\

Neuropsycholo`ia\ Vol[ 25 No[ 01\ pp[ 0202Ð0212\ 0887

Þ 0887 Elsevier Science Ltd[ All rights reserved
Pergamon PII] S9917Ð2821"87#99918Ð2 Printed in Great Britain
9917Ð2821:87 ,08[99¦9[99

Going\ going\ gone [ [ [ < Implicit and explicit tests

of conceptual knowledge in a longitudinal study of
semantic dementia
LORRAINE K[ TYLER% and HELEN E[ MOSS$

 Department of Psychology\ Birkbeck College and $ Department of Psychology\ University of Cambridge

"Received 09 October 0886^ accepted 01 March 0887#

Abstract*Patients su}ering from semantic dementia provide important constraints on theories of the structure and organisation of
semantic memory[ In this article we report one such patient\ AM\ whose progressive deterioration of semantics enables us to address
the much!debated issue of whether conceptual structure is hierarchically organised[ The hierarchical account predicts that brain
damage should impair lower levels of the hierarchy "property information# before a}ecting higher level "category# information
"Warrington and Shallice\ Q[ J[ Exp[ Psychol[ 0864\ 16\ 524Ð546#[ We evaluate this prediction by repeated testing of AM in two
studies*a semantic priming task and a veri_cation task*over an 07 month period\ contrasting the progressive deterioration of
properties "functional and perceptual# and category relations "category co!ordinates and category labels#[ Properties were preserved
longer than category information\ arguing against a hierarchical account of semantic memory[ In addition\ functional properties
were most robust to brain damage\ supporting our claim that functional information plays a special role in semantic representations
"Durrant!Peat_eld et al[\ Proc[ 08th Ann[ Conf[ of the Co`nitive Science Society[ Erlbaum\ Mahwah\ NJ\ 0886\ pp[ 082Ð087[ Tyler et
al[\ Co`nitive Neuropsychol[ 0886\ 03\ 400Ð434#[ Þ 0887 Elsevier Science Ltd[ All rights reserved[

Key Words] semantic memory^ priming^ disorders[

Introduction e}ortlessly and with reasonable grammatical structure\

Patients su}ering from the progressive breakdown of lan!
guage a}ord a unique opportunity for studying the
organisation of the language system[ The ways in which
di}erent aspects of language break down over time and
the associations and dissociations of subcomponents of
the system provide important data about its structure[
Similarly\ within a single sub!component of the language
system\ progressive deterioration can reveal its internal
structure and organisation[ These claims have been amply
supported by the disorder known as {{semantic
dementia||\ in which focal atrophy of temporal neo!
cortical brain regions produces a progressive deterio!
ration of semantic memory ð6\ 24Ł[ The most prominent
aspect of the language disorder in semantic dementia is
degraded word meaning and consequent loss of vocabu!
lary\ both expressive and receptive[ Other aspects of lan!
guage "phonology and syntax# tend to be relatively spared
for a period of time[ Because the patients tend to speak
 Corresponding author[ Centre for Speech and Language\
Department of Psychology\ Birkbeck College\ London WC0E
7HX\ U[K[ E!mail] l[tylerÝpsychology[bbk[ac[uk
though with impoverished content word vocabulary\ this
pattern is also sometimes labelled {{progressive ~uent
aphasia||[ There is debate about whether\ in some cases\
the semantic deterioration a}ects only word meaning^
but the majority of well!documented cases have revealed
semantic de_cits on tasks requiring non!verbal knowl!
edge about people and objects as well as the meanings of
words[ For this reason\ the non!language!speci_c term
semantic dementia is generally preferred[
There is evidence that this disorder is not a generalised
dementia but rather selectively a}ects semantic memory[
These patients typically have well retained abilities in the
domains of perception and problem solving\ as long as
the tasks do not require speci_c knowledge of word! or
object!meaning[ Performance on short!term or working
memory tasks\ such as digit span and visuoÐspatial span\
is typically within normal limits[ There is also relative
preservation of autobiographical and episodic memory\
although recent _ndings suggest that this phenomenon
may be time dependent\ with better memory for recent
events than for more remote autobiographical facts and
episodes ð5\ 25Ł[
Detailed cognitive studies of patients with semantic

