Neurocase (2001) Vol. 7, pp. 189–199

© Oxford University Press 2001

‘I Know your Name but not your Face’: Explaining
Modality-based Differences in Access to Biographical
Knowledge in a Patient with Retrograde Amnesia
Catherine Haslam, Michael Cook1 and Max Coltheart2
School of Psychology, University of Exeter, Exeter EX4 4QG, UK, 1Division of Psychology, The Australian National University,
Canberra and 2School of Behavioural Sciences, Macquarie University, Sydney, Australia

Abstract
In current functional models of person recognition it is proposed that there exists a single common body of semantic
information concerning people accessible from all modalities (e.g. names, faces). In this paper, we present evidence
that challenges this view. Two experiments are discussed investigating residual autobiographical and public knowledge
in a patient suffering from retrograde amnesia. Knowledge about people was investigated in each case by asking a
series of questions, ranging from the very general to the very specific. Experiment 1 examined knowledge about
famous people. The results showed that the patient accessed more information about famous people when cued with
names than when cued with faces. Experiment 2 examined knowledge about people known personally to the patient.
Again, the same pattern of results emerged. While the patient responded accurately to all questions posed with name
cues, her ability to recall the same information when prompted with face cues was clearly inferior. This modality-based
difference in accessing biographical knowledge is discussed in relation to models of person recognition.

Introduction
Biographical information is accessed through a variety of
sources. It can be cued by exposure to a person’s face, name
or even a semantic context. Irrespective of the original cue,
once biographical information is tapped it can be accessed
in response to all cue types. This notion of equal accessibility
is incorporated in current models of person recognition. In
this paper, we provide evidence that challenges this view.
Specifically, we present data from a patient who shows a
differentiation in her knowledge about people when cued
with faces and names. This patient retrieved more semantic
information about people when cued with their names than
with their faces.
There have been relatively few changes to Bruce and
Young’s (1986) original functional model of face recognition,
with perhaps one of the most significant being the introduction
of other modalities, such as names, in the recognition process
(Burton et al., 1990; Valentine et al., 1991; Burton and
Bruce, 1992, 1993; Brédart et al., 1995). Several stages
are involved in the process of recognizing and accessing
information about people. The first stage involves the formation of a representation that contains a visual description of
a seen face. This is compared with all known faces which
are believed to be contained in the face recognition units.

Name recognition units serve a similar function to face
recognition units, but are stored separately. These modalityspecific units contain the information required to identify an
individual’s face, name or even voice as familiar. Information
from these units is then combined in the person-identity
nodes (PINs), allowing recognition of a particular person,
and through this route semantic knowledge is activated.
Name retrieval occurs as a final stage in the process. These
processes are illustrated in Fig. 1.
The implication of this model is that semantic information
available via the face route should also be available via the
name route. However, recent evidence of modality-based
differences in access to semantic information about famous
people challenges this. Eslinger et al. (1996) found different
patterns of residual biographical knowledge in patients with
retrograde amnesia depending on lesion location. One patient,
EK, with more extensive damage in the left temporal region,
was unable to access information about famous people
from name cues but was relatively unimpaired in providing
information in response to faces. The reverse pattern was
evident in another patient, DR, whose damage was largely
confined to the right anterior and medial temporal region. In
this case, the patient was impaired in several tests of famous

Correspondence to: Catherine Haslam, School of Psychology, University of Exeter, Exeter EX4 4QG, UK. Tel: ⫹44 (0)1392 264626;
Fax: ⫹44 (0)1392 264623; e-mail: c.haslam@exeter.ac.uk

190 C. Haslam, M. Cook and M. Coltheart

Fig. 1. Central features of the functional model of person recognition.

face knowledge (including face–name matching with face
prompts and forced choice recognition), but performed normally on all tasks involving proper names. Based on this
evidence, Eslinger et al. (1996) suggested that biographical
information about famous people may be processed asymmetrically in the temporal lobe with the left hemisphere implicated in accessing and storing information associated with
names and the right hemisphere critical in knowledge associated with faces.
There are difficulties in explaining the findings for DR in
particular, using existing models of person recognition. In
the case of EK, a modality-based difference was found in
her familiarity ratings, with face familiarity better than name
familiarity. This suggests that EK’s superior knowledge with
faces may stem from a disconnection between the name
recognition units and PINs and, accordingly, fits the model.
The results for DR, however, are more difficult to explain.
His performance on the face–name matching task is probably
the most pertinent in attempting to identify the locus of
damage. On this task his performance was only impaired
with face cues. The task obviously requires more than mere
familiarity as a particular face must be associated with a
particular name. In fact, there was no discrepancy between
modalities in DR’s familiarity judgements, although both
were somewhat lower than that of the controls. Whether the
damage occurs at the level of the PINs or semantics might
be debated, but this does not change the fundamental difficulty

in explaining this result using the existing model of person
recognition. According to this model, there is no differentiation between modalities once the PINs are reached. Hence,
any modality-based difference that emerges from this level
onwards cannot be explained.
There are other reports of modality-based differences in
access to biographical information. For instance, McCarthy
and Warrington (1992) found a difference in the amount of
semantic information that their patient RFR could provide
about famous people when cued with faces and names. With
face cues, RFR could select the famous face contained in a
set of unknown faces, but failed to provide more detailed
information (e.g. such as identifying their profession). However, the information sourced by names was richer in detail.
Not only was RFR able to recognize famous names as
familiar, he could also provide the superordinate professional
role for the majority of names presented. There are two
possible explanations for RFR’s deficient knowledge in the
case of faces. It might be argued that the route between the
face recognition units and the PINs is damaged. Such an
explanation can be accommodated easily within existing
models of person recognition provided RFR based his fame
judgements on familiarity. Accordingly, any familiar face
would be judged as belonging to a famous person. However,
this is not the only knowledge that RFR may have used to
support such judgements. It could be argued that to ensure
accuracy in fame judgements, access to at least some semantic
information is required. If RFR accessed semantic information
to make fame judgements, then route damage alone would
not explain his performance. In this event, the damage must
lie in the store of semantic knowledge concerning faces and,
as already noted, current models cannot account for modality
differences in access to biographical information.
Evans et al. (1995) reported a similar modality-based
difference in access to biographical information in their
patient VH. On initial examination, VH provided detailed
semantic information about 39 of the 50 famous names
presented. When presented with the faces of these famous
people, VH could only provide identifying information in 11
cases. It was also reported that VH accessed general semantic
information in response to face cues (i.e. concerning a
person’s profession), but only in a small number of cases.
Thus, while VH could provide some identifying semantic
information in response to faces, he provided more detailed
information in response to names. The deficiency in VH’s
case is clearly at the level of the semantic store, but restricted
to a single modality.
In this paper, we present another case of differential access
to biographical knowledge. In this case though, the difference
emerged in both public and autobiographical domains. Two
experiments were conducted to examine knowledge in each
domain. In these experiments the patient was asked to answer
a series of questions about famous and personally familiar
people, ranging from the very general to the very specific.
The modality-based difference that emerged in our patient
adds to the existing evidence that challenges current opinion

‘I know your name but not your face’ 191

about the processes involved in accessing biographical
information.

Table 1. TG’s neuropsychological profile (complete results available in Haslam
et al., 1998)

Case history

General intellectual function
WAIS-R
Verbal IQ
Performance IQ
Full-scale IQ

TG was 36 years old when she was diagnosed with herpes
simplex encephalitis. As a result of her illness, TG developed
a dense retrograde and anterograde amnesia. Cranial magnetic
resonance imaging (MRI) reports noted altered signal intensity within the cortex of both temporal lobes, mainly involving
the medial temporal region, but extending around the anterior
pole of the right temporal lobe on to the lateral cortex. An
EEG report noted slow waves in the right temporal region.
Her case history and neuropsychological profile have been
described in detail previously (Haslam et al., 1998) and thus
only the relevant details will be summarized here. The
relevant test results are presented in Table 1.
TG’s retrograde amnesia in the autobiographical domain
is characterized by a dissociation between semantic and
episodic memory. The deficit to personal semantic (or factual)
knowledge covers a period of at least 20 years, while that to
personal autobiographical (or episodic) knowledge extends
back to childhood. Remote public knowledge was not examined given that the famous faces and events test in existence
was deemed to be inappropriate for an Australian population.
TG’s anterograde memory loss is profound and global. This
was evident in her inability to recall any information at all
following interference or a delay of several minutes. TG’s
deficit to remote and newly acquired knowledge occurs in
the context of preserved semantic memory as noted in her
performance in tests of general word knowledge, comprehension, picture naming and fluency. Visual perception and face
recognition was also preserved.

Experiment 1
The aim of experiment 1 was to determine how much
information TG could provide about famous people in
response to face and name cues. The examination focused
on TG’s remote memory only. A series of direct questions
was asked to probe knowledge for a range of details, from
the very general to the very specific, about these people (see
Fig. 1). Given her intact perception (see Table 1), if the
current model of person recognition is accurate, then the
amount of biographical information accessible from faces
and names should be equal. A finding to the contrary would
be inconsistent with this model and raise questions about the
manner in which biographical details are accessed.
An important point of clarification concerns the knowledge
hierarchy illustrated in Fig. 2. In this hierarchy, names are
the lowest level of specificity, as they represent one of the
most unique features about a person. However, in placing
names at this level, we are not suggesting that they comprise
the lowest level of knowledge hierarchies in general. The
classification of a low entry into a semantic hierarchy is
implied by its properties (e.g. knowing that someone is an

Anterograde memory
WMS-R
Verbal memory index
Visual memory index
General memory index
Attention/concentration index
Delayed recall index
RAVLT
Trials 1–5 ( /15)
Recall after interference
30 min recall
30 min recognition
Rey figure
Copy
3 min delay
30 min delay
Autobiographical remote memory
AMI
Personal semantic
Childhood
Early adult
Recent life
Autobiographical
Childhood
Early adult
Recent life
Modified Crovitz technique
(10 cue words)

101
95
99

Vocabulary
Comprehension

13
13

68
67
61
112
⬍50
4, 6, 6, 5, 5
1(⫹2)/15
0/15
13(⫹27)/15
36
0
0

19
15
10
0
0
0
8/30

Semantic memory (additional tests)
Letter fluency
39
Category fluency
Total animate
35
Total inanimate
36
Picture naming (BORB)
69/76
Comprehension
Token test
163/163
Visual perception
Benton’s facial recognition
(long conversion)
Benton’s judgement line
orientation (total)
BVRT: form discrimination

43/54
28/30
32/32

WAIS-R, Wechsler Adult Intelligence Scale-Revised; WMS-R, Wechsler
Memory Scale-Revised; BVRT, Benton’s Visual Retention Test; RAVLT, Rey
Auditory Verbal Learning Test; AMI, Autobiographical Memory Interview;
BORB, Birmingham Object Recognition Battery.

actor entails knowing that they are an entertainer and that
they are famous). There are no properties intrinsically associated with a name. Each name could belong to any branch in
the hierarchy. In this sense, it could be argued that the
association between a person’s name and their occupation,
for example, is the same as that between their face and
occupation. The decision to include names at the lowest level
was guided by specificity of reference. It does not reflect a
belief that knowledge systems are organized in this manner.
Considered from this point of view, while the hierarchy

192 C. Haslam, M. Cook and M. Coltheart

ated by ventricular enlargement requiring insertion of a
ventriculo-peritoneal shunt. GS is a 33-year-old male who
also suffered a brain haemorrhage, in this case as a result of
an aneurysm in the anterior communicating artery. Two
controls were matched to each patient on the basis of age
and education.

Materials

Fig. 2. The hierarchy of knowledge levels used in experiment 1.

shown in Fig. 1 is a hierarchy of specificity it is only partially
based on semantic associations.
Bearing this in mind, in this experiment names were used
in two ways. In the first instance, when faces were used as
cues, names were included at the most specific level (i.e.
level V) to examine knowledge of face–name associations.
Second, the person’s name was used to cue information about
their profession at various levels (i.e. levels I–IV). In this
second usage, the association between the person’s name and
information about their profession was considered to be
similar to that between the person’s face and information
about their profession. In other words, from this point of
view, the names and faces used in this experiment are
arbitrary stimuli associated with a semantic complex relating
to profession.

Design
The independent variables were cue type and knowledge
level, both of which were manipulated within subjects. Cue
type had two levels, faces and names. Five knowledge levels
were examined with faces. These comprised: fame status,
general occupational category (entertainer/not entertainer),
basic occupational category (actor/sports person), specific
occupational category (subtype of actor and subtype of sports
person) and name. Knowledge at the first four levels was
also examined using name cues. The dependent measure was
memory performance, determined on the basis of responses
to yes/no recognition.

Participants
Nine people took part in the experiment—three amnesic
patients (TG, GS and IP) and six matched controls. Two
patients, GS and IP, suffered only from the anterograde form
of amnesia and were included as additional controls. IP is a
17-year-old male who became amnesic following an intracerebral haemorrhage, due to an aneurysm at the bifurcation
of the left internal carotid artery. His condition was complic-

The materials consisted of a set of face cue cards and two
sets of name cards. The face cues comprised 40 faces, half
of which were famous and the other half were not famous.
The famous people comprised 10 actors and 10 sports people.
In each category, half were male. The names of the famous
people are listed in Appendix 1. They were selected on the
basis of being currently famous and salient to an Australian
population. They were initially chosen on the basis of
responses to a pilot survey requesting the names of famous
individuals (n ⫽ 10) and were subsequently identified as
highly recognizable by a small group of first year psychology
students (n ⫽ 9, mean age ⫽ 21 years, SD ⫽ 2.1). Salient
famous people were selected to maximize the probability
that they were known to patients prior to the onset of amnesia.
The pictures did not contain any features that would provide
clues about the occupation of the subject. They were scanned
into a computer using Adobe Photoshop software to control
for factors such as colour, size and background. All faces
were printed in black and white on a 600 dpi laser printer
and were scaled to a width of 8 cm and a height of between
10 and 12 cm. The 20 non-famous faces were collected from
advertisements in magazines. They were matched to the
famous people as closely as possible in terms of age, sex
and attractiveness and were also scanned and reproduced
using the above procedure.
There were two sets of name cards. One set was used to
test knowledge about famous people. This set consisted of
40 names. Half of these names were those of famous people
and the remainder were unknown. The latter names were
constructed by selecting common surnames from the local
telephone directory (i.e. at least 15 people or more with that
surname) and adding a common first name to each one. The
second set of name cards was used with face cues to examine
knowledge of face–name associations and there were two
types of cards. One type contained a single name and the
other type contained a pair of names. Half of the single name
cards contained correct names and the remainder were
incorrect. The incorrect names were produced by changing
the second syllable of the surname (e.g. ‘Mel Gibson’ was
changed to ‘Mel Gibbons’, see Appendix 1). The name pair
cards contained two famous names from the same gender
and basic occupational category (e.g. two actors or two sports
people). On half of the trials the correct name was the first
item in the name pair and on the remaining trials it was the
second item.

‘I know your name but not your face’ 193

Procedure
Knowledge was tested in two ways—using face cues and
name cues. In the case of patients, knowledge was examined
first with face cues and then with name cues. The order of
cue presentation was counterbalanced across the two controls
matched to each patient. A short break of 10 min was
provided between the two phases of testing.
Five knowledge levels were examined with face cues and
knowledge at each level was elicited by asking questions.
The questions ranged from the very general to the very
specific. Specificity was determined by the number of people
who shared the attribute in question. The attribute was shared
by 20 people at the most general knowledge level, and this
number decreased progressively as the questions became
more specific. The questions presented at each level were
as follows:
level I: ