.
Thus, the conflict over subject selection in the sense of van Oosten
(1984) vanishes entirely as far as Korean is concerned.6
The following examples from Korean and ENHG further support
our argument for the existence of the topic as an independent cate-
gory.
(17) Kokiri-nun ko-ga kil-ta.

Elephant-TOP nose-NOM long-Decl.
Lit. 'As for an elephant, it has a long trunk.' (i.e. An elephant
has a long trunk.)
(18) so sah er, das die Messer, die er geschliffen het, der Rück was
als die Schneid...
'Then he saw that the knives that he had sharpened, the back
was like the blade...' (taken from Shannon 1999)
In the literature, this construction is known as a "double subject con-

struction" in which the left-most constituent in (17) is not a subcate-
gorized argument of the predicate kilta 'to be long'. In ENHG exam-
ple (18) die Messer 'the knives' is also taken to be a free topic, which
is not a part of the verbal valency. Such a construction is not avail-
able in Modern English or in New High German (NHG), since the
grammaticization of a topic into the category of subject took place in
the diachronic development of Germanic (Burridge 1993). Thus, it is
clear that the older stage of modern Germanic languages conforms
more faithfully to the pragmatically determined word order. Judging
from these data, it is fairly obvious that the discourse frame setting
strategy found in TP languages is completely different from that of
SP languages such as English.
146 Sang Hwan Seong
The historical rise of the opaque relationship between "subject cate-

gory" and "topic" in English is also confirmed by the development of
a sister Germanic language such as Scandinavian. Faarlund (1992)
argues that the "subject" category in Scandinavian has changed from
a relatively "non-prototypical" to a "prototypical" one as Scandina-
vian languages have evolved from non-configurationality to configu-
rationality. He assumes that the nominative case in Old Scandinavian
is associated with subjecthood, and that discourse- or reference-
related strategies involving theme-rheme organization are entirely
independent of case marking. Specifically, the nominative-marked
nomináis in modern Scandinavian languages have gained many refer-
ence-related (i.e. topicality in Chafe's term; cf. also Foley and Van
Valin 1984) "subject" qualities that Old Scandinavian nominatives
lacked. For example, while in Modern Scandinavian "the subject is
always definite in some (specifiable) sense", Old Scandinavian
nominatives are often indefinite (Faarlund 1992: 161). Often this
means that "expletives" are used in the modern translation of Old
Scandinavian indefinite nominative subjects. Thus, in Old Scandina-
vian, topicality and primaryhood (cf. Van Oosten 1984) are coded by
completely different formal means.
The basic sentence structure of the SP and TP languages can be
schematically explained in the following diagram (cf. double subject
constructions in [17] and [18]).
Topic Comment
Figure 1. Diagram of topic-comment connection
In this system the category of topic need not be defined in terms of

verbal valency of the lexical predicate. The vertical line in the dia-
gram shows the functional boundary of topic and comment. The di-
rection of the arrow in this diagram also captures the possibility that
the category of topic functions as the core argument of the predicate
as in SP languages. Schematic characterization of the English system
would be the pattern in which the two categories "subject" and
"topic" have converged with each other.
Gapping between unlike constituents with differing grammatical
relations also belongs to the general features of TP languages. Again,
we find a striking typological similarity between Korean and early
Germanic.
(19) a. pasta-nun nae-ka imi mok-ess-jiman ajikdo

TOP I-NOM already eat-pret-but still
0 nama-issta. (Korean)
(subj) left-is
'Even though I already ate of the pasta, there is still (some)
left'
b. Die gefielen dem Herren so wol und [er] sprach zu
DAT [NOM]
ihm ... (ENHG)
'They pleased the Lord very well and [he] spoke to him ...'
(Shannon 1999)
c. Hi haelde mi (object) ende [ _ ] ginc met hem in die taverne
(Middle Dutch)
'He fetched me and [I] went with him into the tavern'
(Burridge 1993: 146)
The Korean construction (19a) suggests that the initial constituent the
pasta can be construed as both the object in the first clause and the
omitted subject in the second clause. In ENHG example (19b) as
well, the "dative-marked noun phrase" dem Herren 'the Lord' in the
first clause functions as the controller of the missing "nominative
nominal" er 'he' in the second clause. The highlighted constituent in
Middle Dutch example (19c) represents the syntactic item omitted in
the conjoined clause. This suggests that in Korean, ENHG and Mid-
dle Dutch the zero noun phrase-anaphor is not syntactically restricted
148 Sang Hwan Seong
and that the structural properties accrued to the category of "subject"

in extreme subject-prominent languages like English are considerably
limited. This means that we should consider two functional roles
"subject" and "topic" in interpreting some core cases of the gram-
matical organization of the subject and topic-prominent languages. If
this is the case, then we are justified in arguing that the structural
topic properties mostly associated with the English subject category,
i.e. grammatical relation changing rules such as passivization and
raising, are highly restricted or not available at all in highly TP lan-
guages. Thus, our argument is based upon a discourse-oriented func-
tional motivation different from the generalization of Plank (1983) in
which the morphology based transparent coding of grammatical rela-
tions is directly linked to the restricted application of the syntactic
rules in German as opposed to Modern English (e.g. raising con-
structions and extractions).
4. Parameterized prototype theory
As we have seen in examples (9-11), the semantic restriction that the

inanimate entities are low in controlling the state of affairs of Korean
transitive constructions predicts the absence of the middle construc-
tion in the Korean voice system. This conceptual difference could be
broadly interpreted so that the selectional restriction for the semantic
content of the subject category in the transitive verbal valency be-
comes stronger when we move from grammatically determined word
order languages like English to pragmatically determined word order
languages like German and Korean (This tent sleeps four / G *Das
Zelt schläft vier, The key opened the doorI G.*Der Schlüssel öffnete
die Tür). This principle also holds for the relationships of verb and
object in German and Korean. The case we have in mind is that in
English, nomináis referring to very unpatient-like semantic roles ap-
pear as direct objects in transitive constructions (Taylor 1995: 212).
(20) a. We laid a carpet in the room,

b. We carpeted the room.
c. He loaded hay onto the truck.

d. He loaded the truck with hay.
In (20b) we note that the patient participant denoting a locative is

incorporated into the verb. This is, however, not a syntactic rule, but
a lexical derivation, i.e. conversion (explanation attributed to R. Dir-
ven). In (20d), the truck is construed as the transitive object of the
same locative verb in (20c). In morphologically transparent systems
like German and Korean, these opaque constructions are not likely to
occur or less developed.
(21) a. Wir haben einen Teppich in das Zimmer gelegt.

we have a carpet in the room laid
b. Wir haben das Zimmer mit einem Teppich ausgelegt.
we have the room with a carpet laid out
c. Er hat das Heu auf den LKW geladen.
d. Er hat den LKW mit dem Heu beladen.
e. *Er hat den LKW mit dem Heu geladen.
f. Hans hat die Wand neu tapeziert.
Hans has the wall new papered
'Hans repapered the wall.' (an opaque instance similar to
English)
In (21b) and (2Id) the German morpho-syntax of the corresponding

clauses shows much clearer lexical choices than in English. The Ko-
rean equivalents in (22a) and (22b) reveal a similar transparent para-
digm.
(22) a. ku-nun cwimcha-ey kuncho-lul sil-ess-ta.

he-TOP truck-LOC hay-ACC load-pret.-Decl.
'He loaded hay unto the truck.'
b. ku-nun kuncho-ro cwimcha-lul chai-u-ess-ta.
he-TOP hay-INSTR truck-ACC fill-caus.-pret.-Decl.
'He loaded the truck with hay.'
150 Sang Hwan Seong
The fact that English grammar entails these opaque grammatical re-
lations tells us that the English transitive construction has undergone
enormous extension due to case syncretism in its history. Languages
differ in the way features of actions are encoded by lexical items.
However, it is no coincidence that German and Korean show a more
transparent overall encoding system for the relationship of verb and
object.
Split intransitive constructions in German and Korean also fit well
into our system of prototype parameters. The experiencer argument
of a mental state is readily encoded as a transitive subject in English
(e.g. The king likes pears). In German and Korean, however, the ex-
istence of the dative case is held to be responsible for encoding most
typically the more topical or semantically salient "experiencer argu-
ment" of a mental state in a valency network of lexical predicate.
(23) a. Ich höre, dass Studenten eine

I hear that students-DAT a
Bahnkarte fehlt.
discount train card-NOM lacks
Lit. Ί hear that students lack a discount train card.'
b. *Ich höre, dass eine Bahnkarte
I hear that a discount train card-NOM
Studenten fehlt.
students-Dat lacks
Lit. Ί hear that students lack a discount train card.'
c. * Thomas glaubt, dass ein Fehler der
Thomas believes that a mistake-NOM the
Mannschaft unterlaufen ist.
team-DAT happened is.
d. Na-eykey/nun ton-i mocara-nta. (Korean)
I-DAT/TOP money-NOM lack-Decl.
Ί lack money.'
The split intransitive constructions (or inversion constructions) de-

noting a mental state with dative experiencer + nominative theme
arguments in the German 'middle field' (Ger. Mittelfeld) and Korean
show that the grammatical encodings specifiable in overtly case-

marked systems are neutralized in English transitive constructions.
This is due to the fact that unlike Modern German, English has be-
come relatively opaque in the declension of case marking and the
comparatively tight association of agent with nominative subject en-
coding. This diachronic change resulted in the integration into the
subject category of formerly subjectless impersonal verbs with expe-
riencer arguments (e.g. OEngl. Him/hine hyngrede 'He is hungry.'
from Plank [1983: 11]). Given the lack of isomorphism between role
and overt coding in a configurational language like Modern English,
Noonan (1977: 377) argues that this system requires the identifica-
tion of a level of grammar not relevant to direct role marking lan-
guages, i.e. a level of grammatical relations (e.g. subject and object).
This is, then, the sense in which Noonan uses the term "subject": "the
highest ranking syntactic slot in an indirect role marking system." In
our prominence typology, the distinctive statuses of the two types of
role-marking systems can be identified with English (indirect role
marking system) and Korean (direct role marking system) respec-
tively. In the English system, then, reliance on traditional grammati-
cal relations to describe the morphological and syntactic process is
most clearly indispensable. In contrast to English, the participant
encoding strategy of pragmatically determined word order languages
such as German and Korean provides a parameterized perspective
with respect to the prototype view of constructions. Thus, the Korean
system might be classified as a highly transparent system in which
some form of overt coding (e.g. case marking) relates directly to se-
mantic roles, without significant neutralization of role distinctions.
Accordingly, we can take the German system to be a relatively trans-
parent system in terms of how the form of overt coding is mapped
onto the role system. By analogy we can also claim that Old Scandi-
navian had a highly transparent verbal encoding system for transitiv-
ity, since this language had two distinctive formal means of marking
nominative argument and topicality, as we confirmed in the preced-
ing section. It is in this sense that Noonan (1977) contrasts the trans-
parent direct-role marking system with an English type "indirect role
152 Sang Hwan Seong
marking systems," in which syntactic slots and argument roles are

arranged hierarchically à la Fillmore (1968: 33).
In a grammatical theory dealing with typologically distinctive lan-
guages, we need to construct a syntactically, semantically and prag-
matically balanced meta-linguistic apparatus according to which the
grammatical properties of each language can be measured and tested.
The following figure schematically demonstrates the relative degree
of transitivity in the verbal encoding system identifiable among lan-
guages investigated via the subject and topic prominence parameter.
< •
more grammaticizing less grammaticizing

subject prominence topic prominence
subject = grammaticized topic subject * topic
This figure indicates that the transitivity and prototype parameter is

constrained such that languages with high topic-prominence strongly
correlate with less grammaticization. Thus, this figure can be under-
stood as a typological frame of reference, a scalar model to explain
each grammaticization process of a given language with respect to
subject and topic parameter. At the extreme left side of our system,
we find an English type language whose properties such as allegedly
looser selectional restrictions for verbs and greater freedom in the
application of raising and passivization may be non-coincidentally

related to the greater "opacity" specific to the system. We argue that
these properties of a broader structural description of the rule are to
be closely linked to the presence of the structural subject position in
English. Since this configurational system requires grammatical rela-
tions to be indispensable to identifying arguments of the clause, the
relationship between overt case-marking and the assignment of se-
mantic roles is expected to be the least isomorphic in English as an
SVO language (e.g. This tent sleeps three. / This bed was slept in. /
G. *Dieses Bett wurde geschlafen in.). At the opposite end of our
system we find a Korean type language in which pragmatically de-
termined word order with a transparency principle motivates a more
direct reflection of semantic roles in the morphology. Thus, the Eng-
lish subject is analyzed as a "grammaticized topic", whereas Korean
as a typical SOV language retains an independent "topic category" in
addition to the notion of subject. Prima facie evidence for the topic
prominence parameter is presented via a historical account involving
Germanic languages. It is clear that there is a typological distinction
between construction-based languages and discourse-based ones.
Recall that in chapter 1, we observed the generalization of Voorst
(1996) that when it comes to the grammatical relations of transitive
constructions in Dutch, French and English, Dutch shows the highest
level of intensity (i.e. strict subcategorization) whereas English
shows the lowest level of intensity (i.e. loose subcategorization).
While Voorst does not provide any explanation as to why such cross-
linguistic variation exists, it is obvious that our typology proposed
here accounts for the cross-linguistic regularities at issue. In this sec-
tion we demonstrated crucially the relative degree of subject and
topic prominence that obtains among Korean, German and English,
and that this regularity provides a better prediction of the grammati-
cization processes involved. In the next section we advance the view
that this principle is directly reflected in the second language acqui-
sition process for Korean.
154 Sang Hwan Seortg
5. Prominence typology and language acquisition
There is substantial evidence manifested in L2 data that, due to the

effects of prior knowledge such as in the form of transfer or interfer-
ence phenomena, native speakers of SP languages experience sys-
tematic difficulties when learning TP languages such as Korean. The
present account of the second language acquisition process for Ko-
rean is based on the part of the research database collected at the Ko-
rean program at the University of Bonn. The initial focus of this
study is centered around the TP features of the target language, such
as topic constructions with topic marker, zero topic (subject and ob-
ject) constructions, and double nominative constructions. The par-
ticipants were level 1 and level 2 German students with German as
their native language. They were asked to describe the participant's
family and what they did last summer. This task evaluated the par-
ticipant's ability to acquire the target features in written production
(free composition). To provide a baseline for each analysis, three
native speakers of Korean also participated as the third group. Thus,
the writing samples are taken from 3 groups at different levels of
proficiency in Korean. See Table 1.
Table 1. Frequency distribution of overt subjects in the written production of

three groups
Korean Proficiency Levels
(number of participants) 1(3) 2(3) 3(2)
Total subjects/ sentences 16 / 20 18 / 36 19 / 44
Careful analysis of the writing samples demonstrates that German

students at the first (beginners') proficiency level tend to write Ko-
rean sentences with overt subjects almost without any exception.
Even in the four cases where empty subjects are found, it is the case
that these empty subjects refer back to overt subjects in the immedi-
ately preceding sentences. Our data show that the tendency to drop
overt subjects increases dramatically when it comes to proficiency
level 2 consisting of second year students. Participants at level 2
dropped only the second sentential subject, which is discourse-bound
to the first subject. On the other hand, level 2 students seem to over-
generalize this principle so that we often find the following ungram-
matical sentence (24b) in which a subject noun with topic marker
should be mentioned sentence-initially for the utterance to be gram-
matical. In the sample text we find the first sentence (24a) with three
subjects (parents, older sister and older brother) which seem to func-
tion ambiguously as the discourse subjects of the sentence in (24b).
(24) a. Pumonim-i keysiko, enni-wa

parents-NOM be-honorif. older sister-with
oppa-ka issupnita
older brother-NOM be.
Ί have parents, an older sister and an older brother.'
b. *samsipo-nyun- june kyulhon-hae-ss-upnita
35 years ago marry-do-pret-honorif. deci
'They married 35 years ago.'
This pattern suggests that German learners of Korean may first over-
produce subjects (cf. level 1) and then gradually learn to drop them as
their proficiency improves. However, it is clear that one cannot freely
drop a subject in a context in which a bare noun with topic marker
pumonim-un 'parents-topic' as the subject should be specified as the
old information as evidenced in (24b).
Among the different types of TP features of Korean, topic con-
structions with the topic marker (n)un seem to be relatively harder to
acquire. Traditionally, it has been observed that the nominative
marker is usually construed with a neutral description, whereas the
topic marker triggers both a thematic and a contrastive reading
(Seong 1999a: 362). The two separate readings become stronger
when the topic particle is used twice in parallel constructions for
conveying the information of equal status. The following example
(25a) directly follows the sentence (24b) in the same sample text
taken from one of level 2 participants.
(25) a. *emeni-ka jubu-ipnita. Apuji-ka kepsuso-eyse

mother-NOM housewife-be father-NOM waterworks-at
156 Sang Hwan Seong
sikmulhaJg'a-ipnita
botanist-be
'My mother is a housewife. My father is a botanist at the
waterworks.'
b. emeni-nun jubu-i-mye apuji-nun
mother-TOP housewife-be-conjunction father-TOP
kepsuso-eyse sikmulhakja-ipnita
waterworks-at botanist-be
'My mother is a housewife and my father is a botanist at the
waterworks.'
While writing about her family, the German female student produced
(25a) instead of (25b). Since the parents of this student are previously
identified information in the preceding context, the two nouns
'mother' and 'father' in (25a) should be marked with topic marker
nun respectively. The expected grammatically correct version of
(25a) would be (25b) in which the first NP marked with nun triggers
a thematic reading, whereas the second NP with nun is associated
with contrastive reading 'on the other hand'. Thus, it is evident that
German students tend to overproduce nominative subject markers
and that the acquisition of Korean by German students is guided by
typological differences. We notice that the typological transfer of
subject prominent features takes place continuously in the acquisition
of Korean and that the pragmatically sensitive grammatical structures
of Korean are cognitively hard to acquire for German native speak-
ers.
As regards the double nominative constructions as in (17) and
(18), not a single instance is attested in the current database for level
1 and level 2 participants. It seems safe at this point to argue that
compared to the other TP features such as null subject and the use of
topic marker, the acquisition of double subject construction takes
place later, although this construction is accessible to the learner as
an early input.
In this section, we have observed that when German students learn
a TP language such as Korean, they rely on known linguistic knowl-
edge before identifying the TP features of Korean. For a SP language
such as English the primary requirement is that a subject should be

specified, and that it should be realized before the verb. In addition,
an object must also be mentioned in the case of transitive construc-
tions (cf. SVO structure). For a TP language such as Korean or Japa-
nese, it is required that the topic should be mentioned first in the
sentence; the need to name a subject is a secondary matter: the sub-
ject is not marked by position and there may be no subject at all. As
we have seen in (16) in section 3, the topical item or the shared in-
formation is usually deleted in the most natural discourse setting. In
this section, we presented L2 evidence that the grammatical structure
of native language controls or influences the output of adult second
language learners considerably. It is clear that successful learning
rests on the identification of the structural properties of the target
language. The analysis of the data in this study also reveals that there
are developmental stages of Korean acquisition by German native
speakers. We have found that German learners of Korean overpro-
duce subjects early on and then gradually learn to drop them as their
proficiency improves, and that topic particles and double nominative
constructions are particularly difficult features for Germans to ac-
quire even in the later stage of learning. Thus, this process of learning
Korean by Germans suggests that learners adopt the L2 acquisitional
strategy of going for the typological indices that simplify to a maxi-
mum the procedure for learning a target-language grammar (cf. also
Jin 1994 for Chinese).
6. Conclusion
In this paper we have developed the idea that the transitivity pa-
rameter should be constrained such that pragmatically determined
word order languages with high semantic transparency proportionally
correlate with less grammaticization. The effects of this theoretical
position are also reflected in the L2 data, such that the kinds of L2
competence that need to be acquired going from languages at each
end of the typological continuum proposed in section 4 are either
predominantly constructional or predominantly discoursal. It is evi-
158 Sang Hwan Seong
dent that SP languages are construction-based whereas TP languages

are discourse-based. We have also seen that linguistic expressions
that signify the same objective circumstances are construed differ-
ently from language to language. As Langacker put it (1988: 11), the
ways in which functionally equivalent linguistic expressions are pre-
sented can be language-specific and in fact may vary from language
to language. But it is not the case that the symbolic elements avail-
able to speakers are completely language specific and differ unpre-
dictably. The idea which the present author wishes to promote is that
the ways in which languages encode things are to a considerable ex-
tent parameterized in a regular fashion when comparable thoughts are
expressed. In this respect we can state that the cognitive grammar
view of languages is not completely relativistic.
Notes
* I thank René Dirven and the other two anonymous reviewers for their valuable
comments and suggestions on this paper. I am also indebted to John Attfield
and Noah Isenberg for carefully reading the manuscript. All responsibility for
errors remains of course my own.
1.As an anonymous reviewer points out, example la (intransitive verb) is of a
different class from lb and lc (transitive verbs). However, this difference is not
so relevant to this discussion because we are dealing with non-agentive subjects
in English.
2. Some examples from the old Germanic dialects are in order (Valentin 1999):
(1) jah warP in jainaim dagam... Gothic, Mk 1, 9 ('and it happened in
those days')
(2) uuerda thin uuilleo obar thesa uuerold alia. Old Saxon, Heliand 1604
('your will be done all over the world')
(3) a. that thi kind giboran...scoldi uuerdan. Old Saxon, Heliand 123
('that the child should be born')
b. nu ist Krist giboran. OS. Heliand 399 ('Christ is now born', 'Christ
has now been born')
(4) thaer hali-ern wearth tha geopenod and tha lac waeron in gebrohte.
Old English ('the sacred house was opened and the offerings were
brought in')
3. Kuno (1973) illustrates the Japanese equivalent of the similar reflexive con-
struction in (11). However, he does not explain why such a grammatical differ-
ence exists between English and Japanese.
4. There is evidence that the serialization of dative argument + accusative argu-
ment in ditransitive constructions should count as basic and unmarked order in
German. Look at the following observation of Lenerz (1977: 44) which in-
volves the focus pattern of dative and accusative objects:
a. Ich habe das Geld dem KASSIERER gegeben.

I have the money the cashier given
b. *Ich habe das GELD dem Kassierer gegeben.
c. Ich habe dem KASSIERER das Geld gegeben.
d. Ich habe dem Kassierer das GELD gegeben.
When the accusative patient precedes the dative recipient, the accusative argu-
ment has to be unfocused or unrhematised for the sentence to be grammatical.
However, no such restriction applies to the reverse order "dative + accusative"
serialization. Thus, the latter order should count as "unmarked" in the ditransi-
tive constructions. This is also confirmed by Eisenberg (1994: 422).
5. Interestingly enough, we find this regularity in the diachronic development of
German and Dutch. Shannon (1999) demonstrates, based on the following in-
stances that the ordering of pronominal object and nominal subject in the mid-
dle field of Middle Dutch (MD) was an unmarked case and that Modern Dutch
has largely given up this structure in favor of the opposite ordering.
MD Doe sette hem sijn vader op sijn peert achter hem.