0202
0203 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [
dementia began in the mid:late 0869|s ð21\ 33Ł\ declined
to some extent during the 0879|s and have come back to
prominence in the last few years ð6\ 09\ 29\ 24\ 25Ł[ Pat!
terns of performance in these single!case investigations
are relevant to many major theoretical issues regarding
semantic memory[ We will focus in this article on one
of these*how the progressive deterioration of di}erent
types of semantic information over time bears on the
claim that conceptual structure is hierarchically organ!
ised ð0\ 33Ł\ from superordinate "a fox is a livin` thin`#\
to basic!level category "a fox is an animal\ similar to
wolves and do`s#\ to speci_c features "a fox is red or sly#[
A number of studies using tasks such as sorting or probe
questions have argued that degraded semantic knowledge
tracks this hierarchy from the bottom up\ with speci_c
features most vulnerable and superordinate information
best preserved ð6\ 33Ł[ This pattern does not require the
view that such hierarchies are explicitly represented] a
network of semantic information in which features essen!
tial for distinguishing between conceptually similar con!
cepts have been degraded should also yield preserved
{{higher!order|| knowledge ð7\ 18Ł[ In this article we test
the hierarchical structure claim by means of a longi!
tudinal comparison of the relative preservation:
deterioration of general category knowledge of an object
concept "e[g[ an elephant is an animal#\ with knowledge
of its speci_c attributes\ including sensory:perceptual
"e[g[ an elephant is `rey and has thick leathery skin# and
functional:associative features\ i[e[ how entities function
and are used in the world "an elephant lives in Africa or
India and is used for transport#[ Comparing patterns of
relative deterioration and preservation over time provides
a strong data!base for evaluating the claim for hier!
archical organisation[
We include the comparison between functional and
perceptual properties because this contrast has been an
important factor in accounting for the representational
di}erences between living and non!living things\ with
some investigators arguing that conceptual knowledge
about living things "e[g[ animals\ fruit# is strongly weigh!
ted in terms of perceptual features\ while functional attri!
butes have equal or even greater weighting\ than
perceptual properties in knowledge about non!living
things "e[g[ furniture and articles of clothing# ð1\ 4\ 04\
34Ł[ Taking a rather di}erent view\ we have recently
argued that visual and functional properties are not
di}erentially important for living and non!living things\
but that what determines category structure is the inter!
correlations between di}erent kinds of properties[ In
addition\ we claim that functional properties are
especially salient and robust in the semantic rep!
resentations of both living and non!living things ð1\ 39\
30Ł\ with di}erent kinds of functional information being
relevant for each[ In two priming studies with normal
subjects we showed that functional properties prime as
robustly as other types of semantic information ð30Ł and
are activated rapidly ð11Ł[ We have also reported data
from brain!damaged patients supporting the claim that
functional information is relatively spared following
brain damage ð30Ł[ The priming and the veri_cation tasks
were therefore designed to compare these two types of
features\ with the prediction that functional properties
should be relatively less vulnerable to the degenerative
process[
The additional interest of our approach lies in the fact
that we use a combination of explicit and implicit tasks[
One way of characterising various cognitive operations
distinguishes between those requiring explicit or delib!
erate processing and those relying on implicit or auto!
matic processing[ Tasks which measure explicit
processing include naming objects\ de_ning words\ sort!
ing pictures or words into conceptual categories\ mat!
ching pictures to their verbal labels\ etc[ Implicit tasks in
the language domain rely on the rapid and automatic
processes by which we comprehend both individual
words and the semantic and syntactic relations between
words\ whether or not explicitly requested to do so ð08\
15\ 17\ 26\ 28\ 31Ł[ These implicit tasks can reveal theor!
etically important features of preserved aspects of pro!
cessing in patients with apparently profound de_cits[
For example\ a patient may fail to name a picture of a
broom or to provide any useful information when asked
to describe a broom and yet show a normal degree of
facilitation when the word {{broom||^ precedes a sem!
antically related word such as {{sweep|| or {{~oor|| to
which the patient is asked to make a lexical decision or
word monitoring response ð15Ł[ Such priming could not
occur unless there were substantial preservation of the
meaning of {{broom||[ Tasks like lexical decision or word
monitoring are considered implicit because\ although
performance requires some explicit understanding of the
task demands\ the subject is not asked explicitly to com!
prehend or to make a deliberate response to the relation!
ship between the primes and targets[ We try to
demonstrate in this article that patterns of impai!
red:preserved comprehension are most successfully
illuminated by a combination of implicit and explicit
approaches[
The patient studied here\ AM\ has already been the
focus of several publications\ designed to assess the
relationship of his semantic de_cit to his cognitive abili!
ties in other domains\ including episodic and auto!
biographical memory ð5Ł and auditoryÐverbal short!term
memory ð02Ł[
Case description
AM is a 53 year old\ right!handed man who had
worked as a senior chemical engineer until his retirement
shortly before this study began[ He was _rst assessed by
Karalyn Patterson and John Hodges in April 0883[ Since
details of performance at this and subsequent assessments
are given in Graham and Hodges "0886#\ Hodges and
Patterson "0886# and Knott et al[ "0886#\ we will only
present a brief summary here[ In April 0883 AM already
 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [
Table 0[ AM] non!semantic and semantic tests at each testing period
Non!semantic
Ravens coloured progressive matrices ð25Ł
Rey copy ð25Ł
Object match ð39Ł
Digit span:forward
Digit span:backwards
Semantic
Confrontation naming ð37Ł
Category ~uency:living
Category ~uency:non!living
had a 4¦ year history of gradually increasing di.culty
in remembering the names of familiar people and objects[
At presentation\ he was fully orientated in time and place
and could give detailed "though anomic# descriptions of
recent events in his life[ His spontaneous speech was
~uent and grammatical with occasional word!_nding
pauses and circumlocutions "especially anomic phrases
like {{one of those things||#\ though at this time he was
still able to convey a considerable amount of information[
He made no phonemic paraphasic errors and only
occasional semantic substitutions[ His wife reported
notable di.culty with comprehension of words that had\
premorbidly\ been common terms to him] for example\
previously a keen gardener\ he could no longer under!
stand the names of ~owers[ Despite these problems\ he
continued to lead a reasonably full life of social and
sporting activities] he played golf and tennis with pro!
_ciency and was still driving his car[
MRI revealed marked focal atrophy of the infer!
olateral temporal lobes\ bilaterally but with especially
striking atrophy on the left side[ There was relative spar!
ing of the hippocampal complex[ A SPECT scan pro!
duced a similar picture\ with severely reduced perfusion
in the anterior and middle portions of the left temporal
lobe and signi_cant though less marked reduction in the
region of the right temporal pole[ Examples of AM|s
MRI at presentation were published in Hodges and Pat!
terson ð8Ł\ which also gives further details of his clinical
picture[
Table 0 presents a pro_le of AM|s performance on
some basic tests over the period of our longitudinal study[
These data were kindly provided by John Hodges and
Karalyn Patterson[ Not all of these tests were given at
every testing stage\ partly because of limited testing time
but also because AM showed the classic pattern of pre!
served and degraded performance reported by the
Hodges and Patterson group for other patients with sem!
antic dementia ð6\ 7\ 09Ł[
 We are grateful to John Hodges and Karalyn Patterson for
providing these data[
0204
April 83 April 84 October 84
* * 29
25 25 24
27 28 27
6 7 4
5 4 4
2 0 5
01 1 *
5 4 9
Preserved[ The tests used to assess AM|s general cog!
nitive abilities included the Ravens Coloured Progressive
Matrices and copying the Rey _gure\ on both of which
he scored well[ His visuoÐperceptual skills were tested
using the object matching test "or {{unusual views||\ ð01Ł#
revealing AM|s excellent perceptual skills[ AuditoryÐver!
bal short!term memory\ as measured by digit span\ was
entirely normal for the _rst year\ but had declined slightly
by October 0884[
De`raded[ By contrast with the _rst set of tests in Table
0\ the remaining tests show signi_cantly impaired and
declining scores[ His performance on a confrontation
naming test\ using the materials from the Hodges and
Patterson ð7Ł semantic test battery\ were severely impaired
in April 0883 and remained so throughout the testing
period[ Similarly\ his ability to produce exemplars of
semantic categories in a category ~uency test were imp!
aired at initial testing and deteriorated further over time[
Each of the living and non!living scores re~ects the com!
bined number of correct exemplars\ given in 0 min
periods\ over four categories at di}erent levels of speci!
_city "for living] animals\ birds\ sea creatures\ breeds of
dog^ for manmade] household objects\ vehicles\ musical
instruments and types of boat#[ A group of 13 normal
controls of AM|s age produced an average of 47 correct
exemplars for the combined set of four living categories
and 44 for non!living categories\ with the lower limit of
the respective normal ranges being 24 and 28 ð7Ł^ AM|s
scores were 01 and 5[
Experimental investigations
Our aim was to evaluate AM|s progressive deterio!
ration of semantic knowledge as measured by means of
both implicit and explicit tasks using the same materials[
The explicit task we chose was veri_cation judgements\
where subjects have to say whether a short statement is
true or false[ The implicit task we used was semantic
priming of lexical decision "LD#\ in which the subject
hears prime!target pairs and makes a lexical decision
0205 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [
to the target[ If the prime word activates its semantic
representation\ then LD latencies should be faster when
the target follows a semantically related prime relative to
a preceding unrelated word[ We used this task to vary
the relationship between the prime word and the target\
in this way probing for di}erent aspects of semantic rep!
resentation[
In the priming and veri_cation studies\ we contrasted
category labels " fox!animal# and category co!ordinates
" fox!horse# with speci_c featural knowledge " fox!red^
fox!sly#[ The two broad classes of features we compared
were] "a# perceptual "or visual#\ i[e[ shape\ colour\ surface
properties\ "e[g[ fox!red#^ "b# functional:associative "e[g[
fox!sly#[ Comparing priming for speci_c featural proper!
ties with that for category labels and category co!ordi!
nates enables us to evaluate the claim that semantic
de_cits impair knowledge of speci_c semantic properties
of a concept before a}ecting more general information
about the object|s category ð33\ 22Ł[ This pattern of pre!
served {{higher!order|| knowledge with loss of speci_c
features predicts that if there is some deterioration of
semantic information for AM\ both category and co!
ordinate relations should prime more robustly than prop!
erty relations and they should maintain their ability to
prime over time even if the detailed features lose their
power to prime[
Semantic priming study
We selected 019 potential prime words "41 living and
57 non!living# from the Snodgrass and Vanderwart ð23Ł
set and entered them into a property generation pre!test
to _nd potential perceptual and functional properties for
each word[ The items were divided into two sets of 59
items and each set was presented to 34 normal controls
"members of the Centre for Speech and Language subject
pool# who were asked to list the properties that {{most
people would generally attribute to the object or thing||[
From this pre!test\ we calculated the percentage of sub!
jects who listed a given property for each stimulus word\
which is standardly taken as a re~ection of the centrality
or salience of that property e[g[ ð16Ł For the priming
study\ we selected properties with high production fre!
quencies "i[e[ highly salient# and then matched the pro!
duction frequency across the four experimental
conditions created by crossing category of object "living
 The mean familiarity of the words denoting living and non!
living things prime words was 3[8 and 40[2\ respectively[
$ We kept the relatedness proportion as low as was practical
without making the experiment too long for the patient to
endure[ A low proportion of related trials reduces the oppor!
tunity for strategic\ rather than automatic priming e}ects[ As a
check on the potential impact of this variable on the magnitude
and pattern of priming\ we compared a 09) and 22) related!
ness proportion for the same set of materials for a group of
young normal subjects and found no di}erence in the pattern
of priming e}ects[
vs non!living# with type of feature "perceptual vs func!
tional^ see Table 1#[ We selected properties which were
expressable in one word\ which could be used as a target
for lexical decision[ In addition\ we chose prime!target
pairs which were not strong associates[ On some
accounts\ associative priming is based on co!occurrence
links between words at the level of lexical form rep!
resentation and does not re~ect semantic relatedness ð10Ł[
Since we were interested in semantic and not associative
priming\ all the prime!target pairs had an association
strength "Postman and Keppel\ 0869^ ð12Ł# of less than
04) in forward and backwards directions[
In addition to the two conditions in which the object
name served as a semantic prime for its perceptual and
its functional feature word "e[g[ fox!red\ fox!sly#\ two
further conditions used the same object name as a related
prime for its category label "e[g[ fox!animal\ house!build!
in`# and for a category co!ordinate " fox!do`\ house!cot!
ta`e# "Table 1#[ The 13 living things and 13 non!living
things used as prime words were matched for familiarity
and the target words for lexical decision were matched as
closely as possible for both familarity and length "see
Table 1#[ According to pre!tests in which we asked nor!
mal subjects to rate the semantic relatedness of prime!
target pairs "on a scale of 0Ð8\ where 8 is highly related]
"see ð13Ł for additional details of this pre!test#\ the degree
of semantic relatedness was high and similar across the
living and non!living things items[ Table 1 lists mean
values on this relatedness scale as well as values for the
other measures on which stimulus terms were matched[
Each prime word was paired with each of its four
targets "e[g[ crocodile] animal\ alli`ator\ `reen\ river#[ LD
response times in these semantically related conditions
were compared to responses to the same targets when
they were preceded by an unrelated control word[ Con!
trol words were matched to prime words on familiarity
and length "e[g[ for crocodile\ mechanic#[ For each prime!
target type\ the related and unrelated pairs were rotated
over two di}erent versions of the materials\ producing
an eight version experiment[ To reduce the number of
testing sessions per subject "and for AM\ per time slice# to
a more manageable four\ we combined half the perceptual
properties with half the category co!ordinates to form
two versions and combined half the category label and
half the functional properties to form two versions[ Each
related and unrelated prime only occurred once per
version[
In addition to the 37 related and 37 unrelated test word
pairs per version\ another 37 unrelated _ller pairs were
inserted to reduce the proportion of semantically related
real words to 22)\$ or 05[4) of the total set including
the non!word LD trials[ To ensure an equal proportion
of words and non!words "yes and no responses#\ we added
033 word:non!word pairs "e[g[ mornin`!hiction\ tulip!
blane#[ All subjects were instructed to respond only to the
second item of each wordÐword or wordÐnon!word pair[
The control subjects made a yes:no button press to each
target[ For AM\ we used a simpli_ed {{go:no!go|| version
 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [ 0206

Table 1[ Semantic priming study] characteristics of the stimuli

Category Category Visual Functional

label coordinate property property
fox!animal fox!do` fox!red fox!sly

Living "n  13#

0
) production frequency:target * * 34 29
1
)association strength "prime!target# 3[4 9 3 2
2
semantic relatedness "prime!target# * 4[84 3[70 4[17
3
familiarity "target# 4[86 3[78 4[44 4[23
4
frequency "target# 78 03 88 49

Non!living "n  13# desk!furniture desk!cabinet desk!wood desk!work

) production frequency:target * * 28 27
)association strength "prime target# 1 0 1 1
semantic relatedness "prime!target# * 5[98 3[84 4[50
familiarity "target# 4[56 4[04 4[34 4[38
frequency "target# 72 37 77 096
0
Pre!tests conducted in our laboratory[
1
Moss and Older "0885# Association norms[
2
Pre!tests conducted in our laboratory[
3
A mixture of the MRC Psycholinguistic data!base and pre!tests in our laboratory[
4
Ho~and and Johanssen "0871# ð00Ł[

of the task where he pressed only the {{yes|| button when related RT#Ł×099#[ These values were as follows] cat!
he considered the target to be a real word[ We tested AM e`ory label 4) "r  0Ð00)#\ cate`ory co!ordinates 8)
on the full experiment at three di}erent times over an 07 "r  6Ð00)#\ functional 6) "2Ð00)# and perceptual 4)
month period[ "1Ð02)#[ These e}ects are shown in Fig[ 0[

Results AM

Elderly controls We analysed AM|s priming data from each time!slice

separately[ In April 0883 his mean lexical decision RT
A group of six elderly normal controls "r  59Ð57 "correct responses to real words# was 822 ms^ although
years# made an average of 0[1) LD errors[ Separate this is outside the normal range of 487Ð716 ms\ his per!
analyses carried out on each of the subtypes "category formance was not excessively slow[ He made 7) LD
label\ category co!ordinate\ perceptual and functional
features# yielded signi_cant priming e}ects and no e}ect
of the living:non!living distinction[ Cate`ory labels pro!
duced a signi_cant 39 ms facilitation e}ect
"F0"0\4#  00[8\ P ³ 9[94^ F1"0\39#  24[9\ P ³ 9[990#
and cate`ory co!ordinates yielded a 58 ms priming e}ect
"F0"0\4#  27[5\ P ³ 9[90^ F1"0\39#  52[1\ P ³ 9[90#[
Functional properties generated a mean facilitation e}ect
of 47 ms "F0"0\4#  20[1\ P ³ 9[90^ F1"0\39#  21[6\
P ³ 9[990# and perceptual properties a 26 ms priming
e}ect "F0"0\4#  06[8\ P ³ 9[90^ F1"0\39#  19[5\
P ³ 9[90#[ An additional overall analysis revealed a sig!
ni_cant priming e}ect of 49 ms "MinF?"0\13#  20[18\
P ³ 9[90# and no other main e}ects or signi_cant inter!
actions[ To obtain an estimate of the magnitude of prim!
ing e}ects in the various conditions\ we calculated the
mean proportion of priming ð"related!unrelated RT:un!

 AM|s LD latencies are not directly comparable to those of

the controls since he made go:no!go responses which are typi! Fig[ 0[ Elderly controls] Proportion of priming for each type of
cally faster than the yes:no responses required of the controls[ semantic relation[
0207 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [
Fig[ 1[ AM] proportion of priming for each type of semantic
relation over time[
errors "control r  9Ð7)#[ By March 0884 his error rate
had increased to 03) and his mean LD latency to 0975
ms[ At the _nal testing session\ his mean RT was 889 ms
with an error rate of 13)[ At each time!slice\ the vast
majority of his errors were false positives[ The absence of a signi_cant interaction between feature
AM|s priming\ pattern di}ered substantially "a# from
that of the controls and "b# across time slices[ Whereas
the controls showed priming for all four types of semantic
relation\ on initial testing AM|s LDs were signi_cantly
speeded only for perceptual and functional properties
"Fig[ 1#[ Separate analyses on each type of semantic
relation showed that whereas perceptual "008 ms facili!
tation^ F"0\23#  4[185\ P  9[91# and functional "092 ms
facilitation^ F"0\27#  4[695\ P  9[91# properties sig!
ni_cantly primed\ category labels "−7 ms^
F"0\39#  9[933\ P ³ 0# and category co!ordinates "31
ms^ F"0\26#  9[421\ P ³ 0# did not[ None of these e}ects
interacted with the living:non!living variable at either this
or any subsequent time slice "all Fs ³ 0#[ Since AM|s
mean LD latency was slightly slower than that of the
controls\ we also calculated proportion of priming as
indicated above for the controls[ This yielded priming
e}ects within the normal control range for both per!
ceptual "00)# and functional "8)# properties but outside
 We assume that the LD response involves two types of
process] the automatic activation of semantic information and
response decision criteria[ We believe that the increase in LD
errors primarily re~ects a general change in response bias[ Over
time\ AM has become increasingly likely to respond positively
in any task\ whether it is LD or veri_cation[
the normal range for both category co!ordinates "3)#
and category labels "−0)#[ These e}ects are shown in
Fig[ 1[
At the second time of testing\ 00 months later\ the
pattern had changed[ Now AM no longer showed prim!
ing for perceptual attributes "proportion of priming]
−0)^ F ³ 0# in addition to continuing to show no prim!
ing for category labels "proportion of priming\ 4)^
F"0\27#  2[05\ P  9[98# or category co!ordinates "pro!
portion of priming] 2)^ F ³ 0#[ However\ functional
properties continued to prime robustly "proportion of
priming] 00)^ F"0\21#  3[33\ P ³ 9[94#[ By the _nal
testing period in October 0884\ none of the four con!
ditions not even functional relations*primed "all Fs ³ 0^
see Fig[ 1#[
These data show that\ when AM was initially tested in
April 0883\ his semantic memory was already impaired\
re~ected in the absence of signi_cant category label or
category co!ordinate priming[ Nonetheless\ salient per!
ceptual and functional features of meaning were appar!
ently preserved at this time and therefore yielded
signi_cant facilitation of LD[ Over time\ these too became
impaired\ although functional properties remained
su.ciently intact to produce signi_cant priming at the
second time slice when perceptual properties had ceased
to prime[ The two notable aspects to these data are]
"0# the contrast between priming for semantic properties
"both functional and perceptual at time 0 and only func!
tional at time 1# but none for category labels or category
co!ordinates at any period^ "1# the relative preservation
of functional information[
type and domain "living vs non!living# might seem sur!
prising since it is widely proposed that perceptual and
functional features\ respectively\ are especially salient in
the representations of living and non!living things[ It
should be noted\ however\ that the contrast between func!
tional and perceptual conditions is based on n  37 per
condition^ crossing this variable with domain halves the
number of observations per condition\ which might ren!
der the data insensitive to a real but small di}erence[
Veri_cation tasks
As a way of probing AM|s ability to make explicit
decisions about the kind of semantic information probed
implicitly in the priming study\ we devised two veri!
_cation tasks\ each requiring AM and control subjects to
make true:false judgements to short statements[ In one
test\ the statements were based on the pairs of words used
in the category label condition of the priming\ study and
in the second statements used the word!pairs from the
two property conditions[ Our question was whether this
task would yield similar or di}erent results to those
observed in priming\ both in the relative preservation
of functional information and in the changing pattern
longitudinally[
 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [
Fig[ 2[ AM] percent errors on property and category veri_cation
over time[
"a# Property veri_cation[ For the property veri_cation
test\ the 13 living and 13 non!living prime words and
their perceptual and functional property target words
were used to construct short true statements[ For exam!
ple\ fox!red from the priming study became foxes are red
and axe!handle became axes have handles[ Each sentence
was the shortest and most straightforward way to express
the property[ Pairing each prime with its two properties
produced 85 true statements and an equal number of false
statements were then constructed by pseudo!random re!
pairing of primes with other targets[ For example\ the
functional property for crocodiles "live in rivers# was
paired with the concept helicopters to produce the false
statement helicopters live in rivers[ We also added _ller
items to the set of true and false properties and the total
set of items was pseudo!randomly ordered[ Each state!
ment was read aloud to subjects who indicated whether
the statement was true or false[
Seven elderly control subjects were very accurate\ only
making between 9Ð6) errors[ Figure 2 shows AM|s per!
formance at three time slices\ approximately equivalent
to those at which the priming test were carried out[ At
the _rst of these "June 0883#\ he already made more errors
"17)# than the control subjects[ What is informative is
 Most of his errors at this and at all subsequent time!slices\
were false positives[
0208
the way in which his error pattern changes over time and
how it relates to the priming data[ First\ his performance
on the veri_cation task in June 0883 is consistent with his
priming data at around this time "April 0883# in which
he showed signi_cant priming e}ects for both perceptual
and functional attributes[ He made 18) errors when
verifying perceptual properties and 16) errors on func!
tional properties "Fig[ 2#[ With respect to the priming\
data\ it is worth noting that in spite of relatively poor
performance on the explicit veri_cation task\ semantic
representations are su.ciently intact to produce su.cient
implicit activation to result in signi_cant priming[
AM was re!tested in December 0883 when his overall
error rate had increased to 27)[ As Fig[ 2 shows\ he now
made more errors on perceptual "31)# than functional
properties "20)#[ Once again\ this is consistent with the
priming data "from three months later\ in March 0884#
which showed robust priming only for functional attri!
butes[ At the _nal timeslice\ in October 0884\ his per!
formance had deteriorated and was no better than chance
"38) error rate# and there was now no di}erence between
perceptual "40) errors# and functional "38) errors#
attributes[ By this time\ he was also not showing any
signi_cant priming e}ects for any type of semantic
relation "Fig[ 1#[ This pattern of deteriorating per!
formance on the veri_cation task was signi_cant for both
perceptual "x1"df[  0#  3[4\ P ³ 9[94# and functional
properties "x1"df[  0#  4[95\ P ³ 9[94#[ At no time was
there a signi_cant di}erence in veri_cation accuracy for
the properties of living vs non!living things[
"b# Cate`ory veri_cation[ In the second veri_cation
task\ probing category knowledge\ subjects were asked
to make true:false judgements to statements about the
category of an object\ such as] a fox is a type of animal\
a house is a type of buildin`[ We constructed short state!
ments using the 37 primes from the priming study\ with
their category label targets "e[g[ fox!animal\ house!build!
in`#[ We pseudo!randomly paired the category co!ordi!
nate targets for each prime word with the category label
to make an equal number of false sentences[ For example\
the co!ordinate target for fox was do`\ which was paired
with the category label toy\ producing the false sentence]
a do` is a type of toy[ The 85 statements were pseudo!
randomly ordered and the list was read to subjects who
indicated whether each statement was true or false[ Con!
trol subjects were very accurate\ making between 9Ð5)
errors[
AM was initially tested on the category veri_cation
task in June 0883\ at which point he was already making
many more errors "04)# than the controls "Fig[ 2#[ Per!
haps surprisingly\ he made a substantial number of errors
even when the category label and the target item crossed
domain[ For example] A drill is a type of ve`etable^ Garlic
is a table of clothin`\ where the item "`arlic^ drill# is from a
di}erent domain "living vs non!living\# than the category
speci_ed in the statement[ Half of his errors were to
statements such as this[ These statements should\ in prin!
ciple\ be quite easy to get right since they only require
0219 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [

that AM should know whether the item is a living or semantic dementia[ This kind of study\ tracking longi!
non!living thing[ The fact that he makes as many errors tudinal decline using several di}erent approaches to
on these as he does to statements such as] A cabinet is a address various theoretical issues\ is still relatively
type of clothin` suggests that the basis upon which he uncommon[ The general discussion will concentrate on
can distinguish between items at this highest level of the main issues outlined in the introduction[
categorisation has already become impaired[
By December 0883 his error rate had increased to 25)
and by August 0884 it was at chance "47)#[ This increase Hierarchical structure
in errors over time was signi_cant "x1"df[  0#  25[4\
P ³ 9[990#[ At each time!slice there was no di}erence in Many discussions of the structure of conceptual knowl!
the error rates for living and non!living things and half edge ð0\ 33Ł suggest the analogy of a downwards!bran!
of his errors were on statements which crossed!domain[ ching hierarchical tree\ with knowledge of features
speci_c to a concept at the bottom "e[g[ a canary is yellow
and sin`s#\ knowledge of its basic category and the
Summary of veri_cation results
properties relevant to all members of that category at a
middle level "a canary is a bird and therefore has feathers
For most aspects of the patterns of results\ there is a
and ~ies# and knowledge pertaining\ to the concept|s
notable match between the priming study and veri_cation
superordinate status plus associated features at the top
tests with the same materials[ The main points of simi!
of the tree "a canary is animate and therefore moves\ `rows\
larity are] "0# the progressive decline in both degree of
etc[#[ As noted in the introduction\ this characterisation
semantic priming and accuracy of veri_cation judgements
might seem to imply that the various levels are explicitly
over an 07 month period\ "1# the strikingly rapid decline
represented\ but such a literal interpretation of hier!
in sensitivity to perceptual features\ which produced both
archical structure is not required[ In connectionist models
signi_cant priming and above chance veri_cation at the
of semantic knowledge\ although higher levels of the tree
_rst testing period but complete absence of priming and
can be labelled by category and superordinate terms\ the
chance!level veri_cation by the next time of testing^ "2#
hierarchy in fact arises from the similarity structure
the relative preservation of knowledge of functional attri!
inherent in the knowledge and not from any explicit
butes in both tasks\ especially at the middle stage of
grouping\ or labelling ð19Ł[ That is\ canaries cluster with
longitudinal testing[
robins\ sparrows and budgerigars not because these are
The one discrepancy between the two measures is what
all represented at an explicit bird node\ but rather because
they suggest about AM|s knowledge of the relationship
the overlap in their feature structure guarantees that they
between an object concept and its category label[ This
will be treated as somewhat similar things[
kind of relationship "e[g[ desk!furniture#\ which yielded
It is frequently claimed that semantic deterioration
robust facilitation for the normal controls\ failed to pro!
re~ects a gradual loss or disruption of speci_c semantic
duce signi_cant implicit priming for AM even from the
features that may leave higher!order knowledge of cat!
very beginning[ By contrast\ his explicit veri_cation
egory and superordinate relations relatively preserved ð7\
judgements of this relationship were well above chance
33Ł[ This characterisation has however been challenged
at initial testing and were still just above chance*though
by Funnell ð3Ł in a case study of a patient with semantic
not by much*at time 1[
dementia\ EP[ EP showed major familiarity:word!fre!
However\ it is clear that AM|s category veri_cations
quency e}ects in her ability to de_ne category names "e[g[
were impaired even at this stage because "i# a 04) error
animal#\ exemplars "e[g[ horse# and speci_c properties
rate is substantially worse than normal and "ii# across
of those exemplars "e[g[ hooves#\ and also a principled
the living:non!living domains\ he made roughly as many
relationship between success in de_ning a speci_c exemp!
cross!domain errors "e[g[ {{a carrot is a type of clothin`||#
lar and verifying its properties[ In none of the experi!
as within!domain errors "e[g[ {{a carrot is a type of
ments\ however\ did Funnell observe any selective
animal||#[ In a strict hierarchical view of conceptual
preservation at {{higher|| levels of a putative conceptual
knowledge\ despite the fact that animals and vegetables
hierarchy[ Data from two cases with a selective de_cit for
di}er in many salient respects\ they share the super!
living things studied by Laiacona et al[ ð03Ł also failed to
ordinate domain of livin` thin` and thus have at least
support the hierarchical pattern] neither patient was more
some properties in common "e[g[ they grow#[ This view
successful in answering category than feature questions
therefore predicts that cross!domain statements should
about living things[
be easier to reject than within!domain false statements\
Our data from AM also seem inconsistent with the
but they were not for AM[
hierarchical assumption[ "i# Although normal control
subjects reveal robust priming from a concept to both its
Discussion category label and to a category co!ordinate\ AM did
not\ at any stage in the longitudinal priming study\ show
We have described the progressive deterioration of signi_cant facilitation in these conditions\ while he did
conceptual knowledge over 0[4 years in a patient with for some time prime speci_c features\ especially the func!
 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [
tional features of the concept[ "ii# In the veri_cation task\
his category membership judgements "as well as per!
ceptual feature judgements# declined more rapidly than
his ability to judge functional features[ ð05Ł that one of the most important bases on which a pre!
Functional vs perceptual knowledge
The data from both the veri_cation and priming tasks
showed the relative preservation of functional infor!
mation as AM|s semantic knowledge deteriorated[ We
observed the same picture in a very di}erent task*when
AM produced de_nitions to spoken words[ On three
occasions\ at roughly the same time as AM was tested on
the priming and veri_cation experiments\ he was given
some of the prime words used in the semantic priming
experiment and asked to say whatever he could about
these words[ In April 0883\ AM could provide some
information about most of the words\ although rarely
enough to identify them[ However\ as time went on\ the
content of his de_nitions became more impoverished and
for many of the words he could not produce any relevant
information[ The more interesting point\ however\ is the
nature of the information that he did manage to provide[
Often this was rather general but\ when he did succeed in
producing more speci_c identifying information\ much
of it was functional rather than perceptual[ We provide
some examples of his de_nitions "from June 0883# to give
the ~avour of the information he produced]
Fox] An animal which is on its own in tree places
and they kill other little animals when they|re near
them and eat them[
Bike] Normally with younger people\ but old people
still use bikes[ A lot use it just to go and get food\
but other people will use them to try and get faster
than others[
Bee] Its normally in all the gardens[ Bees [ [ [ some
