Neurocase (1998) Vol. 4, pp. 311–338
Oxford University Press 1998
Are Living and Non-living Category-specific Deficits
Causally Linked to Impaired Perceptual or Associative
Knowledge? Evidence From a Category-specific Double
Dissociation
Matthew A. Lambon Ralph, David Howard1, Gemma Nightingale2 and Andrew W. Ellis3
MRC Cognition and Brain Sciences Unit, Cambridge, 1Department of Speech, University of Newcastle, 2Department of
Linguistics, University of York and 3Department of Psychology, University of York, UK
Abstract
Perhaps the most influential view of category-specific deficits is one in which the dissociation between
living and non-living kinds reflects differential reliance on, or weighting of visual or associative–functional
attributes. We present data collected from two patients, which question the apparent relationship between
category-specific deficits and loss of specific attribute types. One patient with dementia of Alzheimer’s type
presented with relatively poor performance on living things but failed to show a difference between
knowledge of visual and associative–functional information. The other patient with semantic dementia
demonstrated relatively poor knowledge of visual attributes but failed to exhibit a category-specific
impairment for animate kinds. In fact her comprehension and naming were slightly but significantly better
for living things. The data are discussed with reference to various theories of category-specific impairment.
We suggest that category-specific deficits for living things probably results from a combination of atrophy
to medial and neocortical temporal structures, including the inferior temporal lobe. It is proposed that at the
behavioural level, category-specific deficits arise when both critical identifying attributes of knowledge are
lost and the intercorrelation between features causes disintegration of the category such that each exemplar
‘regresses’ towards a category prototype.
Introduction
Since the publication of Warrington’s (1975) seminal paper
on the selective breakdown of semantic memory, there has
been a steady stream of papers which have reported various
category-specific deficits, that is relatively impaired performance for particular domains of knowledge. The focus
of our present paper is on the distinction between living
and non-living categories. The majority of patients who
show differences between the living and non-living categories have relatively spared comprehension and naming
of non-living or man-made objects but poor ability with
living or animate kinds. This deficit has been noted most
often for patients who have suffered from herpes simplex
virus encephalitis, HSVE (Warrington and Shallice, 1984;
Pietrini et al., 1988; Silveri and Gainotti, 1988; Sartori
and Job, 1988; Sartori et al., 1993a,b; Sheridan and
Humphreys, 1993; De Renzi and Lucchelli, 1994; Wilson
et al., 1995; Funnell and De Mornay Davies, 1996; Gainotti
and Silveri, 1996; Wilson, 1997), although this pattern has
also been reported after head injury (Farah et al.,
1989, 1991; Laiacona et al., 1993; Wilson, 1997), vascular
accidents (Hart et al., 1985; Farah and Wallace, 1992;
Sartori et al., 1993a; Howard et al., 1995; Forde et al.,
1997), dementia of Alzheimer’s type or DAT (Silveri et al.,
1991; Mauri et al., 1994; Montanes et al., 1995; Daum
et al., 1996; Tippett et al., 1996b; Gonnerman et al.,
1997; Garrard et al., 1998), and occasionally in semantic
dementia (Basso et al., 1988; Breedin et al., 1994a;
Barbarotto et al., 1995; Cardebat et al., 1996; Hodges et al.,
1998; Lambon Ralph et al., 1998). The opposite dissociation, better performance on living than non-living kinds,
has been reported in only a handful of studies, normally
as a consequence of cerebrovascular accident (CVA)
(Warrington and McCarthy, 1983, 1987, 1994; Hillis
et al., 1990; Hillis and Caramazza, 1991; Sacchett and
Humphreys, 1992; Breedin et al., 1994a), although it was
exhibited by a group of patients after anterior left temporal
Correspondence to: Dr M. A. Lambon Ralph, MRC Cognition and Brain Sciences Unit, 15, Chaucer Road, Cambridge CB2 2EF, UK. Tel: + 44 1223
355294; Fax: + 44 1223 359062; e-mail: [email protected]
312
M. A. Lambon Ralph et al.
lobectomy (Tippett et al., 1996a), by a single patient
with progressive atrophy of the left temporal and inferior
parietal lobes (Silveri et al., 1997) and occasionally by
individual patients with Alzheimer’s disease (Gonnerman
et al., 1997; Garrard et al., 1998).
Perceptual versus associative knowledge and the
living/non-living category-specific deficits
There are two main theories with regard to the breakdown
of semantic knowledge along the living/non-living
distinction (although there are at least four others; see
Discussion). The first of these is still, perhaps, the
dominant and most influential, namely that the dissociation arises from a selective breakdown of perceptual compared to functional/associative knowledge. Warrington and
Shallice (1984) were the first to suggest that living and
non-living items require rather different types of information to distinguish one exemplar from another. They
argued that man-made objects have ‘clearly defined functions’ and are primarily differentiated by their functional
properties, whereas animals tend to have few functions and
are distinguished in terms of their visual appearance.
Consequently, to differentiate the two broad categories into
their various exemplars, man-made items require detailed
knowledge about function, whereas animate kinds need
more perceptual information. If brain damage leads to a
loss of perceptual knowledge, then living categories will
tend to be affected more, or if there is degraded functional
knowledge then the opposite dissociation should result.
Warrington and Shallice suggested that a division along
perceptual/functional lines was preferable to truly
category-specific deficits as it gave an explanation for the
odd categories of items that tend not to follow the living/
non-living distinction. For instance, Warrington and
Shallice noted that their patients with relatively poor
knowledge of animals were also impaired with respect to
types of cloth and precious stone (which have only one
generic function and are differentiated primarily by colour
or texture) but not with respect to body parts (which have
very different functions).
The hypothesis that a differential weighting of perceptual
and associative/functional attributes underpins the dissociation between living and non-living kinds has been
taken up by various authors. Farah and McClelland (1991)
described a computational model of semantic memory in
which living and non-living items were coded across differing numbers of visual or functional units. They demonstrated that category-specific deficits emerged after selective
‘lesions’ to either the visual or functional units, in line
with Warrington and Shallice’s original hypothesis. Other
studies have supported the ‘differential-weighting hypothesis’ with the observation that the patients impaired for
living categories tend to exhibit poor knowledge about
visual but not functional/associative properties of items
(e.g. De Renzi and Lucchelli, 1994). Some patients demon-
strate an interaction between category and attribute type
resulting in relatively poor performance on tasks tapping
visual knowledge about animate objects (Sartori and Job,
1988; Silveri and Gainotti, 1988; Farah et al., 1989; Hart
and Gordon, 1992; Gainotti and Silveri, 1996; Forde et al.,
1997). For example, Gainotti and Silveri (Silveri and
Gainotti, 1988; Gainotti and Silveri, 1996) showed that
their patient was significantly less likely to provide the
appropriate label to a living thing than a man-made object
when the verbal definition stressed perceptual attributes. If
the definitions stressed functional information, there was
no significant difference between categories (the patient was
equally anomic for the animate and inanimate kinds). In a
thorough analysis of patient SRB, Forde et al. (1997)
gathered evidence favouring the hypothesis that SRB was
impaired on any task that required fine perceptual differentiation. They showed that in addition to exhibiting a
category-specific deficit for living things, SRB was also
impaired when required to name faces and subordinate
categories such as types of car. When asked to put names to
definitions containing either functional–associative or
visual–perceptual information, SRB was only impaired for
the latter type. Another category-specific patient, DM
(Humphreys et al., 1998), has shown the same pattern. It
is interesting to note that more detailed visual processing
is required for living items both for normal subjects
under brief presentation and in discrimination learning for
monkeys (Gaffan and Heywood, 1993).
In addition to the behavioural data in favour of the
differential-weighting hypothesis, some authors have noted
distinct neuroanatomical correlates (Saffran and Schwartz,
1994; Gainotti et al., 1995; Gainotti and Silveri, 1996).
Category-specific deficits for living things tend to be associated with temporal lesions typically involving medial and
inferior temporal areas, while the majority of deficits for
man-made items co-occur with lesions to fronto–parietal
regions (Warrington and McCarthy, 1983; Hillis et al.,
1990; Sacchett and Humphreys, 1992; Breedin et al.,
1994a), although sometimes with temporal involvement as
well (Warrington and McCarthy, 1987, 1994; Silveri et al.,
1997) or just with temporal atrophy (Hillis and Caramazza,
1991; Tippett et al., 1996a). These neural substrates are of
particular interest because they coincide with the ventral
and dorsal visual pathways (Ungerleider and Mishkin,
1982). Thus, a lesion to the inferior temporal lobe should
lead to impaired high-level visual processing and in turn, to
relatively poor performance for animate kinds.
Problems with the differential-weighting hypothesis
The differential-weighting hypothesis provides an appealing and relatively straightforward explanation of the living/
non-living category-specific literature. However, a closer
analysis of the data would seem to leave a number of
important questions unanswered (we concentrate primarily
on one of these below, but for a detailed critique of this
Category-specific deficits
approach see Caramazza and Shelton, 1998). We would
like to sound a note of warning with regard to the
demonstration of an association between category-specific
deficits and impaired perceptual knowledge. Not all
patients with category-specific deficits for animate kinds
exhibit a difference between perceptual and associative
knowledge. Caramazza and Shelton (1998) have noted that
more recent, tightly designed studies have not identified the
association. For example, Laiacona et al. (1993) reported
data from two patients with relatively impaired naming and
comprehension of living things. The patients completed a
semantic feature questionnaire that required access to
perceptual and associative attributes of living and nonliving concepts. Both were equally impaired on the two
attribute types although they still exhibited a significant
category-specific effect overall (for very similar results, see
Caramazza and Shelton, 1998; Moss et al., 1998; Samson
et al., 1998). In their re-analysis of patient JBR (originally
reported by Warrington and Shallice, 1984), Funnell and
De Mornay Davies (1996) also failed to find a difference
between visual and associative knowledge on a semantic
feature questionnaire. SB (Sheridan and Humphreys,
1993), a patient who had suffered from HSVE, demonstrated relatively poor naming and comprehension of
animals and foods in comparison to objects but did not
show any difference between a test of visual knowledge and
verbal-functional verification. Again, like Laiacona et al.’s
patients, SB exhibited a significant category effect in both
tasks (with normal performance for the objects). This
negative result has lead Humphreys and colleagues
(Sheridan and Humphreys, 1993; Humphreys et al., 1995,
1998; Forde et al., 1997) to suggest that category-specific
deficits can result from different functional loci (see also
Gainotti et al., 1995). We will turn to this possibility below
(see Discussion). One of the most problematic observations
for the differential-weighting hypothesis has been made by
Laws et al. (1995). They reported data collected from
patient SE who, after recovering from HSVE, exhibited the
characteristic deficit for animate kinds (and in this case
very near normal levels of naming and comprehension for
manmade items). However across a wide variety of tasks
designed to tap associative-functional and visual knowledge, SE was significantly worse at retrieving associative
than visual knowledge about animals and made few errors
for manmade objects. [It should be noted that Moss et al.
(1997) failed to replicate this pattern in SE. In stark
contrast to Laws et al., they found evidence for a selective
impairment for the visual properties of living things.]
If the comprehension and naming of animals is disproportionately reliant on visual knowledge or even on
increased levels of perceptual processing (Humphreys et al.,
1998), then category-specific deficits should be present in
naming from any modality. This has been shown in some
patients (e.g. Forde et al., 1997). Thus, the dissociation
should hold for naming to definition, even if the descriptions contain functional information alone because the
313
identification of animals, according to this theory, still
requires access to perceptual knowledge. However, patient
LA (Silveri and Gainotti, 1988; Gainotti and Silveri, 1996)
failed to show a category difference on naming to functional definition, while SRB (Forde et al., 1997) and DM
(Humphreys et al., 1998) were able to name the items to
functional definition (both animate and inanimate) as
accurately as control subjects.
Part of the attraction of the differential-weighting
hypothesis is that it provides a unified theory for both
living and non-living dissociations. As far as we are aware,
there has been no investigation of perceptual and associative knowledge in any of the ten cases showing category
specific deficits for non-living kinds (Warrington and
McCarthy, 1983, 1987, 1994; Hillis et al., 1990; Hillis and
Caramazza, 1991; Sacchett and Humphreys, 1992; Breedin
et al., 1994a; Silveri et al., 1997) or for patients with left
anterior temporal lobectomy (Tippett et al., 1996a). The
predicted pattern should be the opposite of living-specific
impairments, namely a deficit to functional–associative
attribute knowledge particularly pronounced for manmade kinds. This hypothesis remains to be tested; some
evidence related to the hypothesis will be presented in this
paper.
Cognitive and psycholinguistic differences between
living and non-living categories
An alternative hypothesis for the category-specific disorders is that they reflect the underlying cognitive and
psycholinguistic properties of the concepts involved. For
instance, animals, particularly foreign animals, tend to be
rated as less familiar than non-living things, and their
names are less frequent. This has lead some researchers to
argue that many apparent category-specific deficits might
be due to poor performance for relatively unfamiliar items
(Funnell and Sheridan, 1992; Stewart et al., 1992; Howard
et al., 1995; Tippett et al., 1996b). This theory also explains
the fact that many category-specific patients are also
impaired for musical instruments, but not for body parts.
Inspection of the familiarity ratings that accompany the
Snodgrass and Vanderwart picture set (Snodgrass and
Vanderwart, 1980) shows that musical instruments and
foreign animals have some of the lowest familiarity ratings
while body parts fall at the other end of the continuum.
Funnell and De Mornay Davies (1996, pp. 469–470) noted
that CW (Wilson et al., 1995) a category-specific HSVE
patient who was a professional musician, demonstrated the
characteristic deficit for living things, but his knowledge for
the category of musical instrument was commensurate with
that of other man-made objects.
There are now a number of reported cases in the
literature that show a significant category effect over and
above the influence of various cognitive and psycholinguistic variables (e.g. Farah et al., 1991, 1996; Barbarotto et al.,
1995; Gainotti and Silveri, 1996; Forde et al., 1997),
314
M. A. Lambon Ralph et al.
including the re-analysis of patient JBR by Funnell and De
Mornay Davies (1996). How many of the claims for
category-specific deficits would remain in the literature if
the studies had controlled for all the appropriate factors?
As familiarity, frequency, etc., affect performance in many
of these patients, these variables cannot be disregarded in
future category-specific cases. In fact, Funnell and De
Mornay Davies (1996) have argued that category-specific
deficits for living things might still be explained in terms of
familiarity if the variable included not only the frequency
but also the quality of interaction with each item (consider
for example the difference between the quality of our
knowledge about the domestic cat versus that of a lion).
The standard line against the cognitive/psycholinguistic
factor hypothesis is that it does not predict the opposite
dissociation (e.g. Warrington and Shallice, 1984; Breedin
et al., 1994a; Gainotti et al., 1995; Silveri et al., 1997). The
double dissociation only proves that the two patient groups
are different in some respect, however. This might be the
separation of living and non-living categories, although it
could be the influence of two different sets of variables
(Howard et al., 1995). For example, normal subjects’ and
aphasic individuals’ speed and accuracy on a variety of
lexical tasks are affected by the age at which people
typically learn the word (Morrison et al., 1992, 1997;
Nickels and Howard, 1994, 1995; Morrison and Ellis, 1995;
Ellis et al., 1996; Lambon Ralph et al., 1998). Age of
acquisition (AoA) is generally lower for animals than
objects (Howard et al., 1995; Morrison et al., 1997), so
some patients might show better performance on animate
kinds because they are relatively early-acquired items.
Howard et al. show that with frequency-matched sets of
animate and inanimate items, the animate items are also
higher in imageability. Some of the reported animacy
effects may be due to such confounds. If, for instance,
patient A was strongly affected by familiarity and patient B
by AoA or imageability, unmatched lists of living and
non-living objects could appear to reveal a double dissociation between the two categories. It would be interesting to know if the patient with progressive left temporal
atrophy reported by Silveri et al. (1997) with a categoryspecific naming deficit favouring animate kinds would still
exhibit the small but significant difference if AoA and
imageability were controlled for. Proper consideration of
the influence of different variables might also lead to a
greater number of non-living impaired case reports. If
familiarity and frequency also affect these patients, a
category-specific deficit for non-living items can be missed
because the impairment is ‘masked’ by the influence of the
cognitive and psycholinguistic variables.
DAT, semantic dementia and category-specificity
We have noted above the neuroanatomical correlates
associated with the living/non-living dissociations (Saffran
and Schwartz, 1994; Gainotti et al., 1995; Gainotti and
Silveri, 1996). In particular, the category-specific living
impairment seems to co-occur most often with medial and
inferior temporal damage (Gainotti et al., 1995; Gainotti
and Silveri, 1996), normally as a result of HSVE or CVA.
We report below data from two patients, one with semantic
dementia and the other with DAT. These two aetiologies
are very interesting with regard to category-specific deficits.
Dementia of Alzheimer’s type is most often associated with
a dense amnesia and at least a degree of semantic impairment (Hodges et al., 1990, 1992b, 1996; Hodges and
Patterson, 1995; Lambon Ralph et al., 1997), which probably results from atrophy that appears first in medial
temporal structures and then in the temporal neocortex
itself (Braak and Braak, 1991; Garrard et al., 1998). Given
the close relationship between the patterns of atrophy seen
in DAT and HSVE one might predict that category-specific
deficits should be common in DAT patients. There have
been a number of group studies that have tested this
prediction. Some have found no evidence for a categoryspecific deficit (Hodges et al., 1992b, 1996). Others have
found relatively poor performance for living things (Silveri
et al., 1991; Montanes et al., 1995; Daum et al., 1996).
Tippett et al. (1996b) replicated this category effect only
with a set of items that controlled for frequency and
prototypicality (the items from Silveri et al., 1991), but
found no effect when two further lists were administered
that also controlled for rated familiarity and visual complexity. Finally, one study has reported better classification
of living than non-living kinds (Montanes et al., 1996),
although the same researchers had previously found better
naming performance for non-living items (Montanes et al.,
1995). It is possible that these mixed findings reflect the
underlying heterogeneity that occurs in DAT, as although a
large number of patients present with amnesia and some
semantic impairment, this is not always the case (e.g. Croot
et al., 1996; Mackenzie Ross et al., 1996). Using a caseseries approach, two studies have been able to identify a
number of individual DAT patients with a category-specific
naming deficit for living things and occasional patients with
the reverse dissociation (Gonnerman et al., 1997; Garrard
et al., 1998), of whom one has come to autopsy showing
biparietal involvement (Garrard et al., 1998). In addition,
there has been a single case study which revealed a significant category-specific deficit for animate objects in naming
and visual knowledge tasks (Mauri et al., 1994).
Semantic dementia, or progressive fluent aphasia
(Hodges et al., 1992a; Snowden et al., 1989) is a disorder
associated with progressive atrophy of the anterior temporal lobes, particularly of the inferior temporal gyrus. The
atrophy typically involves the left side (Hodges et al.,
1992a, 1998), although the disease eventually leads to
bilateral damage (Hodges et al., 1998). The progressive loss
of temporal structures is associated with an inexorable loss
of knowledge about the meanings of words, objects and
concepts. The semantic deficit is present for all sensory
modalities (Snowden et al., 1989; Hodges et al., 1992a) and
Category-specific deficits
is accompanied by a profound anomia. The majority of
patients manifest features of surface dyslexia (although see
Cipolotti and Warrington, 1995). Despite sometimes severe
semantic impairment, the patients perform well on tests of
auditory–verbal short-term memory, non-verbal reasoning,
perceptual and spatial skills, have good single-word
phonology and syntax, and excellent day-to-day (episodic)
memory (Snowden et al., 1989; Hodges et al., 1992a),
although recent research has shown relatively poor recall of
events that occurred in the distant past (Graham and
Hodges, 1997).
We have already noted that previous studies have highlighted the role of inferior temporal structures in the
processing of visual knowledge and the importance this
might have in the representation of living kinds (e.g.
Gainotti et al., 1995). Most, if not all, of the published
category-specific cases for living things have involved
medial and neocortical temporal areas. Patients with
semantic dementia rarely present with atrophy to medial
temporal structures (Hodges et al., 1992a), so they might be
an ideal group of patients to test the apparent neuroanatomical correlation between poor knowledge of living
things and atrophy to the inferior temporal region alone.
There is growing evidence for the first part of the
differential-weighting hypothesis, namely that damage to
the inferior temporal lobes leads to poor visual knowledge.
Patient TOB (McCarthy and Warrington, 1988; Parkin,
1993) seemed to produce definitions to spoken words, or
circumlocutions to pictures that contained detailed functional knowledge with little sensory information. Detailed
analyses of word and picture definitions provided by nine
patients with semantic dementia have confirmed this
pattern (Lambon Ralph et al., unpublished data). Cardebat
et al. (1996) reported that their patient was able to produce
some semantic features regarding function but was unable
to draw either animals or objects. Patient DM (Breedin
et al., 1994b; Srinivas et al., 1997) exhibited a relative
preservation of functional over perceptual attributes and a
similar dissociation has been shown for patient NV (Basso
et al., 1988). Tyler and colleagues (Moss et al., 1995; Tyler
and Moss, 1998) have found significant semantic priming
for functional relationships but no priming for visual or
category associations.
There is little evidence to suggest that this poor visual
knowledge leads to category-specific effects, however,
which is even more surprising given the numerous reports
of frequency and familiarity effects in these patients
(Hodges et al., 1992a, 1994, 1995; Breedin et al., 1994b;
Graham et al., 1994; Patterson et al., 1994; Cipolotti and
Warrington, 1995; Funnell, 1995; Graham et al., 1995;
Hirsh and Funnell, 1995; Lambon Ralph et al., 1998).
Cardebat et al. (1996) found a category effect both in
naming and word–picture matching but it is unclear if the
items were matched for any of the relevant cognitive and
psycholinguistic variables. In a group study of nine
patients, Hodges et al. (1998) found a significant category
315
effect in word-to-picture matching (with items balanced for
familiarity only) but not in naming—although it is unclear
whether any individual patients showed a significant category effect. TOB (McCarthy and Warrington, 1988) was
reported to have a category effect in verbal definition and
NV (Basso et al., 1988) for verbal comprehension and
naming, although in both papers the items were controlled
for frequency alone. TOB’s naming was assessed by Parkin
(1993), who found no category effect if the stimuli were
controlled for frequency and familiarity. Breedin et al.
(1994a,b) found a category-specific effect for DM’s comprehension in favour of tools over animals (the items were
matched for frequency). However, a closer look at the data
reveals that DM was impaired on many other man-made
categories (e.g. vehicles) and he only exhibited the effect in
two out of the three administrations. In addition, on the
visual versus associative attribute test referred to above
(which was matched for frequency and familiarity), DM
showed a significant difference between attribute type but
no difference between living and non-living categories. In
an analysis of the factors which predict naming accuracy in
semantic dementia, Lambon Ralph et al. (1998) found
consistent effects of frequency, familiarity, and AoA across
the nine patients and for the group as a whole. Only one
out of the nine demonstrated a significant effect of category
over and above the influence of the other variables.
There is only one convincing report of a category-specific
deficit in a case of semantic dementia (patient MF:
Barbarotto et al., 1995). In word-to-picture matching,
naming and answering semantic attribute questions, MF
exhibited significantly poorer performance for living categories even when other factors were accounted for (including familiarity, frequency and item difficulty). In fact, for
the first two testing sessions in the longitudinal study, MF’s
scores for the non-living categories remained in the normal
range [i.e. there was a classical dissociation (Shallice, 1988)
between the two sets]. His drawing from memory also
indicated a strong category effect. However, unlike many
other semantic dementia patients, MF failed to show a
significant difference between perceptual and associative
knowledge on the semantic feature questionnaire. It may be
relevant to note that MF was different to other reported
cases of semantic dementia in that his temporal atrophy
involved medial structures including the hippocampus and
parahippocampal gyrus, particularly on the right.
We present two cases studies. The data for these two
patients are drawn from two rather unrelated programmes
of research. Consequently, the patients have not been
tested on exactly the same set of assessments. However,
there is sufficient overlap between the two for powerful
comparisons to be drawn. The first reports data collected
from a patient, DB, with DAT who presented with a
category-specific deficit for living items. This effect was
significant over and above the influence of familiarity,
frequency, visual complexity and AoA (amongst other
variables). On tasks designed to assess her associative and
316
M. A. Lambon Ralph et al.
perceptual knowledge, however, DB exhibited no difference
with regard to attribute type but a consistent category
effect alone. The second case, IW, was a patient with
semantic dementia resulting from progressive atrophy of
her left temporal lobe, including inferior temporal structures. She demonstrated a clear deficit for perceptual over
associative knowledge but in fact, her comprehension and
naming were significantly better for living categories. Thus
DB and IW form two dissociations: the first being a double
dissociation between living and non-living categories;
the second between a category-specific deficit for living
things without differential attribute knowledge on the one
hand, and differential knowledge without a living thing
impairment (in fact the opposite) on the other.
Case study 1: DB—category-specific impairment
without differential attribute knowledge
Background
DB, a right-handed woman, was born in 1910. During her
working life, she was employed in a variety of part-time
domestic posts while she brought up her four children. She
presented in 1993 with a severe recent memory deficit, poor
long-term memory, repetitious conversation and a degree
of anomia. Given the pattern of impairment, a diagnosis of
dementia of Alzheimer’s type (DAT) was made. An MRI
scan (see Fig. 1) revealed generalized cortical atrophy,
including the temporal neocortex and particularly to
medial temporal structures bilaterally. Unfortunately, due
to movement artefacts, we were unable to obtain precise
coronal sections and thus detailed analyses of the temporal
neocortex are impossible.
Although we have been studying DB longitudinally since
May 1995, our experimental investigation with regard to
her category-specific impairment was conducted between
October and December 1996. Her profound amnesia was
very apparent: she would repeat the same stories/
conversation within 5 min of each other, was unable to
remember when she had last been visited, what testing had
been done, and so on. Her poor memory was confirmed by
her scores on the Warrington Recognition Memory Test
(Warrington, 1984). She scored 30/50 on the faces (no
better than chance) and 35/50 on the words (5th percentile)
DB was disorientated in time and place. She achieved 21/30
on the Mini-Mental State Examination (Folstein et al.,
1975). Her forward digit span was five. Her category
fluency was also impoverished (DB produced a total of 40
items across eight categories; normal controls achieve a
mean of 113.7 items; SD 19.4). DB exhibited a mild
impairment on a synonym judgement task (Franklin et al.,
1992), both for concrete (37/50: control range 38–40) and
abstract words (35/40: control range 37–40). Her score of
39/52 on the all-picture version of the Pyramids and Palm
Trees Test (Howard and Patterson, 1992) was outside the
normal range (0–3 errors). DB’s spontaneous speech was
fluent with only occasional word-finding errors. Her
anomia was much more apparent in naming to confrontation. The vast majority of her errors were semantic and
they seemed to occur most often to the animate pictures
(see below). Our detailed analysis of this apparent
category-specific deficit is reported in the next two
sub-sections.
Object recognition and comprehension
DB’s impaired comprehension (on synonym judgement
and the Pyramids and Palm Trees Test) has already been
noted above. Table 1 shows her results on a variety of tests
involving picture stimuli. DB’s performance was outside
the normal range on all the tests except for the two
requiring the subject to match a picture from two different
view-points (minimal feature and foreshortened matching,
Riddoch and Humphreys, 1992). DB’s ability to identify
the silhouette of a real object (VOSP silhouettes;
Warrington and James, 1991) or pick out the real item from
an array of nonsense silhouettes (VOSP object decision;
Warrington and James, 1991) was impaired. Her score for
silhouette identification was significantly lower for the
animate than the inanimate items [÷2(1) 6.7, P = 0.01]. She
was also impaired in object decision when required to sort
real from chimerical objects and animals (Riddoch and
Humphreys, 1992; Graham, 1993). In these tests, DB
typically rejected the correct pictures suggesting that someone had drawn extra parts on another animal (e.g. to the
picture of a deer, DB indicated that someone had ‘stuck
things’ on a ‘horse’s’ head, or that a picture of a zebra was
a ‘horse’ with the ‘wrong colours’). Her score on the TOAD
(Graham, 1993) was significantly worse for the animals
than the objects [÷2(1) 7.44, P = 0.006]. Note the stimuli
in these tests are not controlled for any cognitive psycholinguistic factors.
The within-category word–picture matching tests
(Lambon Ralph, 1997) involve 100 trials, each containing
the target picture together with four category co-ordinate
items (e.g., a goat is presented with donkey, deer, horse and
cow). DB’s scores for the written and spoken versions were
both outside the normal elderly range and were significantly worse on the living than non-living items [spoken:
÷2(1) = 9.1, P = 0.003; written: ÷2(1) = 9.8, P = 0.002]. In
order to ensure that the category effect was not due to the
influence of any potential confounding factors, DB’s
word–picture matching accuracy (including responses from
written and spoken for items that have values on the
various cognitive–psycholinguistic variables) was entered
into a simultaneous logistic regression along with category,
AoA, object familiarity, imageability, phoneme length,
(log) spoken frequency (from the Celex Database, 1993)
and visual complexity. Category was the only factor to
have a significant independent effect on DB’s comprehension score (Wald = 4.05, P = 0.04).
Fig. 1. A horizontal (a) and coronal (b) section from DB’s MRI scan (see text for details).
Category-specific deficits
317
318
M. A. Lambon Ralph et al.
Table 1. DB’s performance on object recognition and comprehension tasks
Test
Living
items
Non-living
items
Total
Control data
BORB minimal feature match
BORB foreshortened matcha
VOSP silhouettesb
VOSP object decisionb
BORB object decisiona
—
—
3/15
—
86/112
—
—
10/15
—
14/16
24/25
23/25
13/30
13/20
100/126
Test of object and animal. Decision (TOAD)c
37/50
47/50
84/100
Within-category written word–picture matchingd
Within category spoken word–picture matchingd
38/48
38/48
52/52
51/52
90/100
89/100
18–25
16–25
>15/30
>14/20
106–126
(mean = 115, SD = 5.7)
89–100
(mean = 97, SD = 3.6)
96–100
96–100
a
Riddoch and Humphreys (1992); bWarrington and James (1991); cGraham (1993); dLambon Ralph (1997).
Naming to confrontation
Three naming tests were administered. In the first, DB was
asked to name 219 simple line drawings which have values
on various psycholinguistic variables (taken from Morrison
et al., 1997). She correctly named 49/84 (58%) of the living
items and 110/135 (81%) of the man-made objects [a
significant difference: ÷2(1)=13.9, P < 0.001]. Her predominant error type was semantic, including superordinate [e.g.,
pineapple<‘some sort of fruit’, snail<‘a creature’, owl<
‘a bird’] and category co-ordinate responses [gorilla<
‘monkey’, ladybird<‘fly’, goat<‘kangaroo’, tiger<‘lion’,
lion<‘sheep’]. The category effect was assessed against the
influence of any potential confounding factors by entering
DB’s naming accuracy into a simultaneous logistic regression along with category, AoA, object familiarity, imageability, phoneme length, (log) spoken frequency (from the
Celex Database, 1993) and visual complexity. Category
(Wald = 5.33, P = 0.02) AoA (Wald = 4.98, P = 0.03)
and object familiarity (Wald = 10.6, P = 0.001) were all
found to have significant independent effects on DB’s
naming accuracy (no other variables reached significance).
The second naming test was taken from Gainotti and
Silveri (1996). It contains items matched pairwise for
familiarity and frequency. The results are shown in Table 2.
Again the majority of DB’s errors were semantic in
nature (15 semantic: e.g. eagle<‘pigeon’, fly<‘wasp’,
tomato<‘apple’; one no response; one visual error:
flute<‘curtain rail’) DB’s scores were significantly better
for man-made than living items [÷2(1) = 9.9, P = 0.002]
and were higher on the more familiar items [÷2(1) = 6.6,
P = 0.01]. If the sets are split into low and high familiarity
groups (at the median value), the animacy effect was
particularly pronounced in the low familiarity set [low
familiarity: ÷2(1)=11.0, P = 0.001; high familiarity: ÷2(1)
< 1, n.s.]. When the interaction is analysed using an exact
text for the homogeneity of odds ratios (Zelen, 1971;
Mehta and Patel, 1995), the null hypothesis (of equal odds
ratios for the effect of animacy in the high and low
familiarity set) cannot be rejected (exact P = 0.36),
although there is a significant effect of category overall
Table 2. DB’s naming performance on the items from Gainotti and Silveri
(1996) split at the median familiarity value
Low Familiarity ( < median)
High Familiarity ( > median)
Total
Living
Non-living
Total
4/15
12/15
16/30
13/15
14/15
27/30
17/30
26/30
43/60
Table 3. DB’s naming performance on the 64 animal and artefact
matched-item set
Familiarity < median
Familiarity > median
Total
Living
Non-living
Total
7/16
12/16
19/32
14/16
15/16
29/32
21/32
27/32
48/64
(odds ratio = 9.98, P = 0.001). Two observations are
worthy of note. First, like patients LA (Gainotti and
Silveri, 1996) and JBR (Funnell and De Mornay Davies,
1996), DB’s category effect seemed to be most pronounced
for the low familiarity group (although the interaction, at
least for DB, was not significant). Second, the three errors
DB made to man-made objects were all to musical instruments (flute<‘curtain rail’, trumpet<‘musical instrument
to blow through’ and guitar<‘violin’).
The third naming test was constructed from a subset of
the 219 items used in the first naming task. Thirty-two
animals and the same number of man-made objects (no
musical instrument) were matched pairwise for object
familiarity, spoken frequency (natural and log values),
phoneme length and imageability. The items were split at
the median familiarity value (low: 1.55–2.57; high 2.59–
4.09). The results are shown in Table 3.
DB’s errors were nearly all semantic (15/16: e.g.
camel<‘animal’,
gorilla<‘monkey’,
bat<‘butterfly’,
whale<‘fish’). The one exception was a circumlocution to
the picture of a tractor. Her scores were very similar to
those found in the previous matched naming test. DB’s
Category-specific deficits
naming was significantly worse for the animals than the
man-made objects [÷2(1) = 8.3, P = 0.04], and there was a
trend in favour of the familiar items [÷2(1) = 3.0, P = 0.08].
If the data are split at the median familiarity value, the
animacy effect was only significant in the low familiarity set
[low: ÷2(1) = 6.8, P = 0.009; high: ÷2(1) < 1, n.s.], although,
again, the interaction was not significant (test for the homogeneity of odds ratio, exact P = 1.0; odds ratio for category
overall = 6.91, P = 0.004).
319
Table 4. DB’s performance on the definition to word matching task
Animals
Artefacts
Total
Perceptual
definition
Associative
definition
Total
11/16
(69%)
12/12
(100%)
23/28
(82%)
11/16
(69%)
12/12
(100%)
23/28
(82%)
22/32
(69%)
24/24
(100%)
46/56
(82%)
Tests of visual and associative knowledge
DB’s comprehension and naming was significantly worse
for living than non-living kinds, even when the effect of
various factors was taken into account. Given her categoryspecific deficit, did DB exhibit differential loss of perceptual
over associative–functional knowledge? The next section
documents the results of various assessments designed to
test this hypothesis.
DB was asked to name 40 simple line drawings and to
indicate (from a choice of ten) the correct colour
(Graham, 1995). DB named 27/40 correctly (all errors were
semantic: e.g., tomato<‘apple’, crocodile<‘fish’), but her
ability to identify the correct colour was even worse, 18/40
correct (normal control mean = 34.8, SD = 2.4; from
Graham, 1995). DB’s poor performance on object decision,
silhouette recognition and colour matching could be taken
as evidence of poor visual processing and thus her
category-specific deficit would be unsurprising. However,
in other tasks where visual and non-visual attribute knowledge is compared, DB failed to demonstrate the difference
that is predicted by the differential-weighting hypothesis
(she appeared to be equally impaired on both types of
knowledge).
DB was given three tests designed to tap knowledge of
associative–functional and perceptual knowledge of living
and non-living things. The first was a revised version of a
naming to definition task (Gainotti and Silveri, 1996).
Rather than requiring an anomic patient to name perceptual or associative definitions, DB was presented with a
written definition (which was also read aloud to her)
together with five written names, the target plus four
‘semantic’ foils (the names of other target items within the
same living or non-living domain). DB was required to
pick the name which best matched the description. The
test retained the same items as the original version but
the definitions were redrafted so that only perceptual or
associative attributes were included within the relevant
descriptions (the definitions, targets and foils are shown in
Appendix 1). It might be suggested that DB’s poor performance on the animate items was due to a confound with
familiarity, or that IW’s poor matching by visual description might be an accentuation of a normal effect. Clearly
this test would be better if it contained equal numbers of
living and non-living things, matched pairwise for all the
relevant factors (instead, we opted to use the same target
Table 5. DB’s mean (and standard deviation) score per item on the
Hodges et al. (1992a) semantic feature question test
Animals
Artefacts
Perceptual
questions
Associative
questions
2.7 (1.0)
2.8 (1.0)
2.7 (1.0)
3.6 (1.0)
Maximum score per item = 4.
Table 6. DB’s mean (and SD) score per item on the semantic feature
questions for the 64 matched-item set
Animals
Artefacts
Perceptual
questions
Associative
questions
3.2 (1.0)
3.7 (0.5)
2.9 (1.1)
3.6 (0.9)
Maximum score per item = 4.
items as those used previously by Gainotti and Silveri,
1996). If the data from DB and IW are considered together,
however, it seems unlikely that DB’s poor performance is
solely due to the effect of familiarity (IW’s performance in
general was strongly affected by familiarity), while if visual
knowledge is relatively harder for normal subjects, we
would have expected to see poor matching by visual
description for both IW and DB.
DB’s scores are shown in Table 4. She was able to match
definitions to names for all 24 non-living items but could
only manage 69% for the animals [a significant difference:
÷2(1) = 7.1, P = 0.008]. There was no effect of definition
type on her performance.
DB was asked to complete two semantic feature
questionnaires. The first includes four perceptual (yes/no)
questions and four associative questions for each of 48
items (24 living, 24 non-living, see Hodges et al., 1992a).
The second set was constructed along identical lines, but
included the 64 matched items used previously to test DB’s
naming to confrontation (see Table 3).
DB’s mean score per item (and standard deviation)
for the two tasks are shown in Tables 5 and 6. On the
items from Hodges et al. (1992), DB exhibited a main effect
of attribute type [F(1,46) = 5.7, P = 0.02], category
320
M. A. Lambon Ralph et al.
[F(1,46) = 6.3, P = 0.02] and an interaction [F(1,46) = 7.0,
P = 0.01], which was due not to poor perceptual information about animals but to relatively good knowledge of
the associative information about artefacts. This result may
be due, at least in part, to the unbalanced nature of the 48
items (the mean object familiarity of the 24 living was 2.28,
versus 3.25 for the 24 artefacts, even though it includes
six musical instruments). On the 64 matched-item set, DB
demonstrated significantly better performance on the
man-made items [category: F(1,60) = 10.6, P = 0.002] but
no effect of attribute type [F(1,60) = 2.4, P = 0.13] nor any
interaction [F(1.60) < 1].
To test DB’s visual knowledge further, we asked her to
draw each of the 64 matched-item set under three different conditions: immediate copy (with target present)
delayed copy and drawing from memory. The delayed
copying technique has been used before (Franklin et al.,
1987) with a patient who was suffering from semantic
dementia. Copying a picture after a delay should require
the subject to use semantic as well as ‘visual’ memory.
Consequently, any impairment to semantic knowledge
should be highlighted in this task (presumably it will be
most sensitive to the status of visual semantic attributes).
In the delayed copying condition, DB was asked to draw
the target picture after it had been removed from sight
(she had been given an unlimited time to study the
picture). For this patient, no delay was required before
seeing a deterioration in her drawing output (perhaps
because of her dense amnesia), although other patients
may required a specific delay (see patient IW below). If
the three conditions are administered together, a number
of methodological problems can be overcome. The immediate copying provides a baseline against which the
other conditions can be measured (e.g., poor draughtsmanship). Drawing from memory requires the subject
both to understand the spoken label and then to produce
a picture. Delayed-copying on the other hand requires no
verbal input or output (and thus one should be able to
eliminate the effects of poor verbal comprehension on
impoverished drawings from memory).
Five example drawings from each of the conditions are
shown in Fig. 2. DB’s immediate copies were the best, while
the delayed copying revealed a clear impairment which was
enhanced when DB was required to draw from memory.
Fifteen naïve subjects were asked to rate, on a scale of 0–6,
how good a representation of the intended target the
picture was (the raters were unaware of which condition
the picture came from and only rated a picture in one of the
three conditions—thus each picture was rated by five
subjects per condition).
The deterioration across conditions was confirmed by
these ratings (see Table 7). A three by two split-plot
ANOVA confirmed a main effect of drawing condition
[F(1,120) = 146.5, P < 0.001] but there was no effect of
category [F(1,60) < 1] and no interaction [F(1,60) < 1].
Each pairwise comparison between drawing conditions was
Fig. 2. Examples of DB’s drawings in each condition.
Table 7. The mean (and SD) rating per item for DB’s drawings of the 64
matched-item set in each condition
Animals
Artefacts
From
memory
Delayed
copying
Immediate
copying
2.29 (1.6)
2.23 (1.6)
2.89 (1.6)
3.23 (1.6)
5.36 (0.8)
5.41 (0.7)
Range of scores per item = 0–6.
significant [immediate versus delayed copying: t = 13.1,
P < 0.001; delayed copying versus drawing from memory:
t = 2.6, P = 0.01; immediate copying versus drawing from
memory: t = 15.2, P < 0.001].
Although there was no category effect present in the
rated accuracy of DB’s pictures, there were qualitative
differences between the categories. The majority of the
animal targets (29/32) were drawn as a though the whole
category had ‘regressed to the mean’, i.e. these items were
drawn with the outline of a highly familiar prototype such
as a cat, dog or horse (see Fig. 3). This ‘prototype effect’
included a number of items that have either two or no legs
[e.g. duck and whale—for the picture of a whale DB
commented, ‘four legs, so it can get out (of the water)’]. For
the remaining items DB produced a semantic drawing error
(e.g., bat<‘bird’, ladybird<‘fly’, caterpillar<‘worm’). The
artefacts were impoverished but were drawn with quite
Category-specific deficits
321
Table 8. The mean (and SD) rating per item for DB’s verbal definitions of
the 64 matched-item set
Familiarity < median
Familiarity > median
Animals
Artefacts
3.3 (1.5)
4.1 (1.1)
4.2 (1.9)
5.2 (0.8)
Range of scores per item = 0–6.
Table 9. The mean number of correct and intrusive (incorrect) perceptual
and associative features produced by DB per item
Correct
Perceptual
Associative
Fig. 3. Examples of DB drawings exhibiting prototypical features for the
animals alone.
different features. We shall return to this observation below
(see Discussion).
The final test administered required DB to give verbal
definitions for each member of the 64 matched-item set. DB
was encouraged to give as much information about each
item as she could. When she failed to produce information
about a certain aspect of the target concept (e.g. what it
looked like), she was given various probe questions including associative–functional and perceptual queries (e.g.
where is it from?, does it walk, fly or swim?, how many legs
does it have?, what colour is it?, what does it eat?, where is
it found?, who uses it?, what is its purpose?). The definitions (a combination of her spontaneous production and
responses to specific questions) were analysed in two ways.
First, five independent subjects were asked to rate the
quality of each definition on the same scale as that used for
the pictures. The mean (and standard deviation) rating
is shown in Table 8. DB’s definitions received higher
ratings for the artefacts than for the animals [category:
Intrusions
Animals
Artefacts
Animals
Artefacts
2.8
2.2
1.7
2.4
1.5
0.6
0.2
0.1
F(1,60) = 7.9, P = 0.007] and for the more familiar kinds
[F(1,60) = 6.2, P = 0.02]. The interaction between these
two factors was not significant [F(1,60) < 1].
In addition, for each item the number of perceptual
and associative–functional attributes produced by DB
were counted. These were split into correct and intrusive
(incorrect) features. The mean number per item is shown in
Table 9.
DB produced more correct information for animals than
artefacts [category: F(1,60) = 3.8, P = 0.05]. There was no
main effect of attribute type [F(1,60) < 1], although there
was a significant interaction between category and attribute
type [F(1,60) = 11.6, P < 0.001], which was due to DB
producing the most information for the visual characteristics of animals. This is surprising, perhaps, given her deficit
for living things on all other tasks. It is possible that normal
subjects might produce even more visual information for
animals in their definitions and so, by comparison, DB
would have a selective deficits for the animate items. DB
also gave more intrusive information for animals than
artefacts [category: F(1,60) = 24.4, P < 0.001] and produced a greater number of visual attribute intrusions
[attribute: F(1,60) = 12.9, P = 0.001], particularly for the
animate items [attribute category: F(1,60) = 7.4,
P = 0.008]. The accuracy ratings from the independent
subjects appeared to be sensitive to both correct (correlation between rated accuracy and number of correct
features per item; r = 0.37, P = 0.003) and intrusive information (correlation between rated accuracy and number of
intrusions per item; r = 0.53, P < 0.001).
Summary
DB, a patient with DAT, presented with a category-specific
naming and comprehension deficit for living things. The
category effect was found to be significant over and above
the effects of other cognitive–psycholinguistic variables,
322
M. A. Lambon Ralph et al.
Table 10. IW’s neuropsychological profile
Test
IW’s score
Normal controls
Boston Naming Testa
Pyramid and Palm Trees Testb
Shallice & McGill’s
Abstract word–picture matchingc
7/60
45/52
14/30
7/15j
8/30j
84/128
59/128
66/80
29/36
64
98
Passed all
Passed all
36/36
43/50
35/50
49–59
49–52
—
—
—
—
—
—
50th percentile
—
—
—
—
—
58th percentile
5th percentile
‘Surface’ Reading Listd
TROGe
Raven’s Progressive Matricesf
WAIS-R
VOSPg
Rey Figure copyh
WRMTi
Concrete
Emotion
Abstract
Regular
Exception
Verbal
Performance
Recognition subtests
Spatial subtests
Faces
Words
a
Kaplan et al. (1976); bHoward and Patterson (1992); cunpublished test; dPatterson and Hodges (1992); eBishop (1989); fRaven (1962); gWarrington and
James (1991); hOsterrieth (1944); iWarrington (1984).
j
No better than chance.
including familiarity, frequency, visual complexity and
age-of-acquisition (AoA). The differential-weighting
hypothesis predicts a greater loss of perceptual than
associative–functional information. This hypothesis was
not supported by DB’s data. DB’s performance on various
tasks designed to tap specific properties was significantly
worse for living than non-living items, but the assessments
failed to find a difference between perceptual and
associative–functional attribute knowledge.
Case study 2: IW—greater loss of perceptual than
attribute knowledge with a category-specific impairment for man-made objects
Background
IW, a right-handed female (although she had always
written with her left hand) was born in 1941. She had been
employed in a variety of retail companies including
managing a successful florist’s in London. IW presented in
1994 after a year or two of ‘poor memory’. Medical
examination at that time noted considerable word-finding
difficulties for both common objects and people’s names. A
MRI scan revealed significant localized atrophy of the left
temporal lobe. Speech therapy assessment confirmed IW’s
expressive dysphasia but also noted an auditory comprehension deficit together with a degree of surface dyslexia
and dysgraphia.
Our investigation began in 1996. A MRI scan taken at
this time showed significant atrophy of the left temporal
lobe, most pronounced at the pole, which included significant reduction of the inferior temporal gyrus. IW’s
right temporal lobe and other cortical structures, including the medial temporal area, appeared to be intact (see
Fig. 4).
IW’s neuropsychological profile is summarized in Table
10. The pattern matches that of semantic dementia (see
Introduction). IW was profoundly anomic and had a
semantic impairment when measured from any modality.
IW’s comprehension of words was significantly worse than
that of pictures. The modality difference has been noted in
other patients with semantic dementia (McCarthy and
Warrington, 1988; M.A. Lambon Ralph et al., unpublished
data). Our detailed investigation of the nature of this
difference is beyond the scope of the present article but it
will be the subject of a future paper.
Although not disorientated in time or place, she scored
poorly on the MMSE (22/30) (Folstein et al., 1975) because
of her expressive and receptive language deficits. She was a
surface dyslexic (better reading of words with regular than
exceptional spelling-to-sound correspondences). Her single
word phonology and syntax were relatively well preserved,
as were her non-verbal, perceptual and spatial skills. Like
other patients with semantic dementia, IW’s day-to-day
memory was extremely good. Our initial assessment also
highlighted an apparent dissociation between visual and
associative knowledge (relatively poor recall of visual information). The formal investigation of this dissociation is
reported in the next section.
Tests of visual and associative knowledge
IW was given the same colour matching test (from
Graham, 1995) as DB. She named 13/40 of the items
correctly (her predominant error type were no response
and attempted circumlocutions) and indicated the correct
colour for 27/40 (normal control mean = 34.8, SD = 2.4;
from Graham, 1995). IW’s performance was very similar to
that found by Graham (1995) in three other patients with
semantic dementia.
Fig. 4. Coronal sections from IW’s MRI scan.
Category-specific deficits
323
324
M. A. Lambon Ralph et al.
Table 11. IW’s performance on the definition to word matching task
Animals
Artefacts
Total
Perceptual
definition
Associative
definition
8/16
(50%)
5/12
(42%)
13/28
(46%)
13/16
(81%)
9/12
(75%)
22/28
(79%)
Table 13. (a) The mean (and SD) number of perceptual and associative
features produced in verbal definition by IW (all items)
Total
21/32
(66%)
14/24
(58%)
35/56
Perceptual
Associative
Animals
Artefacts
Total
Associative
questions
2.5 (1.2)
2.3 (1.0)
2.4 (1.1)
2.6 (0.9)
3.3 (0.8)
2.9 (0.9)
Artefacts
(n = 35)
Total
0.5 (0.8)
1.7 (1.2)
0.2 (0.6)
1.4 (1.4)
0.3 (0.7)
1.5 (1.3)
(b) The mean (and SD) number of perceptual and associative features
produced in verbal definition by IW (including only items for which IW
gave at least some information)
Table 12. IW’s mean (and SD) score per item on the Hodges et al. (1992a)
semantic feature question test
Perceptual
questions
Animals
(n = 15)
Perceptual
Associative
Animals
(n = 14)
Artefacts
(n = 24)
Total
0.6 (0.9)
2.2 (0.9)
0.3 (0.7)
2.0 (1.2)
0.4 (0.8)
2.1 (1.1)
Maximum score per item = 4.
IW’s scores on the definition to word matching task, that
contrasts perceptual and associative definitions, are shown
in Table 11. IW was significantly better when matching
associative than perceptual definitions to words (binomial,
P = 0.02). Despite showing poorer performance on the
perceptual definitions, she was actually slightly better for
the animals than artefact items (although the difference
failed to reach significance).
IW’s performance on the Hodges et al. (1992a) semantic
feature questions is shown in Table 12. IW was significantly
better when answering the associative than perceptual
questions [F(1,46) = 7.6, P = 0.008] but there was no effect
of category [F(1,46) = 1.4, P = 0.24] and her marginally
better performance for the associative questions about the
artefacts approached significance [interaction: F(1,46) =
3.6, P = 0.06].
During the course of a wide variety of testing, IW was
asked to give a definition to 50 spoken words that contained 15 animals and 35 man-made objects. Like DB, IW
was given various prompting questions that included
probes for perceptual and associative–functional knowledge. The mean number of each attribute per item is given
in Table 13a. Although IW only gave small amounts of
information, she nevertheless provided a significantly
greater number of associative than perceptual attributes
[F(1,48) = 45.5, P < 0.001], but was unaffected by category
[F(1,48) = 1.1, n.s.] and did not exhibit an interaction
[F(1,48) < 1]. IW’s poor verbal comprehension has already
been noted, and in this test, involving the definition of
spoken words, she said that she had no idea what a word
meant on 12 trials—1/15 trials with animate items, and
11/35 trials with artefacts. Table 13b shows the mean
number of attributes given per item only for those words
Fig. 5. Examples of IW’s drawings from memory.
for which IW gave at least some information. The pattern
is unchanged: a significant effect of attribute in favour
of associative information [F(1,34) = 77.8, P < 0.001] but
no effect of category [F(1,34) < 1] or an interaction
[F(1,34) < 1]. Unlike DB, IW produced very little intrusive
information.
IW was asked to draw two sets of items (n = 50 and
n = 64) from memory. In order to assist the comprehension
of the spoken name, each item was accompanied by an
associative description (resulting in a drawing for all but
three items). Some example drawings are shown in Fig. 5.
IW’s drawings did not exhibit the same prototypical effect
as demonstrated by DB, rather her pictures tended to miss
out features. Interestingly, in both definition and drawing
tasks IW would complain that she could not either ‘visualize’ the word or ‘remember’ a part of an item (for drawing).
In order to assess whether IW did omit common features,
four aged-matched controls were asked to sketch in
< 1 min, each of the 114 pictures. Two judges used these
pictures to create a list of features that were included by all,
or nearly all, the control subjects (up to four features per
Category-specific deficits
325
Table 14. IW’s object recognition performance
Test
IW
Control data
BORB minimal feature matcha
BORB foreshortened matcha
BORB object decisiona
25/25
25/25
121/126
Test of object and animal
Decision (TOAD)b
VOSP silhouettesc
VOSP object decisionc
Head’s testd
Forced-choice object decisione
98/100
18—25
16–25
106–126
(mean = 115, SD = 5.7)
89–100
(mean = 97, SD = 3.6)
> 15/30
> 14/20
26–32
43–50
17/30
17/20
32/32
48/50
a
Riddoch and Humphreys (1992); bGraham (1993); cWarrington and
James (1991); dRiddoch and Humphreys (1987); eunpublished test, see
text.
Fig. 6. Examples of IW’s delayed copying.
picture) despite the wide variability in their drawing skill.
This scoring scheme was then applied to IW’s drawings
from memory. In the first set the normal subjects produced
178–183 of the 186 features (96–98%). IW only produced
107 (58%). The same pattern was found on the second set:
normals produced 228–231 of the 236 items (97–98%),
while IW produced 119 (50%).
Unfortunately, IW was only given a limited number of
items to copy after a delay. A few examples are shown in
Fig. 6. Unlike DB, IW required up to a minute delay before
her poor visual knowledge began to influence her delayed
copying. However, a similar pattern to that observed in her
drawing from memory emerged. Rather than sketching
very similar pictures of the animals (as in DB’s prototypical
drawings), IW’s pictures demonstrated a loss of visual
features. We shall return to this difference between the two
subjects and their different performance on the definition
task below (see Discussion).
Object recognition
IW was given a number of object recognition tests (the
majority of which DB had also completed, see Table 1) the
results of which are shown in Table 14.
Despite her poor visual knowledge in the drawing and
definition tasks described above, IW completed all the
object recognition assessments achieving in each case scores
slightly above the mean for control subjects. To test her
object recognition even further, IW was given the Head’s
Fig. 7. An example item from the two-alternative forced choice object
decision test.
test (Riddoch and Humphreys, 1987) in which the subject is
required to match the correct head (from four alternative)
to the target body. IW completed this test without error. In
an attempt to make the most sensitive object decision task
possible, IW’s typical drawing omission errors were used
(after redrawing) as the foil in a two-alternative forcedchoice object decision test (an example is shown in Fig. 7).
IW’s performance was as good as university postgraduate
subjects.
Although not the focus of our paper, it is interesting to
note IW’s excellent object decision performance despite her
poor knowledge of visual attributes. This finding runs
counter to the hypothesis that object decision is solely
reliant on the integrity of centrally stored ‘semantic’ knowledge. It is, however, consistent with the hypothesis that
the processes involved with object decision interact with
semantic information (Hodges et al., 1994) such that object
decision performance declines when there is more severe
semantic impairment (see also Hodges et al., 1992a). Other
authors have suggested that object recognition can be
dissociated entirely from semantic knowledge (Riddoch
and Humphreys, 1987), even in cases of semantic dementia
(for a detailed analysis of this issue in a patient with
semantic dementia, see Srinivas et al., 1997). We will be
able to assess these two competing theories as IW’s visual
knowledge deteriorates over time.
326
M. A. Lambon Ralph et al.
Assessment of category-specificity in naming and
comprehension
Table 15. The effect of category on IW’s naming accuracy
The differential-weighting hypothesis predicts that, given
the differential loss of visual knowledge exhibited by IW, a
category-specific deficit should result. This hypothesis was
tested and the results are shown in the next section.
IW’s naming was initially assessed using the same 219
picture set as that administered to DB. She only managed to
name 85 items (39%) correctly. When she could not name
the target picture she tried to give a circumlocution (54%) or
indicated that she could not ‘remember’ the name (31%). IW
also produced a small number of semantic errors (15%: e.g.,
knife<‘butter’, cabbage<‘vegetable’, glove< ‘handbag’).
IW’s naming accuracy was compared to the same seven
variables used in the analysis of DB’s naming (category,
objective AoA, object familiarity, imageability, phoneme
length, (log) spoken frequency and visual complexity).
There were significant simple correlations between IW’s
naming and all the variables except imageability and, interestingly, category. In order to assess the independent effects
of each variable, the items and predictors were entered into
a simultaneous logistic regression (as with DB). The regression revealed significant independent effects of object
familiarity (Wald = 22.1, P < 0.0001), (log) frequency
(Wald = 6.66, P = 0.01), AoA (Wald = 12.6. P = 0.0004)
and better naming of living than non-living objects
(Wald = 4.57, P = 0.03). The difference between the simple
correlations and regression results highlights the importance
of controlling for psycholinguistic factors even in cases with
a category-specific deficit for man-made items (see Introduction). IW appeared to have no category-specific deficit in the
simple correlations because familiarity and frequency also
affected her naming. These two variables tend to favour
man-made items, thus masking the category-specific deficit
(note that patients that are affected by familiarity and frequency but do not have a category-specific deficit can show
a significant simple correlation between naming and category in favour of non-living items which is removed once
familiarity and frequency is controlled for; see Howard
et al., 1995; Nickels and Howard, 1995).
During the course of our study, IW has been given a
number of naming to confrontation tests. In order to
investigate further the factors that influence her naming
accuracy we carried out simultaneous logistic regression
analyses for each individual test and a meta-analysis of the
entire naming corpus (965 naming attempts to the 765
items in six different tests). The analyses only included
items for which values were available on the following
factors: frequency (Celex count), length, rated AoA, rated
word familiarity, rated imageability and animacy (reducing
the original set to 777 naming attempts to 313 different
items). The results with regard to the difference between
living and non-living sets are shown in Table 15.
Across the six tests, IW failed to show a significant
difference between animate and inanimate pictures on three
Naming test
Animate
Inanimate
Odds
ratiob
P valueb
(one tailed)
100 picture naming
192 item set
Nickel’s naming
Animacy set
219 York items
64 matched items
Combined data
0.52
0.42
0.35
0.37
0.42
0.39
0.42
0.31
0.29
0.35
0.3
0.43
0.28
0.33
8.3c
3.5c
1.0
1.8
2.7c
0.3
2.3c
0.0001c
0.02c
n.s.
n.s.
0.04c
n.s.
0.0006c
Raw proportion correcta
a
The proportion correct before any factors are taken into account.
These values are calculated after the effects of the other psycholinguistic
factors have been taken out.
c
The effect is in favour of naming animate items (see text).
b
occasions and on the remainder demonstrated better naming performance for the animate sets. When the items were
combined into a meta-analysis, IW’s naming was found to
be significantly better for the living pictures. In addition,
IW exhibited significant independent effects of object
familiarity, frequency, reverse length (better naming of
pictures with longer names) and AoA. If objective AoA
was substituted for rated AoA, and if visual complexity was
taken into account, the results did not alter.
IW’s comprehension was tested using the withincategory word–picture matching task (Lambon Ralph,
1997) described before for DB. IW scored 82/100 on the
spoken version and 84/100 on the written (normal controls:
96–100). A simultaneous logistic regression analysis of the
factors that predicted her combined word–picture matching
performance (including the same factors as for DB)
revealed significant independent effects of (log) frequency
(Wald = 5.8, P = 0.02) visual complexity (Wald = 5.3,
P = 0.02) and a trend in favour of pictures with longer
names (Wald = 3.4, P = 0.07). There was no effect of
category (Wald = 1.6, n.s.).
IW’s comprehension and naming performance was
assessed, in more detail, across a set of 30 animate items
and 30 artefacts (matched as far as possible for imageability, concreteness, length, frequency, operativity,
familiarity and rated AoA). On separate occasions IW was
asked to define the spoken or written name and to name the
picture (task was counterbalanced across the three sessions). The whole procedure was repeated 6 months later.
IW’s production was scored correct if she named the
picture or if item-specific semantic information was offered
in her definition. The basic data are shown in Table 16.
The effect of category was assessed for the three tasks
using simultaneous logistic regression analyses containing
phoneme length, (log) frequency, familiarity, imageability,
rated AoA, category, and task. IW’s overall performance
was significantly affected by phoneme length (in favour
of longer words: P = 0.001), (log) frequency (P < 0.001),
familiarity (P = 0.002), imageability (P = 0.007), rated
AoA (P = 0.001) and category (in favour of the animate
Category-specific deficits
327
Table 16. The effect of category on IW’s naming and definition performance (raw data)
Picture naming
Animate
Artefacts
Definition of spoken word
Definition of written word
N
Sept 1996
April 1997
Sept 1996
April 1997
Sept 1996
April 1997
30
30
11
9
10
7
13
12
15
7
19
10
19
10
items: P = 0.004). The P values reported here are all
two-tailed. IW’s accuracy may have been affected by different variables in each of the three tasks. This possibility
was analysed by including an interactive term between task
and each variable in the regression equation. When all the
interactive terms were included there was a non-significant
increase in the fit of the model (log likelihood increased
from 179.76 to 195.64: an increase of 15.88, df = 12,
P = 0.20). When the interactive terms were entered
separately, none made a significant contribution (all
P > 0.10). These results suggest that the effects of the
variables were the same in each task, and show that, IW’s
comprehension of both spoken and written words, like
her naming, was significantly better with animate than
inanimate items.
Summary
IW was a patient with semantic dementia resulting from
profound atrophy of her left temporal lobe, most pronounced at the temporal pole. Detailed assessments
revealed a greater loss of visual than associative–functional
knowledge. Despite this dissociation, IW did not exhibit
the predicted category-specific loss for animate things. In
contrast, IW demonstrated better performance on living
kinds than on artefacts in the majority of comprehension
and naming assessments.
Discussion
We have reported data from two patients. DB, a woman
with dementia of Alzheimer’s type, presented with a
category-specific impairment for living things in object
decision, comprehension and naming tasks. Her knowledge
of perceptual and associative attributes about living and
non-living exemplars was assessed by two semantic feature
questionnaires, a definition to word matching task and by
verbal definition (with prompting for visual and associative
information). DB did not shown any difference between
attribute type but nevertheless was worse at retrieving
information in general about living items. IW presented
with semantic dementia. Her knowledge of visual attributes
was significantly worse than her associative–functional
information. Despite this differential loss of visual attribute
knowledge, IW did not show a category-specific deficit for
animate kinds. Rather, her naming and comprehension
favoured the living things. In addition her ability to distinguish between real and chimeric pictures (object decision)
was preserved. Therefore, the two patients form two
important contrasts. The first is a double dissociation
between the living and non-living categories themselves:
DB was significantly worse at naming and understanding
living things compared to non-living items, while IW was
significantly better at naming and understanding living
concepts. The other contrast is between the relationship
between specific-perceptual attribute loss and a categoryspecific deficit for animate things: DB showed no difference
in her ability to access perceptual and functional attributes,
but did show a category-specific impairment for living
things, while IW, who performed poorly with perceptual
attribute judgements, did not show a category-specific
deficit for living things—in fact the reverse.
We would argue that the two patients are not just an
interesting pair of single case studies that might have arisen
from very unusual patterns of brain damage or premorbid
organization, but rather they show patterns of impairment
similar to other patients who have been reported before.
There are other patients of different aetiology, who like DB
exhibit a category specific deficit for animate things that do
not demonstrate a specific difference between attribute
types (Laiacona et al., 1993; Barbarotto et al., 1995;
Funnell and De Mornay Davies, 1996). IW resembles the
pattern shown by the majority of patients with semantic
dementia, not only in terms of her overall neuropsychological profile but also in the way that her semantic impairment
seems to affect visual knowledge much more than
associative–functional information (Basso et al., 1988;
Breedin et al., 1994b; Moss et al., 1995; Cardebat et al.,
1996; Srinivas et al., 1997; Tyler and Moss, 1998; Lambon
Ralph et al., unpublished data). Nevertheless, only one
patient with semantic dementia has been reported with a
specific impairment in understanding and producing the
names of animate items (that also has appropriate control
of cognitive–psycholinguistic factors). This is the case
reported by Barbarotto et al. (1995), who was exceptional
for at least three reasons—he demonstrated a clear (classical) category-specific dissociation but he showed no significant difference between attribute type and he presented
with medial and neocortical damage (see below). We
believe that these data are critical for some of the existing explanations for category-specific deficits. We shall
consider a number of them below.
328
M. A. Lambon Ralph et al.
The differential-weighting hypothesis
We reviewed the ‘differential-weighting’ hypothesis in the
Introduction. This hypothesis covers those theories that
emphasize the importance of certain attribute types for
the differentiation between exemplars within a category.
Warrington and Shallice (1984), amongst others (Farah
and McClelland, 1991; Lambon Ralph et al., 1997), have
suggested that living things are primarily differentiated by
how they look whereas man-made objects tend to have
specific functions. Consequently, impaired knowledge of
visual or associative information should lead to animate or
inanimate deficits, respectively. DB and patients like her
(Laiacona et al., 1993, 1997; Sheridan and Humphreys,
1993; Barbarotto et al., 1995, 1996; Funnell and De
Mornay Davies, 1996; Caramazza and Shelton, 1998;
Moss et al., 1998; Samson et al., 1998) indicate that a
category-specific impairment for living kinds does not
necessarily have to be associated with relatively poor
perceptual knowledge. Conversely, IW’s data would seem
to suggest that a differential loss of visual attributes does
not always lead to poor comprehension and naming of
living things. As far as we are aware IW, a patient with
relatively better performance on living things, is the only
case reported to date for which different attribute knowledge has been investigated. Of course, the results she
provides are the opposite to those predicted by the
differential-weighting hypothesis. When the two patterns,
represented by DB and IW, are taken together the result
would appear to be problematic for the differentialweighting hypothesis.
It could be argued that DB’s null effect for attribute type
could be due to insensitivity within the test materials/
procedures. However, the same materials were used with
IW, who did exhibit a dissociation. In addition, it is
possible that IW might exhibit a category-specific deficit for
animate things after further deterioration of her visual
knowledge, i.e. the current impairment was insufficient to
cause a category-specific effect. This is a hypothesis that we
should be able to test as we study IW’s deterioration over
time. We believe that it is unlikely, however, as there are a
number of other semantic dementia patients with much
greater semantic impairments that do not show a category
effect (Lambon Ralph et al., 1998).
The influence of cognitive and psycholinguistic
variables
Although our patients’ performance was affected by a
number of psycholinguistic variables, including frequency
and familiarity, the influence of category was significant
over and above these other factors. In fact, for one list of
items, IW’s category effect was not observed until the other
factors had been partialled out. This highlights the importance of controlling for the influence of psycholinguistic
variables in category-specific cases of living and non-living
types. Although familiarity and various other variables do
not provide a sufficient explanation for all the categoryspecific cases, it would appear that familiarity plays an
important role (and is probably a contributing factor to the
apparent imbalance in the occurrence of cases, with many
more reports of specific deficits for living things than the
reverse). Funnell and De Mornay Davis (1996) have argued
that a variety of rated familiarity (one that included a
measure of the quality as well as number of interactions
with each object) might prove sufficient for the entire
category-specific effect. Although we have some sympathy
with this notion (consider the difference between our
knowledge of a domestic dog and a jackal), it is possible
that normal subjects do include both quality and frequency
of interaction within their ratings (for further discussion of
the factors that might load onto familiarity ratings see
Lambon Ralph et al., 1998).
Self-organizing semantic networks
An alternative explanation for category-specific deficits can
be found within distributed models of semantic memory.
For example, Ritter and Kohonen (1989) have demonstrated that a self-organizing topographical network can
form areas within the two-dimensional semantic ‘space’
which respond maximally to certain items and categories
(even though category is not explicitly encoded in the
semantic feature input). It is not a large conceptual step to
suggest that if damage was relatively localized the result
could be critical for a specific category of items. It is
possible that the same result could be achieved from
other distributed architectures (e.g. Plaut and Shallice,
1993; Small et al., 1995). Although the representations are
formed in a multi-dimensional hyperspace, items from the
same category are encoded with relatively similar semantic features so that the individual concept attractors are
clustered within the same area of semantic space (see
Plaut and Shallice, 1993). If damage occurs to a selection
of hidden units that is preferentially dominant in the
formation of an area of semantic space then a categoryspecific deficit could result. This does not mean that
certain categories are ‘in’ a specific set of hidden units (or
by extension the neural substrates they are analogous to)
rather this hypothesis suggests that all the units are
involved in the formation of each concept but the correct
activation of a subset is critical for differentiation within
certain categories.
This proposal requires lesioning of a very circumscribed
collection of units for a category-specific deficit to arise
which is likely to be a relatively rare event. This would seem
to fit with the observation that of the many patients with a
semantic impairment few seem to present with a categoryspecific breakdown (see below). However the brain damage
that occurs in category-specific cases is often widespread
(e.g. consider the extent of atrophy that occurs in HSVE)
and this is not easily reconciled with the notion of very
Category-specific deficits
discrete areas of damage within the simulated models.
Another drawback is that the theory makes no clear
predictions with regard to which patients and under what
circumstances a category-specific deficit should arise. Consequently, this hypothesis is little more than a description
of the data and not a falsifiable theory.
Intercorrelation amongst features
A number of recent articles have emphasized the importance of intercorrelations amongst individual semantic
features (e.g. Gonnerman et al., 1997; McRae et al., 1997;
Moss et al., 1998). McRae et al. (1997) have shown that
normal subjects are quicker to confirm or reject attributes if
they are highly correlated with other features (that is these
features tend to co-occur in the formation of individual
concepts). From a statistical analysis of the features that
make up 190 different concepts, McRae et al. found that
there was a much higher intercorrelation between the
features of living things than artefacts. Gonnerman et al.
(1997) have proposed a very similar notion as an explanation for category-specific cases in Alzheimer’s disease.
Their simulation was based on the visual and associative
feature model proposed by Farah and McClelland (1991),
but included the differential intercorrelation that occurs
between the features making up living and non-living
kinds. With increasing levels of damage, their simulation
showed a linear decline in performance for the non-living
things but a non-linear function for the living kinds (see
Fig. 7, p. 274). In early stages of damage, there was little
impact on animate concepts but this was followed by a
catastrophic decline. When the two functions were combined, the model predicted that with small degrees of
impairment patients should exhibit better performance for
the biological kinds. With increasing damage there was a
crossover between the categories such that scores were
better for non-living things (although performance on these
items was not intact either).
This intercorrelation theory is appealing in that it does
not rely on damage to specific subtypes of attribute (visual
or associative) to produce a double dissociation between
living and non-living categories, and thus provides a possible explanation for the category-effects observed here. It is
interesting to consider the nature of DB’s definition and
drawing performance. Although her drawings of nonliving things were not intact, they did preserve the differences amongst the exemplars, which was not the case in
the living domain. In this category 29/32 animals were
drawn as a horizontal, four-legged prototypical creature.
It seems possible that this effect could be explained by
intercorrelations amongst living things. Intercorrelation
may be able to support animate representations under
mild semantic impairment (which is one possible exlanation for IW’s relatively good performance on this
category) but with greater damage item-specific infor-
329
mation will no longer be retrieved, so that damage may
‘pull’ the representations of the unusual living creatures
(e.g. crabs, giraffes, etc.) towards the prototype. (For
further demonstrations of prototype effects in categoryspecific cases see Caramazza and Shelton, 1998; Moss
et al., 1998.)
It should be noted that the theory proposed by
Gonnerman et al. does have its critics. Their model predicts
that the direction of the category-specific effect will depend
on the degree of impairment. Garrard et al. (1998) have
recently completed a cross-sectional study of 58 DAT
patients in which they observed a small number of cases
with category-specific comprehension and naming deficits.
This most commonly favoured non-living over living
things, although there were a small number in the reverse
direction. In contrast to the intercorrelation prediction,
Garrard et al. found no evidence to suggest that the
direction of the category-specific deficit was related to the
degree of impairment. In addition their theory does not
predict why some patients (including our own DB) seem to
exhibit a clear deficit for animals with only mild comprehension impairments. Gonnerman et al., themselves, report
two individual cases studies that do not seem to fit with
their own hypothesis. Patient GP presented with very good
performance for artefacts (95–100%) but was impaired for
living things (70–80%), while patient NB exhibited a difference to approximately the same degree but in favour of the
biological kinds (session three: biological kinds 100%,
artefacts 65%).
Evolutionary-adapted domain-specific knowledge
systems
Caramazza and Shelton (1998) have contrasted the reductionist approaches of the differential-weighting hypothesis
(a modality-specific reductionist theory) and the featureintercorrelation theory (a modality-neutral reductionist
approach) against a domain-specific semantic organization.
We have already mentioned one domain-specific theory,
the self-organizing model (Ritter and Kohonen, 1989),
which provides a mechanism by which category knowledge
can be encoded within a neural system. Caramazza and
Shelton, on the other hand, suggest that evolutionary
pressures have lead to a specialized, distinct neural mechanism for each of three broad categories of knowledge
(animals, plants and artefacts). Category-specific deficits
arise from damage to one of these distinct neural substrates
without a differential impairment of visual or functional
knowledge. Clearly, the data collected from DB and
patients like her, are consistent with this domain-specific
hypothesis. IW’s data suggest that visual knowledge itself
can be differentially impaired (see also Basso et al., 1988;
McCarthy and Warrington, 1988; Parkin, 1993; Breedin
et al., 1994b; Moss et al., 1995; Cardebat et al., 1996;
Srinivas et al., 1997; Tyler and Moss, 1998; Lambon Ralph
330
M. A. Lambon Ralph et al.
et al., unpublished data), implying that semantic knowledge
may be organized by attribute type and in some other way
to produce category-specific deficits.
Caramazza and Shelton note that by assuming neural
substrates for plants and animals, the theory predicts the
relatively high occurrence of category-specific deficits for
these domains of knowledge in comparison to the very
low number of patients reported with an impairment
for artefact knowledge. (It is entirely possible that this
asymmetry is due, in large part, to the effects of cognitive
and psycholinguistic variables, both in false positive
identifications of patients with category-specific deficits
for living things but also missed cases of the reverse dissociation; see Introduction.) They also suggest that patients
seem to have impaired knowledge across the whole of the
affected domain and so, thus far, there have been no reports
of deficits to narrower categories of knowledge (e.g. sea
creatures). Thus, this theory makes the testable hypotheses
that such patients will not be found and that categoryspecific deficits cannot be explained through a more
reductionist approach like the differential-weighting
hypothesis or the intercorrelation theory. Although this
hypothesis makes a clear prediction with regard to the
breadth of category impairments, the mechanism by which
categories are formed (evolutionary pressures) is not obviously open to empirical study. In addition it is not apparent, what if any, evolutionary pressure exists for the artefact
category unless, as Caramazza and Shelton suggest, this
‘category’ is represented by the remaining neural substrates
that do not underpin the conceptual knowledge of animals
and plants.
There are two other aspects that are not currently
addressed by this domain-specific theory. First, Caramazza
and Shelton do not seem to link each specific mechanism
with any particular neural substrate(s) in the brain. Consequently, it is impossible to predict, a priori, which categories should be affected in a patient by considering either
the location of brain damage or performance on other
cognitive tasks; in contrast, for example, the differentialweighting hypothesis predicts that poor knowledge of
living things should follow from impaired visual attributes.
Secondly, we have already noted that the naming of living
things by category-specific patients can be improved to the
same accuracy as that observed for non-living entities when
only functional descriptions are supplied (Silveri and
Gainotti, 1988; Gainotti and Silveri, 1996; Forde et al.,
1997; Humphreys et al., 1998). Caramazza and Shelton are
careful to claim that ‘(everything else being equal)
category-specific deficits should result in comparable
impairments for the visual and functional attributes of a
concept’. By removing a modality-specific basis from the
explanation of category-specific effects (to explain the null
result found in a growing number of category-specific
cases) it then becomes unclear why naming-to-definition
can apparently benefit from the removal of visual
attributes.
Aetiology and neuroanatomical correlates
We noted in the Introduction the debate in the literature as
to whether patients with DAT do show category-specific
impairments. It seems clear that there are a number of
individual cases that do exhibit a category-specific semantic
breakdown (e.g., DB and others: Laiacona et al., 1993;
Gonnerman et al., 1997; Garrard et al., 1998), although
these individual effects can be lost in group studies leading
to rather inconsistent results (Silveri et al., 1991; Hodges
et al., 1992b; Montanes et al., 1995; Daum et al., 1996;
Hodges et al., 1996; Tippett et al., 1996b). Category-specific
effects seem to be a rare event in semantic dementia
(Lambon Ralph et al., 1998) if psycholinguistic factors are
properly taken into account. Because many of these
patients have more severe impairments to visual than
attribute knowledge, one might predict a selective difficulty
with animate kinds. However, our patient, IW, exhibited
no effect on some materials while better performance for
living things on other measures (and a significant effect in
this direction overall). This pattern has been reported
before for a patient with progressive left temporal atrophy
(Silveri et al., 1997), although in both cases the difference
between the domains was significant but not profound, and
the dissociation was not a classical one. Although we have
noted the apparent rarity of category-specific impairment
in semantic dementia and DAT, it is not clear that the rate
of occurrence in other aetiologies is any greater. Howard
et al. (1995) studied 18 aphasic subjects following CVA and
found a clear category-specific deficit in only one patient.
Category-specific deficits for animate things have been
reported most often in patients who have recovered from
HSVE. Although HSVE is a common aetiology in
category-specific cases, this does not mean that categoryspecific impairment is common in HSVE patients. Kapur
et al. (1994) have recently reported MRI and neuropsychology findings for a group of 10 HSVE patients. Although all
were amnesic, only four of the ten were noted to have a
naming deficit (the authors do not report how many of
these exhibited a category-specific effect). Barbarotto et al.
(1995) investigated seven HSVE patients and found a
significant effect of category (over and above other psycholinguistic variables) in four individuals. However,
Barbarotto et al. noted that the patients were selected
primarily for their cognitive deficits and thus were not fully
representative of the whole HSVE population. We are
unaware of any study that looks for the presence of
category-specific in an unselected sample of HSVE cases.
Semantic dementia patients would seem to be an ideal set
of patients in which to find category-specific impairments.
They all present with semantic impairment that is graded
by familiarity (Funnell, 1995; Hodges et al., 1992a, 1995) as
a result of progressive atrophy of the temporal neocortex,
including the inferior temporal gyrus. A number of authors
have noted the possibility that visual knowledge may be
underpinned by this particular area of the temporal lobe
Category-specific deficits
and as a consequence category-specific deficits for living
things might result (Saffran and Schwartz, 1994; Gainotti
et al., 1995; Gainotti and Silveri, 1996). In addition, some
functional imaging studies have investigated the relative
activation of brain regions for living and non-living items.
The results from these PET studies differ in the exact
number and location of specific activated regions (which
may be due, at least in part, to the different experimental
tasks employed). However, a number have highlighted the
role of the inferior temporal region in the processing of
living kinds (Perani et al., 1995; Damasio et al., 1996;
Martin et al., 1996). Like other patients with semantic
dementia, IW did have inferior temporal atrophy and
worse knowledge for the visual attributes but she did not
exhibit relatively poor knowledge of living things. Patients
with HSVE or DAT who do show a category-specific deficit
can have rather widespread atrophy that includes both the
inferior temporal gyrus and medial temporal structures.
Although damage to medial temporal structures is associated with the amnesic syndrome, temporolimbic structures
could also make an important contribution to the understanding of category-specific deficits for living things.
Gainotti and Silveri (1995) note that these structures
receive direct and indirect input from all sensory modalities. Category-specific deficits for living things may arise
from a combination of medial and neocortical temporal
areas (including inferior temporal structures). There seems
to be evidence to suggest that damage to either region alone
does not lead to category-specific impairments (not all
HSVE amnesic patients have comprehension deficits, let
alone category-specific semantic impairments). The possible importance for a combination of atrophy is highlighted
by the comparison between MF (Barbarotto et al., 1995)
and other patients with semantic dementia. MF’s classical
dissociation between categories resulted from atrophy
to medial and inferior temporal structures (somewhat
more like the pattern of atrophy seen in HSVE) whereas
other patients with semantic dementia do not have medial
temporal damage (Hodges et al., 1992a).
An attempt at a unified theory
We have discussed various neurological and neuropsychological theories of category-specific impairment. Are we
any closer to an adequate explanation for these effects? The
differential-weighting hypothesis provides an explanation
for differential impairment of both living and non-living
domains, although our study here (and others in which
category-specific impairment is not accompanied by differential attribute knowledge) seems to have undermined this
approach. The influence of psycholinguistic factors in these
cases is clearly important, although the variables identified
thus far do not provide a complete explanation for all
category-specific patients. Self-organizing networks demonstrate category-specific organization of the internal
semantic space without explicit encoding of category within
331
the input. This could lead to category-specific impairments
although it is not clear how this converts into the type of
impairment and pattern of atrophy seen in categoryspecific patients (nor into any falsifiable predictions). The
notion of intercorrelated features allows for both types of
category-specific impairment, but under the current
description there is supposed to be a clear relationship
between degree of impairment and the direction of the
impairment, a relationship for which there is little positive
evidence. Caramazza and Shelton’s (1998) domain-specific
theory would also provide an alternative explanation for
much of the category-specific literature, although we have
noted one or two aspects which still require attention
before this theory is complete.
The notion of intercorrelations seems to go some way
towards capturing the confusion between exemplars that
occurs in category-specific cases. We have already noted
the prototypical drawings of DB (see also De Renzi and
Lucchelli, 1994), the prototype effect found in two other
category-specific patients (Caramazza and Shelton, 1998;
Moss et al., 1998), and the relatively high proportion of
intrusive information produced in DB’s definitions (this has
been observed in other category-specific patients; see Forde
et al., 1997). Although small category differences are predicted by the effect of intercorrelated features (Gonnerman
et al., 1997), it does not provide an explanation for the
larger differences that can be observed in patients at the
early stages of decline (e.g. DB, but see also patients GP
and NB: Gonnerman et al., 1997).
We believe that a possible explanation is one that
includes both an inability to activate critical uniquely
identifying features correctly and an abnormal degree of
confusion between the exemplars of a category. If semantic
knowledge is made up (at least in part) by an amalgamation of the information supported by the distributed sensory association areas (Allport, 1985; Warrington and
McCarthy, 1987), then the many direct and indirect projections to and from the temporolimbic structures may encode
the relationships within and between sensory-derived
attributes (see also Gainotti and Silveri, 1996). Each concept will be represented by the combined activation across
each association cortex (perhaps in differing amounts
according to which senses the items are experienced in) and
the connections between them. This complex semantic
network will be sensitive to the intercorrelation amongst
features. Damage to individual types of attribute
knowledge (e.g. visual) may not necessarily lead to
category-specific impairments because (a) other critical
uniquely-identifying features could be directly activated
from sensory input and (b) the interconnections with other
intact parts of the semantic system can help to maintain
sufficient correct activation of any remaining knowledge
to differentiate between category exemplars. Isolated
impairment to the flow of information through the interconnections may not result in a comprehension deficit as
long as the underlying information is still activated by
332
M. A. Lambon Ralph et al.
direct sensory input and by any remaining activation
passing between association areas. When damage involves
both an underlying set of attributes and the interconnections between areas, then there may be no way for individual exemplars within a category to be differentiated from
one another. If there is damage to the visual association
areas a patient may appear to lose knowledge about the
appearance of all concrete objects (cf. IW’s drawing and
definitions). When the remaining visual information can
combine with the activation projected between association
areas, animate exemplars may still be differentiated from
one another. If impaired visual information is combined
with poor interconnections between the association cortices
then the differences between animate concepts may not be
maintained. Unfamiliar concepts will tend to ‘regress’
towards the prototype (perhaps as a result of the influence
of intercorrelations) and the patient will produce intrusive
information. It is possible that the ensuing semantic ‘confusion’ will propagate from malfunctioning perceptual
regions to ‘intact’ areas supporting different types of information (e.g. associative–functional) (note that this is similar to the notion of ‘critical mass’; see Experiment 3, Farah
and McClelland, 1991). If the propagated noise or ‘confusion’ is relatively mild a category by modality interaction
may be observed (e.g. poor visual knowledge of living
things). However, with much greater degrees of noise a
patient might not be able to produce the correct nonperceptual attributes and thus may not exhibit a difference
between attribute types. If a definition containing functional information alone is used as input, the undamaged
regions that underpin this knowledge will be able to
differentiate between the animate exemplars (note DB’s
good associative definition-to-word matching or the
absence of category-specific effects when naming functional
definitions observed by Gainotti and Silveri, 1996; Forde
et al., 1997). Although this information will not necessarily
activate the correct perceptual knowledge, the object will be
uniquely identified from consideration of the functional–
associative information itself. Consequently, the patient
can either match or provide the correct name for a nonvisual definition, even for living entities.
We have had little to say with regard to category-specific
effects for non-living things. The small number of reported
cases hinders the development of detailed theories for this
type of category-specific impairment. In these cases there
has been no attempt to probe the status of the underlying
information and to control for the full range of psycholinguistic factors (apart from patient IW reported here and
Howard et al., 1995), while the neuroanatomical correlates
are based on the examination of CT scans. It is possible
that the ‘unified’ theory described above could be extended
to this type of patient. Assuming that artefacts are represented by a relatively different distribution/weighting of
critical features than living concepts (this could include
functional information together with other differences such
as biological versus non-biological motion), then if the
neural structures that support these critical attributes are
damaged along with the interconnections between areas of
the semantic network then exemplars within non-living
categories may show the same type of ‘confusion’ as that
seen in category-specific impairments for living things (note
that patients with category-specific deficits for artefacts
tend to make within-category semantic errors both in
comprehension and naming: Hillis et al., 1990; Hillis and
Caramazza, 1991; Sacchett and Humphreys, 1992).
Caramazza and Shelton argue that there seems to be a
tripartite semantic organisation. They note that some
patients demonstrate selective impairment to knowledge
for animals or fruit/vegetables (Hart et al., 1985; Hillis and
Caramazza, 1991; Caramazza and Shelton, 1998). The
unified hypothesis could accommodate selective impairments for fruit/vegetables assuming that for these concepts
there is a relative importance of smell and taste (Gainotti
and Silveri, 1996; note that olfactory and gustatory information is projected into the temporolimbic systems; see
also Warrington and McCarthy, 1987). Thus if there is a
disruption of this type of knowledge together with poor
inter-association area activation, a category-specific deficit
for vegetables and fruit could arise. Although this hypothesis can address the three broad categories suggested by
Caramazza and Shelton, it is unlikely that this reductionist
approach could be extended to increasingly fine category
distinctions. This follows from the nature of reductionist
explanations of psychological phenomena in which differential performance is modelled through relatively graded
properties and thus, is incompatible with very narrow
category-specific deficits.
Acknowledgements
The research was supported by the Engineering and Physical Sciences Research Council (MALR). We would like to
thank Drs Richard Wise and Peter Garrard for the analysis
of the MRI results. We are indebted to I. W. and D. B. for
the time and patience they have given.
References
Allport DA. Distributed memory, modular subsystems and dysphasia. In:
Newman S, Epstein R, editors. Current perspectives in dysphasia.
Edinburgh: Churchill Livingstone, 1985.
Barbarotto R, Capitani E, Spinnler H, Trivelli C. Slowly progressive
semantic impairment with category specificity. Neurocase 1995; 1:
107–19.
Barbarotto R, Capitani E, Laiacona M. Naming deficit in herpes-simplex
encephalitis. Acta Neurologica Scandinavica 1996; 93: 272–80.
Basso A, Capitani E, Laiacona M. Progressive language impairment
without dementia: a case with isolated category specific semantic defect.
Journal of Neurology, Neurosurgery and Psychiatry 1988; 51: 1201–7.
Bishop DVM. Test for the reception of grammar. London: Medical
Research Council, 1989.
Braak H, Braak E. Neuropathological staging of Alzheimer-related
changes. Acta Neuropathologica 1991; 82: 239–59.
Breedin SD, Martin N, Saffran EM. Category-specific semantic
impairments: an infrequent occurrence. Brain and Language 1994a; 47:
383–6.
Category-specific deficits
Breedin SD, Saffran EM, Coslett HB. Reversal of the concreteness effect
in a patient with semantic dementia. Cognitive Neuropsychology 1994b;
11: 617–60.
Caramazza A, Shelton JR. Domain-specific knowledge systems in the
brain: the animate–inanimate distinction. Journal of Cognitive Neuroscience 1998; 10: 1–34.
Cardebat D, Demonet JF, Celsis P, Puel M. Living/non-living dissociation
in a case of semantic dementia: a SPECT activation study. Neuropsychologia 1996; 34: 1175–9.
Centre for Lexical Information. The Celex Lexical Database. Nijmegen,
1993.
Cipolotti L, Warrington EK. Semantic memory and reading abilities: a
case report. Journal of the International Neuropsychological Society
1995; 1: 104–10.
Croot K, Patterson K, Hodges JR. Phonological disruption in atypical
dementia of the Alzheimer’s type. Brain and Cognition 1996; 32:
186–90.
Damasio H, Grabowski TJ, Tranel D, Hichwa RD, Damasio AR. A
neural basis for lexical retrieval. Nature 1996; 380: 499–505.
Daum I, Riesch G, Sartori G, Birbaumer N. Semantic memory impairment in Alzheimer’s disease. Journal of Clinical and Experimental
Neuropsychology 1996; 18: 648–65.
De Renzi E, Lucchelli F. Are semantic systems separately represented in
the brain? The case of living category impairment. Cortex 1994; 30:
3–25.
Ellis AW, Lum C, Lambon Ralph MA. On the use of regression
techniques for the analysis of single-case aphasic data. Journal of
Neurolinguistics 1996; 9: 165–74.
Farah MJ, McClelland JL. A computational model of semantic memory
impairment: modality specificity and emergent category specificity.
Journal of Experimental Psychology General 1991; 120: 339–57.
Farah MJ, Wallace MA. Semantically-bounded anomia: implications for
the neural implementation of naming. Neuropsychologia 1992; 30:
609–21.
Farah MJ, Hammond KM, Mehta Z, Ratcliff G. Category-specificity and
modality-specificity in semantic memory. Neuropsychologia 1989; 27:
193–200.
Farah MJ, McMullen PA, Meyer MM. Can recognition of living things be
selectively impaired? Neuropsychologia 1991; 29: 185–93.
Farah MJ, Meyer MM, McMullen PA. The living/nonliving dissociation
is not an artifact: giving an a priori implausible hypothesis a strong test.
Cognitive Neuropsychology 1996; 13: 137–54.
Folstein MF, Folstein SE, McHugh PR. Mini-mental state: a practical
method for grading the cognitive state of patients for the clinician.
Journal of Psychiatric Research 1975; 12: 189–98.
Forde EME, Francis D, Riddoch MJ, Rumiati RI, Humphreys GW. On
the links between visual knowledge and naming: a single case study of a
patient with a category-specific impairment for living things. Cognitive
Neuropsychology 1997; 14: 403–58.
Franklin S, van Sommers P, Howard D. Drawing without meaning?
Dissociations in the graphic performance of an agnosic artist. In:
Campbell R, editor. Mental lives: case studies in cognition. Oxford:
Blackwell, 1987.
Franklin S, Turner JM, Ellis AW. The ADA comprehension battery.
London: Action for Dysphasic Adults, 1992.
Funnell E. Objects and properties: a study of the breakdown of semantic
memory. Memory 1995; 3: 497–518.
Funnell E, Sheridan J. Categories of knowledge? Unfamiliar aspects of
living and nonliving things. Cognitive Neuropsychology 1992; 9: 135–
53.
Funnell E, De Mornay Davies P. JBR: a reassessment of concept
familiarity and a category-specific disorder for living things. Neurocase
1996; 2: 461–74.
Gaffan D, Heywood CA. A spurious category-specific visual agnosia for
living things in normal human and nonhuman-primates. Journal of
Cognitive Neuroscience 1993; 5: 118–28.
Gainotti G, Silveri MC. Cognitive and anatomical locus of lesion in a
patient with a category-specific semantic impairment for living beings.
Cognitive Neuropsychology 1996; 13: 357–90.
Gainotti G, Silveri MC, Daniele A, Giustolisi L. Neuroanatomical
correlates of category-specific semantic disorders: a critical survey.
Memory 1995; 3: 247–64.
Garrard P, Patterson K, Watson PC, Hodges JR. Category-specific
semantic loss in dementia of Alzheimer’s type. Brain 1998; 121: 633–46.
333
Gonnerman LM, Andersen ES, Devlin JT, Kempler D, Seidenberg MS.
Double dissociation of semantic categories in Alzheimer’s disease. Brain
and Language 1997; 57: 254–79.
Graham KS. TOAD: test of object and animal decision. Unpublished test,
1993.
Graham KS. Semantics memory impairment and anomia in progressive
fluent aphasia. Cambridge: unpublished thesis, 1995.
Graham KS, Hodges JR. Differentiating 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. Neuropsychology 1997; 11: 77–89.
Graham KS, Hodges JR, Patterson KE. The relationship between comprehension and oral reading in progressive fluent aphasia. Neuropsychologia 1994; 32: 299–316.
Graham K, Patterson K, Hodges JR. Progressive pure anomia: insufficient activation of phonology by meaning. Neurocase 1995; 1: 25–38.
Hart J, Gordon B. Neural subsystems for object knowledge. Nature 1992;
359: 60–4.
Hart J, Berndt RS, Caramazza A. Category-specific naming deficit
following cerebral infarction. Nature 1985; 316: 439–40.
Hillis AE, Caramazza A. Category-specific naming and comprehension
impairment: a double dissociation. Brain 1991; 114: 2081–94.
Hillis AE, Rapp B, Romani C, Caramazza A. Selective impairment of
semantic in lexical processing. Cognitive Neuropsychology 1990; 7:
191–243.
Hirsh KW, Funnell E. Those old, familiar things: age of acquisition,
familiarity and lexical access in progressive aphasia. Journal of Neurolinguistics 1995; 9: 23–32.
Hodges JR, Patterson K. Is semantic memory consistently impaired early
in the course of Alzheimer’s disease? Neuroanatomical and diagnostic
implications. Neuropsychologia 1995; 33: 441–59.
Hodges JR, Salmon DP, Butters N. Differential impairment of semantic
and episodic memory in Alzheimer’s and Huntington’s diseases: a
controlled prospective study. Journal of Neurology, Neurosurgery and
Psychiatry 1990; 53: 1089–95.
Hodges JR, Patterson K, Oxbury S, Funnell E. Semantic dementia:
progressive fluent aphasia with temporal lobe atrophy. Brain 1992a;
115: 1783–1806.
Hodges JR, Salmon DP, Butters N. Semantic memory impairment
in Alzheimer’s disease: failure of access or degraded knowledge?
Neuropsychologia 1992b; 30: 301–14.
Hodges JR, Patterson K, Tyler LK. Loss of semantic memory: implications for the modularity of mind. Cognitive Neuropsychology 1994;
11: 505–42.
Hodges JR, Graham N, Patterson K. Charting the progression of
semantic dementia: implications for the organisation of semantic
memory. Memory 1995; 3: 463–95.
Hodges JR, Patterson K, Graham N, Dawson K. Naming and knowing in
dementia of Alzheimer’s type. Brain and Language 1996; 54: 302–25.
Hodges JR, Garrard P, Patterson K. Semantic dementia. In: Kertesz A,
Munoz DG, editors. Pick’s disease and Pick complex. New York:
Wylie–Liss, 1998.
Howard D, Patterson K. The Pyramids and Palm Trees Test: a test of
semantic access from words and pictures. Bury St Edmunds: Thames
Valley Test Company, 1992.
Howard D, Best W, Bruce C, Gatehouse C. Operativity and animacy
effects in aphasic naming. Special Issue in honour of Ruth Lesser.
European Journal of Disorders of Communication 1995; 30: 286–302.
Humphreys GW, Lamote C, Lloyd-Jones TJ. An interactive activation
approach to object processing: effects of structural similarity, name
frequency, and task in normality and pathology. Memory 1995; 3:
535–86.
Humphreys GW, Riddoch MJ, Price CJ. Top-down processes in object
identification: Evidence from experimental psychology, neuropsychology and functional anatomy. Proceedings of the Royal Society, Series
B, 1998; in press.
Kaplan E, Goodglass H, Weintraub S. The Boston Naming Test. Boston:
Veteran’s Administration, 1976.
Kapur N, Barker S, Burrows EH, Ellison D, Brice J, Illis LS et al.
Herpes-simplex encephalitis: Long-term magnetic-resonance-imaging
and neuropsychological profile. Journal of Neurology, Neurosurgery
and Psychiatry 1994; 57: 1334–42.
334
M. A. Lambon Ralph et al.
Laiacona M, Barbarotto R, Capitani E. Perceptual and associative
knowledge in category specific impairment of semantic memory: a study
of two cases. Cortex 1993; 29: 727–40.
Laiacona M, Capitani E, Barbarotto R. Semantic category dissociations:
a longitudinal study of two cases. Cortex 1997; 33: 441–61.
Lambon Ralph MA. The role of semantics in reading aloud. York
University: Unpublished thesis, 1997.
Lambon Ralph MA, Patterson K, Hodges JR. Longitudinal study of
naming and knowing in Dementia of Alzheimer type. Neuropsychologia 1997; 35: 1251–60.
Lambon Ralph MA, Graham KS, Ellis AW, Hodges JR. Naming in
semantic dementia—what matters? Neuropsychologia 1998; in press.
Laws KR, Evans JJ, Hodges JR, McCarthy RA. Naming without
knowing and appearance without associations: evidence for constructive processes in semantic memory? Memory 1995; 3: 409–33.
Mackenzie Ross SJ, Graham N, Stuart-Green L, Prins M, Xuereb J,
Patterson K et al. Progressive biparietal atrophy: an atypical presentation of Alzheimer’s disease. Journal of Neurology, Neurosurgery and
Psychiatry 1996; 61: 388–95.
Martin A, Wiggs CL, Ungerleider LG, Haxby JV. Neural correlates of
category-specific knowledge. Nature 1996; 379: 649–52.
Mauri A, Daum I, Sartori G, Riesch G et al. Category-specific semantic
impairment in Alzheimer’s disease and temporal lobe dysfunction: a
comparative study. Journal of Clinical and Experimental Neuropsychology 1994; 16: 689–701.
McCarthy RA, Warrington EK. Evidence for modality-specific meaning
systems in the brain. Nature 1988; 334: 428–30.
McRae K, de Sa VR, Seidenberg MS. On the nature and scope of featural
representation of word meaning. Journal of Experimental Psychology
General 1997; 126: 99–130.
Mehta C, Patel N. StatXact 3 for Windows. Cambridge, Mass.: Cytel
Software Corporation, 1995.
Montanes P, Goldbaum MC, Boller F. The naming impairment of living
and nonliving items in Alzheimer’s disease. Journal of the International
Neuropsychological Society 1995; 1: 39–48.
Montanes P, Goldblum MC, Boller F. Classification deficits in
Alzheimer’s disease with special reference to living and nonliving things.
Brain and Language 1996; 54: 335–58.
Morrison CM, Ellis AW. Roles of word frequency and age of acquisition
in word naming and lexical decision. Journal of Experimental Psychology Learning Memory and Cognition 1995; 21: 116–33.
Morrison CM, Ellis AW, Quinlan PT. Age of acquisition, not word
frequency, affects object naming, not object recognition. Memory and
Cognition 1992; 20: 705–14.
Morrison CM, Chappell TD, Ellis AW. Age of acquisition norms for a
large set of object names and their relation to adult estimates and other
variables. Quarterly Journal of Experimental Psychology 1997; 50:
528–59.
Moss HE, Tyler LK, Hodges JR, Patterson K. Exploring the loss of
semantic memory in semantic dementia: evidence from a primed
monitoring study. Neuropsychology 1995; 9: 16–26.
Moss HE, Tyler LK, Jennings F. When leopards lose their spots:
knowledge of visual properties in category-specific deficits for living
things. Cognitive Neuropsychology 1997; 14: 901–50.
Moss HE, Tyler LK, Durrant-Peatfield M, Bunn EM. Two eyes of a
see-through: impaired and intact semantic knowledge in a case of
selective deficit for living things. Neurocase 1998; 291–310.
Nickels L, Howard D. A frequent occurrence? Factors affecting the
production of semantic errors in aphasic naming. Cognitive Neuropsychology 1994; 11: 289–320.
Nickels L, Howard D. Aphasic naming: what matters? Neuropsychologia
1995; 33: 1281–303.
Osterrieth P. Le test de copie d’une figure complexe. Archives de Psychologie 1944; 30: 205–550.
Parkin AJ. Progressive aphasia without dementia: a clinical and cognitive
neuropsychological analysis. Brain and Language 1993; 44: 201–20.
Patterson K, Hodges JR. Deterioration of word meaning: implications for
reading. Neuropsychologia 1992; 30: 1025–40.
Patterson KE, Graham N, Hodges JR. The impact of semantic memory
loss on phonological representations. Journal of Cognitive Neuroscience 1994; 6: 57–69.
Perani D, Cappa SF, Bettinardi V, Bressi S, Gornotempini M, Matarrese
M et al. Different neural systems for the recognition of animals and
man-made tools. Neuroreport 1995; 6: 1637–41.
Pietrini V, Nertempi P, Vaglia A, Revello MG, Pinna V, Ferromilone F.
Recovery from herpes-simplex encephalitis—selective impairment of
specific semantic categories with neuroradiological correlation. Journal
of Neurology, Neurosurgery and Psychiatry 1988; 51: 1284–93.
Plaut DC, Shallice T. Deep dyslexia: a case study of connectionist
neuropsychology. Cognitive Neuropsychology 1993; 10: 377–500.
Raven JC. Coloured progressive matrices: sets A, AB, B. London:
HK Lewis, 1962.
Riddoch MJ, Humphreys GW. Visual object processing in optic aphasia:
a case study of semantic access agnosia. Cognitive Neuropsychology
1987; 4: 131–85.
Riddoch MJ, Humphreys GW. The Birmingham Object Recognition
Battery (BORB). Hove: Lawrence Erlbaum Associates, 1992.
Ritter H, Kohonen T. Self-organizing semantic maps. Biological
Cybernetics 1989; 61: 241–54.
Sacchett C, Humphreys GW. Calling a squirrel a squirrel but a canoe a
wigwam: a category-specific deficit for artefactual objects and body
parts. Cognitive Neuropsychology 1992; 9: 73–86.
Saffran EM, Schwartz MF. Of cabbages and things: semantic memory
from a neuropsychological perspective—a tutorial review. Attention
and Performance 1994; 25: 507–36.
Samson D, De Wilde V, Pillon A. Impaired knowledge of visual and
nonvisual attributes in a patient with a naming impairment for living
entities: a case of a true category-specific deficit. Neurocase 1998;
273–90.
Sartori G, Job R. The oyster with four legs: a neuropsychological study
on the interaction of visual and semantic information. Cognitive
Neuropsychology 1988; 5: 105–32.
