Attached.

Running head: CASE STUDY KNOWLEDGE REPRESENTATION

Case Study Knowledge Representation
Student
Institution

`1

CASE STUDY KNOWLEDGE REPRESENTATION

2

From the table output, the result is compared from the drawing line together with both
variables; naming line drawings and the naming photographs as well. The ER average was able
to record correctly on 33% of the animals and the overall percentage of 30% for fruits or
vegetables despite the 71% correct of the approximated naming of the line drawings. This was
due to the dependence of the ER performance’s category that shows both deficits regarding
animals and fruits or vegetable artefacts. The comparison for control variables was displayed on
the average naming of the animals representing 98% and the vegetables and fruit was measured
approximately 97.5% (Kolinsky et al., 2002). However, the recording indicated 99% of the
artefacts concerning line drawing.
On the naming of the photographs, ER was examined to be below the average with an
approximate 50% for animals and 39% for vegetables or fruits. This ER was also done except for
the body parts and figures, resulting in naming. On the contrary, the total artefacts recorded was
82%. Controls however performed exceptionally well with 97% of the fruits or vegetables and
100% score for animals. The lowest score was 94% under the variable as the total recording was
99%. This result was important since it highlighted the ER’s fit behavior for another model
concerning semantic memory to consider the organisation category.
The ER is, therefore confabulated with certain kinds of information since most of the
confabulators issue nuanced explanations as well as the accompaniments of facts. Moreover, the
researchers’ goal was to report a case of brain damage to patients who expressed the consistent
remarks of a certain deficit of living things. Finally, the relevance of the control groups as per the
researchers was to elaborate the unchanged correlation between variables.

CASE STUDY KNOWLEDGE REPRESENTATION
Reference
Kolinsky, R., Fery, P., Messina, D., Peretz, I., Evinck, S., Ventura, P., & Morais, J. (2002). The
fur of the crocodile and the mooing sheep: A study of a patient with a category-specific
impairment for biological things. Cognitive Neuropsychology, 19(4), 301-342.

3


COGNITIVE NEUROPSY CHOLOGY, 2002, 19 (4), 301–342

THE FUR OF THE CROCODILE AND THE MOOING
SHEEP: A STUDY OF A PATIENT WITH A CATEGORY SPECIFIC IMPAIRMENT FOR BIOLOGICAL THINGS
Régine Kolinsky
Université Libre de Bruxelles, Belgium and Fonds National de la Recherche Scientifique, Belgium

Patrick Fery and Diana Messina
Université Libre de Bruxelles, Belgium

Isabelle Peretz
Université de Montréal, Canada

Sylvie Evinck
Université Libre de Bruxelles, Belgium

Paulo Ventura
University of Lisbon, Portugal

José Morais
Université Libre de Bruxelles, Belgium

We report a single case study of a brain-damaged patient, ER, who showed a remarkably consistent category-specific deficit for living things. His impairment was observed across tasks (naming, definition,
matching, drawing from memory, questionnaires), input modalities (visual, verbal, nonverbal auditory),
and output modalities (verbal vs. pointing or visual matching responses) as well as for different types of
knowledge. Although visual knowledge of living things was severely affected, his category-specific
impairment in nonverbal sound recognition is inconsistent with models of category-specific deficits
based on pre-semantic visual descriptions. ER’s deficit cannot fully be explained by item typicality,
word frequency, visual complexity, homomorphy, age of acquisition, value to perceiver, or modality of
transaction. Furthermore, in ER, contextual cues were even slightly detrimental for the recognition of
animals. ER’s naming and recognition errors were constrained by the categorical structure of the knowledge base: In most cases they respected both the second- and first-order superordinates. In particular,
ER’s knowledge of shared categorical properties related to biological function was almost spared. This
result is compatible with the idea that, for living things, shared functional properties and shared perceptual properties are strongly correlated. Feature-based models assuming perceptual vs. functional semantic components cannot account for ER’s deficit, since for living things he was impaired on both kinds of
features to a similar extent. ER’s behaviour is quite consistent with the notion that conceptual knowledge is organised categorically in the brain, with one or several specialised subsystems for biologically
related entities.
Requests for reprints should be addressed to Régine Kolinsky, Unité de Recherche en Neurosciences Cognitives (UNESCOG),
Université Libre de Bruxelles, CP 191, Av. F.D. Roosevelt, 50, B-1050 Brussels, Belgium (Fax: 32-2-650.22.09; Email:
rkolins@ulb.ac.be).
We wish to thank Tim Shallice for very helpful discussions on a former version of the manuscript, as well as the Editor and two
anonymous reviewers for their very constructive comments. We also thank all the people who participated in the present study, including the members of the Research Unit in Cognitive Neuroscience and of the Laboratoire de Psychologie Expérimentale of the Université
Libre de Bruxelles, who were prepared to spend several hours filling in questionnaires. This work was supported by A.F.R.F.C.–
F.N.R.S. grant.
Ó 2002 Psychology Press Ltd
http://www.tandf.co.uk/journals/pp/02643294.html

301

DOI:10.1080/02643290143000196

KOLINSKY ET AL.

The observation of specific semantic memory disorders, i.e., that some patients exhibit a selective
loss of knowledge of items from certain categories,
for example biological (or living) entities, by comparison with nonbiological (or non-living) objects,
has generated a large debate about the internal
organisation of the semantic system (see recent
reviews, e.g., in Caramazza, 1998; Caramazza &
Shelton, 1998; Forde & Humphreys, 1999;
Humphreys & Forde, 2001; Saffran & Sholl, 1999;
Shelton & Caramazza, 1999).
According to Caramazza and Shelton (1998, p.
1), most present models reflect the “dominant,
reductionist theory of category-specific deficits,”
which holds that the categorical nature of the deficits is the result of damage to non-categorically
organised semantic representations. Indeed, several
explanations share the notion that the apparent
boundary between biological and nonbiological
things should not be taken at face value, but as
reflecting a second dimension of stimulus variation
correlated with the biological/nonbiological
dimension. These explanations can be divided into
those that consider the category-specific deficits as
artefacts of uncontrolled variables associated with
the impaired categories and those that explain the
phenomenon as an emergent property of noncategorical attributes of the members of the
categories.
According to the artefact view, impairments of
the knowledge of biological things result from the
confounding of category with other variables,
namely word frequency, concept familiarity, and
visual complexity (Funnell & Sheridan, 1992;
Hillis & Caramazza, 1995; Stewart, Parkin, &
Hunkin, 1992). Closely related to the artefact
account is the notion of visually crowded categories
(e.g., Damasio, 1990; Gaffan & Heywood, 1993),
put forward by Humphreys, Riddoch and their colleagues (e.g., Humphreys & Riddoch, 1987, 1988;
Humphreys, Riddoch, & Quinlan, 1988; Riddoch,
Humphreys, Coltheart, & Funnell, 1988) as well as
by other authors (e.g., Sartori & Job, 1988). Since
biological categories such as mammals or edible
fruits have a high degree of intracategory visual
similarity, a same amount of noise (or visual processing impairment) would affect the recognition of

302

COGNITIVE NEUROPSYCHOLOGY , 2002, 19 (4)

biological objects to a larger extent than the recognition of nonbiological ones. Arguin and colleagues
(Arguin, Bub, & Dudek, 1996; Dixon, Bub, &
Arguin, 1997) developed a similar idea but involving both visual and semantic dimensions.
The second type of explanation, namely the
emergent property account of category-specific deficits, assumes that conceptual knowledge and/or the
computational and structural properties of semantic categories are organised in such a way that selective damage to a particular brain area will result in a
category-like effect even though knowledge is not
organised categorically.
For example, according to the sensory-functional
theory (henceforth, SFT), introduced by
Warrington and McCarthy (1983) and
Warrington and Shallice (1984), category-specific
impairments reflect an opposition between
sensorial and nonsensorial features. This opposition is related to the notion of distinguishing features, which serve to differentiate items belonging
to the same category (e.g., Gonnerman, Andersen,
Devlin, Kempler, & Seidenberg, 1997). Sensorial
features (shape, colour, texture, sound, etc.) would
tend to differentiate living things from one another
(e.g., the feature differentiates tigers
from lions), whereas motor and/or associative and
functional features would allow one to distinguish
among non-biological objects (what they are used
for, how they are made, etc.).
Computational variants of the SFT (Farah &
McClelland, 1991) have shown how a system in
which the ratio of visual to functional features is
higher for living than nonliving things exhibits
worse performance in naming living things when
the visual semantic component is damaged. The
evidence provided by fMRI studies (Chao, Haxby,
& Martin, 1999; Thompson-Schill, Aguirre,
D’Esposito, & Farah, 1999), showing that areas
assumed to be involved in storing information
about object form were activated when subjects
answered both visual and functional questions
about animals, might argue for the interactive
nature of the two semantic representations (but see
Caramazza, 2000, for a different view).
The differential diagnostic value of sensorial vs.
functional-associative knowledge for living relative

CATEGORY -SPECIFIC DEFICIT

to nonliving things has been incorporated both
within multiple (e.g., Warrington & McCarthy,
1983; Warrington & Shallice, 1984) and single
semantic system models (e.g., Forde, Francis,
Riddoch, Rumiati, & Humphreys, 1997;
Humphreys, Lamonte, & Lloyd-Jones, 1995;
Humphreys & Riddoch, 1987, 1988; Humphreys
et al., 1988; Lloyd-Jones & Humphreys, 1997;
Riddoch et al., 1988). The main difference between
multiple vs. single semantic system models is the
level at which the sensorial vs. nonsensorial features’ opposition takes place, namely, within the
semantic system itself, or not. Warrington,
McCarthy, and Shallice have argued for both input
modality and informational content specificity1
within the semantic store. By contrast, Humphreys
and colleagues reserve the term semantic memory
for nonsensorial features and propose access to the
amodal semantic memory store to be accomplished
via pre-semantic systems that hold modalitycongruent sensorial knowledge.
The Organised Unitary Content Hypothesis, or
OUCH, does not introduce such an opposition
between modalities (sensorial vs. nonsensorial features); nor does it resort to any artefact explanation
(e.g., Caramazza, Hillis, Rapp, & Romani, 1990;
Hillis & Caramazza, 1991; Hillis, Rapp, &
Caramazza, 1995; Rapp, Hillis, & Caramazza,
1993). According to OUCH, although there are no
explicit category boundaries within the semantic
system, a category-like structure nevertheless
emerges because the semantic properties of concepts of a given category tend to cluster together.
Indeed, members of a semantic category share
many semantic features (e.g., all animals breathe),
and some properties of an object tend to be highly
intercorrelated (e.g., things that breathe also tend
to be made of certain kinds of substance). In addition there are highly specific correlations between
distinctive sensorial and functional features of
objects, because the action pattern associated with
them (e.g., “used for cutting”) is determined considerably by their shape (e.g., “has a blade”; see also

Capitani, Laiacona, Barbarotto, & Trivelli, 1993;
De Renzi & Luchelli, 1994). An implication of
these assumptions is that the semantic system is not
homogeneous but “lumpy”: Some regions are
densely packed and others are sparse. Since the
properties that cluster together will also tend to be
damaged together, a lesion affecting a densely
occupied region will lead to a category-like impairment, which would not be the case for a lesion
affecting a less densely occupied region.
Other accounts of category-specific deficits
share with OUCH the assumption that the properties of members of semantic categories tend to cluster together and, therefore, to be damaged together.
However, they also assume an interaction with the
different weighting of shared and intercorrelated
features. According to these models, biological
things share many more highly correlated semantic
(mainly perceptual) features than nonbiological
objects (e.g., Devlin, Gonnerman, Andersen, &
Seidenberg, 1998). For example the features and are intercorrelated because
they tend to appear in the same basic-level concepts: cats, dogs, tigers, lions, etc., which have fur,
also have whiskers. Since many animals present
these properties, they are not informative enough to
distinguish between alternative items. In contrast,
nonbiological objects have proportionally more distinctive properties. McRae, De Sa, and Seidenberg
(1997) suggested that this difference between the
two categories might reflect constraints on the
structure of the two types of objects (geneticevolutionary principles vs. functional, aesthetic,
economic, etc.).
Moss, Tyler, and colleagues (Moss & Tyler,
2000; Moss, Tyler, Durrant-Peatfield, & Bunn,
1998; Moss, Tyler, Hodges, & Patterson, 1995;
Tyler & Moss, 1997; Tyler, Moss, DurrantPeatfield, & Levy, 2000) argued that living and
nonliving things differ both in degree and type of
correlation among features. As in Devlin et al.
(1998), they assume that living and nonliving
things differ in terms of distinctiveness and correla-