people kept loads of bees beside them[ What do
they use it for<
AM is not the only semantically impaired patient who
has demonstrated rather selective preservation of func!
tional information[ A previous patient with semantic
dementia\ PP\ showed chance performance on most
explicit semantic tests even at initial assessment ð09Ł[ Non!
etheless\ an implicit word!monitoring task revealed sig!
ni_cant priming for PP\ but only if the target word was
functionally related to the preceding item in the word
sequence ð15Ł[ Similarly\ DJ\ a patient with a mild sem!
antic de_cit following herpes simplex encephalitis\
showed no priming for categories or co!ordinates\ but
signi_cant priming for functional properties ð14Ł[
 One potential problem here is that the feature terms might
have been easier to understand than the category labels "3#[ We
tried to avoid this problem by matching the familiarity and
frequency of the target words used in the lexical decision and
veri_cation tasks[
0210
This pattern supports our claim that functional infor!
mation plays a special role in semantic representations
ð30Ł[ Consistent with this idea is the claim by Mandler
verbal infant begins to distinguish between animate and
inanimate objects is that only instances of the former
class initiate movement and interact with people from
a distance[ This distinction arises from the functional
behaviour of these two classes of object\ not from their
visual appearances[ If this kind of functional information
is early learned and prominent in conceptual knowledge\
then it might be expected to be more resistant to the
ravages of brain disease[ Furthermore\ we have recently
developed a computational model of semantic memory
in which progressive damage to the network resulted in
an advantage for functional relative to other types of
information ð1\ 39Ł[ This occurred because functional
information was always intercorrelated with perceptual
information in the model\ unlike perceptual properties
which were not always intercorrelated[ Since inter!
correlated properties are relatively robust to brain dam!
age because they can be reinstated by the preserved
features with which they are associated\ functional
properties are invariably well!preserved[ On this view\
the dissociation between functional and perceptual infor!
mation should be a single\ not a double\ dissociation[
With substantial loss of semantic memory\ functional
information will of course eventually degrade\ but it
should never be more impaired than perceptual infor!
mation[
Implicit vs explicit tasks
It has been well established in cases from many di}er!
ent aetiologies "cerebrovascular accidents\ neuro!
degenerative diseases like semantic dementia\ devel!
opmental disorders like Williams syndrome\ etc[# that
implicit tasks can be sensitive to preserved knowledge
even when performance on explicit tasks is severely
abnormal[ The development of such implicit paradigms
has been a major advance in cognitive neuropsychology
over the last decade or so and has produced important
evidence on the nature of processing in disorders like
blindsight ð35Ł\ memory ð20Ł and prosopagnosia ð32Ł as
well as aphasia ð15\ 26Ð28Ł[
The sensitivity of implicit tasks to aspects of preserved
processing might lead to the conclusion that they are
{{better|| measures^ but explicit tasks also make a valuable
contribution to detailed accounts of disorders and their
implications for models of processing[ As in most con!
trasts between di}erent approaches\ the {{best|| measure
depends on the question one is asking[ It seems clear
that implicit measures are more sensitive indicators of
underlying mechanism^ but the automatic processes that
they tap are not the whole story of comprehension[ It is
our goal to understand not only these automatic mech!
anisms but also how they serve every day behaviour
0211 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [
requiring explicit responses[ This point was forcefully
made by Marcel ð06\ 07Ł in his empirical and theoretical
descriptions of conscious and non!conscious perceptual
processing[ This work is most often cited as a dem!
onstration of the nature and importance of implicit\ non!
conscious processing^ but in fact Marcel emphasised that
much of what we do in the world requires a conscious
percept[ Signi_cant facilitation by the word broom of a
subsequent response to sweep or ~oor does indeed estab!
lish that the word broom has activated its meaning[
Suppose\ however\ that a patient who shows such facili!
tation is asked {{Can you use a broom to sweep the ~oor<||
and responds {{What|s a broom<|| ð09\ 15Ł[ This suggests
that\ even when automatic semantic activation has
occurred\ it does not necessarily support all of the aspects
of processing required for normal language com!
prehension ð27\ 28Ł[ This raises the important\ but neg!
lected question\ of how these two aspects of the
comprehension process*the conscious and non!con!
scious*interface[ How do the products of implicit pro!
cessing become available for conscious re~ection<
Answers to this question will depend upon a sustained
and detailed investigation of the relationship between
implicit and explicit tasks[
In the present research\ an interesting outcome was the
extent to which the results from implicit and explicit
techniques dovetailed[ The majority of studies using both
techniques have emphasised preserved implicit pro!
cessing in patients who fail explicit tests[ This remains a
key advantage of implicit tests^ but the assumption that
this is a ubiquitous pattern might lead to the conclusion
that all apparent de_cits of semantic memory re~ect only
an impaired ability to use "in an explicit way# information
which is still representationally intact[ The current study
establishes that the structure of semantic knowledge
itself\ even when assessed by the most sensitive priming
paradigms\ can be severely degraded[ This is especially
important in the context of a longitudinal study\ where
initially normal levels of priming for some sorts of infor!
mation can be shown to evaporate as the disease
progresses[
Conclusions
We suggest that advances in understanding semantic
memory and its disorders will require the kind of mul!
tiplicity of techniques included in this study[ The picture
that emerges from such a variety of approaches may be
not as simple or clear as the conclusions sometimes drawn
from a single type of experiment^ but it seems more likely
to inform us about the genuine complexities of the struc!
ture of conceptual knowledge and its organisation in the
brain[
Acknowled`ements*First\ we would like to thank AM for
enduring testing sessions so cheerfully and willingly[ We are also
grateful to Karalyn Patterson and John Hodges for providing us
with some of their unpublished data on AM[ KP also read and
substantially revised an earlier version of this article\ for which
improvements we are indebted[ Various members of the Centre
for Speech and Language contributed to this research by testing
AM and analysing his data[ Amongst these are Mary Cooke\
Julie Morris and J[ Kate Voice[ This research was funded by
an MRC programme grant to William Marslen!Wilson and
LKT\ an MRC project grant to LKT\ John Hodges and Kar!
alyn Patterson and by a British Academy Fellowship to HM[
References
0[ Collins\ A[ M[ and Quillian\ M[ R[\ Retrieval time
from semantic memory[ Journal of Verbal Learnin`
and Verbal Behaviour\ 0858\ 7\ 139Ð136[
1[ Durrant!Peat_eld\ M[\ Tyler\ L[ K[\ Moss\ H[ E[ and
Levy\ J[\ The distinctiveness of form and function in