Sartori G, Job R, Miozzo M, Zago S, Marchiori G. Category-specific
form-knowledge deficit in a patient with herpes-simplex virus encephalitis. Journal of Clinical and Experimental Neuropsychology 1993a; 15:
280–99.
Sartori G, Miozzo M, Job R. Category-specific naming impairments? Yes.
Quarterly Journal of Experimental Psychology Human Experimental
Psychology 1993b; 46a: 489–504.
Shallice T. From neuropsychology to mental structure. Cambridge:
Cambridge University Press, 1988.
Sheridan J, Humphreys GW. A verbal–semantic category-specific recognition impairment. Cognitive Neuropsychology 1993; 10: 143–84.
Silveri MC, Gainotti G. Interaction between vision and language in
category-specific semantic impairment. Cognitive Neuropsychology
1988; 5: 677–709.
Silveri MC, Daniele A, Giustolisi L, Gainotti G et al. Dissociation
between knowledge of living and nonliving things in dementia of the
Alzheimer type. Neurology 1991; 41: 545–6.
Silveri MC, Gainotti G, Perani D, Cappelletti JY, Carbone G, Fazio F.
Naming deficit for non-living items: neuropsychological and PET study.
Neuropsychologia 1997; 35: 359–67.
Small SL, Hart J, Nguyen T, Gordon B. Distributed representations of
semantic knowledge in the brain. Brain 1995; 118: 441–53.
Snodgrass JG, Vanderwart M. A standardised set of 260 pictures: norms
for name agreement, image agreement, familiarity, and visual complexity. Journal of Experimental Psychology: Human Learning
and Memory 1980; 6: 174–215.
Snowden JS, Goulding PJ, Neary D. Semantic dementia: a form
of circumscribed cerebral atrophy. Behavioural Neurology 1989; 2:
167–82.
Srinivas K, Breedin SD, Coslett HB, Saffran EM. Intact perceptual
priming in a patient with damage to the anterior inferior temporal
lobes. Journal of Cognitive Neuroscience 1997; 9: 490–511.
Stewart F, Parkin AJ, Hunkin NM. Naming impairments following
recovery from herpes simplex encephalitis: category-specific? Quarterly
Journal of Experimental Psychology Human Experimental Psychology
1992; 44a: 261–84.
Tippett LJ, Glosser G, Farah MJ. A category-specific naming impairment
after temporal lobectomy. Neuropsychologia 1996a; 34: 139–46.
Tippett LJ, Grossman M, Farah MJ. The semantic memory impairment
of Alzheimer’s disease: category-specific? Cortex 1996b; 32: 143–53.
Tyler LK, Moss HE. Functional properties of concepts: studies of normal
and brain-damaged patients. Cognitive Neuropsychology 1997; 14:
511–45.
Ungerleider LG, Mishkin M. Two cortical visual systems. In: Ingle DG,
Goodale MA, Mansfield RJW, editors. Analysis of visual behaviour.
Cambridge, MA: MIT Press, 1982.
Category-specific deficits
Warrington EK. The selective impairment of semantic memory. Quarterly
Journal of Experimental Psychology 1975; 27: 635–57.
Warrington EK. Recognition memory test. Windsor: NFER-Nelson,
1984.
Warrington EK, James M. The visual object and space perception battery.
Bury St Edmunds: Thames Valley Test Company, 1991.
Warrington EK, McCarthy R. Category specific access dysphasia. Brain
1983; 106: 859–78.
Warrington EK, McCarthy R. Categories of knowledge: further fractionations and an attempted integration. Brain 1987; 110: 1273–96.
Warrington EK, McCarthy RA. Multiple meaning systems in the brain: a
case for visual semantics. Neuropsychologia 1994; 32: 1465–73.
Warrington EK, Shallice T. Category specific semantic impairments.
Brain 1984; 107: 829–54.
Wilson BA. Semantic memory impairments following non-progressive
brain injury: a study of four cases. Brain Injury 1997; 11: 259–69.
Wilson BA, Baddeley AD, Kapur N. Dense amnesia in a professional
musician following herpes simplex virus encephalitis. Journal of Clinical
and Experimental Neuropsychology 1995; 17: 668–81.
Zelen M. The analysis of several 2 2 contingency tables. Biometrika
1971; 58: 129–37.
Received on 12 September, 1997; resubmitted on 13 January, 1998;
accepted on 19 March, 1998
335
A little black insect that buzzes
An insect that is often found in the house flying around food and is killed
using a spray or swat
A little black or red insect with six legs and two antennae
An insect that lives in great numbers and work together to carry food
A fat insect with a striped black and yellow body
An insect that makes honey and wax
A large yellow-brown animal with a long neck and two humps on its back
An animal that can live in the desert for long periods without drinking
water
A four-legged animal with a tail and whiskers that maiows and purrs
An animal that lives in the home, catches mice and likes being stroked
A brown feathered bird with long tail feathers and a red comb on the top
of its head
The male of the bird commonly kept on farms which crows at dawn
A four-legged animal that has udders and moos
A farm animal that supplies us with milk and meat
A large long reptile with a long, big mouth full of sharp teeth
A reptile that lives in the river whose skin is used to make expensive
handbags, belts and shoes
A large grey animal with large ears, long tusks and a long trunk
An animal that lives in India and Africa from which ivory is collected
A large animal sometimes brown, black or white with a long mane and tail
An animal kept on the land or farms that can be used for riding or for
pulling vehicles
A fierce animal with yellowish fur and dark spots
A foreign animal whose fur is used to make coats
A large four-legged animal with sharp teeth, claws and a mane. It has
yellow fur
A fierce animal that eats meat, lives in Africa, in groups called ‘prides’ and
is known as the ‘king of the animals’
A small brown animal with whiskers and a tail that makes a squeak
An animal that eats cheese and can live under the skirting-boards
A very colourful bird with a hooked beak
A foreign bird that was carried by pirates and can repeat human speech
FLY
FLY
MOUSE
MOUSE
PARROT
PARROT
LION
LEOPARD
LEOPARD
LION
ELEPHANT
ELEPHANT
HORSE
HORSE
COCKEREL
COW
COW
CROCODILE
CROCODILE
CAT
CAT
COCKEREL
ANT
ANT
BEE
BEE
CAMEL
CAMEL
Definition
Item
Appendix 1. Items from the associative versus perceptual definition-to-word matching test
MOUSE
MOUSE
PARROT
PARROT
LION
LION
LION
LION
CROCODILE
CROCODILE
COW
COW
PARROT
ELEPHANT
ELEPHANT
CROCODILE
CROCODILE
LEOPARD
LEOPARD
PARROT
MOUSE
MOUSE
PARROT
PARROT
LEOPARD
LEOPARD
COCKEREL
COCKEREL
Choice 1
CAT
CAT
FLY
FLY
ELEPHANT
ELEPHANT
ELEPHANT
ELEPHANT
CAMEL
CAMEL
HORSE
HORSE
COCKEREL
COW
COW
CAMEL
CAMEL
COW
COW
COCKEREL
CAT
CAT
FLY
FLY
LION
LION
BEE
BEE
Choice 2
FLY
FLY
SWALLOW
SWALLOW
CROCODILE
CAT
CAT
CROCODILE
ELEPHANT
ELEPHANT
PIG
PIG
FLY
HORSE
HORSE
ELEPHANT
ELEPHANT
HORSE
HORSE
FLY
FLY
FLY
COCKEREL
COCKEREL
CROCODILE
CROCODILE
PARROT
PARROT
Choice 3
ANT
ANT
BEE
BEE
LEOPARD
LEOPARD
LEOPARD
LEOPARD
LEOPARD
LEOPARD
CAT
CAT
SWALLOW
CAT
CAT
LEOPARD
LEOPARD
PIG
PIG
SWALLOW
ANT
ANT
BEE
BEE
CAMEL
CAMEL
FLY
FLY
Choice 4
BEE
BEE
COCKEREL
COCKEREL
CAMEL
CAMEL
CAMEL
CAMEL
LION
LION
LEOPARD
LEOPARD
BEE
PIG
PIG
LION
LION
CAT
CAT
BEE
BEE
BEE
ANT
ANT
ELEPHANT
ELEPHANT
ANT
ANT
Choice 5
336
M. A. Lambon Ralph et al.
A fat pink animal, with four short legs and a curled tail
An animal kept on the farm for pork, bacon and ham
A small black bird with white breast-feathers and a long forked-tail
A bird that migrates to Britain in the summer, which catches and eats
insects whilst flying
A vehicle with two wheels, a saddle and two pedals
A vehicle for one person which does not have an engine
A green or brown container with a narrow neck
A container made of glass used to keep wine in
An instrument made from of a block with many teeth
Something used to keep the hair tidy and keep it in place
An instrument with a handle and a single sharp blade
A tool used for cutting your food
A weapon with a handle, a short barrel and a trigger
A weapon used by cowboys to shoot
A long handled tool with a series of prongs at the end
A tool used for gathering up leaves or grass clippings
A tool made from a long blade with a toothed edge
A tool used for cutting wood
A tool made from two sharp blades fixed in the middle
A tool used for cutting paper and by tailors to cut cloth
Something made up of a frame and two glass lenses
Something you wear to improve your eyesight
A tool you use with a shallow bowl on a handle
Something you use to drink soup and eat dessert
An instrument made from a long sharp blade that is fixed in a hilt
An instrument which used to be used in battles for armed combat
A device with red, amber and green lights
A mechanical device used to control the vehicles at crossroads
PIG
PIG
SWALLOW
SWALLOW
BICYCLE
BICYCLE
BOTTLE
BOTTLE
COMB
COMB
KNIFE
KNIFE
PISTOL
PISTOL
RAKE
RAKE
SAW
SAW
SCISSORS
SCISSORS
SPECTACLES
SPECTACLES
SPOON
SPOON
SWORD
SWORD
TRAFFIC-LIGHTS
TRAFFIC-LIGHTS
Definition
Item
Appendix 1. Continued.
SAW
SAW
KNIFE
KNIFE
PISTOL
PISTOL
SAW
SAW
KNIFE
KNIFE
RAKE
RAKE
SAW
SAW
SCISSORS
SCISSORS
COMB
COMB
KNIFE
KNIFE
SAW
SAW
SPECTACLES
SPECTACLES
HORSE
HORSE
PARROT
PARROT
Choice 1
BICYCLE
BICYCLE
BOTTLE
BOTTLE
SPECTACLES
SPECTACLES
KNIFE
KNIFE
SCISSORS
SCISSORS
COMB
COMB
SWORD
SWORD
SWORD
SWORD
PISTOL
PISTOL
SCISSORS
SCISSORS
KNIFE
KNIFE
BICYCLE
BICYCLE
PIG
PIG
FLY
FLY
Choice 2
PISTOL
PISTOL
SPOON
SPOON
COMB
COMB
SCISSORS
SCISSORS
SWORD
SWORD
KNIFE
KNIFE
SCISSORS
SCISSORS
SPOON
SPOON
SPECTACLES
SPECTACLES
SWORD
SWORD
SCISSORS
SCISSORS
SWORD
SWORD
CAT
CAT
SWALLOW
SWALLOW
Choice 3
TRAFFIC-LIGHTS
TRAFFIC-LIGHTS
SCISSORS
SCISSORS
TRAFFIC-LIGHTS
TRAFFIC-LIGHTS
SWORD
SWORD
PISTOL
PISTOL
SCISSORS
SCISSORS
KNIFE
KNIFE
SAW
SAW
TRAFFIC-LIGHTS
TRAFFIC-LIGHTS
SPOON
SPOON
SWORD
SWORD
TRAFFIC-LIGHTS
TRAFFIC-LIGHTS
COW
COW
BEE
BEE
Choice 4
SPECTACLES
SPECTACLES
BICYCLE
BICYCLE
BOTTLE
BOTTLE
SPOON
SPOON
RAKE
RAKE
SAW
SAW
SPOON
SPOON
KNIFE
KNIFE
BOTTLE
BOTTLE
SAW
SAW
SPOON
SPOON
RAKE
RAKE
LEOPARD
LEOPARD
COCKEREL
COCKEREL
Choice 5
Category-specific deficits
337
338
M. A. Lambon Ralph et al.
Are living and non-living category-specific
deficits causally linked to impaired
perceptual or associative knowledge?
Evidence from a category-specific double
dissociation
Other assessment
M. A. Lambon Ralph, D. Howard,
G. Nightingale and A. W. Ellis
Lesion type
Abstract
Perhaps the most influential view of category-specific deficits is one in
which the dissociation between living and non-living kinds reflects differential reliance on, or weighting of visual or associative–functional
attributes. We present data collected from two patients, which question
the apparent relationship between category-specific deficits and loss of
specific attribute types. One patient with dementia of Alzheimer’s type
presented with relatively poor performance on living things but failed to
show a difference between knowledge of visual and associative–functional
information. The other patient with semantic dementia demonstrated
relatively poor knowledge of visual attributes but failed to exhibit a
category-specific impairment for animate kinds. In fact her comprehension and naming were slightly but significantly better for living things. The
data are discussed with reference to various theories of category-specific
impairment. We suggest that category-specific deficits for living things
probably results from a combination of atrophy to medial and neocortical
temporal structures, including the inferior temporal lobe. It is proposed
that at the behavioural level, category-specific deficits arise when both
critical identifying attributes of knowledge are lost and the intercorrelation between features causes disintegration of the category such that each
exemplar ‘regresses’ towards a category prototype.
Journal
Neurocase 1998; 4: 311–38
Neurocase Reference Number:
O122
Primary diagnosis of interest
Category specific deficits
Author’s designation of the case
IW and DB
Key theoretical issue
The functional and neuroanatomical correlates of category-specific
deficits for living things. Previous theories have suggested that
category-specific deficits are caused by loss of visual knowledge following damage to inferotemporal structures. Two patients are described
whose data seem to question this hypothesis. DB presented with a
category-specific deficit without differential attribute loss while IW
presented with a relative loss of visual knowledge (following atrophy
including the inferotemporal gyrus on the left) without a categoryspecific effect for living things (in fact the reverse)
Key words: semantic dementia; dementia of Alzheimer’s type; category
specificity; semantic memory; anomia
Scan, EEG and related measures
MRI
Standardized assessment
Visual Object and Space Perception Battery, Boston Naming Test,
Raven’s Progressive Matrices, WAIS-R, Warrington Recognition
Memory Test, Rey figure copy
Additional tests: Pyramid and Palm Trees, Test for the Reception of
Grammar, various picture naming tasks, picture and word comprehension
tests, attribute production and comprehension tests, object decision,
drawing from memory, delayed copying
None
Lesion location
DB: generalized cortical atrophy including the temporal neocortex and
particularly to medial temporal structures bilaterally.
IW: significant atrophy of the left temporal lobe, most pronounced at
the pole, which included significant reduction of the inferior temporal
gyrus. IW’s right temporal lobe and other cortical structures, including
the medial temporal area, appeared to be intact.
Progressive atrophy
Language
English