D Toen zette zijn vader hem op zijn paard, achter hem.
'Then his father set him on his horse, in back of him.'
Shannon also adds that German has preserved this pragmatically determined
word order considerably, but has seen a substantial decrease.
6. An identical grammatical phenomenon is confirmed in Japanese (Y. Kasai
[p.c.]). For details, see Seong (2000).
References
Abraham, Werner
1995 Diathesis: The middle, particularly in West-Germanic. In: Werner
Abraham, Talmy Givón, and Sandra Thompson (eds.), Discourse
Grammar and Typology, Papers in Honor of John W. M. Ver-
haar, 3-47. Amsterdam: John Benjamins.
160 Sang Hwan Seong
Burridge, Kate
1993 Syntactic Change in Germanic. Amsterdam: John Benjamins.
Chafe, Wallace
1976 Givenness, contrastiveness, definiteness, subjects, topics, and
point of view. In: Charles Li (ed.), Subject and Topic, 25-55.
New York: Academic Press.
Drosdowski, Günther (ed.)
1995 Duden. Grammatik der deutschen Gegenwartssprache.
Mannheim: Bibliographisches Institut.
Eisenberg, Peter
1994 Grundriss der deutschen Grammatik. Stuttgart: J. B. Metzler.
Faarlund, Jan T.
1992 The subject as a thematic category in the history of Scandinavian.
Folia Linguistica. XXVI. 1(2): 151-68.
Fillmore, Charles
1968 The case for case. In: Emmon Bach and Robert T. Harms (eds.),
Universals in Linguistic Theory, 1-88. New York: Holt, Rinehart
& Winston.
Firbas, Jan
1992 Functional Sentence Perspective in Written and Spoken Commu-
nication. Cambridge: Cambridge University Press.
Foley, William A. and Robert D. Van Valin
1984 Functional Syntax and Universal Grammar. Cambridge: Cam-
bridge University Press.
Givón, Talmy
1979 On Understanding Grammar. New York: Academic Press.
Goodluck, Helen
1991 Language Acquisition. Oxford and Cambridge: Blackwell.
Gundel, Jeanette
1988 Universals of topic-comment structure. In: Michael Hammond,
Edith Moravcsik, and Jessica Wirth (eds.), Studies in Syntactic
Typology, 209-242. Amsterdam: John Benjamins.
Haider, Hubert
1985 The case of German. In: Jindrich Toman (ed.), Studies in German
Grammar, 65-101. Dordrecht: Foris.
Hawkins, John A.
1986 A Comparative Typology of English and German". Unifying the
Contrasts. Austin: University of Texas Press.
Helbig, Gerhard and Joachim Buscha
1994 Deutsche Grammatik: Ein Handbuch för den Ausländer-
unterricht. Leipzig: Langenscheidt.
Hopper, Paul and Sandra Thompson

1980 Transitivity in grammar and discourse. Language 56: 251-99.
Jin, Hong G.
1994 Topic-prominence and subject-prominence in L2 acquisition:
evidence of English-to-Chinese typological transfer. Language
Learning 44(1): 101-22.
Klaiman, Miriam H.
1988 Affectedness and control: a typology of voice systems. In: Masa-
yoshi Shibatani (ed.), Passive and Voice, 25-83. Amsterdam:
John Benjamins.
Kuno, Susumu
1973 The Structure of the Japanese Language. Cambridge: The MIT
Press.
Langacker, Ronald
1988 An overview of cognitive grammar In: Brygida Rudzka-Ostyn
(ed.), Topics in Cognitive Linguistics, 3-48. Amsterdam: Ben-
jamins.
Lenerz, Jürgen
1977 Zur Abfolge nominaler Satzglieder im Deutschen. Tübingen:
Narr.
Li, Charles and Sandra A. Thompson
1976 Subject and topic: A new typology of language. In: Charles N. Li
(ed.), Subject and Topic, 457-89. New York: Academic Press.
Müller-Gotama, F.
1994 Grammatical Relations: A Cross-Linguistic Perspective on their
Syntax and Semantics. Berlin: Mouton de Gruyter.
Nam, Ki-Sim and Koh, Young-Kun
1993 Standard Korean Grammar. Seoul: Top Publishing (in Korean).
Noonan, Michael
1977 On subjects and topics. Proceedings of the Berkeley Linguistic
Society 3: 372-85.
O'Grady, William
1991 Categories and Case. Amsterdam: John Benjamins.
Plank, Frans
1983 Transparent versus functional encoding of grammatical relations:
a parameter for syntactic change and typology. In: Linguistische
Berichte. 86: 1-13.
Seong, Sang H.
1999a On word order in German and Korean. In: Gerald F. Carr, Wayne
Harbert, and Lihua Zhang (eds.): Interdigitations·. Essays for Ir-
mengard Rauch, 355-366. New York: Peter Lang.
162 Sang Hwan Seong
1999b Semantic transparency and its implications: with special reference

to German and Korean as SOV type. In: Irmengard Rauch and
Gerald F. Carr (eds.), New Insights in Germanic Linguistics I,
233-260. New York: Peter Lang.
1999c Are Behaghel's Laws heterogeneous ones? To appear In:
Sprachwissenschaft auf dem Weg ins nächste Jahrtausend: the
Proceedings of the 34th Colloquium of Linguistics at the Univer-
sity of Mainz, Germersheim.
2000 Wie hätte Behaghel Koreanisch und Japanisch analysiert? Paper
presented at the Ost-West Kolloquium für Sprachwissenschaft (3.
17-18,2000) at the Humboldt University in Berlin. (MS)
Shannon, Thomas F.
1987 On some recent claims of relational grammar. In: Aske, Jon, Na-
tasha Beery, Laura Michaelis and Hana Filip (eds.), Proceedings
of the Berkeley Linguistic Society 13: 247-262.
1999 German word order change at the Millennium. To appear in:
Sprachwissenschaft auf dem Weg ins nächste Jahrtausend: the
Proceedings of the 34th Colloquium of Linguistics at the Univer-
sity of Mainz, Germersheim.
Shibatani, Masayoshi
1985 Passives and related constructions: A prototype approach. Lan-
guage 61: 821-848.
Sohn, Ho-Min
1994 Korean. London: Routledge.
Talmy, Leonard
1988 Force dynamics in language and cognition, Cognitive Science 12:
49-100.
Taylor, John
1995 Linguistic Categorization: Prototypes in Linguistic Theory. Ox-
ford: Clarendon Press.
Traugott, Elisabeth C. and Bernd Heine (eds.)
1991 Approaches to Grammaticalization. Amsterdam: John Ben-
jamins.
Valentin, Paul
1999 Wirdhu and the Germanic passive. In: Gerald F. Carr, Wayne
Harbert, and Lihua Zhang (eds.), Interdigitations: Essays for Ir-
mengard Rauch, 141-46. New York: Peter Lang.
Van Oosten, Jeanne H.
1984 The nature of subjects, topics and agents: A cognitive explana-
tion. Unpublished Ph.D. Dissertation. University of California,
Berkeley.
Voorst, Jan G. van

1996 Some systematic differences between the Dutch, French and
English transitive construction, Language Science, Vol. 18, Nos
1-2: 227-245.
Section 3
Neurocognitive and cognitive issues

of language acquisition in general
Learning syntax - a neurocognitive approach
Sydney M. Lamb
Not only every language, but every lexeme of a language,

is an entire world in itself.
Igor Mel'chuk
Like other papers in this collection, this one is concerned with learn-
ing, in particular, with the learning of syntax. It addresses the ques-
tion of how children, or adults learning a second language, learn to
handle what appear to be the syntactic categories needed for using a
language. And in order to talk about this question there is a very im-
portant prerequisite: We need to understand just what it is that is be-
ing learned.
According to a traditional view of the learning of syntax, the child,
or the adult second-language learner, must somehow acquire syntac-
tic rules. In one manner of speaking it is assumed that a speaker has
"internalized" such rules. The assumption is that since syntactic rules
are useful in descriptions of sentences, they must be present within
the system that produces them. This mode of thinking comes from an
unstated assumption, that patterns observable in linguistic data repre-
sent knowledge in the minds of those who produce such data. Is this
assumption supported by any evidence? I have a hard time finding
any basis for it. It is somewhat like supposing that since we can de-
vise equations for describing the movements of planets around the
sun, those planets must have internalized such equations. If we can
find other sources for the patterns found, there is no reason to adopt
this assumption (Lamb 1999: 227-247).
168 Sydney M. Lamb
Of course, there has to be some internal system that accounts for

what people are able to say and to understand. But it need not be as-
sumed to have the form of rules used in classifying utterances, nor is
there any a priori reason for assuming that it contains the categories
and other devices that may be employed by those attempting to de-
scribe them. And such attempts are particularly suspect as formulated
by those who attempt to describe them in the most economical possi-
ble way, hence with the broadest possible categories and what have
erroneously been called "linguistically significant generalizations".
The discrepancies between the results of such taxonomic thinking
and a more realistic view of the cognitive reality may be seen at all
levels of linguistic structure, but they are nowhere more evident than
in the area of syntax.
An alternative is to treat most syntactic information, or even all of
it, as attached to individual lexical items. In that case, the acquisition
of syntactic knowledge is part and parcel of the acquisition of lexical
knowledge and therefore occurs little by little as individual lexemes
are learned. This view has become increasingly attractive in recent
years.
Yet there is also a lot of evidence for the existence in our cogni-
tive systems of some kind of constructions, for example the argu-
ment-structure constructions described by Goldberg (1995). Such
constructions, to be considered below, evidently make use of syntac-
tic categories. And so we have a problem.
Tomasello and Brooks (1999), who accept the cognitive existence
of constructions of this kind (cf. Tomasello 1998) identify areas that
require further investigation for developing a viable theory of syntac-
tic learning, stating that "the various psychological processes in-
volved in early syntactic development [...] need to be identified and
characterised" (1999: 185). Of three such processes they identify, I
would like to focus on this essential one: "[...] children's early skills
to categorise not only isolated bits of language into item-based cate-
gories, but also their skills at categorising larger linguistic units into
the various syntactic schémas and constructions that underlie much
the productivity of human language" (1999: 185).
Learning syntax - a neurocognitive approach 169
To put it briefly: If there are syntactic constructions in a neurocogni-

tive system, it is necessary to consider how they might be acquired by
the child. Categories as such are not presented to the child for learn-
ing, only actual expressions.
1. What is it that has to be learned?
Of course, before we can consider learning we need to understand

what it is that is being learned. We need to consider the question of
just what kind of information is involved. We commonly think of a
category as a combination of objects - in this case linguistic objects.
Two considerations make a difference here: First, what kind of ob-
jects? Second, what kind of combination? We can bypass such ques-
tions only at the danger of adopting unwarranted assumptions.
One approach, too simple, would have it that the objects are mor-
phemes. That is clearly to be rejected, as syntactic categories often
have sequences as members. Moreover, the proper basic unit for
syntax is not the morpheme but the lexeme. The term lexeme was
coined by Benjamin Lee Whorf over half a century ago, but has yet to
become widely used, despite what seems to me its obvious useful-
ness, indeed its indispensability. Part of the reason is that linguists
have been so influenced by analytical rather than cognitive concerns.
In the analytical tradition one attempts to analyze as much as possi-
ble, hence down to minimal units; for grammar, down to the mor-
phemes. By contrast, the cognitive view of what a person learns,
taking into account the workings of the brain, recognizes that people
quite commonly learn larger combinations as units, and not just for
language. For example, the lexeme activity is surely learned and used
as a unit by English speakers despite the fact that it can be analyzed
into three morphemes. This principle applies much more broadly as
well. We learn whole phrases and clauses as units (hence, phrasal and
clausal lexemes), like it doesn't take a brain surgeon to realize that
... and tell it like it is.
Another common misconception is that words are the units with
which syntax is concerned. But a lexeme can consist of multiple
170 Sydney M. Lamb
words or of just part of a word, for example, the plural ending -s or

the past tense ending of verbs; they occur freely even with newly
coined nouns and verbs.
The other question posed above is of greater interest: what kind of
combination is a category? Is it a set of lexemes and combinations of
lexemes? For example, do we need to know a category Noun Phrase
as a list whose members are all the noun phrases of English? Clearly
not, for there are many possible noun phrases that, while perfectly
acceptable once received in a suitable context, have never yet been
uttered. We might say, adopting a commonly held point of view, that
a syntactic category is a set consisting partly of individual forms
(lexemes) and partly of combinations generated by rules. Such rules,
representing constructions, of course use categories themselves, and
their categories are similarly defined. According to this view, the
knowledge one needs to learn consists of the rules for combinations
plus a list of the individual members of the categories (like the nouns,
verbs, etc.) not specified by rules. But I reject such a view, not only
as too simple minded but, more important, as cognitively implausi-
ble. In this paper I develop an alternative view.
Still under the heading of considering what it is that must be
learned in order to have a command of the syntax of a language, let
us next be explicit that syntax is concerned with combinations of
lexemes and indeed with a hierarchy of combinations - phrases,
clauses, sentences, etc. So we need to take a look at the kinds of
combinations a person must be able to command, starting with the
simplest, a "combination" of one. And of course we take a neurocog-
nitive point of view.
First, then, we have the lexeme. But I would like to suggest that
we should recognize many more units as lexemes than are usually
considered. The cognitive orientation forces us to accept that people
learn as units any combination that has occurred with sufficient fre-
quency or to which sufficient attention has been given, as a conse-
quence of the brain's natural tendency to "absorb" repeatedly occur-
ring phenomena. If a locution attracts enough attention, it will be
learned on the basis of very few occurrences. For example, at the
time of the Persian Gulf War, a new lexeme was introduced into
English: the mother of all battles. It arose from a statement of Sad-

dam Hussein, or rather from the English translation of his statement,
which in Arabic was less colorful. It only took that one statement,
heard perhaps once or a very few times by Americans in newscasts
over the next few days, for that lexeme to be learned and, for a while,
widely used. This one also illustrates that some lexemes have a rather
transitory existence, as its use gradually declined over the next sev-
eral months to the point at which ten years later one encounters it
only rarely if at all.
So we have simple lexemes and complex lexemes. The latter are
not to be conflated with idioms, by the way. Idioms are those com-
plex lexemes whose meaning is not clear from the meanings of their
constituents. That is, a lexeme can be transparent or opaque, and
these are not two values but the endpoints of a continuous scale.
Some lexemes are quite transparent, like plastic bag, others are quite
opaque, like red herring. A lexeme like blackboard is relatively
transparent, but a blackboard doesn't have to be black. Another con-
tinuous scale relating to lexemes is that of entrenchment. A lexeme
becomes more entrenched with more use: The neurocognitive path-
ways which support it become stronger the more they are traveled
(Lamb 1999: 164-166). This is the first of several observations in
this paper which support the hypothesis that a neurocognitive system
has the form of a network.
Transparent lexemes can be interpreted in two ways: either via the
constituents or via the whole. This is no problem for a network ap-
proach. In fact, the fact that both kinds of interpretation do occur,
presumably in parallel, constitutes further evidence in favor of the
network model (cf. Lamb 1999: 167).
It is surprising how much ordinary English text is made up of
complex lexemes. This observation is important for the study of the
cognitive operations relating to syntax since combinations which are
"stored" in memory as units don't have to be constructed for their
production or understanding. Consider the following illustrative ex-
amples of lexemes in English:
172 Sydney M. Lamb
Relatively short:
horse sense strictly speaking painfully obvious
no laughing matter a people person a no-brainer
not written in stone the bottom line a dumb question
as clear as mud a New York minute right then and there
Relatively longer:
round up the usual suspects it ain't over till it's over
if it ain't broken don't fix it you know what I mean
you can bet your bottom dollar the truth of the matter is
between a rock and a hard place been there, done that
But of course there is a great deal that cannot be accounted for so

simply. At the next level of complexity, I would like to suggest, we
have the complex lexeme with a variable constituent. Any complex
lexeme can be said to be composed of constituents; for example, red
herring has red and herring as its constituents. For very complex
lexemes, like a stitch in time saves nine, we can also distinguish im-
mediate constituents from ultimate constituents. The variable con-
stituent can be illustrated by what happened to the lexeme the mother
of all battles soon after it was introduced into English. It developed
into what can be called a mutable lexeme (Lamb 1999: 263-266), as
the constituent battle mutated into others, like meteor. A story in the
New York Times describes a spectacular meteor that was seen in the
sky in the Eastern United States at around the time of the Persian
Gulf War. An airline pilot who witnessed it was quoted as calling it
the mother of all meteors.
Moving on to the next level of complexity, we have what could be
called the lexeme with more than one variable constituent. As exam-
ples, we have
you don't have to be a <X> to <Y> that

X: brain surgeon, rocket scientist
Y: understand, appreciate, see
it comes as no [great] surprise that
The constituent enclosed in square brackets is an optional constitu-

ent, a type of variable in that it is a constituent varying with zero.
2. Constructions
Pushing this notion further, we arrive at the limiting case, in which

all constituents are variable and in which all have a wide range of
values. This is what is more commonly called the construction, in
which the variable constituents are called syntactic categories. We
have as examples the argument structure constructions treated by
Adele Goldberg (1998). Following Goldberg, we may identify these
constructions:
intransitive motion The fly buzzed into the room

ditransitive he faxed Bill a letter
caused motion she pushed the pencil off the table
resultative they wiped the counter clean
conative she kicked at Henry
According to Goldberg, "Constructions which correspond to basic

sentence types encode as their central senses event types that are ba-
sic to human experience." (1995: 39).
I differ with Goldberg's treatment in considering all of these to
represent verb phrase constructions rather than sentence types. This is
a minor difference and does not require us to change Goldberg's
names for them. The treatment as verb phrases is needed to account
for their occurrence in the infinitive form, for example, (ditransitive)
to fax Bill a letter (that is, with no subject expressed), as in The boss
asked her to fax Bill a letter, or as participles, as in (intransitive mo-
tion) Sauntering into the room, she cast a glance my way. None of
them have anything special in the relationship of the subject to the
verb phrase.
Instead, I believe we need to reckon with another basic construc-
tion with very broad scope, the Actor-Action construction (so-called
by Bloomfield 1933: 172).
174 Sydney M. Lamb
Additionally, we need to consider, along with the verb phrase con-

structions listed above, the more basic one,
transitive she smacked Helen (cf. Goldberg 1995: 117-119)
In this illustrative example, I include the subject she not because it is

a part of the construction but just to keep the example parallel to
those listed above. As with them, the addition of the subject comes
about in keeping with the operation of the actor-action construction.
Related to the transitive construction, in fact dependent on it, is
the
passive Helen was smacked [by her]
I take it as reasonable to assert that all combinations of lexemes, that

is all of syntax, can be accounted for in terms of these types dis-
cussed above: complex lexemes, complex lexemes with variable con-
stituents, and constructions (which could be called complex lexemes
with multiple variable constituents, except that the term construction
is more established and more convenient).
3. Syntactic categories and their members
Now we are ready for our basic question, that concerning the nature
of categories. We have to ask just what information must a child (or
second-language learner) acquire in order to handle the syntactic
categories of the language. Does such knowledge consist, for exam-
ple, of a listing of the membership of the simple categories (like
noun, verb) together with a set of rules for generating the more com-
plex ones. If not, what? It is important to recognize that the notion of
category comes from analytical linguistics, an essentially non-
cognitive endeavor. As with other tools of analytical linguistics, we
are not obliged to suppose that they are internalized in the mental
systems of speakers.
In keeping with what has been observed so far, we could rephrase the
question in terms of variable constituents rather than categories: How
does a language learner learn how to use variable constituents? But
for now, let us continue with the more traditional notion of syntactic
categories.
The first observation, an important one, is that as syntactic catego-
ries have been traditionally employed, they don't work. They are just
rough approximations - not wholly useless, just not cognitively plau-
sible. Approaching them with prototypicality notions helps, but ulti-
mately, the closer we look at any syntactic category, the more it
breaks down, until we get down to the individual lexeme.
Consider, for example, the category of prepositions. If they do
constitute a syntactic category, they behave alike syntactically. If they
do not, how can the category specify their combinations? So let us
take a look: We have in love but not *on love, yet we have on vaca-
tion but not *in vacation. The following are examples of the use of
basic prepositions with various objects for describing states that one
can find oneself in or in which one can do things. We have:
Preposition Object
*at, *by, ?in, *out of, ?under, with assurance
*at, *by, ?in, *out of, *under, with candor
?at, by, ?in, *out of, *under, *with chance
?at, *by, in, out of, *under, ?with danger
*at, *by, in, out of, *under, ?with desperation
*at, *by, in, out of, ?under, ?with doubt
*at, *by, in, ?out of, *under, ?with fear
*at, *by, in, ?out of, *under, with insecurity
at, *by, *in, *out of, *under, *with leisure
*at, *by, in, out of, *under, with love
*at, *by, in, out of, *under, *with pain
at, *by, in, out of, *under, ?with play
?at, *by, ?in, *out of, *under, with pleasure
*at, *by, *in, *out of, under, *with pressure
at, *by, ?in, *out of, *under, *with rest
*at, by, in, *out of, *under, *with thought
*at, *by, *in, *out of, *under, *with vacation
176 Sydney M. Lamb
Notice that the last noun listed, vacation, does not occur with any of
the prepositions considered here, but does occur with on, and that on
does not occur with any of the other nouns in the list. By the way, the
judgements presented are mine and they may differ in some details
from those of other native speakers. For a neurocognitive approach,
that is no problem, since the object of investigation in neurocognitive
linguistics is the neurocognitive system of the individual speaker, not
some disembodied "language"; and we recognize that the system of
every speaker differs from that of every other.
Given findings like those listed above, what cognitive sense can
we make of the notion that there is a construction utilizing the cate-
gories Preposition and Noun Phrase?
Syntactic categories are based upon an old tradition, that of the
"parts of speech", which goes back to the ancient Greeks. According
to the doctrine of parts of speech, every word must belong to one or
more of these categories - and there is a catch-all category "adverb"
for the difficult cases. Now, what about tantamount! According to
the ordinary dictionary, which treats the parts-of-speech myth as re-
ality, it is classed as an adjective, as is the equally unique akimbo.
Also of clearly unique distribution, but classed as a preposition, is
between.
Let's take a brief look at verbs. From the point of view of their
syntactic distribution there are clearly many different kinds. We
might be tempted to suppose that if we subcategorize to a sufficient
degree we will get down to subcategories whose members behave
alike. We might, for example, consider just the subcategory of verbs
of perception: see, hear, listen, smell, etc. But only a little observa-
tion makes it clear that even this tiny subcategory doesn't help us to
define what can and can't occur syntactically. Apart from the fact that
we see visible objects but hear audible things, we have the different
aspectual possibilities: You see and hear punctually and comple-
tively, but you listen [to] and look at duratively. We don't have to
look very far to see that each of these verbs, and indeed every verb of
any other subcategory, has its own distribution. And the same can be
observed about members of any of the other parts of speech.
And so we conclude that every lexeme has its own syntax. Now that
is a conclusion which puts the validity of the concept of syntactic
category into considerable doubt. How can there be syntactic catego-
ries, as constituting part of the information used by speakers of a lan-
guage, if they don't work? And if they are useless, how can we ex-
plain why they come up so regularly in discussions of syntax? The
answer to this question is that such categories result from applying
the taxonomic techniques of analytical linguistics. Analytical lin-
guistics is concerned with analyzing and categorizing and describing
patterns found in things that people say. It is natural in such a pursuit
to classify things that appear to share properties. In such classifica-
tion it is easy to overlook that the things categorized together do not
really behave alike. In any case, there is no justification for assuming
that concepts of analytical linguistics can be taken over directly into
an understanding of the cognitive basis of language.
An alternative to the taxonomic approach is that most syntactic in-
formation, or even all of it, is attached to individual lexical items. In
that case, most of the syntactic generalizations that can be formulated
in rules are epiphenomenal, and the actual internal information that
gives rise to them is widely distributed, among thousands of separate
items. According to such a view the acquisition of syntactic knowl-
edge is the acquisition of lexical knowledge.
4. Learning syntax as (mostly) learning lexicon
If it is the case that every lexeme has its own syntax, then it follows
that the only way to learn syntax is to learn lexicon. If this is so, then
what seemed to be a process of learning syntax is really just the
learning of vocabulary, a process that occurs one lexeme at a time.
This view makes sense not only because every lexeme has its own
syntax, but also because, as a consequence, you can't know how to
use a lexeme without knowing how it connects with other lexemes.
This conclusion is strongly supported by findings of Elizabeth
Bates et al. (In press), who have examined the correlation between
development of grammatical complexity and vocabulary size in chil-
178 Sydney M. Lamb
dren. They have found these two aspects of language development

are very strongly correlated. As they write,
... the relationship between grammar and vocabulary development has

turned out to be one of the strongest associations that has ever been ob-
served in any aspect of human development (with the possible exception of
height and weight!). ... the relationship holds at every point from 50-600
words (covering the period from 16-30 months of age). One certainly might
have expected a relationship at the lower end of the distribution, simply be-
cause one cannot combine words until there is something to combine. We
might also have expected some kind of "trigger" or "threshold" relationship
between vocabulary and grammar, e.g. a critical number of words that need
to be acquired for grammar to get off the ground. What we find instead is a
continuous and accelerating function that holds at every point across this pe-
riod of development.... The powerful relationship holds for very late talkers,
and very early talkers as well.
Yet to conclude that learning syntax is just part of learning vocabu-

lary leaves us with a lot to explain: What about the broad similarities
in ordering: Subject before predicate (in the unmarked clause),
preposition before noun phrase, etc.? And what about the general
constructions like the argument structure constructions described by
Goldberg? They surely use categories. And what about the ability,
even of young children, to produce and to understand new combina-
tions?
5. Participant roles
Let's consider the case of a simple verb like eat, accepting the obser-
vation that its syntactic distribution is unique. That being the case, we
are tempted to conclude that the operative knowledge used in pro-
ducing a sentence like Mommy's eating an apple, with its ordering of
the two participants in relation to the process - the agent before and
the patient after the verb - is in large part information connected di-
rectly with the lexeme eat. Yet it seems also to be the case that the
actor-action construction and the do-smthg-to-patient construction
(i.e. the transitive construction) are also involved, even if their in-
volvement can be seen as peripheral in relation to that of <EATER>