1

As discussed by Caramazza and Shelton (1998, pp. 29–30), the notion of modality specificity is “multiply ambiguous.” We use
here the term “modality” to refer only to type of knowledge (e.g., sensorial vs. functional). Modality of presentation of the stimulus
(e.g., visual vs. auditory) is referred to as “modality of input.”
COGNITIVE NEUROPSYCHOLOGY, 2002, 19 (4)

303

KOLINSKY ET AL.

tion among properties. But, unlike Devlin et al.,
they propose that, in the case of living things,
shared properties concerning biological actions
(e.g., eating, walking, growing, etc.), have a more
special status than specific uses (e.g., riding horses,
etc.). Moss et al. also claim that shared biological
functional properties are densely correlated with
other properties, and therefore less vulnerable to
damage. Functional information for living and
nonliving things is relatively resistant to damage
because it is supported by correlations with shared
(living) and distinctive (nonliving) sensorial features (Tyler & Moss, 1998).
The models based on the reductionist hypothesis refer to features rather than to raw categories and
explain category-specific impairments as resulting
from noncategorical properties of semantic representations. Several predictions from these models
do conflict with the patients’ data. For example,
there seems to be no systematic relation between
modality and category (see discussion, e.g., in
Caramazza & Shelton, 1998) nor between severity
of damage and nature of the affected category
(Garrard, Patterson, Watson, & Hodges, 1998; see
discussions in Devlin et al., 1998; Moss et al.,
1998). Additionally, there have been reports of
patients with living entities deficits but no selective
impairment for sensorial features (e.g., Laiacona,
Capitani, & Barbarotto, 1997; Lambon Ralph,
Howard, Nightingale, & Ellis, 1998; Moss et al.,
1998), as well as patients with poor knowledge of
visual information but no disproportionate deficit
for living over nonliving things (e.g., Lambon
Ralph et al., 1998).
More generally, all reductionist models fail to
predict which categories may be specifically
impaired and to explain why very narrow categoryspecific deficits, like birds but not fish, are not
observed (cf., Caramazza, 1998), and “why it is that
by far the most prevalent form of category-specific
deficit involves the category of living things”
(Caramazza & Shelton, 1998, p. 9).
To be sure, artefact accounts, including those
referring to visual complexity, do offer an explanation for the higher incidence of category-specific
deficits for biological entities: they assume that
these deficits are due to differences in the degree of

304

COGNITIVE NEUROPSYCHOLOGY , 2002, 19 (4)

processing demands, biological categories being
supposedly more difficult to process (more “visually
crowded,” and/or less familiar, etc.). However,
artefact explanations cannot account for some
attested patterns of impairment, like selective
impairment for biological categories in patients displaying intact structural knowledge (e.g., SB, studied by Sheridan & Humphreys, 1993), and selective
impairment for nonbiological items, supposedly
easier to process (Sacchett & Humphreys, 1992;
Silveri et al., 1997; Tippett, Glosser, & Farah,
1996; Warrington & McCarthy, 1983, 1987). In
particular, observations of opposite patterns of category deficits on the same items (Gonnerman et al.,
1997; Hillis & Caramazza, 1991) strongly reinforce
the notion that these deficits cannot be due to differences in the degree of processing demand. Some
authors have recently proposed that nonliving
things may even be harder to process (Laws, 2000;
Laws & Neve, 1999; Turnbull & Laws, 2000). The
greater within-item structural diversity of nonliving
things (e.g., one telephone may look quite dissimilar to another telephone) implies that more structural information may be needed in order to
recognise nonliving than living items, which are
more within-item structurally redundant.
In addition, several studies controlling for the
effects of numerous variables that were thought to
be correlated with the biological/nonbiological
dimension (e.g., familiarity, visual complexity,
name length and frequency, image agreement,
within-category visual similarity, control subjects’
performance, etc.) or entering such variables in
regression analyses have shown that the “biologicalnonbiological” variable is an important explicative
factor in accounting for patients’ performance (e.g.,
Farah, McMullen, & Meyer; 1991; Gaffan & Heywood, 1993; Kurbat, 1997; Kurbat & Farah, 1998).
Thus, category-specific deficits cannot be
reduced in a simple way either to a core disorder of
sensory or functional-associative semantics per se
or to uncontrolled factors. Other propositions, like
for example the idea that motor-kinaesthetic integration may be especially relevant for the identification of man-made objects (Warrington &
McCarthy, 1987; see also Damasio, 1989, 1990),
also seem challenged by the data (e.g., Gaffan &