category structure] A connectionist model[ In Pro!
ceedin`s of the Nineteenth Annual Conference of the
Co`nitive Science Society[ Erlbaum\ Mahwah\ NJ\
0886\ pp[ 082Ð087[
2[ Farah\ M[ J[ and McClelland\ J[ L[\ A computational
model of semantic memory impairment] modality
speci_city and emergent category speci_city[ Journal
of Experimental Psycholo`y] General\ 0880\ 019\ 228Ð
246[
3[ Funnell\ E[\ Objects and properties] a study of the
breakdown of semantic memory[ Memory\ 0884\ 2\
386Ð407[
4[ Gainotti\ G[\ Silveri\ M[ C[\ Daniele\ A[ and Gius!
tolisi\ L[\ Neuroanatomical correlates of category!
speci_c semantic disorders] A critical survey[
Memory\ 0884\ 2\ 136Ð153[
5[ Graham\ K[ and Hodges\ J[\ Di}erentiating the roles
of the hippocampal complex and the neocortex in
long!term memory storage] Evidence from the study
of semantic dementia and Alzheimer|s disease\ Neu!
ropsycholo`y\ 0886\ 00\ 66Ð78[
6[ Hodges\ J[\ Patterson\ K[\ Oxbury\ S[ and Funnell\
E[\ Semantic dementia] Progressive ~uent aphasia
with temporal lobe atrophy[ Brain\ 0881\ 004\ 0672Ð
0795[
7[ Hodges\ J[ and Patterson\ K[\ Is semantic memory
consistently impaired early in the course of Alzh!
eimer|s disease< Neuroanatomical and diagnostic
implications[ Neuropsycholo`ia\ 0884\ 22\ 330Ð348[
8[ Hodges\ J[ and Patterson\ K[\ Non~uent progressive
aphasia and semantic dementia] A comparative neu!
ropsychological study[ Journal of the International
Neuropsycholo`y Society\ 0886[
09[ Hodges\ J[\ Patterson\ K[ and Tyler\ L[ K[\ Loss of
semantic memory] Implications for the modularity
of mind[ Co`nitive Neuropsycholo`y\ 0883\ 00\ 494Ð
431[
00[ Ho~and\ K[ and Johanssen\ S[\ Word frequencies in
British and American En`lish[ Longman\ Harlow\
0871[
01[ Humphreys\ G[ and Riddoch\ M[\ The fractionation
of visual agnosia[ In Visual Object Processin` eds
G[ W[ Humphreys and M[ J[ Riddoch[ Lawrence
Erlbaum Associates\ London\ 0876\ pp[ 170Ð295[
 L[ K[ Tyler and H[ E[ Moss:Going\ going\ gone [ [ [
02[ Knott\ R[\ Patterson\ K[ and Hodges\ J[ R[\ Lexical
and semantic binding e}ects in short!term memory]
Evidence from semantic dementia[ "submitted# 0886[
03[ Laiacona\ M[\ Barbarotto\ R[ and Capitani\ E[\ Per!
ceptual and associative knowledge in category spec!
i_c impairment of semantic memory] A study of two
cases[ Cortex\ 0882\ 18\ 616Ð639[
04[ Lambon Ralph\ M[\ Patterson\ K[ and Hodges\ J[\
The relationship between naming and semantic
knowledge for di}erent categories in dementia of
Alzheimer|s type[ Neuropsycholo`ia\ 0886\ 24\ 0140Ð
0159[
05[ Mandler\ J[\ How to build a baby II] Conceptual
primitives[ Psycholo`ical Review\ 0881\ 88\ 476Ð593[
06[ Marcel\ A[ J[\ Conscious and unconscious per!
ception] Experiments on visual masking and word
recognition[ Co`nitive Psycholo`y\ 0872a\ 04\ 086Ð
126[
07[ Marcel\ A[ J[\ Conscious and unconscious per!
ception] An approach to the relations between
phenomenal experience and perceptual processes[
Co`nitive Psycholo`y\ 0872b\ 04\ 127Ð299[
08[ Marslen!Wilson\ W[ D[ and Tyler\ L[ K[\ The tem!
poral structure of spoken language understanding[
Co`nition\ 0879\ 7\ 0Ð60[
19[ McClelland\ J[ McNaughton\ B[ and O|Reilly\ R[\
Why are there complementary learning systems in
the hippocampus and neocortex] Insights from the
success and failures of connectionist models of learn!
ing and memory[ Psycholo`ical Review\ 0884\ 091\
308Ð346[
10[ Moss\ H[\ Hare\ M[\ Day\ P[ and Tyler\ L[ K[\ A
distributed memory model of the associative boost
in semantic priming[ Co`nitive Science\ 0883\ 50\
302Ð316[
11[ Moss\ H[ E[\ McCormick\ S[ and Tyler\ L[ K[\ The
time!course of activation of semantic information
during spoken word recognition] Function precedes
form[ Lan`ua`e and Co`nitive Processes\ 0886\ 01\
584Ð622[
12[ Moss\ H[ and Older\ L[\ Birkbeck Word Association
Norms[ LEA\ Hove\ England\ 0885[
13[ Moss\ H[ E[\ Tyler\ L[ K[ and Jennings\ F[\ When
leopards lose their spots] Knowledge of visual
properties in category!speci_c de_cits for living
things[ Co`nitive Neuropsycholo`y\ 0886\ 03\ 890Ð
849[
14[ Moss\ H[ E[\ Tyler\ L[ K[ and Kopelman\ M[\ Knowl!
edge of general and speci_c information following
damage to semantic memory "submitted#\ 0886[
15[ Moss\ H[\ Tyler\ L[ K[\ Patterson\ K[ and Hodges\ J[\
Exploring the loss of semantic memory in semantic
dementia] Evidence from a primed monitoring study[
Neuropsycholo`y\ 0884\ 8\ 05Ð16[
16[ Neebes\ R[ D[ and Brady\ C[ B[\ Integrity of semantic
_elds in Alzheimer|s Disease[ Cortex\ 0877\ 13\ 180Ð
299[
17[ Ostrin\ R[ and Tyler[ L[ K[\ Automatic access to
lexical semantics in aphasia] Evidence from semantic
and associative priming[ Brain and Lan`ua`e\ 0882\
34\ 036Ð059[
18[ Rapp\ B[ and Caramazza\ A[\ General to speci_c
0212
access to word meaning] A claim re!examined[ Co`!
nitive Neuropsycholo`y\ 0878\ 5\ 140Ð161[
29[ Sa}ran\ E[ M[ and Schwartz\ M[ F[\ Of cabbages and
things] semantic memory from a neuropsychological
perspective*a tutorial view[ In Attention and Per!
formance XV\ eds C[ Umilta and M[ Moscovitch[
Lawrence Erlbaum\ Hove\ UK\ 0883\ pp[ 496Ð424[
20[ Schacter\ D[\ Consciousness and Awareness in mem!
ory and amnesia] Critical issues[ In The Neu!
ropsycholo`y of Consciousness\ eds A[ D[ Milner and
M[ D[ Rugg[ Academic Press\ London\ 0881[
21[ Schwartz\ M[\ Marin\ O[ and Sa}ran\ E[\ Dis!
sociation of language function in dementia] A case
study[ Brain and Lan`ua`e\ 0868\ 6\ 106Ð295[
22[ Shallice\ T[\ From neuropsycholo`y to mental struc!
ture[ Cambridge University Press\ Cambridge\ UK\
0877[
23[ Snodgrass\ J[ G[ and Vanderwart\ M[\ A stan!
dardised set of 159 pictures] norms for name agree!
ment\ image agreement\ familiarity and visual
complexity[ Journal of Human Experimental Psy!
cholo`y] Learnin` and Memory\ 0879\ 5\ 063Ð104[
24[ Snowden\ J[\ Goulding\ P[ J[ and Neary\ D[\ Sem!
antic dementia] A form of circumscribed cerebral
atrophy[ Behavioural Neurolo`y\ 0878\ 1\ 056Ð071[
25[ Snowden\ J[\ Gri.ths\ H[ and Neary\ D[\ Semantic
dementia] Autobiographical contribution to pres!
ervation of meaning[ Co`nitive Neuropsycholo`y\
0883\ 00\ 154Ð177[
26[ Tyler\ L[ K[ Real!time comprehension processes in
agrammatism] a case study[ Brain and Lan`ua`e\
0874\ 15\ 148Ð164[
27[ Tyler\ L[ K[\ The distinction between implicit and
explicit language function] Evidence from aphasia[
In Neuropsycholo`y of Consciousness\ eds M[ Rugg
and D[ Milner[ Academic Press\ London\0881a[
28[ Tyler\ L[ K[\ Spoken Lan`ua`e Comprehension] An
experimental approach to normal and disordered pro!
cessin`[ MIT Press\ Cambridge\ Mass\ 0881b[
39[ Tyler\ L[ K[\ Durrant!Peat_eld\ M[\ Levy\ J[\ Voice\
J[ K[ and Moss\ H[ E[\ Distinctiveness and cor!
relations in the structure of categories] Behavioural
data and a connectionist model[ Brain and Lan`ua`e\
0885\ 44\ 78Ð81[
30[ Tyler\ L[ K[ and Moss\ H[ E[\ Functional properties
of word meanings] studies of normal and brain!dam!
aged patients[ Co`nitive Neuropsycholo`y\ 0886\ 03\
400Ð434[
31[ Tyler\ L[ K[\ Moss\ H[ E[\ Patterson\ K[ and Hodges\
J[\ The gradual deterioration of syntax and semantics
in a patient with progressive aphasia[ Brain and Lan!
`ua`e\ 0886\ 45\ 315Ð366[
32[ Young\ A[ and De Haan\ E[\ Face recognition and
awareness after brain injury[ In The Neuropsycholo`y
of Consciousness\ eds A[ D[ Milner and M[ D[ Rugg[
Academic Press\ London\ 0881[
33[ Warrington\ E[\ The selective impairment of sem!
antic memory[ Quarterly Journal of Experimental
Psycholo`y\ 0864\ 16\ 524Ð546[
34[ Warrington\ E[ and Shallice\ T[\ Category speci_c
semantic impairments[ Brain\ 0873\ 096\ 718Ð743[
35[ Weiskrantz\ L[\ Blindsi`ht[ Oxford University Press\
New York\ 0875[