EAT <FOOD>. Ordinary analytical linguists will balk at the sugges-
tion of such a proposal, since the two methods of handling the com-
bination appear to be in competition with each other. Why have two
methods where one will suffice? But the cognitive considerations are
essential here, as elsewhere. We have evidence that both kinds of
knowledge may be presumed to be present in the cognitive system.
For example, without the actor-action construction and the transitive
construction, how can we account for the passive construction? How
two seemingly competing methods of producing the combination can
coexist is no problem for a network approach. The network allows
multiple structures to operate in parallel (cf. Lamb 1999: 233-236).
The presence of redundancy is likewise no problem in a cognitively
realistic approach, as we need to recognize that our brains do learn
many kinds of things redundantly.
As a transitive verb, eat co-occurs with an agent and a patient, and
so it entails two categories: that of its possible agents and that of its
possible patients. We have two questions to consider: First, what
does the information specifying the membership of each of these
categories consist of, how is it organized, how represented in the
mind of the child? Second, how is this information learned? How
does the child (or the grown-up learning a second language) learn the
ordering of these major constituents of the clause, often seemingly
after having heard just a few examples?
What, then, is the status of <EATER>? IS it a category? If it is, it is
a semantic category. And as such it would seem to consist of all the
possible agents of EAT. But the more relevant question is: What must
a child learn, what must a speaker of English know, in order to use
EAT productively? First, we have to understand that it is not possible
to understand what eating is apart from eaters and eatees. Knowledge
of a range of possible participants is part and parcel of the knowledge
of the process - not something separate, as the lexemes Mommy and
apple are separate from eat in the linguistic expression of the process
MOMMY EAT APPLE. You can't have eating without <EATER> and
<EATEE> - it is simply impossible. Therefore, the participant infor-
mation must be attached to individual processes. The same observa-
180 Sydney M. Lamb
tions can be made for perception: Your visual system apprehends a

person walking; the walking simply cannot be perceived without per-
ceiving the person (or other legged creature) performing the process.
Now we can explain why it is that a child doesn't need more than
a very few examples like this to learn how to express instances of
eating together with its participant categories. The child already
knows what eating is through his/her own experience of eating and
his/her observation of others. That is, she already has a subnetwork
of semantic information, with connections in the motor and kines-
thetic areas of her cortex, representing her own experience, together
with visual and perhaps also auditory connections, based on observa-
tion of others. It is just a question of connecting this semantic sub-
network to a means of linguistic expression. As with lexemes gener-
ally, a child (or grown-up - in fact anyone except an analytical lin-
guist) focuses on the meaning that a lexemic node is connected to
more than on the lexeme itself. And you simply can't know the
meaning of Leat without knowing that the process requires eater and
food; and so the participant categories are simply EATER and EATEE.
We don't need to ask that these categories be more clearly defined,
nor that they be defined in some general way that could apply for the
whole speech community. They will be defined for each child ac-
cording to that child's semological system, according to that child's
knowledge (already present) of what eating is, as it exists at the time
of learning of the word eat. Later, as the child learns more, it may
change its understanding of the categories, as an automatic conse-
quence of changing her understanding of what constitutes eating.
So what the child actually needs to learn is how to form linguistic
expression for the already existing semantic information. The expres-
sion includes not only the phonological realization of eat but the se-
quencing information: The realization of the <EATER> comes first,
then eat, then the realization of the <EATEE> (unless some other con-
struction like passive intervenes). It only takes one example, if clearly
understood, perhaps then supported by one or two additional exam-
ples for confirmation and reinforcement, for the child to have control
of the lexemic realization of the process.
In other words, since there are two participants in the (already

known) process, the child has to learn not only a phonological ex-
pression for eat but also the ordering of the participants in relation to
it. Until this ordering information has been learned the child knows
from experience and observation that eating involves two participants
but doesn't know their relative order in the linguistic expression of
the process. As soon as she hears (and pays attention to) an example,
like Mommy's eating an apple she knows, since she recognizes
MOMMY as eater and APPLE as eatee. That is, she now knows that
<EATER> precedes eat and <EATEE> follows. So now, without any
additional information that is not already in her network, the child is
able to use the word eat more or less like older members of the
speech community.
In the future use of this verb, a child (or adult) will produce and
understand it in some ways that depart very little from those encoun-
tered earliest and most often, and in others that are more different -
perhaps some day for a car "eating" gasoline and even for acid "eat-
ing" pipe. It depends entirely on whether the process under consid-
eration is construed as an instance of eating.
And so <EATER> is not a category whose membership has to be
learned. Rather, the value of <EATER> depends entirely upon the
knowledge of what constitutes eating. For the producer of an occur-
rence of <EATER> EAT <FOOD>, it is entirely a question of whether or
not the action under consideration can be construed as eating. If so,
then it has an agent, as it must, since eating cannot occur without an
eater. This point is at once very simple and altogether essential for an
understanding of the knowledge underlying syntactic performance.
Its simplicity makes its import easy to overlook, so I shall emphasize
it by repeating, in different words: If the process under consideration
is construed as eating, then whoever/whatever is doing it is the
<EATER>, and by virtue of just that fact becomes at that moment a
member of the "category" <EATER>. The consequence of this simple
observation is that the child does not have to learn the membership of
the category as such. Rather, the learning needed is just the learning
of what eating is. And that knowledge is already present as concep-
tual and perceptual and motor network structures - even before the
182 Sydney M. Lamb
child learns the lexeme Leat. For any normal child has such struc-
tures, and therefore a concept of what eating is, from early on in its
life. To be sure, the knowledge of what eating is may well undergo
further development as the child matures and learns more about the
culture, the mythology, stories, and so forth. Eventually a person may
come to accept, or to reject, such notions as an automobile eating
gasoline or a furnace eating coal. He may well come to appreciate the
riddle What has three legs and eats marbles? (Answer: A three-
legged marble-eater). Whether accepted or rejected, it depends not
upon the "category" <EATER> as such but on EAT. If the process can
be called eating, then whatever is performing it is the <EATER>. NO
separate knowledge of that "category" is needed.
The foregoing observations definitely simplify the task of the cog-
nitive syntactician, for they eliminate the whole problem of deter-
mining what knowledge must be learned to learn such „categories".
For the answer is that no additional knowledge is needed beyond
knowledge of the process itself. To be sure, there is still plenty to
investigate: What form does the knowledge of eating have in a per-
son's brain, and how is that knowledge acquired? But note well that
the problem of answering those questions was a problem already pre-
sent for cognitive neuroscience anyway. What I am claiming is that
no further knowledge beyond that is needed for syntactic purposes.
In the preceding two paragraphs I have started to put category in
quotation marks, because it is apparent that we are no longer talking
about what the term has commonly meant in discourse about syntac-
tic categories. The difference is clearly seen in the context of the
question of what information in the cognitive system gives rise to the
appearance of categories in analytical linguistics. More commonly
that information would be seen as, in one way or another, specifying
the membership of the category. That notion of category, as involving
one-and-many, however useful it may be in analytical linguistics, is
now seen to represent an illusion from the neurocognitive point of
view. For this reason I prefer the term variable, free from such con-
notations, and so I shall use it from now on.
6. Syntactic presupposition and variable constituents
And so, as has been observed many times in the past, we have verbs
presupposing participants while nouns do not. And we have not just
theoretical reasons based on analysis of linguistic data for such an
assertion. It is surely related to the fact that Broca's aphasies, whose
area of damage is in the frontal lobe, typically have trouble not only
with phonological production and with grammar, but also with verbs,
much more so than with nouns. And they also have trouble with
prepositions and with "function words" generally.
All lexemes other than nouns evidently presuppose some other
constituent or constituents with which they normally co-occur, just as
eat presupposes <EATER> and <EATEE>. We have for example clause
introducers like clearly, which presuppose a following clause or
<ASSERTION>. Like it are other assertion introducers, including those
described above as lexemes with variable constituents. We now see
that what was written above in citing them was incomplete and that
they should be written as follows, to include the presupposed con-
stituent <ASSERTI0N> (with Lclearly also shown):
you don't have to be a <X> to <Y> that <ASSERTION>

X: brain surgeon, rocket scientist
Y: understand, appreciate, see
it comes as no [great] surprise that <ASSERTION>
clearly, <ASSERTION>
In the first of these examples we now see that there are three variable
constituents, not just <X> and <7>. We also observe that a variable
can range over a small number of values, like
<X> and <Y> of this example, as well as over a large number of val-
ues, like <ASSERTION>, <EATER>, and <EATEE>.
Similarly, the "mother of all ..." construction, identified above as a

mutable lexeme, is seen, if we look at its syntactic properties, to be a
nominal modifier; that is, it presupposes a noun or noun phrase, a
184 Sydney M. Lamb
realization of a member of the conceptual category <THING>. Using

notation like the above, we can write it as follows:
the mother of all <THING> -S
Likewise, the passive construction may be written
<PATIENT> b e <ACTION> - e d [ b y <AGENT>].
As the use of small caps is intended to indicate, the pertinent infor-

mation for these variables is evidently at the semantic-conceptual
level rather the lexemic. It's not categories like noun, verb, etc., but
semantic information. On the other hand, the variables <X> and <Y>
above range over lexemes, not semantic units (except insofar as lex-
emes are connected to semantic units).
We can use the "mother of all..." lexeme to consider the process
by which a lexeme becomes mutable. This change occurred in adult
systems but is doubtless like that which occurs much more often in
children and which accounts for much of their syntactic development.
For the process to occur, the lexeme the mother of all battles has first
to be segmented - in particular, battle has to be mentally segmented
from the rest - and then other things are seen to be substitutable for
battle. The segmentation of battle from the rest at the level of expres-
sion has a corresponding segmentation at the semantic level, so that
the mother of all is seen to have some meaning of its own, perhaps
paraphrasable as 'spectacularly large'. Such segmentation is auto-
matic, since battle is recognized along with the whole lexeme, as are
its other constituents. That is, the lexeme as a whole and also its con-
stituents are activated in parallel in the neurocognitive network.
When battle is segmented, the remainder of whatever meaning is
connected to the mother of all battles is assigned to the remainder of
the expression, i.e. the mother of all ...-s. The conditions for the oc-
currence of this process, for example, for the substitution of meteor
for battle, are presumably that the situation involving the meteor was
sufficiently similar to that involving battle to call up that (heretofore)
fixed lexeme. This makes it easy to substitute meteor and other nouns
for battle to describe situations in which something (anything) is

spectacularly large or important. And when this happens, the mother
of all has become another noun modifier. Note that the similarity
involved here is semantic. And the range of values of the variable is
exactly those things which can be construed as spectacularly large. It
is an ad hoc variable, specific to this mutable lexeme.
Garry Trudeau (the author of Doonesbury), more creative in his
use of English than most, took it one step further in a column in the
New York Times, March 21, 1991, in writing about the appearance
of a can of Diet Pepsi on the table in front of General Norman
Schwartzkopf, commander of the victorious allied forces in the Per-
sian Gulf war, in the official pool photo of the cease-fire ceremony,
beamed around the world to thousands of newspapers. He wrote, "It
was the mother, father and next of kin of all product placements - too
insanely great to be true."
7. Constructions as lexemes with variable constituents
Now, finally, we have to consider the more general constructions,

like Goldberg's argument constructions and the actor-action con-
struction. These seem to be quite different from <EATER> EAT
<FOOD>, since they are quite general in their coverage while the latter
is tied to a specific lexeme.
First, we can ask why they need to be recognized at all, if each of
the constituents in any instance has its own syntactic information in
accordance with the principle that every lexeme has its own syntax.
The answer comes from the consideration of cases like Goldberg's
example Frank sneezed the tissue off the table. The ability of sneeze
to occur in such an expression is surely not a syntactic property we
want to attribute to sneeze, normally an intransitive verb (Goldberg
1995: 152). Rather, its occurrence here is accounted for by the
caused-motion construction itself, which evidently accommodates
any verb whose performance can be construed under any circum-
stances as causing motion. Similarly anomalous from the point of
view of its own syntactic properties is the use of conserve in But can
186 Sydney M. Lamb
Californians conserve their way out of the energy crisis?, heard on

NBC Nightly News, January 26, 2001. Now conserve is a transitive
verb, but it takes as patient something conserved, and in this exam-
ple, way is not to be construed as such a patient. Rather, this usage is
accounted for by the way-construction, seen also in such examples as
he dug his way out of the prison and she fought her way into the
room (Goldberg 1995:16 )
The caused motion construction (omitting the subject, which I do
not consider part of it, see above) can be written <MOVE> <PATIENT>
<DIR>, where <DIR> stands for "directional", covering such possi-
bilities as away, into the room, off the table, out of here, etc. Here,
<MOVE> represents any action that can cause motion, and so it per-
mits any verb that can be so construed, even sneeze, to impart motion
to a suitable patient - that is, anything construable as a suitable pa-
tient. Notice that we are talking semantics and not about syntax as
traditionally treated, which won't work. As soon as a speaker can
construe any action as an instance of caused motion, the values of the
variables are those which fit this construal of the situation. It's the
same principle as that discussed above in connection with <EATER>
<EAT> <EATEE>.
The question we must now ask is: How can such a construction,
with all of its constituents variable, get learned, since the input to any
child consists of specific utterances, not generalizations like this con-
struction? A reasonable answer seems to be that the process is like
that sketched above for the mutable lexeme the mother of all
<THING> -s. The child may be presumed to start with a specific in-
stance, say push it away or push it o f f , acquired first as a lexeme, and
then to make it mutable by substituting other constituents for one of
these three. Later, he substitutes also for another constituent, and then
for the third. At this point all three constituents will have become
variable. Notice that, as before, none of these variable constituents
needs to be seen as a category whose membership has to be learned
or stored. How could we ever devise a category definition that could
accommodate in advance all of the objects that can participate in this
construction?
And so the difference between the construction and the lexeme with
variable constituents is not that great. In fact a construction might be
considered a lexeme with more than one variable constituent. Some
constructions also include one or more fixed constituents, others do
not. Among those which do are the way-construction (they made
their way to the door) and the passive construction, which includes
the verb be asa fixed constituent. And as we have seen, there are also
variables which have a very small number of values, which can be
enumerated, for example, it doesn't take a X to Y that <ASSERTION>,
in which both X and 7 have just a few possible values.
8. Learning syntax in a second language
The above rough descriptions of what I take to be reasonable scenar-

ios for learning syntax are in the context of the developing syntax in
one's native language. What about second-language learning? It hap-
pens that this same technique, as developed by Leonard Bloomfield
and his colleagues, was being used in some of the second language
teaching at Yale University during the time when I was an under-
graduate there. It was used in my classes in German and Russian. The
method was very simple: The course consisted of a series of lessons,
each of which began with a number of sentences to be memorized.
They were whole sentences, not just words or phrases. Each of them
had a function that was both known and useful to the student. Since
they were memorized (and the students were given a weekly test in
which they had to reproduce them verbatim), they became lexemes
within the student's cognitive system. Then each lesson guides the
student in making substitutions for one or more of the constituents of
these sentence-length lexemes. I still remember some of the early
Russian sentences, over fifty years later: U mjinja yestj karandash Ί
have a pencil'. U minja njet karandasha. Ί don't have a pencil', U
mjinja estj pjiro Ί have a pen', U mjinja njet pjira Ί don't have a
pen' (cf. Cornyn 1950). Since we were adult college students, the
lessons also included a little grammar section in which were imparted
some of the grammatical principles involved in the sentences we
188 Sydney M. Lamb
were memorizing and learning how to develop into mutable lexemes.

But the grammar presentation was secondary.
Another feature of this method was that we were not introduced to
the writing system of either language during the first portion of the
course, not until several weeks into the term, and we had to learn to
pronounce with a high degree of accuracy. Native speaker coaches
would make us repeat endlessly until we got the right pronunciation.
Our learning thus had to utilize the auditory and motor cortices, for
phonological recognition and production respectively. Had we been
introduced early to writing, we would have used the college student's
natural reliance upon written forms and would thus have been at-
tempting to learn the forms using the visual cortex rather than the
auditory and motor cortices.
This method of language instruction seems to me to be based
upon the same principles used by children in their native language
development, and I don't believe that any better method of second
language teaching has ever been devised.
9. Concluding observations
Let me sum up by mentioning some of the principles that have

guided this study and some of the conclusions reached. First, it is
apparent that neurocognitive and analytical approaches to language
have different concerns from each other. What I call analytical lin-
guistics is concerned with analyzing linguistic data, utterances, sen-
tences, and the like, and with finding patterns in such data, often
guided by theoretical concerns that have little cognitive basis and
usually no neurological basis at all. In neurocognitive linguistics, by
contrast, while such data is still examined, the object of study is the
neurocognitive system of the individual. This difference of focus has
a number of consequences, not least of which is the recognition and
acceptance that the system of every individual is different from that
of every other. More important for the concerns of this paper is that
by taking the natural operation of the human brain into consideration
we recognize that linguistic information, like other kinds of informa-
tion, is often represented redundantly in the neurocognitive system.

The tendency of analytical linguists to seek out the most economical
possible means of handling a given body of data is seen to be lacking
any neurocognitive motivation.
Second, I have argued that syntactic categories of the kind usually
recognized are artifacts of analytical linguistics, based on approxi-
mations and motivated by the desire for economical description.
They appear not to have any direct cognitive basis. The same can be
said for some of the forms in which constructions based on such
categories have often been conceived..
The human cognitive system represents information as connec-
tivity in a network. It operates by means of widely distributed repre-
sentations and parallel processing. As a consequence, linguistic forms
can be recognized or produced by means of different structures oper-
ating in parallel. So for example, both a general construction and the
information pertaining to lexemes may operate jointly to produce a
given sentence or phrase. In contrast to the usual approach in analyti-
cal linguistics, we need not suppose that people use just a single uni-
fied strategy for producing and understanding combinations of lexe-
mes. The brain thrives on redundancy and on multiplicity of strate-
gies.
I have also argued that every lexeme has its own syntax, and that
learning syntax is mainly a matter of learning lexemes. This is a one-
by-one process. And even the general constructions can be seen as
lexical items, to be learned one at a time. There is a lexical scale
from the fixed lexeme at one end, to the lexeme with a variable con-
stituent of limited range, to the lexeme with variable constituent with
broad range, to that with multiple variable constituents, and at the
right end of the scale, the construction, with all constituents variable.
Moreover, it is plausible that such constructions are acquired by a
process that begins as the learning of a fixed lexeme, which then be-
comes mutable as segmentation occurs, leading to the possibility of
substitution, whereupon a constituent becomes variable. And such
segmentation is automatic, as soon as such a constituent is recog-
nized as a unit. It is recognized as a unit while the lexeme as a whole
190 Sydney M. Lamb
is also recognized, in accordance with the parallel processing of ac-

tivity in the network.
Also, the knowledge needed to manage variable constituents like
the participant roles of processes (e.g. <EATEE>) does not include any
information other than that pertaining to the process itself (e.g.
<EAT>).
Finally, I suggest that the best method of second language instruc-
tion may be that which comes as close as possible to utilizing these
same techniques, presumably those used by children in native lan-
guage development.
References
Bates, Elizabeth, Donna Thai, Barbara Finlay and Barbara Clancy

In press Early language development and its neural correlates. In: I. Rapin
and S. Segalowitz (eds.), Handbook of Neuropsychology, Vol. 7:
Child Neurology [2nd ed.]. Amsterdam: Elsevier.
Bloomfield, Leonard
1933 Language. New York: Henry Holt
Cornyn, William S.
1950 Beginning Russian. New Haven: Yale University Press.
Goldberg, Adele
1995 Constructions: A Construction Grammar Approach to Argument
Structure. Chicago: University of Chicago Press.
Hockett, Charles F.
1997 Approaches to syntax. Lingua 100: 151-170.
Lamb, Sydney M.
1999 Pathways of the Brain: The Neurocognitive Basis of Language.
Amsterdam/Philadelphia: John Benjamins.
2000 Neurocognitive structure in the interplay of language and thought.
In: Martin Pütz and Marjolijn Verspoor (eds.), Explorations in
Linguistic Relativity, 173-196. Amsterdam/Philadelphia: John
Benjamins.
Mel'chuk, Igor
1988 Dependency Syntax. Albany: SUNY Press.
1998 The meaning-text approach to the study of natural language and
linguistic functional models. LACUS Forum XXIV: 5-19.
Tomasello, Michael
1998 The return of constructions. Journal of Child Language 25: 431-
442.
Tomasello, Michael and Patricia J. Brooks
1999 Early syntactic development: A construction grammar approach.
In: Martyn Barrett (ed.), The Development of Language. Hove,
UK: Psychology Press.
Conceptual primes in
early language development
Cliff Goddard
1. Introduction
In his inspirational book Acts of Meaning, the esteemed psychologist

Jerome Bruner (1990: 72) argues that: "how we 'enter language'
must rest upon a selective set of prelinguistic 'readiness for mean-
ing'. That is to say, there are certain classes of meaning to which hu-
man beings are innately tuned and for which they actively search."
Bruner here articulates a nativist approach to language acquisition,
but a conceptual/semantic nativism rather than the syntactic nativism
of Chomskyan linguistics.1 Bruner continues: "Prior to language,
these exist in primitive form as protolinguistic representations of the
world whose full realization depends upon the cultural tool of lan-
guage".
The present study explores certain hypotheses about the nature and
identities of the innate concepts which may underpin language acqui-
sition. These hypotheses have arisen from one of the most promising
and productive approaches to cognitive semantics - the natural se-
mantic metalanguage (NSM) approach originated by Anna Wierz-
bicka (1972, 1980, 1996, cf. Goddard 1998a, Goddard and Wierz-
bicka eds. 1994, forthcoming). Though the NSM approach has been
responsible for literally hundreds of descriptive studies in lexical and
grammatical semantics and pragmatics across a wide range of lan-
guages,2 it has not been applied very extensively to language acquisi-
tion. The only previous studies, Wierzbicka (1995) and Tien (1999),
are unpublished. I hope to show, however, that the NSM approach
generates interesting research hypotheses on language acquisition and
allows for increased precision and testability in the notoriously diffi-
194 Cliff Goddard
cult area of child language semantics. In particular, it enables one to

propose concrete and constrained semantic analyses of early "child
meanings", proposals of a kind which are surprisingly sparse in the
otherwise abundant literature on early lexical development (e.g.
Bates et al. 1988, Bloom 1973, Bowerman 1978, Braine 1976, Clark
1993, Dromi 1987, Gopnik and Meltzoff 1997, Halliday 1975).
The general outline of the NSM theory is well-known (but see
Goddard 1998b for a discussion of popular misconceptions), so I will
not go into it in great detail here. The essential claims are: (i) that at
the heart of any language there is a specifiable set of simple, indefin-
able concepts, i.e. conceptual primes, also known as semantic primes,
(ii) that these conceptual primes can be found embodied as the
meanings of ordinary lexical items (words, bound morphemes, or
phrasemes) in all languages (Goddard and Wierzbicka 1994), (iii)
that conceptual primes have an inherent "conceptual syntax", in the
sense that they can be combined in certain specifiable ways which
make sense; such combinations, furthermore, can be expected to have
equivalents in all languages, (iv) that conceptual primes and their
combinatorial syntax3 can be thought of a comprising a "mini-
language" in terms of which all other meanings (whether expressed
by words, grammatical constructions or prosodies) can be explicated.
Most NSM researchers do not regard the natural semantic metalan-
guage merely as a useful analytical tool but rather as a conceptual and
psychological reality. Conceptual primes are supposed to be, literally,
elements of comprehension, and semantic explications framed in
primes are supposed to be, literally, models of the conceptual mean-
ings encoded in and expressed by linguistic forms.
The current inventory of primes can be listed (using English expo-
nents) as in Table 1. There are, of course, many aspects of the meta-
language and its realization in the English language which cannot be
displayed in a summary presentation. For example, many of the
words listed are polysemous, but in each case only one, specifiable
sense is being proposed as a conceptual prime. The organization of
Table 1, i.e. the grouping and categories, is intended to reflect some
of the functional and combinatorial relationships among primes.
Conceptual primes in early language development 195
Table 1. Proposed semantic primes (after Goddard and Wierzbicka in press)
Substantives and I, YOU, SOMEONE/PERSON, SOMETHING/

substantive relations: THING, PEOPLE, BODY, PART OF, KIND OF
Determiners and quantifiers: THIS, THE SAME, OTHER, ONE, TWO, ALL,
MUCH/MANY, SOME
Descriptors and evaluators: GOOD, BAD, BIG, SMALL
Intensifier: VERY
Mental predicates: WANT, FEEL, THINK, KNOW, SEE, HEAR
Speech: SAY, WORD, TRUE
Actions and events: DO, HAPPEN, MOVE, TOUCH
Existence and possession: THERE IS, HAVE
Life: LIVE, DIE
Augmenter: MORE
Logical concepts: NOT, MAYBE, IF, CAN, BECAUSE
Time: WHEN/TIME, NOW, BEFORE, AFTER,
A SHORT TIME, A LONG TIME,
FOR SOME TIME
Space: WHERE/PLACE, HERE, BELOW, ABOVE,
INSIDE, SIDE, FAR, NEAR
Similarity: LIKE
The claim of NSM researchers is that the primes indicated in Table 1

are necessaiy and sufficient to explicate all the meaning content ex-
pressible through the words, grammar and intonation patterns of the
adult language. To what extent the same can be claimed for the lan-
guage of young children is an open question, and one of the main
concerns of this paper. A point I would like to emphasize, however,
is that to the extent that the proposed primes (or a subset of them)
turn out to be necessaiy and sufficient for modeling early child
meanings, they will escape a potent criticism which has been leveled
at other approaches to the semantics of child language. This criticism
was first voiced by Susan Carey (1982). She pointed out that the
majority of semantic descriptors used in child language research are
highly abstract and theory-laden. Terms such as 'recurrence', 'agen-
tive', 'dative', 'factitive', 'instrumental', 'regulatory', 'implication',
196 CliffGoddard
'dimension', 'polarity', and so on, come to mind. According to

Carey, such "elegant and abstract" terms represent "a sophisticated
schematization of knowledge by linguists". As such, they may be
useful for certain purposes, but they are not plausible candidates for
what Carey calls "developmental primitives", in the sense of "innate
... or at least very early-acquired concepts, out of which all other con-
cepts are built" (p. 351).
This criticism cannot be leveled at the NSM system, or at least,
not with anything like the same force, because the NSM primes are
not "abstract terms from theories which the child has not yet en-
countered". On the contrary, they are plain words and expressions of
ordinary language, which the child is hearing every day, and which,
in many cases, exist in the child's own active vocabulary. If we wish
to fashion a system of semantic-conceptual representation for the
child, then the NSM system is, at the least, a very good place to start.
1.1. Diary study of Pete
Many of the examples discussed in this paper are drawn from diary
notes of the language development of my son Pete (a pseudonym),
who was born in August 1996. The notes were made by myself and
by my wife, Mee Wun Lee, commencing (in earnest) from the time
just before the child's second birthday. Our procedure was not as
rigorous or systematic as some other diary studies, but it did yield a
continuous sampling of the child's utterances over an 18-month pe-
riod. I was the main record-taker. I tried to make a point, whenever I
was with the child, of having some notepaper and a pencil with me,
and I simply jotted down any utterance which seemed either typical
of the way the child was speaking at that time, or interesting in the
sense of showing "emergent" meanings or structures. Often some
notes on the context or apparent intention of the child were also nec-
essary. At different times, observations were made all over the house
- at the breakfast table, in the bath, in the living room, in Pete's bed-
room, and in the backyard, as well as in the car, at friends' houses, on
shopping trips, at playgrounds, and so on. This is, admittedly, a rather
haphazard and intuitive technique but it is not different from the pro-
cedure followed by many linguists in taking field notes of an indige-
nous language in a naturalistic setting.
Diary studies are most useful when they selectively focus on spe-
cific issues of interest to the investigators (Mervis et al. 1992). My
observations were guided firstly by an interest in semantic primes,
and secondly by an interest in general syntactic development. The
fact that one is taking a day-by-day record tends to make one sensi-
tive to new developments and thus more likely to record them; but on
the other hand there were many hours each day when the child was
not being observed, so the diary records cannot be taken to indicate
the earliest occurrences of any word or structure. In most cases they
probably pick up features which have been present for an unknown
previous time. (We did not record utterances which were simply
repetitions, in whole or part, of something which had just been said
by an adult.) My wife, who is a native speaker of Cantonese, often
spoke to the child in Cantonese when they were alone together; and
especially in the early days I also used to address the child using my
own rudimentary command of this language. Not surprisingly, a good
proportion of Pete's earliest words were Cantonese, but for the pur-
poses of this paper I have usually given English equivalents.
Unlike several children who feature prominently in the child lan-
guage literature, Pete is not an "early talker" but his general language
development seems fairly typical of a child growing up in a middle-
class Western household. From an examination of the corpus I have
divided his early language development into the following rough
stages.
Prelinguistic stage
Stage I (from about 14 months): single words
Stage //(about 21 to early-26): two word combinations
Stage III (early-26 to mid-29): some multi-word sentences, usually
limited to a simple clause with an adjunct; also, only one utter-
ance at a time.
Stage IV (mid-29 to early-32): two or more related sentences in a
row; begins to have conversational exchanges.
198 CliffGoddard
Stage V (early-32 to 36 months): onset of complex sentences (want-

complements, when-clauses, clausal linking with and), early
morphology (-ing, -s, -er) and grammatical verbs (e.g. in-
choative ¿^-construction, have to, make-causative).
Stage VI (36 to 41 months): past tense -ed, if-clauses, clitic copula -s.
The present paper focuses primarily on the very early stages, i.e. the
Prelinguistic stage, Stage I (one-word stage) and Stage II (two-word
stage).
1.2. Semantic primes in the production vocabulary vs. the

"conceptual vocabulary"
It is useful at the onset to summarize the emergence, in Pete's pro-

duction vocabulary, of lexical exponents of the proposed conceptual
primes. This is done in Table 2, in which each column represents one
of the Stages just described. Within the columns, the order does not
represent chronological sequence but rather the groupings used ear-
lier in Table 1. A couple of items (PART and MOVE) appear earlier
than I would have expected from the research literature (cf. also Tien
1999: 115-141), but otherwise there is little in this sequence which
will surprise child language researchers.
The onset times given in Table 2 assume that certain primes are
first expounded by words which are different to those of the adult
language (cf. Wierzbicka 1995). For example, SMALL first appears as
little, A SHORT TIME as a little while, BEFORE as first, AFTER as later,
THE SAME as too, and WORD as called. It is not possible here to dis-
cuss and justify these decisions adequately, but in any case the adult
exponents of these primes all appear by Pete's third birthday. The
main general observation is simply that most of the proposed primes
are clearly present by the child's third birthday; the remaining ones
follow within a further six months or so (Goddard Forthc. a). Of
course, one could certainly not expect all children to follow Pete's
sequence, given the sizable individual differences in early lexical
development even among English-speaking children (cf. Bates et al.
1988). Aside from personality differences, factors influencing the

production sequence are individual preferences in "lexical selection"
on phonological grounds (especially in the early stages) and the na-
ture of the language input, e.g. Schwartz and Leonard (1982), Ninio
(1992). Looking cross-culturally, differences in interactional style
with infants is another potent factor (cf. Schieffelin and Ochs 1986).
Table 2. Acquisition sequence of lexical exponents of primes in Pete's production

vocabulary, up to 41 months
Stage I Stage II Stage III Stage IV Stage V Stage VI
I YOU
PERSON PEOPLE BODY
THING/WHAT PART KIND
THIS OTHER MUCH/MANY ONE
THE SAME(TOO) TWO ALL SOME
GOOD
BAD
BIG
SMALL(LITTLE)
VERY(SO)
WANT WORD(CALLED) KNOW
SEE MOVE HAPPEN SAY HEAR
DO TOUCH THINK FEEL
HAVE THERE IS DIE
MORE NOT(NO) CAN MAYBE IF
BECAUSE(COS)
NOW
BEFORE(FIRST) AFTER(LATER) A SHORT TIME A LONG TIME
(LITTLE WHILE)
WHERE/PLACE
HERE NEAR(NEXT TO) SIDE
INSIDE(LN) FAR
ABOVE(UP)
BELOW(DOWN)
LIKE
I hasten to note that the presence of a lexical exponent does not mean
that a prime is "fully acquired", in the adult sense. The child may
have active command over only a small part of the prime's syntactic
possibilities (as they exist in the adult language), with the result that
200 Cliff Goddard
its range of use may be highly restricted in comparison with adult

usage. For example, though the word DO appears early in the third
year it is a long time before the child is able to use it with the full
range of complement and valency options, e.g. DO SOMETHING ΤΟ
SOMETHING, DO SOMETHING WITH SOMETHING, DO SOMETHING TO
SOMETHING WITH SOMETHING. Nevertheless, the presence of the lexi-
cal item, used with apparently the same meaning as its adult counter-
part, is concrete linguistic evidence that the corresponding concept
has some foothold, so to speak, in the child's mind.
With this as background, the key issue to be addressed in this
study can now be approached. With a small number of notable ex-
ceptions, most of the child's early words are not exponents of con-
ceptual primes - for example, Stage I words such as mama, papa,
bath, bird, duck, nana (banana), ball, and oh-oh, Stage II words such
as mouth, hand, wheel, door, o f f , broke, made, noise, and many oth-
ers. Dozens of early words appear whose semantic structure cannot
be explicated in terms of the child's inventory of lexicalized primes.
What can the NSM theory say about this situation? One response
would be to observe that most of the child's early words must have
simpler meanings than the corresponding words in an adult's vo-
cabulary. This may be true enough, but it still seems impossible that
dozens of Stage I words (for example) could be explicated in terms of
a small handful of conceptual primes. The only possibility is that the
child is making use of a larger set of primes, including some which
are present conceptually but which still lack lexical exponents.4 As
Bruner (1990) put it, there must be meanings which are present "in
protolinguistic form" prior to their emergence in language proper.
From a theoretical point of view, this deduction is hardly surpris-
ing. It is well-known that children's comprehension runs well in ad-
vance of their production vocabulary. Synthesizing results from vari-
ous studies, Ingram (1989: 140-143) suggests that, as a norm, about
100 words are understood (in some fashion) even before the first rec-
ognizable words are produced. In Pete's case, a parental "self-
interview" identified about 60 words which we believed the child
could understand (in some fashion) at 15 months. These included the
following (or their Cantonese equivalents): dog, cannot, drink, eat,
water, go out, hurry up, milk, hand, foot, fly, ball, pick up, teddy, wet,
hot, don't want, see, come, nappy, light, this, and many others. From
the NSM point of view, if something is understood (in a linguistic
sense) then it is necessarily understood in terms of some conceptual
primes. In short, the child must have a certain "conceptual vocabu-
lary" of prime concepts even before the onset of intelligible words.
Theoretically, this is all well and good, but from a methodological
point of view it is problematical. How can we identify conceptual
primes which may be active in the child's mind, in the absence of
tangible surface exponents? As far as I can see, there are two possible
sources of linguistic evidence: (i) semantic analysis of the child's
production vocabulary, and (ii) semantic analysis of the child's com-
prehension capabilities. The second option presents even more meth-
odological difficulties than the first, and I have nothing to say about it
in this paper. What I will try to do is to undertake semantic analysis
of words in the child's non-prime production vocabulary, for if we
can determine the meaning of these words this would furnish direct
evidence of the conceptual vocabulary of the child at that age. In this
I am taking a lead from Tien (1999), who, as far as I know, was the
first to argue that conceptual primes may be "latent" in a child's early
lexicon, in the sense of being hidden or implicit in the meanings of
other, non-prime words.
The task may seem like a daunting one, but it is not altogether dif-
ferent from that which faces a field linguist who undertakes semantic
analysis of an unknown adult language from an unfamiliar culture. In
either case we have to begin with close naturalistic observation of
usage: documenting the range of contexts in which a certain expres-
sion is used - and not used - and comparing the usage of alternative
expressions which can be found in different contexts. Then we ex-
periment to discover the most economical semantic explications
which match the attested range of usage. Obviously one must guard
against the assumption that the child's meaning for a particular form
corresponds to the adult meaning, i.e. against "adultocentrism", the
child language analogue of ethnocentrism.
In this paper I deal only with a fairly small number of early child
words and utterances, and the fine details of the analyses in many
202 CliffGoddard
cases remain open to question. Even so, the exercise supports some
highly specific proposals about which primes emerge first at a con-
ceptual level and about the time lag between conceptual acquisition
and lexicalization.
2. Semantic primes in early "conceptual vocabulary"
2.1. Explicating "proto-linguistic " meanings
By general agreement, a child's first recognizable words are not his

or her first attempts at intentional communication. The first words
come after a lengthy period of prelinguistic (or proto-linguistic) ex-
pression by means of (consistent) signals composed of sounds,
prosodies, gestures and gazing. For example, according to Halliday
(1975), his son Nigel was producing meaningful proto-linguistic sig-
nals (sound-meaning pairings) from as early as 9 months, with this
system of signals expanding in size and function for about six months
prior to the child's first truly language-like utterances. At 13 months,
for example, Nigel used a form ?nä Pnä ?nä ?nä to indicate some-
thing like Ί want that', and another form yi yi yi yi yi (high level
tone) to respond when asked if he wanted a particular object to be
given to him, meaning something like 'Yes. I want that' (Halliday
1975: 24). Halliday emphasizes that expressions like these are not
"word-like" in the adult sense, but are direct form-meaning pairings;
and also that they are for the most part not derived from forms of the
adult language. Nevertheless they are intentional and systematic, they
represent a child's long experience with his or her own system of
expression, and they are fundamental in readying the child for lan-
guage development in the adult sense. Similar conclusions have been
reached by Carter (1978, 1979), Dore (1975), Dore et al. (1976), and
Scollon (1976,1979).
Regrettably I did not attempt to document and analyze Pete's
proto-linguistic system. Even so, I think it will be useful to articulate
some hypotheses about the kinds of meaning he was expressing in
this period. For this purpose, I have studied Halliday's account of

Nigel and attempted to interpret some of what he says into NSM
terms - with apologies to Halliday. The following ideas are ordered
roughly according to Nigel's sequence of development and my confi-
dence in them (the two orders happening to coincide). First, the
primes WANT, SEE, DO, and THIS are involved in very early messages
like the following, which Halliday assigns to the instrumental, regu-
latory, and interpersonal functions, respectively:
WANT THIS
DO THIS
SEE THIS
Later, WANT and SEE form the basis of a "proto-mood" distinction in

Nigel's speech, indicated by the contrast between high tone (i WANT)
and falling tone (i SEE), respectively. This appears to correspond to
the proposal by Bates et al. (1979: 115) that early child speech shows
a distinction between proto-imperatives and proto-declaratives.
Proto-imperatives - "the child's intentional use of the listener as an
agent or tool in achieving some end" - are framed or introduced by
the component ι WANT - . Proto-declaratives - "a preverbal effort to
direct the adult's attention to some event or object in the world" - are
framed or introduced by the component ι SEE - . In adult speech, of
course, the declarative mood is associated, prototypically, with a
more complex illocutionary component based not on SEE, but on
KNOW, namely I WANT YOU TO KNOW SOMETHING. (Some researchers
apply the proto-imperative vs. proto-declarative distinction also to
the use of pointing by infant children, cf. Reddy 1999: 43.)
Still at an early stage, the elements NOW and HERE are implied in
some of Nigel's prelinguistic signals, in combinations such as the
following:
DO THIS NOW differs from DO THIS by "an additional feature

of urgency", Halliday's gloss 'Do that right
now'
204 CliffGoddard
SOMEONE HERE NOW a component of one of Nigel's "initiating

expressions ... used typically when another
person comes newly to his attention"
Before long, it seems that MORE enters the picture, in combinations

like:
DO THIS MORE - Halliday's gloss 'do that (again)'

SEE MORE - Halliday's gloss 'look, that's interesting'
Interestingly, the prelinguistic Nigel clearly seems to express the

message DON'Τ-WANT. On an adult semantic analysis, this involves
negation (NOT) in combination with WANT, but there is no direct evi-
dence in Nigel's early vocalizations for negation as a separable com-
ponent. Possibly DON'T-WANT is a single unanalyzable meaning for
him at this stage.
During the proto-linguistic stage Nigel also appears clearly to ex-
press some simple messages about his "feelings". From fairly early
on, there are expressions glossed by Halliday as 'that's nice' and
'nice to see you' (pleasure). Towards the end of the period, Halliday
lists two expressions which seem to express negative feelings; he
glosses them as 'let's be sad' (shared regret) and 'I'm fed up' (com-
plaint). In adult language, one would assign components such as I
FEEL SOMETHING GOOD and Ι FEEL SOMETHING BAD but these seem too
complex for the young Nigel, and there is no evidence for the pres-
ence of FEEL, GOOD, or BAD as separate items. I therefore suggest that
in the early stages of development, there are two global and undiffer-
entiated meanings which we can represent as: FEEL-GOOD and FEEL-
BAD.
In summary, therefore, a consideration of proto-linguistic mes-
sages suggests that the following set of meanings may be present, in
some incipient fashion, even prior to the first ordinary word: WANT,
SEE, DO, THIS, SOMEONE, NOW, HERE, MORE, DON'T-WANT, FEEL-
GOOD a n d FEEL-BAD.5
2.2. Explicating early Stage I words
I will focus on the following small set of (non-prime) words from

early in Pete's Stage I. There are indications (Dromi 1987, 1999) that
the early and late periods of the one-word phrase may be qualitatively
rather different, in terms of the child's word-learning strategies.
mama, papa
bath!
nati 'hot/cold' (from Cantonese 'hot')
oh-oh!
bird! (Cantonese)
duck! (Cantonese)
broom-broom! 'car'
What can we infer about the meanings of these words, when Pete was
18-20 months of age? Obviously we cannot attribute to the child
anything like the semantic complexity of the comparable words in the
adult lexicon. Nevertheless it is necessary to attribute some meanings
to them, presumably the simplest conceivable meanings which are
compatible with their range of use in his speech.
The words mama and papa surely involve - minimally - the ele-
ment SOMEONE(PERSON), presumably in combination with THIS. What
else? One possibility would be to interpret mama and papa as, so to
speak, proto-names; i.e. to attribute a semantic structure along the
following lines (the use of inverted commas around mama in the ex-
plication is intended to refer to the sound of the word only).
mama =
this person
this person is called "mama" (i.e. word for this person is "mama")
papa =
this person
this person is called "papa" (i.e. word for this person is "papa")
206 CliffGoddard
According to these explications, the child's early words mama (and

papa) have no descriptive content. They simply embody a recogni-
tion of a particular individual who is associated with a particular "la-
bel". Accepting this, however, means accepting that even very young
children have some rudimentary notion of WORD. Some child lan-
guage researchers would no doubt find this hard to swallow, for chil-
dren at this stage; but others would not. Halliday (1975), for example,
interprets the transition from the prelinguistic to the truly linguistic in
terms precisely of the emergence of the "word" as a unit in the sys-
tem: "The adult language is not a two-level system but a three-level
system; it is composed not merely of meanings and sounds, but has
another level of coding in between, one which, using folk-linguistic
terminology, we may refer to as the level of wording' (p. 50, cf. also
p. 34).
A great deal has been written about the early onset of the so-called
"naming insight", which some researchers see as explaining the vo-
cabulary spurt which many children experience later in the one-word
stage (cf. McShane 1979, Goldfield and Reznick 1990, Harris et al.
1988, Harris et al. 1995). I am inclined to agree with Harris et al.
(1995), however, that the vocabulary spurt "might better be thought
of as marking a transition from the view that some things have names
to a view that everything has a name" [my emphasis]. It seems en-
tirely plausible that Pete's earliest awareness of "names" or "labels",
i.e. in NSM terms WORDS, would be in relation to those special
"someones", his mama and papa.6
Various researchers have noted that very young children often use
certain words in ways which defy the adult expectation of a distinc-
tion between "nouns" (words for things) and "verbs" (words for ac-
tions and events). Rather such words seem to stand for, or even to be
part of, an "activity script" featuring the child him or herself in some
daily routine such as going to bed, changing nappy, or having a bath.
Early in Stage I, the word bath was like this for Pete. He would utter
it upon seeing the bathtub filling up. The following explication pres-
ents bath! more as an exclamation than a noun or a verb; and like
exclamations in adult speech (in this respect) it relies heavily on de-
ictic elements like THIS, HERE, and NOW. Also implied is the loca-
tional (WHERE) relationship.
bath! =
I see something
something is happening here
I (can) be in water (this 'stuff) now
An important question which I will leave untouched is the status in

the explication of the word water. I am prepared to take for granted
that, experientially, water is directly recognizable to the child; but I
am at a loss to know whether, and to what extent, the child has an
articulated concept of water, deserving of its own "mini-explication".
Pete's word nat derives from the Cantonese word for 'hot'. He
often heard us say it to warn him about hot foods and drink, and also
about the heater in our living room, but for a while Pete used his
word nat not only about hot things, but also about very cold things -
mainly about ice cubes and stuff taken directly from the freezer. An
explication along the following lines seems to be called for. Notice
the apparent necessity for the term TOUCH - a recent, and as yet
largely untested, candidate for the prime inventory.
nat! ('hot/cold') =
I don't want to touch this
when I touch this, I feel something bad (or: I feel-bad)
In his one-word period, Pete uttered the word oh-oh! when he saw
that 'something bad' had just happened to something, typically that
something had fallen over, been broken, dropped or spilled. Needless
to say, this usage was heavily modeled for him by his parents, but
Pete's oh-oh! embodied a simpler semantic structure than adult us-
age.7
oh-oh! =
something bad happened to something now
208 CliffGoddard
Coming now to bird, duck, and car, we see what first appear to be
clear examples of "nominal" terms, in the sense of words which are
clearly and only used about things. In their very earliest uses, how-
ever, such words are used essentially to make an observation about
immediate experience: the illocutionary frame of the one-word utter-
ance is Ί see (Consistent with this, Halliday [1975] placed
Nigel's early one-word utterances of this type under the "Personal"
(expressive) function.) As parents too, we tend to use words like
these to young children precisely to point out things: 'Look baby -
(a) bird!'.
Clearly these words can have only a very simple structure com-
pared to the enormous complexity of the adult words (cf. Wierzbicka
1985, 1996). Significantly, all three words refer to things that move.
For bird and duck, I suggest a further salient feature is, so to speak,
the locus of movement. Birds move up in the air; ducks move in the
water. As for broom-broom 'car', I suggest its salient feature is that
there is a person inside it.
bird! =
I see something
it can move up-high
this kind of thing is called "bird" (i.e. word for this thing is "bird")
duck! =
I see something
it can move in water
this kind of thing is called "duck" (i.e. word for this thing is "duck")
broom-broom! 'car' =
I see something
it can move
someone can be inside it
this kind of thing is called "broom-broom" (i.e. word for this thing is
"broom-broom")
Notice that these explications require the element KIND OF, even
though the word kind does not "surface" till Stage V. Otherwise, they
would be depicting the words bird, duck, and broom-broom as proper
nouns, rather than as designations for recognizable classes of things.
This is consistent with mounting psycholinguistic evidence for
genuine categorization, i.e. categorization by kinds rather than simply
by perceptual similarity, in very young children (Gopnik and
Meltzoff 1997: Chapter 6, cf. Markman 1989, Keil 1989, Mervis
1987).
Even from this small sample, then, we can see the implied pres-
ence of a dozen-and-a-half primes which do not surface as words in
their own right for some months to come: I, SOMEONE(PERSON),
SOMETHING(THING), KIND, THIS, HERE, NOW, SEE, WANT, DON'T-
WANT, BAD, HAPPEN, MOVE, TOUCH, CAN, ABOVE(UP), INSIDE, and
WORD(CALLED). TO this list we can add DO, MORE, and FEEL-GOOD,
which were indicated already in the proto-linguistic period (even
though they have not turned up in the handful of Stage I words we
have just looked at). At the one-word stage, it seems, Pete has a con-
ceptual vocabulary of at least 20 semantic primes, about one-third of
the eventual adult inventory.
2.3. Explicating Stage II words
In the three-month period between about 26.07 and 29.15 Pete learnt
to say a large number of new words, and increasingly to produce two-
word combinations (though he also continued to use a lot of single-
word utterances). As set out in Table 2, by the end of Stage II the
child had about 21 exponents of primes in his production vocabulary.
This set is not sufficient, however, to plausibly explicate the wow-
prime vocabulary of Stage II. Rather, semantic analysis of his non-
prime vocabulary suggests that Pete's conceptual vocabulary at this
period already included most of the primes posited to appear in Stage
III, plus several others which would not appear till subsequent stages.
First, it is worth pointing out a few fairly simple "semantic mole-
cules" (cf. Wierzbicka 1995):
210 Cliff Goddard
dark (here) = a person can't see here

(this is) stuck = this thing can't move here
(it 's) all gone = it was here before, it isn't here anymore
A noticeable fact about Stage II is the appearance of sets of related

nominal words (cf. Clark 1993, 1995), such as clothing words, body-
part terms, words for vehicles and for animals, and words for things
to eat and drink. Clothing words were among Pete's first nouns, the
first of them (pants, shirt) appearing before any of the body-part
terms. Actually the words in question were Cantonese "baby talk"
words. Fu-fu means roughly the same as 'pants', and saam takes in
shirts and t-shirts. The distinction is between the top part of the body
(saam) and the bottom part of the body (fu-fu). As far as I can see, to
get plausible "child-level" explications for these and other clothing
words we need not only to refer to certain parts of the body, but also
to the idea that people 'do something with' these items and end up, as
a result, with parts of their body inside them. For example:
fu-fu (pants) =
something
a person does something with it
afterwards the bottom part of a person is inside it
saam (top) =
something
a person does something with it
afterwards the top part of a person is inside it
The terms 'top part' and 'bottom part' are obviously based on the
primes ABOVE(UP) and BELOW(DOWN). The simplest thing would be
to regard 'top' and 'bottom' as simply "adjectival" variants of ABOVE
(UP) and BELOW(DOWN), respectively.
The need for the prime PART is surely clear from the proliferation
of body-part terms, such as mouth, eyes, nose, ear, head, and foot.
However, I do not think we have to posit the prime BODY at this
stage; it seems enough to explicate hands, for example, as 'parts of a
person' rather than as 'parts of a person's body'. Notice that many of

the body-part terms {eyes, ears, hands, feet) are explicated here "in
the dual", so to speak. This reflects the view, which can be traced
back to von Humboldt (cf. Plank 1989), that the gestalt of a "pair" is
impressed even upon an infant child from the experience of having,
and seeing, dual body-parts. In the case of the eyes, in particular,
their "two-ness" literally stares one in the face.
eyes - mouth =
two things part of a person
they are part of a person things can 'go' inside a person there
because of these things, a person can see a person can do things (to things)
with this part
hands = head =
two things one thing
they are parts of a person it is part of a person
a person can do many things with them it is above the other parts
feet/legs (kiok) = ears =

two things two things
they are parts of a person they are parts of a person
they are long one is on one side of the head
they are parts of the bottom of a person one is on the other side of the head
a person can do some things with them
Another couple of Pete's early nominal words are wheel and door.
The former is really a favorite word for Pete at age two, and perhaps
would qualify as the prototypical part-term in relation to physical
objects. An interest point about both words (especially clear in the
case of door) is that they seem to require the semantic component
SIDE. 8
wheel = door =
part of something something
it is round (i.e. when you see it, it is in a place
it is the same on all sides) someone can be on one side of it
it moves after this, it moves
when it moves, the other thing moves after this someone can be on the other
side of it
212 CliffGoddard
Like many other children, in his two-word stage Pete began to use a
largish number of animal names, such as horsie, dog, pig, cow, seal,
cat, and monkey. To analyze these meanings in detail would be a fas-
cinating project, which I believe would show that, even at this early
stage, such words can involve semantic components describing the
animal's size, some salient body-part features, characteristic sound,
reference to habitat and (at least in some cases) reference to its typi-
cal food. This kind of study is beyond the scope of the present paper,
however, and for present purposes I mention these words only to
make the point that they further attest to the conceptual presence of
the notion of KINDS. I will move instead to some explications for a set
of "verbal" words. As with the nomináis, this listing is not exhaustive
but it is a broad enough sample to indicate the range of semantic
components which are needed.
Perhaps the expression come here! is not entitled to be termed
fully "verbal", since it is, at this stage, essentially an imperative for-
mula; and the same applies to help me! Even simple structures such
as the following make it plain that the prime YOU is called for. In-
deed, one could claim that any example of genuinely "addressee-
directed" speech implies YOU - for YOU is what an "addressee" is.9
come here! = help me! =

I want you to do something I want you to do something
you are not here I want to do something (to this)
I want you to be here I can't do it
The expressions fall down and bump both seem to imply the element
HAPPEN. At age two, Pete used fall down not only about himself (af-
ter he fell), but also about something he threw or dropped. Bump not
only seems to imply HAPPEN, but also TOUCH (without TOUCH, it is
hard to see how the "contact" aspect of the event could be captured).
The explications also seem more plausible with an explicit causal
component BECAUSE OF THIS (as shown).
fall down =
something happened to this thing/person
because of this, it is down now
bump (head) =
something happened to me
part of me (my head) touched something now
because of this, I feel something bad now
In connection with TOUCH, it is interesting to look at the word off

one of the most common "verbs" in Pete's Stage II speech. He would
say things like off shoe (after taking his shoes off), o f f f u or off socks
(to describe what was happening, or to ask for it to be done). At this
time, we used to have him sleep in a "baby sleeping-bag", rather than
covering him with blankets in the cold Annidale winter. In the
mornings, he would say off bag. He would also use the word off
about his nappy, or (in the two-word mode) say off nappy. Similar
verb-like uses of ochave been reported by Tomasello (1987), among
others. The following explication for off has a kind of causative
structure.
off ('take off/remove/get out') =

this thing was touching me somewhere before
after this someone did something
because of this, it's not touching me like this any more
The word broke is an interesting one. It was first used in situations

where one part of something, e.g. a toy, fell off, or when a Lego con-
struction came apart. On one occasion he used it after the cap had
fallen off a marker pen. The explication not only employs HAPPEN but
also the element ONE, which is not strongly in evidence elsewhere at
Stage II.
broke =
something happened to it
because of this, it is not one thing any more
214 CliffGoddard
Give and make are an interesting pair. They both seem to demand an
explicit "before-and-after" scenario: the situation as it was before is
changed as a result of someone doing something. In the case of give,
the change concerns possession (i.e. someone having something); in
the case of make the change concerns the existence of something.10
(I) gave it (to someone) =

this person didn't have it before
after this I did something with it
because of this, this someone else has it now
I make house =
there wasn't a house here before
after this I did something with some things
because of this, there is a house here now
Verbs of physical manipulation, such as put, seem to call for a similar

overall structure. The verb eat involves "doing" and a resultative
component (including the component INSIDE).
(someone) put (it) here =

it wasn't here before
after this someone did something with it
because of this, it is here now
(I) ate it =
something was in my mouth
I did something to it with my mouth
because of this, it is inside me now
Finally in this quick sweep through some of Pete's Stage II words,

some evidence for the conceptual presence of HEAR (cf. Tien 1999:
104-111). First, Pete used the word noise in month 25, for example,
to remark on the noise of water draining away in the bath. Similarly,
he used the words cry and sing when not much older than two years,
e.g. to comment on another child crying. The meanings of these
words surely call for HEAR. Second, this was a time when Pete was
already quite interested in the characteristic sounds made by animals
and birds: "quack-quack" for ducks, "tweet-tweet" for birds, "moo-
moo" for cow, "oink-oink" for pigs, and so on. He could produce any
of these sounds at around two years of age. Surely this implies
something like (for example):
"quack-quack" =
ducks do something
when they do it, a person can hear something like this: "quack-
quack"
There are also two primes, specifically, KNOW and SAY, whose pres-
ence is implied by functional (illocutionary) facts. They do not ap-
pear in Pete's production lexicon for three or four months, but their
conceptual presence is implied by the fact that the child begins, in
Stage II, to ask simple information-seeking questions (both polar and
wh-), as shown below, and also to respond appropriately to questions
asked by adults. The illocutionary intention of questions, conveyed
by intonation, involves the components: Ι WANT TO KNOW SOME-
THING, I WANT YOU ΤΟ SAY SOMETHING. (Interestingly, all the re-
corded examples of Pete's early questions concern location. I am not
sure what to make of this: perhaps that is just the main kind of thing
he wanted to know about.)
25.26 Moon there? Roo there? (re. a photo, i.e. is that the moon
there? is that a kangaroo there?)
26.12 (CG told Pete they were going to the coast tonight, to the
sea) Pete: Seal there? CG: No, no seals. Pete: Boat there?
26.17 Mama slippers where?
26.18 CG: Pete, we're going to Jaew tse-tse's place for dinner.
Pete: Baby there?
28.06 Where plane?
In summary, from the sample of explications offered here one can

deduce that the following set of elements are present as conceptual
216 CliffGoddard
primes at Stage II, despite the fact that they have not yet surfaced as
individual words. They are presented in three groups: those which are
destined to appear as words in the next stage (Stage III), i.e. within a
couple of months; and those which do not appear until subsequent
stages.
To appear in Stage III: PEOPLE, PART, OTHER, TWO, HAPPEN, TOUCH,

HAVE, CAN, NOW, AFTER(LATER), LIKE
To appear in Stage IV: YOU, SAY, WORD(CALLED), THERE IS, SIDE
To appear in Stage V: ONE, KNOW, HEAR, BECAUSE
In the case of the Stage III group, what we are seeing is virtually the
entire list of primes which appear at Stage III (with two exceptions:
NEAR(NEXT TO) and FAR). This hardly seems like a surprising result
any more. We saw the same pattern in relation to Stage I; and, as
mentioned earlier, it is a pattern which is consistent with the fact that
comprehension competence runs several months ahead of production
competence. The conceptual presence of primes which only surface
lexically four, five, or six months afterwards, however, does seem to
call for some special explanation (see below).
3. Discussion
The findings of this study of Pete's early semantic and conceptual

development can be summarized as follows. All but a handful of the
proposed NSM semantic primes have lexical exponents by 41
months, i.e. by age three-and-a-half (cf. Goddard Forthc. a). Semantic
analysis of Pete's «on-prime words and utterances shows, however,
that many conceptual primes are present well prior to acquiring sur-
face lexical exponents. The general pattern is that the production vo-
cabulary of primes lags several months behind the conceptual vo-
cabulary. About 10 primes seem to be present in the conceptual vo-
cabulary even in the proto-linguistic period, prior to the first recog-
nizable word as such. They are: SOMEONE, THIS, WANT, DON'T-WANT,
SEE, DO, MORE, NOW, HERE, and possibly a pair of "proto-primes"
FEEL-GOOD and FEEL-BAD. By the end of the one-word stage, Pete's

conceptual vocabulary had doubled to 20 or so primes, about one-
third of the adult inventory, even though at this time only a small
handful of primes had lexical exponents of their own. By the end of
the two-word stage, Pete's conceptual vocabulary had again doubled,
to account for over two-thirds of the adult inventory.
Perhaps the most important question raised by this study is to what
extent Pete's acquisition sequence is typical - in terms of production
vocabulary, in terms of conceptual vocabulary, and in terms of the
relationship between them. As mentioned earlier, one might expect
there to be significant differences in the production sequence across
children and across languages, which it would be interesting to map
out and try to explain. 1 In relation to the acquisition sequence of
primes in the conceptual vocabulary, I would like to advance the hy-
pothesis that this is much more stable and consistent across children
and across languages. These issues, it seems to me, have the potential
to sustain a wide-ranging research program.
A second research area concerns a range of specific questions
about the developmental sequence of conceptual primes. For exam-
ple: What explains which primes appear early and which appear late?
Why don't more primes appear earlier in production, during the one-
word and two-word periods? Why do some primes take longer than
others to "bridge the gap" between their first appearance in the con-
ceptual vocabulary and their eventual appearance in production?
What explains the late lexicalization of certain primes? And so on.
Developmental, functional and syntactic factors may all play a role
here. For example, MORE, NO, and THIS may occur so early because
they can be used as single words and still convey a functionally use-
ful message, KNOW, HEAR, and BECAUSE may appear only relatively
late in production because to manipulate these terms in production
requires command over complement and adjunct structures, which
may present too much of a processing problem for the child.
A third area for further research concerns syntactic aspects of lan-
guage development. Conceptual primes have valency, complementa-
tion and combinatorial properties (cf. Goddard and Wierzbicka in
press) which, it can be argued, are the basis of syntactic structure at
218 Cliff Goddard
large in the adult language. As mentioned earlier, however, the

child's initial acquisition of a lexical prime does not normally con-
stitute "full" acquisition because initially the child may command
only a small part of the prime's syntactic possibilities. Thus, one
could study the syntactic development of primes, both in production
and in the conceptual vocabulary, and the relationship between this
and general syntactic development.
It should be clear that paraphrase in terms of the NSM semantic
primes can provide a practical, yet highly constrained, method for
formulating testable hypotheses about early child meanings, i.e. a
serviceable method for actually "doing" child semantics. Further-
more, to echo the concluding words of Wierzbicka (1995), the ap-
proach generates new and interesting research questions about lan-
guage acquisition.
In the Introduction I mentioned that there have been surprisingly
few studies which propose a substantial body of semantic analyses of
"child meanings". One of the notable exceptions to this generaliza-
tion is Michael Tomasello's (1992) book First Verbs. In his opening
chapters, Tomasello (1992: 33) makes the point repeatedly that any
such project requires "some form of cognitive-developmental theory
to help us reconstruct the child's point of view from our point of
view as we examine the contexts in which she used a particular
word". His own study began with the assumption that the child "lived
in a sensory-motor world of objects and their spatial, temporal, and
causal relations. Also, she may be assumed to know a good deal
about other persons and how they work ..." (p. 36). As a mode of
representation he adopted image-schematic diagrams, somewhat in
the style of Langacker (1987), and explicitly linked his empirical
study with one of the main currents in contemporary cognitive lin-
guistics.12 Needless to say, the present study can also be seen as an
exercise in cognitive linguistics.
There may be questions raised, however, about whether the NSM
framework does allow us to "reconstruct the child's point of view",
in the sense of safeguarding us against adultocentrism. Doesn't the
framework start, in effect, from adult language primes and impose
them on children? How do we know that young children don't have
their own conceptual primes which are different from, or even in-
commensurable with, the adult system? I will conclude by saying
something about these theoretical questions.
One cannot a priori rule out the possibility that the young child's
semantic system is incommensurable with the adult system. On the
other hand, practical experience shows that there can be a high de-
gree of mutual understanding between young children and adults. A
radical incommensurability thesis in relation to young children would
also face the problem of accounting for the developmental continuity
of child and adult understanding. In my view, the issue must remain
open for the time being. One thing is certain, however. We will never
get anywhere at resolving the issue unless we are willing to knuckle
down and attempt serious semantic analyses of early child language.
If we try, and fail, using a representational system which assumes
some limited conceptual continuity with the adult system, then this
will be evidence for incommensurability - and vice versa if we suc-
ceed. Above all, we have to try; and I submit that the NSM system
offers a highly facilitative framework for doing just that.
On the second point, it is not really correct to say that the NSM
system uses adult primes to explicate child language speech (thus
imposing an adult point of view). From the present study it would
seem that the meanings of a child's early words can be adequately
paraphrased in terms of the language of the same child - as it will be
a few months down the track. That is, even when the explications
cannot be framed entirely within the child's own production vocabu-
lary (which may be the case until the child is, say, four years of age),
they can at least be framed in terms which will soon be part of that
child's production vocabulary. Another way of putting it is that al-
though the language of toddlers and young children is not yet "meta-
semantically adequate" (as, apparently, all adult languages are) it
does not stay that way for long. There is every indication that the
entire NSM metalexicon exists in the production competence of a
four-year old child. If so, the continuity/incommensurability issue
should not be pitched in terms of child vs. adult, but in terms of two-
year old vs. four-year old. In this respect, the NSM primes retain a
significant advantage over other representational systems such as
220 CliffGoddard
complex, technical descriptors ('recurrence', 'instrumental', 'joint

attention', and the like) and complex, technical diagrams, which truly
are beyond the grasp of children at any age.
Acknowledgements
Mee Wun Lee has made a substantive contribution to this paper, es-
pecially in relation to the semantic analyses of Pete's early words.
Vicki Knox also made a number of helpful suggestions. I am grateful
to Anna Wierzbicka, René Dirven and Nick Enfield, who read and
made valuable comments on earlier drafts, to participants in the
LAUD Symposium held at Landau, Germany, in March 2000, and to
an anonymous reviewer.
Notes
1.A similar position was long held by Dan Slobin (1985), Melissa Bowerman
(1985), and others, but in recent years they have begun to repudiate their earlier
emphasis on an innate prelinguistic conceptual basis for language acquisition, in
favor of an emphasis on the characteristics of the adult "input" language, espe-
cially its language-specific aspects (cf. Bowerman 1996, Slobin 1997). In my
view there is no necessary conflict between these two positions, but it is not
possible to pursue the matter here.
2. The NSM bibliography is extensive and cannot be reviewed here for reasons of
space. Aside from works cited elsewhere in the chapter, representative works
include: Ameka (1990), Chappell (1986), Goddard (1996, 1997), Harkins
(1990,1996), Hasada (1994), Peeters (1993), Travis (1998), Wilkins (1986).
3. It is true that in the NSM model it can make sense to speak of "universal syn-
tax" or "universal grammar" (cf. Goddard and Wierzbicka forthcoming). For
some readers such expressions may carry a connotation of autonomous syntax,
so it is perhaps worth stressing that what is intended is literally a "conceptual
syntax". What we are trying to say is that certain combinations of primes neces-
sarily make sense and should be expressible in all languages, e.g. 'something',
'do', and 'good' can combine to form 'do something good'; 'say', 'something'
and 'someone' can combine to form 'say something to someone'. Importantly,
the potential for these combinations is inherent in and springs from the mean-
ings themselves. Thus the combinatorial syntax of primes is not autonomous

(from meaning) - quite the opposite in fact.
4. Why, it might be asked, could the young child not be working in terms of "con-
ceptual gestalts", in the sense of meaning complexes which are grasped as
whole configurations without any apprehension of their individual conceptual
constituents? In principle I have nothing against this suggestion, especially in
relation to very young children (see my comments at the end of Section 2.1).
However, the following analysis indicates that the same conceptual components
(e.g. WANT, SEE, DO, THIS, SOMEONE, SOMETHING) apparently recur in m a n y a n d
varied early child words; and furthermore, that such concepts generally acquire
surface lexical exponents within a few subsequent months. To my mind, these
facts are most easily explained on the assumption that the concepts in question
exist in a discrete fashion, so to speak, in the young child's mind.
5. If children's proto-linguistic signals can be explicated, and turn out to express
combinations of a certain small set of conceptual primes, perhaps the same ap-
proach can be used about intentional communication by non-human primates
(cf. Cheney and Seyfarth 1990, Tomasello and Call 1997). For example, the
three alarm calls used by vervet monkeys to indicate the presence of three dif-
ferent predators (eagles, snakes, leopards) could perhaps be explicated as mes-
sages that 'something bad can happen, because something is moving up-above
(for eagles)/down-below (for snakes)/near here (for leopards)'. See Jones
(1999) for a discussion of conceptual primes in an evolutionary, phylogenetic
perspective.
6. For some early talkers it might be that mama functions in its veiy earliest uses
merely as a "calling device". In Pete's case, however, he used the word to indi-
cate recognition of his mother, as well as to call her; and as the discussion in
§2.1 indicated, there are reasons to believe that the "proto-declarative" function
(with the illocutionary frame I SEE) is already present in the prelinguistic stage.
With this in mind, the only alternative explication for mama (and papa) would
seem to be as shown below. This explication does without WORD, but at the cost
of attributing to the child something like a self-conscious generalization, which
seems less than plausible.
mama =
this person
when I see/want this person, I say "mama"
7. Adult Oh-oh! can be used in a broader range of situations; for example, I could
say Oh-oh! upon reaching into my pocket for my keys and realizing that they
are not there. Aside from conveying the idea that something bad and unforeseen
is imminent, there is also a component of "minimisation" (Goddard 1998a:
190).
222 CliffGoddard
8. The prime SIDE is intended to be "spatial-relational" in character. That is, it

describes the positional relationship between two entities: 'X is on this (one,
two, all, etc.) side(s) of Y \ In some languages (including English) the same
word can also function as a nominal, but this nominal usage is semantically
complex (roughly, 'one side of X = one part of X, this part is on one side of all
the other parts of X') and non-universal. The proposed explication for wheel,
therefore, is not intended to imply that a wheel has sides.
9. If a young child does not comprehend Τ and 'you' that child would be quite at
a loss to understand a lot of conversation between adults. It seems likely that
children do understand the shifting referents in overheard speech long before
they are able to carry out the same manipulations in their own speech. Oshima-
Takane (1999) makes an argument that observing (and understanding) adults
using pronouns with one another is highly facilitative of the child developing
his or her own command over shifters.
10. Re. HAVE, it is interesting to note the importance of "offer" gestures, and of
"giving" and "taking" generally, in toddler play, cf. Bronson (1981). It is also
worth noting that "giving-and-taking" a small object is a routine which many
adults play with a young, preverbal child.
11. Esther Dromi's (1987) daughter Keren was different in many ways to Pete. Not
only was she acquiring Hebrew rather than English, she was an early and pro-
lific talker. Her first word came at 10 months and by the onset of the two-word
stage, late in month 17, she had acquired an impressive 337 different one-word
utterances. Dromi (1987: 171-179) lists Keren's entire one-word lexicon and it
is interesting to extract from this list the words which appear to be exponents of
semantic primes: (in order of appearance) HERE, WHAT, THIS, MORE, UP, WHERE,
NO, WANT, BIG, VERY, SMALL, GOOD, MOVE, SEE, YOU. Admittedly, some of
these putative identifications may be mistaken, mainly because the English
"glosses" relate to the adult meanings of the words, rather than necessarily re-
flecting the child's own usage. Nevertheless, there is a very substantial overlap
between this list and Pete's early vocabulary of primes.
12. The diagrams in Tomasello (1992) are sometimes so complex that they cannot
be understood without verbal explication, which makes it debatable whether
they really succeed in detaching the representation from adult language (for dis-
cussion, see Goddard Forthc. b).
References
Ameka, Felix
1990 The grammatical packaging of experiencers in Ewe: A study in
the semantics of syntax. Australian Journal of Linguistics 10(2):
139-181.
Bates, Elizabeth, Inge Bretherton and Lynn Snyder

1988 From First Words to Grammar. Individual Differences and Dis-
sociable Mechanisms. Cambridge: Cambridge University Press.
Bates, Elizabeth, Luigia Camaioni and Virginia Volterra
1979 The acquisition of performatives prior to speech. In: Elinor Ochs
and Bambi B. Schieffelin (eds.), Developmental Pragmatics,
111-130. New York: Academic Press.
Bloom, Lois
1973 One Word at a Time. The Hague: Mouton.
1991 Language Development from Two to Three. Cambridge: Cam-
bridge University Press.
Braine, Martin D. S.
1976 Children's first word combinations. Monographs of the Society
for Reseach in Child Development 44(1): 1-97.
Bowerman, Melissa
1978 The acquisition of word meaning: An investigation into some cur-
rent conflicts. In: N. Waterman and C. Snow (eds.), The Devel-
opment of Communication, 263-287. New York: John Wiley.
1996 The origins of children's spatial semantic categories: Cognitive
vs. linguistic determinants. In: John J. Gumperz and Stephen C.
Levinson (eds.), Rethinking Linguistic Relativity, 145-176. Cam-
bridge: Cambridge University Press.
Branson, Wanda C.
1981 Toddlers' Behaviors with Agemates: Issues of Interaction, Cog-
nition, and Affect. Norwood, NJ: Ablex.
Bruner, Jerome
1990 Acts of Meaning. Cambridge, MA: Harvard University Press.
Carey, Susan
1982 Semantic development: The state of the art. In: E. Wanner and L.
R. Gleitman (eds.), Language Acquisition: The state of the art,
347-379. Cambridge: Cambridge University Press.
Carter, Anne Lindsay
1978 The development of systematic vocalizations prior to words: A
case study. In: N. Waterson and C. Snow (eds.), The Develop-
ment of Communication, 127-138. London: John Wiley.
1979 The disappearance schema: Case study of a second-year commu-
nicative behavior. In: Elinor Ochs and Bambi B. Schieffelin
(eds.), Developmental Pragmatics, 131-158. New York: Aca-
demic Press.
Chappell, Hilary
1986 The passive of bodily effect in Chinese. Studies in Language
10(2): 271-296.
224 Cliff Goddard
Cheney, Dorothy L. and Robert M. Seyfarth

1990 How Monkeys See the World: Inside the Mind of Another Spe-
cies. Chicago: University of Chicago Press.
Clark, Eve V.
1993 The Lexicon in Acquisition. Cambridge: Cambridge University
Press.
1995 Later lexical development and word formation. In: Paul Fletcher
and Brian MacWhinney (eds.), The Handbook of Child Lan-
guage, 393-412. Oxford: Basil Blackwell.
Dore, John
1975 Holophrases, speech acts and language universals. Journal of
Child Language 2:21-40.
Dore, John, Margery B. Franklin, Robert T. Miller and Andrya L. H. Ramer
1976 Transitional phenomena in early language acquisition. Journal of
Child Language 3: 13-28.
Dromi, Esther
1987 Early Lexical Development. Cambridge: Cambridge University
Press.
1999 Early lexical development. In: Martyn Barrett (ed.), The Devel-
opment of Language, 99-132. Hove, UK: Psychology Press.
Goddard, Cliff
1996 The "social emotions" of Malay (Bahasa Melayu). Ethos 24(3):
426-464.
1997 Contrastive semantics and cultural psychology: 'Surprise' in Ma-
lay and English. Culture & Psychology 3(2): 153-181.
1998a Semantic Analysis: A Practical Introduction. Oxford: Oxford
University Press.
1998b Bad arguments against semantic primitives. Theoretical Linguis-
tics 24(2/3): 129-156.
Forthc. a Semantic and conceptual primes in early language development:
A diary study.
Forthc. b Verbal explication and the place of NSM semantics in cognitive
linguistics. In: June Luchjenbroers (ed.), Cognitive Linguistics:
Investigations across Languages, Fields, and Philosophical
Boundaries. Amsterdam: John Benjamins.
Goddard, Cliff and Anna Wierzbicka (eds.)
1994 Semantic and Lexical Universals - Theory and Empirical Find-
ings. Amsterdam: John Benjamins.
Forthcoming Meaning and Universal Grammar - Theory and Empirical
Findings. Amsterdam: John Benjamins.
Goldfield, Beverly A. and J. Steven Reznick