CATEGORY -SPECIFIC DEFICIT

Heywood, 1993). The available evidence thus
undermines any simplistic model in which a single
division (e.g., “perceptual vs. functional,” “manipulable vs. non-manipulable,” or “structurally similar
vs. structurally dissimilar”) exactly reflects the partition between categories like animals and manufactured objects.
An alternative explanation of category-specific
impairments could be that these deficits reflect an
underlying organisation in which knowledge about
different categories would be represented by different subsystems. According to this category-specific
explanation, the impairment of knowledge of biological things is the result of a damage to a subsystem of semantic memory dedicated to “biological”
or “living” things.
This explanation was originally proposed by
Nielsen (1946) and recently rehabilitated by
Caramazza and Shelton (1998; see also Caramazza,
1998). Evolutionary pressures (e.g., need for nurture, protection, and food, and avoidance of predators) resulted in dedicated neural mechanisms for
processing specific kinds of objects like animals and
plant life (and also conspecifics, according to Kay &
Hanley, 1999; Miceli et al., 2000; Shelton, Fouch,
& Caramazza, 1998). Nonliving things need not be
a specific domain knowledge emerging from evolutionary pressures (Barbarotto, Capitani, &
Laiacona, 2001; Caramazza & Shelton, 1998;
Laiacona & Capitani, 2001) for category-specific
deficits to occur as a result of brain damage. Such a
view is supported, e.g., by concept acquisition studies: The concepts of animal vs. artefact seem to be
acquired through the operation of innate domainspecific mechanisms (e.g., Mandler, 1994; Spelke,
Phillips, & Woodward, 1995).
According to this “domain-specific knowledge
hypothesis,” there are separate semantic representations for different categories, but there is no
separation of features within each category representation: dedicated neural circuits for distinct
categories are not committed to the processing of a
specific kind of information (but see Coltheart et
al., 1998; Thompson-Schill & Gabrieli, 1999, for
different positions on this matter).
In the present paper, we provide further evidence against the SFT theory and other

reductionist accounts of category-specific deficits.
We report an 8-year examination of ER, a case of
category-specific deficit, who displayed a strong
deficit for the naming and recognition of most
items belonging to biological categories. First, we
show that ER’s visual knowledge and visual access
to living things visual information were impaired.
Then, we demonstrate that his category-specific
deficit was not restricted to visual knowledge and
that his knowledge of auditory properties was also
selectively affected for living things. Third, we provide evidence that the integration of sound, vision
and contextual cues did not help ER to access living
things semantic information. Finally, we show that
ER’s deficit for living things comprises functional/
associative attributes as well as sensorial ones, and
that his category-specific deficit does not extend
to shared properties specifying superordinate
categories.

CASE REPORT
ER, a right-handed man born in 1932, suffered a
herpes encephalitis in 1969. He was examined at
the Neurology Department of Erasmus Hospital
(Brussels) in 1985 and in 1986. A CT scan showed
bilateral hippocampal lesions with areas of
hypodensity in the uncus of both temporal lobes
(see Figure 1). Neurological examination was normal and remained unchanged throughout investigation. Since ER was almost totally unable to recall
any event subsequent to the onset of his illness and
since he had been abandoned by his family and
placed in a psycho-geriatric hospital, no other
information relative to his medical history could be
obtained. At the end of his professional life, he was
working as a computer programmer. ER died in
1996.
ER’s IQ score, measured with the WAIS,
decreased from 97 in 1986 and 1989 to 91 in 1991
(Verbal IQ was 95, 94, and 88, and Performance IQ
was 101, 97, and 97, respectively). ER’s comprehension and expressive speech were normal, except
for some minor word-finding difficulties. Reading
and writing were accomplished without difficulty.
When asked to say antonyms and synonyms of
COGNITIVE NEUROPSYCHOLOGY, 2002, 19 (4)

305

KOLINSKY ET AL.

Figure 1. CT scans of ER.

words, he obtained 98% and 78% correct responses,
respectively. Performance was normal at the Wisconsin Card Sorting Test.
Memory testing showed a dense amnesic syndrome. ER was severely impaired when he had to
recall ongoing events and was disoriented in time
and place. However, recall of premorbid autobiographical events seemed accurate and without confabulation. ER’s performance was normal for Digit
Span but extremely poor for the recall of the Rey
figure (3/36) and for the logical memory (3 and 2/
24) and the paired associate learning (2/30) subtests
of the Wechsler Memory Scale. The Benton Visual
Retention Test was performed at chance level (4/15
2

on recognition). The Rivermead Behavioral Memory Test revealed screening scores as low as 0 and 1
(out of 12).
ER’s ocular examination yielded results within
the normal range, considering the age of the
patient. Visual acuity was perfect for both eyes; slit
lamp examination, pupillary sizes and reflexes, and
ocular motility were also normal, as were pattern
visual evoked potentials. Colour perception was
roughly intact. Normal results were obtained on the
Ishiara’s (1979) test (one error for each eye). Some
errors were observed on the Gelb Goldstein colour
sorting test (Goldstein & Scheerer, 1941), but
some browns and oranges could not be sorted; beige
was classified as pale green, then as white; dark and
pale pinks were separated. On the second part of
this test, ER correctly sorted the skeins when categories were selected by the examiner. The
Farnsworth-Munsell 100-hue test (Farnsworth,
1943) was performed in the normal time and led to
scores of 96 and 129 for the right and left eye,
respectively, without definite confusion zone.
The other visual perceptual abilities were also
intact. ER proved to be able to discriminate the
individual components of three to five overlapping
figures, consisting of either geometric forms or
object drawings. Discrimination of unfamiliar
faces, through changes of either orientation or
expression, was almost perfect. He was able to copy
perfectly simple as well as complex drawings of
objects (e.g., the Rey figure, Osterrieth, 1944).
Visual discrimination abilities were investigated
using the four tasks designed by Humphreys and
Riddoch (1984) to evaluate length, size, orientation, and position discrimination. Compared to
four matched controls2, who obtained on the average 92% correct responses, ranging from 85% to
99%, ER was in the normal range, with 85% correct
(95% for length, 85% for size, 75% for orientation,
and 85% for position discrimination).
Routine investigation of naming showed some
minor difficulties to name real (3-D) artefacts presented either haptically (80% correct naming; but
83% of the unnamed objects were identified as
demonstrated by correct usage) or visually (95%

Henceforth, Controls refer to normal subjects of similar age and sociocultural background as ER.

306

COGNITIVE NEUROPSYCHOLOGY , 2002, 19 (4)

CATEGORY -SPECIFIC DEFICIT

correct naming, the function of the unnamed object
being correctly identified). However, when presented visually with plastic 3-D representations of
fruits and animals, ER only recognised 65% of
them. With line drawings, ER recognised 86% of
the artefacts, but only 54% of the biological ones.
When presented in a further session with the spoken names of the same drawings and asked to give a
definition or to describe them, ER obtained 89%
correct responses for the artefacts but only 58% for
the biological ones. This was most surprising,
because ER mentioned that he lived in a farm (and
thus knew many animals). Interestingly, ER
showed no difficulty with body parts: he pointed on
himself 29 out of 30 parts named by the examiner
and correctly named his own body parts when
touched by the examiner (24 and 30/30 with eyes
open vs. closed).
Thus, ER’s neuropsychological assessment
revealed two main impairments: a severe amnesic
syndrome and, apparently, a selective trouble in the
recognition of biological items. The following
investigation, carried out in 1986, 1988, and 1991,
is concerned with the second point.

NEUROPSYCHOLOGICAL
INVESTIGATION
ER’s visual knowledge and access to
semantic information about living things are
impaired
The first set of tests was aimed at determining
whether or not ER suffered from a specific impairment for biological objects when presented with
visual input. This question was inspected both
through drawing and picture naming and through
tests that examine visual access to semantic knowledge without requiring a verbal response (word-