1990 Early lexical acquisition: Rate, content, and the vocabulary spurt.
Journal of Child Language 17: 171-183.
Gopnik, Alison and Andrew N. Meltzoff
1997 Words, Thoughts and Theories. Cambridge, M.A.: MIT Press.
Halliday, M.A.K.
1975 Learning How to Mean - Explorations in the Development of
Language. London: Edward Arnold.
Harkins, Jean
1996 Linguistic and cultural differences in concepts of shame. In: D.
Parker, R. Dalziell and I. Wright (eds.), Shame and the Modern
Self, 84-96. Melbourne: Australian Scholarly Publishing.
Harris, Margaret, Martyn Barrett, David Jones, and Susan Brookes
1988 Linguistic input and early word meaning. Journal of Child Lan-
guage 15: 77-94.
Harris, Margaret, Caroline Yeeles, Joan Chasin, and Yvonne Oakley
1995 Symmetries and asymmetries in early lexical comprehension and
production. Journal of Child Language 22: 1-18.
Hasada, Rie
1996 Some aspects of Japanese cultural ethos embedded in nonverbal
communicative behaviour. In: Fernando Poyatos (ed.), Nonverbal
Communication in Translation, 83-103. Amsterdam: John Ben-
jamins.
Ingram, David
1989 First Language Acquisition. Cambridge: Cambridge University
Press.
Jones, Doug
1999 Evolutionary psychology. Annual Review of Anthropology 28:
553-575.
Keil, Frank C.
1989 Concepts, Kinds, and Cognitive Development. Cambridge, MA:
MIT Press.
1987 Foundations of Cognitive Grammar, Volume 1. Stanford, CA:
Stanford University Press.
Markman, Ellen M.
1989 Categorization and Naming in Children: Problems of Induction.
Cambridge, MA: MIT Press.
McShane, John
1979 The development of naming. Linguistics 17: 879-905.
226 CliffGoddard
Mervis, Carolyn B.
1987 Child-basic object categories and early lexical development. In:
U. Niesser (ed.), Concepts and Conceptual Development: Eco-
logical and intellectual factors in categorization, 201-233. Cam-
bridge: Cambridge University Press.
Mervis, Carolyn B., Cynthia A. Mervis, Kathy E. Johnson and Jacquelyn Bertrand
1992 Studying early lexical development: The value of the systematic
diary method. In: Carolyn Rovee-Collier and Lewis P. Lipstitt
(eds.), Advances in Infancy Research, Vol. 7, 292-378. Nor-
wood, NJ: Ablex.
Ninio, Anat
1992 The relation of children's single word utterances to single word
utterances in the input. Journal of Child Language 19: 87-110.
Oshima-Takane, Yuriko
1999 The learning of first and second person pronouns in English. In:
R. Jackendoff, P. Bloom and K. Wynn (eds.), Language, Logic
and Concepts, 373-410. Cambridge, M.A.: MIT Press.
Peeters, Bert
1993 Commencer et se mettre à: une description axiologico-concep-
tuelle. Langue française 98: 24-47.
Plank, Frans
1989 On Humboldt on the dual. In: R. Corrigan, F. Eckman and M.
Noonan (eds.), Linguistic Categorization, 293-333. Amsterdam:
John Benjamins.
Reddy, Vasudevi
1999 Prelinguistic communication. In: Martyn Barrett (ed.), The De-
velopment of Language, 25-50. Hove, UK: Psychology Press.
Schieffelin, Bambi B. and Elinor Ochs (eds.)
1986 Language Socialization across Cultures. Cambridge: Cambridge
University Press.
Schwartz, R. and L. Leonard
1982 Do children pick and choose: an examination of phonological
selection and avoidance in early lexical acquisition. Journal of
Child Language 9: 319-36.
Scollon, Ronald
1976 Conversations with a One-Year-Old: A Case Study of the Devel-
opmental Foundation of Syntax. Honolulu: University Press of
Hawaii.
1979 A real early stage: An unzippered condensation of a dissertation
on child language. In: Elinor Ochs and Bambi B. Schieffelin
(eds.), Developmental Pragmatics, 215-228. New York: Aca-
demic Press.
Slobin, Dan I.
1985 Crosslinguistic evidence for the language-making capacity. In:
Dan I. Slobin (ed.), The Crosslinguistic Study of Language Ac-
quisition, Vol. 2: Theoretical issues, 1157-1256. Hillsdale, New
Jersey: Lawrence Erlbaum.
1997 The origins of grammaticizable notions: Beyond the individual
mind. In: Dan I. Slobin (ed.), The Crosslinguistic Study of Lan-
guage Acquisition, Vol. 5: Expanding the Contexts, 265-323.
Mahwah, New Jersey: Lawrence Erlbaum.
Tien, Adrian
1999 Early lexical exponents and 'related' lexical items as manifesta-
tions of conceptual/semantic primitives in child language. MA
Thesis. Australian National University.
Tomasello, Michael
1987 Learning to use prepositions: A case study. Journal of Child Lan-
guage 14: 79-98.
1992 First Verbs: A Case Study of Early Grammatical Development.
Cambridge: Cambridge University Press.
Tomasello, Michael and Josep Call
1997 Primate Cognition. Oxford: Oxford University Press.
Travis, Catherine
1998 Omoiyari as a core Japanese value: Japanese-style empathy? In:
Angeliki Athanasiadou and Elzbieta Tabakowska (eds.), Speaking
of Emotions: Conceptualization and Expression, 55-82. Berlin:
Mouton de Gruyter.
Wierzbicka, Anna
1972 Semantic Primitives. Translated by A. Wierzbicka and J. Bese-
meres. Frankfurt/M.: Athenäum Verlag.
1980 Lingua Mentalis: The Semantics of Natural Language. Sydney:
Academic Press.
1985 Lexicography and Conceptual Analysis. Ann Arbor: Karoma.
1995 Universal semantic primitives as a tool for the study of language
acquisition. Unpublished MS.
1996 Semantics, Primes and Universals. Oxford: Oxford University
Press.
Wilkins, David
1986 Particles/clitics for criticism and complaint in Mparntwe Arrernte
(Aranda). Journal of Pragmatics 10(5): 575-596.
No preposition required.
The role of prepositions for the understanding of
spatial relations in language acquisition
1. Introduction
In cognitive semantics locative prepositions play an important role

for the processing of spatial relations. By using a locative preposi-
tion, the speaker's understanding of a spatial relation is marked. One
of the most striking issues for cognitive science is whether there are
universal relations that are first acquired and then marked via a
preposition associated to them. Some studies (Clark 1973, Grieve,
Hoogenraad and Murray 1977, Sinha et al. 1999) from developmental
psychology indicate that this may be the case. In English and Ger-
man, an order of acquisition has been observed, whereby the relations
IN and ON are acquired earlier than e.g. UNDER. It is probable that
these relations IN and ON are universal and therefore acquired first.
On the other hand Bowerman's (1996a) results indicate a high degree
of language specificity for spatial relations acquisition. Choi and
Bowerman (1991) conducted a study, in which they collected and
compared data of children learning English and Korean. This study is
especially important, because the languages differ significantly in
how they classify spatial configurations: the English IN-relation is
matched by two different Korean relations, depending on whether the
relation is a tight fit (like a video cassette in its box) or whether the
relation is a loose fit (like an apple in a bowl). This different spatial
configuration originates from a different categorization: "English
makes a distinction between actions resulting in containment (put in)
versus support or surface attachment (put on), while Korean makes a
cross-cutting distinction between tight-fit relations (kkita) versus
230 Katharina J. Rohlfing
loose-fit or other contact relations (various verbs). In particular, the

Korean verb kkita refers to actions resulting in a tight fit relation re-
gardless of containment or support" (Choi et al. 1999: 241). The re-
sults from the study (Choi and Bowerman 1991) showed that infants
distinguish very early between different forms in Korean for the one
English IN-relation and acquire them without giving preference to
any universal form. The acquisition of prepositions is, according to
Bowerman (1996a), language-specific and leaves little opportunity to
bear out universal relations. These results suggest that children do not
necessarily develop preverbal concepts, but can construct concepts on
the basis of the linguistic input (Bowerman 1996b).
A recent study by Choi, McDonough, Bowerman and Mandler
(1999), however, shows that children's sensitivity to language-
specific spatial categories begins at the age of 18 to 23 months. But
children acquire an extensive knowledge about space, objects and
events already before the age of 18 months. Another possibility there-
fore is that "children do map spatial words directly to pre-established
spatial concepts - it is just that this set of notions is larger and more
varied than has typically been supposed; for example, in addition to
'containment' and 'support', children might also have a prelinguistic
notion of 'tight fit'" (Choi et al. 1999: 264). It is therefore important
to elaborate which role the language-specific spatial categories ex-
actly play in the child's primary process of learning about space.
The psycholinguistic study I have undertaken examines the role of
prepositions for understanding instructions with spatial relations in
language acquisition. It examines not only the extent to which prepo-
sitions are important for understanding spatial relations, but also
whether there are other, non-linguistic strategies for understanding
that make universale at a non-linguistic level conceivable after all.
Thus, the aim of the study was to examine whether the infants ex-
plicitly respond to prepositions and receive specific information
through them.
As I will show, the non-linguistic strategies contain culture-
specific aspects and the input of language could therefore be found
there already. However, even if we assume that these strategies are
like an equation for a behavior with one culture-specific variable,
No preposition required 231
they can nevertheless let us get an insight into the learning processes:
not the language-specific content of the spatial knowledge is more
focused, but the (probably) universal processes which include some
cultural-dependent aspects.
2. Non-linguistic strategies
My study touches on a discussion about non-linguistic strategies that

are said to be responsible for understanding locative instructions.
According to the discussion, infants' understanding is based on non-
linguistic strategies and incomplete semantic knowledge, although
the incompleteness of semantic knowledge concedes that children
have and use semantic knowledge at this age. It is, moreover, thought
that non-linguistic strategies make infants much more sensitive to the
context than to the meaning of linguistic utterances. It is not my in-
tention to deny that non-linguistic strategies are culture-dependent,
rather, my main point is that these strategies are grounded in the
context, i.e. in the environment of the child.
The main task for the children in the study was to perform a rela-
tion between two objects, i.e. to set the trajector to the landmark.
Note that already at the beginning of the relational performance, cul-
tural aspects are crucial: the canonical orientation (Sinha 1982), i.e.
in which position the object is linked to an activity or how an object
is held to achieve something, plays an important role for the child
handling the objects. The canonical orientation of an object is estab-
lished by canonical rules1. The canonical rules, however, are pro-
vided by human social interest and value (Sinha 1983: 276). Per-
forming the task, the child determines moreover which object is tra-
jector and which landmark, decides about the proper relation between
these two objects, about the appropriate motion of the trajector to the
landmark, and about the direction or the kind of the motion. The re-
sult of my study suggests that non-linguistic strategies help children
in their decision.
Eve Clark (1973) mentions non-linguistic strategies in her studies
with on average two-year-old infants. She claims that infants' be-
232 Katharina J. Rohlfìng
havior is based on non-linguistic strategies and incomplete semantic

knowledge. Concerning non-linguistic strategies, infants may be
guided by the features of objects. The physical properties of an object
provoke a certain type of behavior and are the reason why the ac-
companying instructions are not taken into account. E.g. if an object
includes a surface, the infants' behavior is influenced insofar as they
will always put something on this surface, without following the se-
mantic meaning of the instruction correctly. Clark (1973: 168) ob-
served the same type of influenced behavior when the landmark-
object1 (LM-object) is a container. In conclusion, Clark claimed two
rules of non-linguistic knowledge based on physical properties of
objects:
Rule 1: If the LM-object is a container, the TR-object (trajector-

object2) is inside it
Rule 2: If the LM-object has a horizontal surface, the TR-object is
on it.
Note, however, that there is already a culture-dependent aspect in this

physical-property strategy. When infants see a new object (e.g. a
bridge), they associate it with their knowledge of the orientational
and functional properties of an object they already know well (e.g. a
table). In my study, only one child had some problems in placing the
bridge in an upright orientation and placed it upside-down, as a crib.
Thomas Thiel (1985) criticized Clark's non-linguistic strategies.
In his study, he focussed the "relational character of objects [relation-
aler Charakter eines Objektes]" (Thiel 1985: 192), i.e. the infants link
a certain activity to a type of object in its canonical orientation (Sinha
1982). Thus, the relational character is culture-dependent and subjec-
tive because of the child's previous experiences, i.e. which activity
the child has repeatedly observed with a certain object. The relational
character consists of two parameters: active - passive and one or
more spatial relations. If a child observes that objects are put on a
plate, the plate will have a passive relational character involving ON-
relations (as a result, the objects are put on it), because the plate will
repeatedly serve as a landmark-object supporting others. The objects
that are put on the plate (food or cups) obtain an active relational
character.
Chris Sinha (1982, et al. 1999) determined the relational character
of objects on the basis of cultural differences. He focuses on the ca-
nonical orientation of objects as artefacts in a given culture. If, for
example, an object is used as a container, the canonical orientation
will be as follows: the object is supported by a surface, its top is open
and the infants are inclined to put things in this object. Therefore it is
very unusual for infants to use this object upside-down as a support-
ing object and to put things on the bottom of it. The canonical rules
help children to handle the objects, i.e. they know in which orienta-
tion to use the objects to achieve something (e.g. how to hold a pen
to write). Observing this, children develop an extensive knowledge
about objects and the activities associated with them and try to pre-
serve it. For the task in my study it means that as soon as the child
recognizes an object (the trajector and/or the landmark), correspond-
ing background knowledge (Sinha 1983) is activated. This back-
ground knowledge about one object (e.g. the landmark) can dominate
over the relational behavior in the sense that the object serves as a
target and motivates the motion needed in order to reach this target in
a relation, while it does not matter what the trajector is and if it suits
this relation (e.g. the TR-object is too big for the LM-object).
To summarize, the non-linguistic strategies discussed here have
the claim in common that they more strongly determine infants' be-
havior than their lexical knowledge3 does. To examine how strong
the influence of language is, the following study was designed.
3. Experiment
The study focuses on the claim of non-linguistic strategies and is

based on two types of instructions, which have been designed to ex-
amine whether the infants explicitly react to prepositions and receive
specific information through them: one type consists in a syntacti-
cally correct request relating two objects to each other in view of a
234 Katharina J. Rohlfmg
certain spatial relation; the other also consists in a request, but in this
case it is syntactically incorrect: the preposition is omitted.
The choice of preposition was guided by the order of prepositional
acquisition in German (e.g. Thiel 1985) and in English (e.g. Clark
1973, Sinha et al. 1999). According to the order, the preposition IN is
acquired first, the preposition ON / AUF second. This can also be
observed in Polish for the prepositions DO (IN) and NA (ON)
(Rohlfing 1998: pilot study). In Polish, however, there are two prepo-
sitions equivalent to English ON: a dynamic and a static one. Both
are verb-derived but the latter depends on the motion of the verb and
describes the spatial movement, whereby the former stands for a
fixed location or a state. The dynamic or static character of a prepo-
sition determines the inflection of the cases (see below); in the case
of the IN-preposition there are also two different lexemes: DO for a
dynamic IN-preposition and NA for a static one. In this study only
the dynamic form NA has been used, because the static form involves
more syntactic complexity in the instructions4. In the following it will
be marked with "->" to distinguish it from the static form.
The other motivation for the choice of this preposition is the fea-
ture of inflection in Polish. Spatial relations are created with the aid
of a preposition and, depending on the object's gender, with inflec-
tion of the cases. An instruction which contains this relation is:
(1) Daj konika NA—> most

Give horse-AKK ON bridge
'Put the horse on the bridge'
(2) Daj piticç NA-> kulç

Give Ball-AKK ON Sphere-AKK]
'Put the ball on the sphere'
A syntactically incomplete instruction is:

(3) a. Daj konika _ most

Give horse-AKK_ bridge]
'Put the horse the bridge'
One can argue, referring to Slobin's (1973: 191) organizations' prin-

ciples, that the infants know which preposition is omitted because of
the inflection. One needs to keep in mind, however, that syntactically
it is also possible to complete the incorrect instruction with other
prepositions:
b. Daj konika POD-+ most

Give horse-AKK UNDER bridge
c. Daj konika NAD-> most
Give horse-AKK OVER bridge
d. Daj konika ZA—> most
Give horse-AKK BEHIND bridge
e. Daj konika PRZED-> most
Give horse-AKK IN FRONT OF bridge
f. *Daj konika W—> most
Give horse-AKK INTO bridge
Therefore, there must be more than a syntactical indicator leading

infants to a correct understanding. According to these multiple possi-
ble solutions to complete an instruction syntactically, it is especially
interesting to observe infants' behavior when they react to the in-
struction without a preposition.
3.1. Stimuli
In two different situations, infants were instructed to relate two ob-

jects in an ON-relation. In a well-known situation, objects and toys
were used that the infants knew from their everyday situations. The
trials in the well-known situation consisted of three sets:
Set 1: Pot (Trajector-object) and Table (Landmark-object);

Set 2: Cup (TR) and Plate (LM);
Set 3: LEGO®-Horse (TR) and Bridge (LM).
In the second situation in which infants were tested the infants were
confronted with an abstract wooden construction. The construction
can be said to be abstract or 'not yet specified' because the child has
never seen it before. Furthermore, it was designed not to provoke a
certain relation with its physical properties and to give as little con-
text as possible coming from the relational character of the objects
used. Especially for the choice of relational objects in this wooden
construction, it was important for them not to be associated with a
relation, as e.g. an object with a surface associates the relation ON.
Thus, two balls were used: the first was installed in a wooden con-
struction named HiK5 and covered with Velcro®; the second was mo-
bile and could be related (and attached) to the first one. Balls suggest
few spatial relations, because they have neither a horizontal surface
nor are they containers. They suggest only the well-known activity of
"rolling".
Figure 1. The ^ - S i t u a t i o n
3.2. Design and procedure
In two situations (an abstract ///^-situation and a well-known situa-

tion), instructions were given without and with a preposition:
Table 1. Design of the Experiment
SITUATION
well-known abstract HiK
without preposition (INS„p) AKA, AKA 3
INSTRUCTION
with preposition (INSNA->) AKA 2 AKA 4
AKA: Number of correct performances according to the instruction
The order of the set (HiK, then sets 1, 2, 3 from the well-known
situation) corresponded to the progressive interest of the children. For
many infants the //zX-situation was too abstract and it was doubtful
whether they would have wanted to continue to play with it, after
they had seen the sets 1-3. Consequently, every child was seated in
front of the HiK first. To attract her/his attention, questions like
"What is it?" were asked. Then the names for the objects (sphere and
ball) were introduced. Finally, instructions were given that were neu-
tral in their formulation. They always began with Daj... 'Give...'.
This formulation is semantically correct in Polish. Other verbs like
Postaw... 'Put...' imply a state in which one object is supported by
another (e.g. lies on a table).
3.3. Predictions
These two situations correspond to two hypotheses. For the well-

known situation the first hypothesis can be formulated as follows: if
the infants master both tasks (instruction with and without preposi-
tion) with the same success, it means that in a well-known situation
infants understand instructions without prepositions as well as with
prepositions (AKAi = AKA2), because the relational character of
objects is crucial here. The focus will be on the child's first response
to an instruction without preposition: if she or he carried out the ex-
pected relation, it would not be due to the preposition.
For the abstract situation the second hypothesis is formed: if the
preposition supplies important information, then the infants will pay
more attention to it in an abstract situation. This means that the in-

structions with prepositions will lead to better performance than in-
structions without prepositions (AKA3 < AKA4). In this case, the
focus will be on the child's second response, i.e. to the instruction
with preposition NA-» - can she or he specify the location according
to this preposition?
Finally, one can expect that the relational character of objects is a
component of infants' understanding insofar as their behavior is
guided by the relational character of objects, and it is easier for them
to perform instructions that refer to well-known or typical situations.
This means that toy-objects such as a table or a bridge imply the re-
lation ON and, even when performing incomplete instructions, in-
fants will choose this relation (in spite of the fact that linguistically
there are other options to complete this instruction).
3.4. Participants
Twenty-four Polish-speaking infants (thirteen females and eleven

males) with an average age of 23 months participated in this study.
All the children were native speakers of Polish. The youngest partici-
pant was 20, the oldest 26 months old. With the aid of two pediatri-
cians and one pharmacist the parents were invited to participate and
contacted later. The experiment was carried out in three different
places: in two doctors' surgeries and a parish room. Each subject was
interviewed separately. The parents were asked to make no com-
ments during the experimental trials, and to encourage the infants
whenever they turned to their parents for help.
3.5. Scoring
The following table shows the infants' performance before scoring,

i.e. how many infants performed the NA-» relation as suggested by
the objects regarding different sets. The line 'other relations' in the
table refers to how many infants performed different relations than
expected (e.g. POD [UNDER] instead of NA->). The line 'used dif-
ferent toy' shows how many children performed the NA-» relation
with different toys as provided by the set (e.g. they put the horse ON
the table instead of ON the bridge) — this performance was excluded
from the final scoring. The term 'anticipation' refers to a situation in
which a child performed the NA-» relation before the instruction
could be expressed (as soon as both objects were presented). This
reaction is also omitted for the final scoring.
Table 2. Infants' performance in a well-known situation (before scoring)
set 1 set 2 set 3

INSwp pot/table cup/plate horse/bridge
INSTRUCTION
performed NA-> 14 13 14
INSwp other relations 0 1 0
used different 6 3 3
invalid 4 7 7
performed NA-> 13 9 16
INSna_» Anticipation 2 5 2
other relations 0 1+26 0
used different 3 4 3
invalid 6 3 3
Referring to the ///^-situation, the table shows how many infants

performed a contact between the ball and the sphere due to the in-
struction without a preposition. In the case of an instruction with a
preposition, infants were expected to specify the location of the ball
on the sphere. If they could not specify the location due to the in-
struction, they mostly repeated the already performed relation - how
many infants behaved this way can be viewed under 'other relation'.
The line 'another action' refers to responses in which infants did not
make a contact between the ball and the sphere but carried out other
actions like throwing the ball away or trying to take the sphere down.
Table 3. Infants' performance in an abstract ///^-situation (before scoring)
performed contact 14
INS
"τ other relations 5
INSTRUCTION performed N A - > 4~
INSNA-> another action 4
another action 11
The final scoring (presented in Table 4 and 5) concentrates on a

comparison between the different tasks' forms (with and without a
preposition). Every infant's performance in the well-known situation
containing an instruction with a preposition, e.g. in set 1, was com-
pared to her/his performance resulting from an instruction without a
preposition from another set, e.g. set 2. Altogether six comparisons
were added up to establish the total result, but the performance in the
well-known situation was not compared to the //zX-situation. The
infants' performance due to instructions was coded as a solved or
non-solved task: if a child performed a relation NA-> between the
LM and TR-object in a well-known situation due to an instruction
without preposition (INSwp) as well as due to an instruction with
preposition (INSNA->) from a different set, this task was coded as a
solved task. If a child performed another relation than NA-» between
the LM and TR-object in a well-known situation, this response was
coded as a non-solved task. In the case when the child performed the
relation with a different object as provided by the set or her/his rela-
tional reaction came before the instruction could be uttered, this re-
sponse was excluded from scoring.
In an abstract situation, in the case of an instruction without any
preposition a task was coded as solved if a child produced a contact
between the ball and the sphere. In the case of an instruction with a
preposition, a task was only coded as solved if a child located the ball
in the correct NA-» position on the sphere (merely to produce a
contact was not sufficient). Any other reaction, e.g. to put the ball on
the bottom of the //zX-construction or on its vertical wall, was coded
as non-solved.
4. Results
The first result is my personal observation and concerns the infants'

responses to the instructions without preposition. They were not sur-
prised when the incomplete instruction was uttered. Rather they
seemed to react as if it were a complete instruction and did not show
any uncertainties in their behavior. The results of the experiment in
the first situation are:
Table 4. Results in the well-known situation
INSwp (AKA,)
solved non-solved
INS na ^ solved 48 1
(AKA2) non-solved 1 0
According to the McNemar-test for dependent samples the tasks are

equally difficult when the number of persons who solved the first
task (INSwp) and did not solve the second task (INSna-*) is equal to
the number of persons who did not solve the first but solved the sec-
ond task. In the table above, the number is 1 in both cases. The re-
sults indicate that the tasks were equally difficult, and the hypothesis
can be confirmed: AKAi = AKA2, i.e. that in a well-known situation,
infants understand instructions without prepositions as well as in-
structions with prepositions.
Note, that in the case of an instruction without a preposition, a
certain relation (in this case: NA-») emerges especially from the ob-
jects.
For the abstract ///.^-situation, the results are as shown in Table 5.
Table 5. Results in the abstract ///^-situation

INSwp (AKA3)
solved non-solved
INSna_> solved 2 3
(AKA4) non-solved 10 4
Because the sum of the interesting fields for the McNemar-test for
dependent samples is 13 and therefore smaller than 60, the Binomial-
test was applied to this data. Assuming a significance level of 0.05
the results of the Binomial-test indicate that the tasks were equally
difficult. Moreover, the results suggest a tendency (-3 = 0.035) against
the second hypothesis made for the abstract situation (if the preposi-
tions supply important information, then the infants will pay more
attention to them and the task INSNA-> will be easier for them to per-
form). In the abstract situation, it seemed to be easier for the infants
to react to an instruction without preposition because this type of
instruction implies only a contact between the sphere and the ball. In
contrast, the type of instruction implying a concrete location of the
ball on the sphere was difficult for the infants to carry out. Thus, the
second hypothesis that infants need prepositions to understand loca-
tive instructions in an abstract situation cannot be borne out. On the
contrary, the results tend to mean just the opposite: in the HiK-
situation the location and the appropriate preposition of the ball on
the sphere do not have a meaning; only the activity of attaching mat-
ters to the infants.
5. Discussion
The results from the well-known situation show that the ΝΑ-»
preposition does not have a prior role in understanding locative in-
structions. To solve the tasks, the relational character of objects used
in the trials was crucial. As a consequence of the instructions without
prepositions, infants carried out the NA—>· (ON)-relation as implied
by the LM-object. Only in two cases during the cup / plate-set did the
infants show other reactions. Their 'mistakes' were interesting in
their own way. Marcel (23 months) received the instruction to put
"the cup plate" (saucer) and he put it the other way around, i.e. the
plate (saucer) ON the cup instead. I wondered about his reaction:
according to Thiel's relational character, should not the plate suggest
a passive ON-relation? After consultation with his mother it turned
out that she always makes tea that way: she puts the tea in a cup and
then covers it up with a saucer to stew. Thus, while solving the task,
this behavior imitated a situation that was well-known. In this exam-
ple, the non-linguistic strategy was established by the typicality of the
situation, depending on certain experiences of a child. According to
my observations during the study, the typicality of a situation consists
of a canonical spatial relation implying a canonical orientation of
objects as noticed by Sinha (1982: 140). The child perceives the
situation task-dependently, and the emerging relation is linked to a
certain activity and a canonical relation (in Marcel's case: plate ON
cup) with a canonical orientation (plate inverted, cup upright) of ob-
jects that depends on that situation. From the task-oriented situation
follows the relational character of objects, which, as described by
Thiel (1985), is dominated by the aspect of canonical orientation of
an object. So, if the child observes a specific kind of tea-making and
gets used to it (typicality of a situation), the child will associate the
plate with the activity of covering, and the relational character of the
plate is in this case: active, ON.
The results presented here confirm Clark's partial semantics hy-
pothesis to the extent that, at this stage, the infants' behavior is based
on the combination of partial semantic knowledge and non-linguistic
strategy. However, more than just the two rules formulated by Clark
(1973) that are based on physical properties of objects are involved in
the non-linguistic strategies; the typicality of a situation also belongs
to them, which is an extension of the relational character of objects
formulated by Thiel (1985) and includes cultural aspects as the ca-
nonical orientation of objects described by Sinha (1982).
The tendency to refute the second hypothesis on the basis of the
results from the abstract situation suggests that in a new and abstract
situation the prepositions in an instruction do not supply any addi-
tional information. Moreover, in this procedure, a task in which chil-
dren were expected to respond to a preposition directly and to specify
the location accordingly, seemed to be more difficult. Nevertheless, a
longer familiarization period could help to find out more about in-
fants' behavior in the ///X-situation. For the time being, the results
suggest that infants employ non-linguistic strategies to understand
locative instructions, and a linguistic instruction cannot be under-
stood context-independently at this age. Infants seem to rely on their

experience of what an object does - they cannot perform only on the
basis of a linguistic instruction or its lexical meaning.
The results from the well-known situation show that infants un-
derstand instructions containing the spatial relation NA-> (ON) very
well. However, their understanding of instructions does not correlate
with the lexical knowledge about this spatial relation. Many factors
are responsible for infants' understanding of instructions in such a
situation: non-linguistic strategies according to Clark (1973), the re-
lational character of objects (Thiel 1985) including their canonical
aspects (Sinha 1982) and the typicality of a situation. There seems to
be a variety of strategies that establish the process of acquiring the
lexical meaning of the spatial relation. Considering the observations
presented here, some consequences can be drawn for studying con-
ceptual and lexical knowledge: conceptual knowledge should be ex-
plored more task- and context-dependently, because the non-
linguistic strategies are a part of the infants' understanding and they
establish the lexical meaning of spatial relations. Nonetheless some
spatial relations like NA are easier to understand with the aid of non-
linguistic strategies than others (e.g. POD [UNDER]). Why? It seems
plausible to speculate that the NA-relation can be mapped more di-
rectly to pre-established spatial concepts configured with the aid of
non-linguistic strategies. In this sense, the NA-relation is supposedly
more basic than universal, because it is based on non-linguistic
strategies. Yet it is also culture-dependent, because non-linguistic
strategies include culture-dependent aspects like canonicity, which
corresponds to cultural values and experiences and therefore cannot
be viewed as universal.
Acknowledgements
The research reported in this paper took place in the Graduate Pro-
gram "Task-Oriented Communication" (GK 256) at the University of
Bielefeld, Germany and was supported by the Deutsche Forschungs-
gemeinschaft.
I am grateful to Dan I. Slobin for the idea of instructions without

prepositions - and to Chris Sinha and René Dirven for their com-
ments on this paper.
I would like to thank Joanna Kosiedowska for her help in organ-
izing this study in Poland, all the children and parents who partici-
pated, and Marc Rohlfing for support and videotaping.
Notes
1. Canonical rules, in general, are what govern relations between endostructure

and exostructure in non-natural kinds, and their motivation is to be sought in the
way in which human interest and value is both shaped by, and shapes, a natural
environment. "Cups, for example, do not occur naturally, but must be produced;
[...] Basically, a cup, to be a drinking vessel, must be a container; and contain-
ers must possess a certain (endo)structure. In this sense we may refer to the cup
as a canonical container." (Sinha 1983: 276)
2. The terminology is adapted from Langacker ( 1991: 6, 1998: 5)
3. In the analysis of my study, the term lexical knowledge is applied task-
dependently, concerning the main question of how much information a child re-
ceives through a certain preposition. Thus, by lexical knowledge I refer to the
child's cognitive ability to abstract the particular meaning of an ON-relation and
transfer it to every conceivable situation even if she or he has not seen this
situation in their every day life. I distinguish lexical knowledge from conceptual
knowledge, because I am convinced that conceptual knowledge develops much
earlier than lexical knowledge can be determined.
4. An instruction including a static preposition could be constructed as e.g. Zrób
tak, zeby konik bylna moscie 'Do this way that the horse is on [NA] the bridge-
LOK]' - which is more complicated than the instruction Daj konika na most
'Put the horse on [NA-»] the bridge'.
5. The name HiK stands for "Heidelberger interessante Konstruktion", and refers
to the fact that the idea for this construction was born in Heidelberg.
6. In this case only one child performed the converse relation (i.e. the plate on the
cup), the other two put the cup on the table (on which the objects were pre-
sented, not on the toy-table).
References
Bowerman, Melissa
1996a The origins of children's spatial semantic categories: cognitive
versus linguistic determinants. In: John J. Gumperz, and Stephen
C. Levinson (eds.), Rethinking Linguistic Relativity, 145-176.
Cambridge: Cambridge University Press.
1996b Learning how to structure space for language: A crosslinguistic
perspective. In: P. Bloom, M. Peterson, L. Nadel and M. Garrett,
(eds.), Language and Space, 385-436. Cambridge, Mass.: MIT
Press.
Choi, Soonja and Melissa Bowerman
1991 Learning to express motion events in English and Korean: The in-
fluence of language-specific lexicalization patterns. Cognition 41:
83-121.
Choi, Soonja, Laraine McDonough, Melissa Bowerman and Jean M. Mandler
1999 Early sensitivity to language-specific spatial categories in English
and Korean. Cognitive Development 14: 241-268.
Clark, Eve V.
1973 Non-linguistic strategies and the acquisition of word meanings.
Cognition3: 161-182.
Grieve, Robert, Robert Hoogenraad and Diarmid Murray
1977 On the young child's use of lexis and syntax in understanding
locative instructions. Cognition 5:235-250.
1991 Concept, Image, and Symbol. The Cognitive Basis of Grammar.
Berlin and New York: de Gruyter.
1998 Topic, subject, and possessor. Linguistic Notes from La Jolla 19:
1-28.
Sinha, Chris
1982 Representational development and the structure of action. In: G.
Butterworth and P. Light (eds.), Social Cognition: Studies of the
Development of Understanding, 137-162. Brighton: Harvester.
1983 Background knowledge, presupposition and canonicality. In: T.
Seiler and W. Wannenmacher (eds.), Concept Development and
the Development of Word Meaning, 269-296. Berlin: Springer.
Sinha, Chris, Lis Thorseng, Mariko Hayashi and Kim Plunkett
1999 Spatial language acquisition in Danish, English and Japanese. In:
P. Broeder and J. Murre (eds.), Language and Thought in Devel-
opment. Cross Linguistic Studies, 95-125. Tübingen: Gunter
Narr.
Slobin, Dan I.
1973 Cognitive prerequisites for the development of grammar. In:
Charles A. Ferguson and Dan I. Slobin (eds.), Studies of Child
Language Development Crosslinguistic. Evidence for the Lan-
guage-Making Capacity, 175-276. New York: Holt, Rinehart and
Winston, Inc.
Thiel, Thomas
1985 Räumliches Denken und Verständnis von Lokativen beim Spra-
cherwerb [Spatial thinking and understanding of locatives in lan-
guage acquisition]. In: Harro Schweizer (ed.), Sprache und Raum
[Language and Space], 184-208. Stuttgart: Metzler.
Wilcox, Stephen and David S. Palermo
1974 "In", "on", and "under" revisited. Cognition 3: 245-254.
The 'Graded Salience Hypothesis' in
second language acquisition
Istvan Kecskes
1. Objectives
This paper discusses the validity of the 'graded salience hypothesis'

(GSH) in second language acquisition based on a cross-sectional
study conducted with 30 native speakers of English (NSs) and 51
non-native speakers of English (NNSs) from 12 countries who stud-
ied English as a foreign language in their home country for at least
four years and have spent from six months to two years in the target
language environment.
2. The Graded Salience Hypothesis
The GHS claims that in LI processing figurative and literal language

use is governed by a general principle of salience. Salient meanings
(e.g. conventional, frequent, familiar, predictable, enhanced by prior
context) are processed first, and parallel processing is induced when
more than one meaning is salient (e.g. Gibbs 1980, 1984, Blasko and
Connine 1993, Giora 1997, forthcoming). The hypothesis requires
that the standard pragmatic model be revised: Instead of postulating
the priority of literal meaning, the priority of salient (e.g. conven-
tional, familiar, frequent) meaning should be postulated (Giora 1997,
forthcoming). Consequently, it is not the figurative versus literal
split, but the salient versus non-salient continuum that really counts
when processing the meaning of words or utterances in the LI.
Salient interpretation has unconditional priority over less salient
interpretation. The most salient meaning of a word or utterance is
250 Istvan Kecskes
always activated, and is always activated initially. The more salient

the meaning is, the more difficult it is to reject as the intended
meaning. In the case of conventional metaphors or expressions,
whose figurative and literal meanings are as salient, both are pro-
cessed initially. Three types of processing can be distinguished
(Giora 1997):
i. Direct processing applies when highly salient meanings are in-

tended. For example, the salient figurative meaning of highly con-
ventionalized idioms or implicatures.
ii. Parallel processing occurs when alternative meanings are equally
salient (conventional metaphors whose metaphoric and literal
meanings are equally salient).
ili..Sequential processing applies when language is used innovatively,
as in the case of literal uses of highly conventional idioms. The
more salient meaning is processed initially, before the intended
meaning is derived.
3. The Study
3.1. Selection of pragmatic units
A study was conducted to investigate the use of situation-bound ut-

terances (Kecskes 1999,2000a, 2000b) and implicatures (Grice 1968,
Bouton 1994) by Non-Native Speakers of English (NNSs). The se-
lection of these pragmatic units over other lexical phrases and ex-
pressions needs explanation:
i. Implicatures and situation-bound utterances (SBUs) are especially

suitable for checking the validity of GSH in second language ac-
quisition because their interpretation is dependent on the context
of a particular language use. However, they differ in the obligato-
riness of occurrence. In institutionalized contexts the choice be-
tween actions that are theoretically available to the individual is
much more restricted than in informal contexts. Reactions to
The 'GradedSalience Hypothesis' 251
situations are designed beforehand, and in many cases it is the

SBUs that serve as a means of providing the required next move
in the course of action (cf. Coulmas 1979). Some SBUs are very
specifically bound to one type of situation while others are poly-
functional in usage. Implicatures differ from SBUs: they rarely oc-
cur in institutionalized contexts. If they do, there is a specific rea-
son for that. Implicatures are usually not ready-made, prefabri-
cated expressions but generated ad hoc for a particular situation.
Some conventional implicatures, however, may be "routinized"
and behave like SBUs. But as we will see later, even if they are
"routinized" it is a type of conversation rather than a situational
frame that they are tied to.
ii. The proper application of SBUs and conversational implicatures
in speech is one of the best signs of idiomatic language use and
conceptual fluency (Kecskes 2000a, 2000b). NNSs' language pro-
duction is usually not idiomatic enough for several reasons: (1) In
principle every communicative function can be fulfilled in any
language, but this does not imply that this is done in a like manner
in different languages. (2) Types of social situations are not cross-
culturally invariant. (3) Without significant immersion in the tar-
get language culture NNSs do not have access to "conventional-
ized conceptualizations" of the L2 so they rely on the conceptual
base of their LI. In the cognitive approach semantic representa-
tions which constitute the semantic pole of a linguistic sign are
equated with "conventionalized conceptualizations" (Langacker
1988). Taylor argued that if the meanings of linguistic forms are
equated with conceptualizations, and these conceptualizations are
conventionalized in a language, then the conceptualizations are
made available to speakers of a language by the language system
that they have learned (Taylor 1993: 212). When acquiring a non-
primary language, students have to learn not only the forms of that
particular language, but also the conceptual structures associated
with those forms. It is not only grammatical competence but also
conceptual fluency that governs the proper use of language
(Danesi 1992, Kecskes 1999, 2000b, Kecskes and Papp 2000b).
252 Istvan Kecskes
3.1.1. Implicatures
Mey (1993: 99) argued that an implicature is something that is im-

plied in conversation, that is, something which is left implicit in ac-
tual language use. Grice (1968, 1975) made a distinction between
conversational implicatures and conventional implicatures. Conver-
sational implicatures are created ad hoc in the course of conversation
and have to be calculated by the interlocutor. They are generated by a
particular utterance for a particular receiver in a particular situation,
so what is conversationally implicated is somewhat indeterminate
and cancelable. The conventional implicature of an expression, on
the other hand, is quite specific, and usually means the same for
members of a speech community so its content is hardly cancelable.
Conventional implicatures may become routine formulas and are
used almost automatically. These expressions are routinely inter-
preted in a derived (i.e. non-literal) sense. For instance, "Please help
yourself' is often used by Americans at the table to urge their guests
to start to eat or take some more food. The original cognitive mecha-
nism responsible for the situational meaning of the expression is no
longer maintained consciously, and no inferential reasoning is neces-
sary to find out that the speaker asks you to "help yourself' not be-
cause he does not want to. The linguistic form has acquired a prag-
matically motivated sense which became conventionalized.
3.1.2. Situation-Bound Utterances (SBUs)
SBUs are highly conventionalized, prefabricated pragmatic units

whose occurrence is tied to standardized communicative situations
(Kecskes 1999, 2000a). If, according to their obligatoriness and pre-
dictability in social situations, formulaic expressions are placed on a
continuum where obligatoriness increases to the right, situation-
bound utterances will take the rightmost place because their use is
highly predetermined by the situation. The acquisition of these units
in an L2 requires the knowledge of the socio-cultural background of
the target language because SBUs are functional units whose mean-
ing can be explained only as functions of habitual usage (Kecskes

2000a). The pragmatic functions are usually not encoded in these
linguistic units, therefore SBUs often receive their "charge" from the
situation they are used in. It is generally this situational charge that
distinguishes SBUs from their freely generated counterparts. Many of
them can have both their figurative and the literal meaning as salient.
However, the most salient meaning is usually the figurative one.
3.2. Data collection
The participants were given a test which consisted of two parts. The
first part focused on implicatures while the second part tested the use
of SBUs.
3.2.1. Implicatures
The first task presented the students with ten dialogs that included
implicatures with their immediate context whose meaning the stu-
dents were expected to explain.
1/1. Two teachers are talking about a student's paper:

Mr. R: Have you finished with Mark's term paper yet?
Mr. M: Yeah, I read it last night.
Mr. R: What did you think of it?
Mr. M: Well, I thought it was well typed.
1/2. Amy: Don't you think Jim drinks a bit too much?
Billy: Is the Pope Catholic?
1/3. Pam: Do you need sugar in your coffee?

Betty: About as much as I needed that last piece of cheese cake.
VA. Wife: Have you seen my lipstick?

Husband: Your daughter was in your room this morning.
254 Istvan Kecskes
1/5. Ann: How do you like your teacher?

Bill: Well, he dresses nicely.
1/6. Alan: Do you think Mr. Herbert will give us a quiz today?
Bob: Does the sun rise in the East?
1/7. Mary: Do you know how my car got scratched?

Steve: I have a meeting to go to and I'm late.
1/8. Adam: How is Bill doing at his new job in the bank?
Jim: Oh, quite well, I think. He hasn 't been thrown in prison
yet.
1/9. Tim: Where's my ham sandwich?

Bill: Well, the dog looks happy.
1/10. Mark: Does Joe have a girlfriend these days?

Tom: He's been making a lot of visits to Billings lately.
3.2.2. Situation-Bound Utterances
In the second part students were expected to demonstrate their com-

petence in the use of situation-bound utterances (SBUs) in three dif-
ferent tests:
Test 1
In Test 1 students were asked to describe the meaning of situation-
bound utterances without context:
S/l. Take it easy.

S/2. Piece of cake.
S/3. Get out of here.
S/4. Why don't you sit down?
S/5. I'll talk to you later.
S/6. Welcome aboard.
S/7. Stick around.

S/8. Tim is going out with that blonde.
S/9. Please hear me out.
S/10.1 am sorry, I didn't get your name.
S/11. Hang on for a second.
S/12. You are welcome.
S/13. Help yourself.
S/14. Give me a break.
S/15. Give me a hand.
Test 2
In Test 2 students were given dialogs with SBUs most of which have
both a literal and a figurative meaning as salient. The participants
were expected to recognize the most salient meaning in the given
situation. Some expressions from the previous list (S/2, S/3, S/5, S/6)
were repeated here to find out how student responses change if the
expressions are used in a situational context.
S/16. - Bill, I don't think I can agree with you.

- OK, shoot.
S/17. - Frank, I think you really deserved that prize.

- Get out of here.
S/l 8. - Hurry up. We will miss the train.

- Come on, we have plenty of time.
S/l 9. - Jim, do you think you can repair this coffee machine?
- Piece of cake.
S/20. - John, please clean up this mess from the desk.

- Oh, give me a break, will you?
S/21. - David, let me introduce you to our new secretary, Sally.

- Nice to meet you, Sally. Welcome aboard.
256 Istvan Kecskes
S/22. - 1 must be going now. See you soon.

- OK, I'll talk to you later.
S/23. - 1 think John was really rude to you yesterday.

- You tell me about it.
S/24. - Thank you very much for your kindness.

- You bet.
Test 3
The last part of the test focused on SBUs in which the figurative
meaning is usually the most salient meaning. Some of these expres-
sions (S/25, S/29) are rarely used in their literal meaning, others
(S/26, S/27), however, quite frequently keep their original meaning.
In the dialogs below these formulae were given in their literal mean-
ing, and participants were expected to consider the literal meaning of
the expressions as intended meaning in the given situation.
S/25. - Bill, you must show me how to use this word processor.
- Sorry, I am busy. Help yourself on that.
S/26. - Jim, can I spend the night in your apartment? It's, kind
of, too late to go home now.
- You know jOH are always welcome.
S/27. - Mary, why don't you sit down? There is a chair over there.
- OK, OK, I prefer to stand.
S/28. - Andy, can you tell me why you left me in the store without a
word?
- Not now. Don't you see I am busy. I'll talk to you later.
S/29. - What do you want me to do?

- Stick around here until I get back.
The 'Graded Salience Hypothesis' 257
S/30. - I feel like going to town.

- I'm not going out with you now. I have a lot to do.
S/31. - Tom, get me out of this hole, will you?

- OK, give me a hand.
4. Hypothesis
It was hypothesized that salience is not only a linguistic but also a

socio-cultural phenomenon which is based on particular knowledge
structures that are language- and culture-specific and depends on
conceptual fluency in the target language. Not having those knowl-
edge structures in their Common Underlying Conceptual Base1 (Kec-
skes 1998; Kecskes and Papp 2000a) NNSs have difficulty identify-
ing salient meaning of an expression in the L2, and usually consider
the compositional and literal meaning of the utterance as salient. The
study attempted to answer the following questions:
Can NNSs apply the principle of salience in their L2?

What are the characteristics of conceptual fluency in the L2?
In what cases do NNSs rely on their LI salience?
How language-specific is the principle of salience?
5. Data analysis
5.1. Implicatures
Data demonstrated that in most cases students had more difficulty

interpreting conventional implicatures (1/1; 1/2:1/5; 1/6) than conver-
sational implicatures (1/4; 1/8; 1/9; 1/10). This difficulty derives from
the fact that conventional implicatures usually require more specific
background knowledge and are more typical for a particular culture
than conversational implicatures which are usually based on common
sense.
258 Istvan Kecskes
Table 1. Implicatures
IMPLICATURE CORRECT MISINTER. NO IDEA

Well, I thought it was well 35% 61% 4%
typed.
Is the Pope Catholic? 33% 14% 53%
About as much as I needed that 29% little much 21%
last piece of cheesecake. 25% 26%
Your daughter was in your room 86% 4% 10%
this morning.
Well, he dresses nicely. 41% well- good 10%
dressed teacher
25% 24%
Does the sun rise in the East? 61% 22% 17%
I have a meeting to go to and 63% I did it I didn't 11%
I'm late. 14% 12%
Oh, quite well, I think. He hasn't 88% 12% 0%
been thrown in prison yet.
Well, the dog looks happy. 96% 4% 0%
He's been making a lot of visits 80% 18% 2%
to Billings.
There is a special group of conventional implicatures that are espe-

cially problematic for NNSs: relevance implicatures. The uniqueness
of this group is in that it is usually not the expression itself that is
conventionalized but the function and/or the structure. When, for
instance, there is a "yes/no" question, or a statement and the obvious
answer from the perspective of the interlocutor is "yes", a relevance
implicature may be used in a question format as in 1/1 :
Amy: Don't you think Jim drinks a bit too much?

Billy: Is the Pope Catholic?
One and the same function ("Isn't this obvious for you?") can be
lexicalized in a lot of different ways. This kind of implicature is not
necessarily culture-specific but the way it is lexicalized can be, and
this may result in an absolute misinterpretation of the implicature in
L2 processing. The pragmatic unit "Is the Pope Catholic?" is a good
example of how a conventional implicature becomes a fixed expres-

sion.
Some of the explanations from respondents of different LI back-
ground and culture demonstrated very clearly that students relied on
their LI socio-cultural background when processing relevance and
misinterpreted the relevance implicature when its content was too
specific in the American-English culture was too specific. For in-
stance:
1/4. Russian student (2 years in US, female, age 23): "Catholics

drink too much."
1/5. Spanish student (6 months in US, female, age 27): "Billy thinks
that Catholic people drink too much."
1/6. Spanish student (8 months in US, male, age 22): "Billy con-
fused Catholic with alcoholic."
When the relevance implicature was expressed by an utterance whose

processing required less specific knowledge the results showed that
the connection was easier for NNSs to make. (See numbers in Table
1.) For instance in 1/2:
Alan: Do you think Mr. Herbert will give us a quiz today?

Bob: Does the sun rise in the East?
Another interesting example for the relevance implicature is 1/1, and

1/5. What is conventionalized here is not the expression but the func-
tion: "if you do not want to say anything negative about some-
thing/someone, talk about things that are positive about that person or
thing". The obvious irrelevance of response is expected to indicate
negative opinion to the interlocutor as in 1/1 and 1/5:
1/1. Mr. R: Have you finished with Mark's term paper yet?
Mr. M: Yeah, I read it last night.
Mr. R: What did you think of it?
Mr. M: Well, I thought it was well typed.
260 Istvan Kecskes
1/5. Ann: How do you like your teacher?

Bill: Well, he dresses nicely.
The numbers in Table 1. (Correct: 1/1= 35%; 1/5.= 41%) demonstrate

that students relying on the propositional meaning of the expressions
had difficulty interpreting the intended meaning in these situations. In
1/5 many students noticed only the positive propositional meaning of
the expression and concluded either that Bill had positive opinion
about the way the teacher dresses (25%), or that Bill thought that the
teacher was a good one (24%).
Conversational implicatures were easier for NNSs to process be-
cause the salient meaning of the utterance is the propositional mean-
ing that did not require any culture-specific interpretation. For in-
stance:
1/8. Adam: How is Bill doing at his new job in the bank?
Jim: Oh, quite well, I think. He hasn't been thrown in
prison yet.
1/9. Tim: Where's my ham sandwich?

Bill: Well, the dog looks happy.
However, when the conversational implicature contained some cul-

ture-specific reference such as, for instance, a connection between
cheesecake and proper diet, the full understanding of the proposi-
tional meaning of the utterance did not necessarily result in correct
interpretation of the expression. (See numbers in Table 1.) For in-
stance in 1/3:
Pam: Do you need sugar in your coffee?

Betty: About as much as I needed that last piece of cheesecake.
5.2. Situation-Bound Utterances
Testi
In the first test SBUs were listed without context. According to the
GSH for information to be salient, i.e. to be foremost on one's mind,
it needs to be stored or coded in the mental lexicon (Giora 1997,
forthcoming). Stored information enjoys cognitive priority over un-
stored information such as new information or information inferable
from context. Consequently, it is the most salient sense of a word or
expression that is directly computable from the mental lexicon before
any extra inferences based on contextual information are put to play .
This approach claims that context has a limited role in the activation
of salient meanings because salience is a matter of convention, fre-
quency, and familiarity. Research (Rayner, Pacht and Duffy 1994,
Giora 1997) suggested that even when prior context is heavily biased
in favor of the less salient (e.g. less frequent, familiar and conven-
tional) meaning of an ambiguous expression, salient meaning is acti-
vated first. So salient meanings cannot be bypassed in LI processing.
Familiar metaphors and frequent SBUs often have both a figura-
tive and a literal meaning as salient. They activate both their literal
and figurative meanings simultaneously if both meanings are equally
salient. However, one of the meanings is canceled based on contex-
tual clues. In the first test the expressions could have more than one
salient meaning. However, in most cases the figurative meaning was
expected to be considered salient.
Both the NS and NNS responses support GSH. The majority of re-
spondents found the most salient meaning without context. The num-
bers demonstrate how important frequency of encounters with the
expression is for both NSs and NNSs. Even the NSs made their deci-
sions based on frequency when two salient meanings were possible.
For instance: "Take it easy", "I'll talk to you later". Where parallel
processing would be needed because two meanings are equally sali-
ent (for instance: "Get out of here", "Give me a break"), NSs are di-
vided in their decisions since there are no contextual cues to rely on.
NNSs, however, prefer literal meaning to figurative meaning, which
makes sense, and supports our hypothesis about NNS language proc-
262 Istvan Kecskes
essing which usually prioritize literal meaning. The expression "Wel-

come aboard" is a good example to demonstrate how much NNSs
rely on their experience, and how important familiarity with the ex-
pression is, when there is no context. A significant number of NNSs
(22%) identified "Welcome aboard" with the expression they hear
from the captain on an airplane.
Table 2. SBUs without context
SBUs NSs NNSs

FIGURA- LITERAL OR FIGURA- LITERAL MIS-
TIVE F2 TIVE OR F2 INTERPR
Take it easy relax bye relax bye
83% 17% 75% 18% 7%
Piece of cake 100% 0% 90% 6% 4%
Get out of here kidding leave kidding leave
47% 53% 8% 82% 10%
Why don't you sit down let's talk
sit down. 83% 17% 92% 0% 8%
I'll talk to you bye else
later 70% 30% 75% 18% 7%
Welcome aboard soc. accept. ship soc. accept. airp.
90% 10% 61% 22% 17%
Stick around 100% 0% 73% 25% 2%
Tim is going out dating derog.
with that blonde 80% 20% 69%6% 25%
Please hear me listen don't interr.
out 80% 20% 80%0% 20%
I am sorry, I 96% 4% 76%10% 14%
didn't get your
name
Hang on for a 100% 0% 88% 4% 8%
second
You are welcome no problem guest no problem guest
83% 17% 75% 6% 19%
Help yourself eat do yourself
87% 13% 41% 29% 30%
Give me a break disagree stop kidding rest leave me
47% 53% 31% 27% 42%
Give me a hand 98% 2% 96% 2% 2%
In some cases there are clear differences between NSs and NNSs.
The expression "Give me a break" was interpreted by NSs as "I dis-
agree with you" (47%), or "Stop kidding" (53%). The NNSs, how-
ever, thought that the intended meaning was either "Let me rest"
(31%), or "Leave me alone" (27%). This also demonstrates that
NNSs are usually more comfortable with a meaning that is close to
the compositional meaning of the expression.
The fact that NSs and NNSs do not share a common socio-cultural
background also led to different interpretations of one and the same
expression. Quite a number of the NSs (20%) said that the expression
"Tim is going out with that blonde" was derogative. None of the
NNSs thought that way.
Test 2
In the second test context was biased for the figurative meaning of
SBUs no matter whether the figurative meaning or the literal mean-
ing of the expression was salient such as, for instance, in S/16 ("OK,
shoot"), S/17 ("Get out of here"), S/18 ("Come on").
Table 3. SBUs withfigurativemeaning
SBUs FIGURATIVE LITERAL OR MISINTERPRE-

F2 TATION
OK, shoot 55% 0% 45%
Get out of here shut up leave me
39% 24% 47%
Come on 76% 10% 14%
Piece of cake 100% 0% 0%
Oh, give me a don't bother want rest
break, will you? 61% 18% 21%
Welcome aboard 70% 5% 25%
I'll talk to you 84% 14% 2%
later
Tell me about it agree talk about it
49% 22% 29%
You bet 92% 0% 8%
264 Istvan Kecskes
There is absolutely no difference in how NSs processed the expres-

sions. When parallel processing applied because both the literal and
figurative meanings were equally salient, context played a decisive
role. However, this was not necessarily the case for NNSs. The table
below shows that their interpretation was based on frequency ("Piece
of cake", which is very frequently used in everyday conversation was
processed 100% correctly), and contextual cues not always helped
them to find the correct salient meaning if it was not the literal
meaning. For instance: S/16 ("OK, shoot"), and S/17 ("Get out of
here").
It is interesting to compare the responses in two situations in Test
1 and Test 2. Context changed the interpretation of these two expres-
sions significantly.
"Get out of here"

Test 1. S/3. "Kidding": 8% "Leave": 82% Misinterpr.: 10%
Test 2. S/17. "Shut up": 39% "Leave": 24% Misinterpr.: 27%
"Give me a break"
Test 1. S/14. "Rest": 31% "Leave me": 27% Misinterpr.: 42%
Test 2. S/20. "Don't bother me": 61% "Want rest": 18% Misinterpr.:
21%
Test 3
In the third test SBUs were used in their less salient meaning which
was the original literal meaning. Usually it is very difficult to find
situations where an SBU is used in its original literal meaning. This
is what Gibbs (1980) called "literal uses of highly conventional ex-
pressions." Responses of NSs demonstrated that the less salient
meaning was processed sequentially with no problem if the context
was clear. In a couple of cases, however, where the situation was
open-ended, and the context was unclear sequential processing did
not always occur in the responses of participants because some NSs
and many NNSs processed the most salient meaning directly. From
this respect the S/31 "Give me a hand" situation deserves attention.
S/31 - Tom, get me out of this hole, will you?

- OK, give me a hand.
The most salient meaning of "Give me a hand" is "Help me", that is

to say, help me, the speaker. 24% of the NS respondents and 45%
NNSs interpreted "Give me a hand" in that way: "Help me so that I
can help you". Most of the NSs (76%), however, felt that the situ-
ational meaning was "Extend a hand (and I will pull you out of the
hole)", which is not the most salient meaning of the expression.
In the case of the expression "I'll talk to you later" both the literal
and figurative meaning can be salient. A part of the respondents in-
terpreted the SBU literally: "When I have time I will explain to you
why I left you in the store" while the other part understood it as say-
ing "Don't bother me now".
Table 4. SBUs with literal meaning
NSs NNSs
SBUs Literal Figurative Literal Figurative Misinter.
Help yourself 100% - 100% - -
You are always wel- 100% - 96% - 4%

come
Why don't you sit 100% - 100% - -
down?
I'll talk to you later explain don't bother explain don't bother
57% 43% 57% 35% 8%
Stick around 100% 100%
I am not going out busy
with you now 100% 84% 10% 6%
OK, give me a hand extend a help extend a help
hand hand
76% 24% 41% 45% 14%
266 Istvan Kecskes
6. Conclusions
6.1. Salience is culture-specific
The study confirmed the validity of the hypothesis which claimed

that salience is language- and culture-specific and derives from cer-
tain knowledge structures that are essential parts of the conceptual
base of native speakers. The more frequent, familiar, conventional
and prototypical a certain piece of information in a particular linguis-
tic community is the more salient it is in that community. Salience is
a very important part of the socio-cultural heritage of NSs. It is
something that reveals a unique feature of the mind: Giora (forth-
coming) argues that rather than accessing the context-compatible
information first, our mind opts for the most accessible information.
When it fails to integrate with the context, we access the next avail-
able interpretation. It is this most accessible information that is cul-
ture-specific and makes language processing difficult for NNSs be-
cause they do not have direct access to the most accessible informa-
tion since it is not stored or coded in their LI-dominated mental lexi-
con. What is stored and encoded there is what is salient in the native
tongue and culture of language learners, and that usually does not
work in the target language.
6.2. Low level of conceptual fluency in L2
NNSs can hardly apply the principle of salience in the target lan-
guage. The lack or low level of conceptual fluency in the L2 forces
NNS to rely on linguistic signs rather than conceptualizations while
processing L2. The study demonstrated this very well. NNSs usually
had no difficulty identifying the compositional meaning of SBUs and
implicatures. Problems occurred when literal meaning was not the
most salient meaning. Then NNSs mapped target language expres-
sions on LI conceptualizations, which often resulted in misinterpre-
tation of expressions. Using the principle of salience, LI speakers
processed figurative meanings directly without falling back on literal
meanings. Adult L2 speakers, however, could hardly ignore the literal

meaning.
6.3. The priority of contextual cues in L2
The GSH claims that context affects comprehension after highly sali-
ent information has been accessed (Giora 1997, forthcoming). There
is empirical research that supports the hypothesis that available in-
formation is accessed initially in LI processing, regardless of con-
textual fit or speaker's intent (Keysar 1998; Keysar, Barr, Balin and
Paek 1998). This is not so in L2 processing where contextual cues
seem to have priority over salience for reasons discussed above. It is
usually the linguistic context that NNSs rely on, and this is a direct
consequence of the compositional interpretation of words and ex-
pressions in the target language.
6.4. Application of GSH in second language acquisition
The GSH was developed to explain LI processing. However, its ap-

plication in second language acquisition can also be very useful since
it may reveal and explain several important features of L2 language
use and processing, which can contribute to better understanding of
what goes on in the multilingual mind.
Note
1. Bi- and multilingual development results in the emergence of a Common Un-

derlying Conceptual Base that is responsible for the operation of each language
channel (Kecskes 1998, Kecskes and Papp 2000a).
268 Istvan Kecskes
References
Blasko, Dawn G. and Cynthia Connine

1993 Effects of familiarity and aptness on metaphor processing. Jour-
nal of Experimental Psychology: Learning, Memory and Cogni-
tion. 19 (2): 295-308.
Bouton, Larry
1994 Can NNS skill in interpreting implicatures in American English
be improved through explicit instruction? - A Pilot Study. In:
Lany Bouton and J. Kachru (eds.), Pragmatics and Language
Learning. Volume 5, 88-10. Urbana-Champaign: University of
Illinois Press.
Coulmas, Florian
1979 On the sociolinguistic relevance of routine formulae. Journal of
Pragmatics 3: 239-266.
Danesi, Marcel
1992 Metaphorical competence in second language acquisition and
second language teaching: The neglected dimension. In: James E.
Alatis (ed.), Georgetown University Round Table on Languages
and Linguistics, 489-500. Washington, DC: Georgetown Univer-
sity Press.
Gibbs, Ray W. Jr.
1980 Spilling the bean on understanding: Comprehension and appre-
ciation. Journal of Experimental Psychology : Learning, Memory
and Cognition 9 (3): 667-675.
1984 Literal meaning and psychological theory. Cognitive Science 8:
575-604.
Giora, Rachel
1997 Understanding figurative and literal meaning: The graded sali-
ence hypothesis. Cognitive Linguistics 8 (3): 183-206.
Forthcoming On Our Mind: Salience, Context, and Figurative Language. New
York: Oxford University Press.
Grice, H. Paul
1968 Studies in the Way of Words. Cambridge, M.A.: Harvard Univer-
sity Press.
1975 Logic and conversation. In: P. Cole and J. L. Morgan (eds.),
Syntax and Semantics. Vol. 3: Speech Acts, 41-58. New York:
Academic Press.
Kecskes, Istvan
1998 The state of LI knowledge in foreign language learners. WORD,
Journal of the International Linguistics Association, Volume 49,
No. 3: 321-340.
1999 Situation-Bound Utterances from an interlanguage perspective.

In: Jef Verschueren (ed.), Pragmatics in 1998: Selected Papers
from the 6th International Pragmatics Conference, Volume 2,
299-310. Antwerp: International Pragmatics Association.
2000a A cognitive-pragmatic approach to situation-bound utterances.
Journal of Pragmatics 32: 605-625.
2000b Conceptual fluency and situation-bound utterances. In: Mia Vic-
tori (ed.), Autonomy in L2 Language Learning. Links & Letters.
2000. Issue 7: 143-158. Barcelona: UAB.
Kecskes, Istvan and Tunde Papp
2000a Foreign Language and Mother Tongue. Hillsdale, NJ: Lawrence
Erlbaum.
2000b Metaphorical competence in trilingual language production. In: J.
Cenoz and U. Jessner (eds.), Acquisition of English as a Third
Language, 99-120. Clevedon: Multilingual Matters.
Keysar, Boaz
1998 Language users as problem solvers. Just what ambiguity problem
do they solve? In: Susan R. Fussell and Roger Kreuz (eds.), So-
cial and Cognitive Approaches to Interpersonal Communication,
175-200. Mahwah, NJ: Lawrence Erlbaum.
Keysar, Boaz, Dale Barr, Jennifer A. Balin and Timothy S. Paek
1998 Definite reference and mutual knowledge. Process models of
common ground in comprehension. Journal of Memory and Lan-
guage. 39: 1-20.
Langacker, Ronald
1988 A usage-based model. In: Brygida Rudzka-Ostyn (ed.), Topics in
Cognitive Linguistics, 127-161. Amsterdam: John Benjamins.
Mey, Jacob
1993 Pragmatics. Oxford: Blackwell.
Rayner, Keith, Jeremy M. Pacht and Susan Dufly
1994 Effect of prior encounter and global discourse bias on the proc-
essing of lexically ambiguous words: Evidence from eye fixation.
Journal of Memory and Language. 33: 527-544.
Taylor, John R.
1993 Some pedagogical implications of cognitive linguistics. In: Rich-
ard A. Geiger and Brygida Rudzka-Ostyn (eds.), Conceptualiza-
tions and Mental Processing in Language, 200-223. Berlin/New
York: Mouton de Gruyter.
Subject Index
acquisition 44, 131, 153f, 156f, 168, decontextualized language skill 59

177, 193-222, 229f, 234, 249f, delaying function 114, 124
267 direct processing 250
Actor-Action construction 173f, 178f, direct role marking system 151
185 discourse 5, 15, 42-44, 46f, 50, 58,
actuality 34f, 44,63,65f, 69f, 73f, 77, 63, 92, 97, 109-112, 116f, 119f,
80, 85, 88f, 93 f, 96-100 123-127, 139, 141, 143-146, 148,
adultocentrism 201,218 153-155, 157f, 182
analytical linguistics 174, 177, 182, - function 123f
188f discourse- or reference-related strate-
argument structure 131, 173, 178 gies 146
attenuation 64, 66, 69, 70, 73, 77, 80, distal viewpoint 42f, 46
86-89, 94-97, 99 Early New High German (ENHG)
background knowledge 233,357 144f, 147
canonical orientation 231-233,243 English tense 63, 65-67, 69f, 91, 9 6 -
canonical rules 231, 233,245 98
categorization 209,229 experiential correlation 68, 70, 78-
child language 44,193-222 84, 87, 89, 98f
cognitive grammar 3 - 6 , 9 , 1 3 I f , 158 explication 194, 201, 205-207, 209f,
Common Underlying Conceptual 212f, 215,219,221f
Base 257,267 family resemblance 109,112,125
conceptual fictive 15, 17, 19-21, 30, 36
-fluency 251,257,266 fragments 11 If, 124-126
- prime 193-222 grammaticalization 124
- substrate 15-18,21 ground 5, 16, 22, 24, 30, 41, 46, 98,
- vocabulary 201, 202, 209, 216- 178, 231
218 habitual(s) 33f, 43,49f, 56f, 59,253
conceptualization 8, 10, 15, 42, 58, //¿^-situation 236f, 239-243
81,251,266 idiom 171,250f
construal 6-8, lOf, 13f, 19, 29, 37, image concepts 68,70, 75-78, 88
69f, 186 immediate scope 9f, 12, 14, 22-24,
construction 4f, l l f , 15, 21, 33f, 42f, 26, 69
50, 93f, 97, 109, 111-113, 117, imperfective 1 If, 14,23-26,37,43f
122-127, 132f, 135-141, 144- implicature(s) 19, 69, 94, 250-253,
159, 168-170, 173f, 176, 178- 257-260,266
180, 183-189, 194, 198,213, 236, indirect role marking system 151
240, 245 information 16, 18f, 42, 52, 55, 69,
mental-21,33f 72, 74, 76-78, 81, 89, 91-94, 97f,
corpus 109, 111, 115,126f, 197 lOOf, 120, 125, 155-157, 168f,
272 Subject index
174, 177-182, 184f, 188-190, non-îinguistic strategies 230-233,

215, 230, 233, 237, 242f, 245, 243f
261, 266f Old Scandinavian 146,151
- n e w 120,261 one-word stage 198,206,209-217
- o l d 155 parallel processing 189f, 249f, 261,
intimacy 63-66, 69-71, 74, 77, 80, 264
83f, 88-91,96f, 99 participant role 178,190
Korean 57, 131, 135-145, 147-150, passives 134-138,140f
153-157,229f perfective(s) 10-14, 21, 23-29, 33,
language 42-44,47, 136f
- acquisition 153f, 193, 218, 220, performative 18,26-28
229f, 249f, 267 perspectival shift 44,58
- learnability 131 pragmatic strengthening 65, 69, 80,
language teaching 5, 63, 66, 97, 99, 89, 9If, 98f
187f pragmatically determined word order
lexeme 81, 93, 167-172, 174f, 177- 131, 142, 145, 148,151, 153, 157,
180, 182-189 159
mutable - 172,183,185f, 188 present tense 11, 21-30, 32-37, 45,
lexical knowledge 168, 177, 233, 47-50, 52-54, 56, 58, 67, 69-72,
244f 89f, 92-94, 97,100
linguistic variation 13If, 153 presupposed constituent 183
literacy 44,46, 59,126 presupposition 112, 124,183
- events 46, 59 preverbal concepts) 230
literal language 249 primaryhood 144, 146
locative preposition 229 primates 221
meaning extension 65, 69, 98 prime(s)
mental construction 21, 33f semantic - 194f, 197f, 202, 209,
metaphor 15, 19, 68, 74, 123, 133, 216,218,222
250, 261 conceptual - 193-222
multilingual development 257 principle of salience 249,257,266
mutable lexeme 172,183,185f, 188 processing
naming insight 206 direct - 250
narrative voice 48, 50, 53 parallel - 189f, 249f, 261,264
narrativity 42 production vocabulary 198-201, 209,
natural semantic metalanguage 216f, 219
(NSM) 193f, 196, 200f, 203, 206, profile 9-12, 14, 16, 20, 22-24, 26f,
216,218-220 32-35
network 4, 65, 150, 171, 179-181, progressive 5, 11-14, 22, 24-26, 42-
184, 189f 44,46f, 50-54, 56, 58f
neurocognitive 167, 169-171, 176, - present 44, 50f, 53f, 56, 58f
182, 184,188-190 proto-linguistic stage 204-206
Subject Index ITS
prototype 109, 123, 125-127, 129, speech event 18, 22-27, 32, 36, 43,
13 If, 134, 137, 148, 150-152 68f
proximal-distal spatial relations 65, spoken discourse 44, 110-112, 121,
78, 80, 82f, 86, 88f, 95, 99 125
pseudocleft 109-117,119f, 122-125 subject and topic prominence 139,
-piece 112-116,119-121,123f 142,152f
relational character of objects 232f, syntactic category 167-70, 173-177,
236-238,242-244 182, 189
response concepts 68-70, 75-80, 83, syntax 46, 58, 109, 142, 149, 167-
88f, 99 171, 174, 177f, 185-187, 189,
role marking system 194, 220f
-direct 151 temporality 57
-indirect 151 tense 3, 1 If, 16, 21-30, 32-37, 41-
salience 63-66, 69f, 72, 74, 77, 80, 43,45-59, 63-74, 80-83, 88-100,
84, 88f, 91-93, 96f, 99, 249, 257, 114, 170, 198
261,266f English - 63, 65-457, 69f, 91, 96-
principle o f - 249,257,266 98
salient meaning 249f, 253, 255-257, present - 11, 21-30, 32-37, 45,
260f, 264-266 47-50, 52-54, 56, 58, 67, 69-72,
second language acquisition 153f, 89f, 92-94, 97, 100
249f, 267 time-reference 63-66, 69, 71, 77f,
segmentation 184,189 80-82, 88-92, 94-96, 99
semantic transitivity parameter 131, 141, 152,
- prime(s) 194f, 197f, 202, 209, 157
216,218,222 transitivity parameter and prominence
-transparency 131,141,157 typology 131,152
sequential processing 250,264 two-word stage 198,212,217,222
situation-bound utterances (SBU) typicality of the situation 243f
250-256,261-266 understanding of instructions 244
socio-cultural background 252, 259, universal relation 229f
263 usage-based model 3, 5
spatial variable constituent 172-175, 183,
- configuration 13,229 185-187, 189f
- relation 65, 229f, 232, 234, 236, viewing arrangement 15-17, 19, 21,
243f 26f, 29f, 32-37,41-46, 69f, 93
Edited by René Dirven, Ronald W. Langacker and
John R. Taylor
Mouton de Gruyter · Berlin · New York
This series offers a forum for the presentation of research within the per-
spective of "cognitive linguistics". This rubric subsumes a variety of con-
cerns and broadly compatible theoretical approaches that have a common
basic outlook: that language is an integral facet of cognition which reflects
the interaction of social, cultural, psychological, communicative and func-
tional considerations, and which can only be understood in the context of
a realistic view of acquisition, cognitive development and mental process-
ing. Cognitive linguistics thus eschews the imposition of artificial bound-
aries, both internal and external. Internally, it seeks a unified account of
language structure that avoids such problematic dichotomies as lexicon vs.
grammar, morphology vs. syntax, semantics vs. pragmatics, and synchrony
vs. diachrony. Externally, it seeks insofar as possible to explicate language
structure in terms of the other facets of cognition on which it draws, as well
as the communicative function it serves. Linguistic analysis can therefore
profit from the insights of neighboring and overlapping disciplines such as
sociology, cultural anthropology, neuroscience, philosophy, psychology, and
cognitive science.
1 Ronald W. Langacker, Concept, Image, and Symbol. The Cognitive

Basis of Grammar. 1990.
2 Paul D. Deane, Grammar in Mind and Brain. Explorations in Cognitive
Syntax. 1992.
3 Conceptualizations and Mental Processing in Language. Edited by
Richard A. Geiger and Brygida Rudzka-Ostyn. 1993.
4 Laura A. Janda, A Geography of Case Semantics. The Czech Dative
and the Russian Instrumental. 1993.
5 Dirk Geeraerts, Stefan Grondelaers and Peter Bakema, The Structure of
Lexical Variation. Meaning, Naming, and Context. 1994.
6 Cognitive Linguistics in the Redwoods. The Expansion of a New Para-
digm in Linguistics. Edited by Eugene H. Casad. 1996.
7 John Newman, Give. A Cognitive Linguistic Study. 1996.
Edited by René Dirven, Ronald W. Langacker and
John R. Taylor
Mouton de Gruyter · Berlin · New York
8 The Construal of Space in Language and Thought. Edited by Martin

Pütz and René Dirven. 1996.
9 Ewa D^browska, Cognitive Semantics and the Polish Dative. 1997.
10 Speaking of Emotions: Conceptualisation and Expression. Edited by
Angeliki Athanasiadou and Elzbieta Tabakowska. 1998.
11 Michel Achard, Representation of Cognitive Structures. 1998.
12 Issues in Cognitive Linguistics. 1993 Proceedings of the International
Cognitive Linguistics Conference. Edited by Leon de Stadler and
Christoph Eyrich. 1999.
13 Historical Semantics and Cognition. Edited by Andreas Blank and
Peter Koch. 1999.
14 Ronald W. Langacker, Grammar and Conceptualization. 1999.
15 Cognitive Linguistics: Foundations, Scope, and Methodology. Edited
by Theo Janssen and Gisela Redeker. 1999.
16 A Cognitive Approach to the Verb. Morphological and Constructional
Perspectives. Edited by Hanne Gram Simonsen and Rolf Theil Endre-
sen. 2001.
17 Emotions in Crosslinguistic Perspective. Edited by Jean Harkins and
Anna Wierzbicka. 2001.
18 Cognitive Linguistics and Non-Indo-European Languages. Edited by
Eugene Casad and Gary B. Palmer. Forthcoming.

(Cognitive Linguistics Research, 19.1) Susanne Niemeier, Martin Pütz - Theory and Language Acquisition-De Gruyter Mouton (2001)

Uploaded by

Copyright:

Available Formats

You might also like

(Cognitive Linguistics Research, 19.1) Susanne Niemeier, Martin Pütz - Theory and Language Acquisition-De Gruyter Mouton (2001)

Uploaded by

Document Information

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

(Cognitive Linguistics Research, 19.1) Susanne Niemeier, Martin Pütz - Theory and Language Acquisition-De Gruyter Mouton (2001)

Uploaded by

Copyright:

Available Formats

Applied Cognitive Linguistics I: Theory and Language Acquisition

© Printed on acid-free paper

Library of Congress Cataloging-in-Publication Data

Applied cognitive linguistics / edited by Martin Pütz, Susanne

Die Deutsche Bibliothek — Cataloging-in-Publication Data

Applied cognitive linguistics / ed. by Martin Pütz .... - Berlin ;

Most of the present contributions were originally presented at the 29th

Applied Cognitive Linguistics. Vol. I. Theory and Language Acquisi-

Applied Cognitive Linguistics. Vol. II. Language Pedagogy

While compiling the two volumes we have incurred a number of

We would like to thank the organizing staff of the symposium, in

University of California, Santa Barbara, U.S. A.

Georgetown University, Washington, D.C., U.S.A.

University of New England, Annidale, Australia

Carnegie Mellon University, Pittsburgh, U.S.A.

State University of New York at Albany, U.S.A.

University of California, Santa Barbara, U.S.A.

Sydney M. Lamb Houston, Tx., U.S.A.

List of Contributors vii

Section 1: Cognitive approaches to the English tense system

Cognitive linguistics, language pedagogy, and the

Pretend play: trial ground for the simple present 41

The relation between experience, conceptual structure

Section 2: Facets of prototypes in grammatical constructions

Grammatical constructions and their discourse origins:

Transitivity parameter and prominence typology:

Section 3: Neurocognitive and cognitive issues of

Learning syntax - a neurocognitive approach

Conceptual primes in early language development

No preposition required. The role of prepositions for the

The 'Graded Salience Hypothesis' in second language

List of Contributors vii

Section 1: Bottom-up approaches: Phrasal verbs and

English phrasal verbs: theory and didactic application 3

Teaching English phrasal verbs: a cognitive approach 29

A usage-based approach to modeling and teaching

Section 2: Top-down approaches:

A cognitive linguistic view of learning idioms in

On the systematic contrastive analysis of conceptual metaphors:

Section 3: Systematical order instead of chaos

A conceptual analysis of English -er nomináis 149

Basicness and conceptual hierarchies in foreign language

Section 4: Cultural models in education

The African cultural model of community in English language

Subject Index 261

Martin Pütz, René Dirven and Susanne Niemeier

1. As one domain of human cognition, language is intimately linked

5. Given the interaction among language subcomponents as well as

As a dynamic usage-based model (Langacker 2000: 9Iff.), cognitive

views. Although Langacker does not draw specific conclusions about

structure and meaning: Non temporal uses of tense and language

to explain to the learner why the foreign language should be as it is

foreign language teaching. In general, teachers should pay attention

distinction is based on the admittedly disputable claim that analytical

tive prepositions, but also on non-linguistic strategies, such as an

Niemeier, Susanne and Michel Achard (eds.)

Cognitive approaches to the English

Although I cannot claim any experience or substantial knowledge in

2. Implications of a usage-based model