drawing matching and semantic association of
drawings). Structural knowledge of both artefacts
and biological things as well as knowledge of visual
details of animals were also examined.
Category-specific impairment in visual naming
Naming of standardised line drawings. A set of 151
line drawings taken from Snodgrass and
Vanderwart (1980) was presented twice to ER and
12 controls. It included 76 artefacts, 52 animals,
and 23 fruits or vegetables3.
On average4, ER was able to name correctly only
33% of the animals and 30% of the fruits or vegetables, although he reached 71% correct naming for
the artefacts. ER’s performance thus depended on
category, c2 (2) = 23.09, p < .001, showing a deficit
for both animals and fruits or vegetables in comparison to artefacts. The Controls displayed, on average, 98% correct naming for animals, 97.5% for
fruits or vegetables, and 99% for artefacts (lowest
scores: 90%, 91%, and 96%, respectively).
ER provided a correct description of the target
several times, without being able to name it. For
example, the anchor was described as “what you
need on a boat, that you throw into the sea to make
the boat stop.” Since we wanted to assess the importance of ER’s semantic rather than lexical access
deficit, we also considered a recognition score based
on naming and/or description5. ER still displayed
a strong category effect, with about 36.5% correct
recognition for animals, 35% for fruits or
vegetables, and 85.5% for artefacts, c2 (2) = 38.91,
p < .001.
In order to examine the “category-specific”
explanation in more detail, we checked whether ER
was impaired for some artefact subcategories. This
was not the case: ER’s recognition performance was
uniformly higher for artefact than for biological
subcategories (see Table 1). Even the two worst

3

The words corresponding to the to-be-named pictures served in a word definition task that is presented later. For this reason the
stimuli were divided into two sets, each including approximately the same number of biological and nonbiological objects, and the two
tasks (naming and word definition) were distributed between two sessions according to an AB-BA design.
2
4
Since ER’s performance was stable and consistent, r0 = .82, p < .0001; McNemar test of change: c (1) = 1, we averaged the two
session scores.
5

The item was accepted as having been correctly recognised if it could be identified on the basis of ER’s verbal response by three
independent judges.
COGNITIVE NEUROPSYCHOLOGY, 2002, 19 (4)

307

KOLINSKY ET AL.

Table 1. ER’s correct recognition scores (in percentages: CR) for
various biological and artefact subcategories
Sub-category
Biological
Insects
Vegetables
Birds
Fish and molluscs
Reptiles and batrachians
Fruits
Mammals
Artefact
Musical instruments
Toys
Furniture and household items
Buildings (or parts of them)
Means of transport (or parts of them)
Clothes and jewellery
Containers
Tools and weapons
Other artefacts

% correct
12.5
20.0
22.0
33.0
40.0
46.0
48.0

72.0
72.0
79.0
86.0
89.0
91.0
92.0
100.0
100.0

artefact subcategories (musical instruments and
toys) were far more successful than the best biological subcategory (mammals). Category specificity
was thus quite strong in this patient. Besides, since
ER’s performance on the animate artefacts (transport means) was very good, we discard the idea that
his impairment resulted from a damage to the
feature “motility” (e.g., Hillis & Caramazza, 1991).
The items that ER was consistently unable to
recognise were concentrated among the less typical,
infrequent, and unfamiliar items6. Mean item
familiarity and mean word frequency were significantly lower for the items that he did not recognise
consistently (2.5 and 29.45, respectively) than those
that he did (3.1 and 64.36), F(1, 147) = 13.62, p <
.0005; and, F(1, 119) = 5.64, p < .025, for the two
measures, respectively; this effect did not interact
with item category (F < 1 in both cases). Typicality

seemed to affect only biological items, (interaction
with category F(1, 128) = 3.54, p =.06), successful
biological items showing higher typicality than the
failures (34.36 vs. 17.22, on the average; for
artefacts: 19.89 for successful items and 18.59 for
failures). Visual complexity7 was, on average,
higher for biological items (3.48) than for artefacts
(2.93), F(1, 147) = 7.97, p < .005, but did not differ
between successful and failed items (F @ 1).
To better estimate the possible effect of uncontrolled variables on ER’s recognition performance,
we checked whether the whole set of animals, fruits,
and vegetables differed from the artefact items on
the 4 mentioned variables as well as on 13 other
variables. These are image agreement7, similarity to
most similar other object 8, contour overlap9, which
measures intra-category object visual similarity
(Humphreys et al., 1988), age of acquisition10,
French name length in terms of number of letters and
number of syllables, proportion of internal details, and,
at both coarse and fine spatial scales, proportion of
straight contours, curvature variability, and total
number of concavities (Kurbat, 1997). Averaged values of these material characteristics are presented in
Appendix A. ANOVAs were performed on these
measures (see Appendix A for the results of the
Scheffé’s tests), showing that categories differed in
terms of image agreement, familiarity, visual complexity, straight contours, curvature variability and
total number of concavities at both scales, contour
overlap, typicality (p always at least £ .01), similarity
to the most similar other object...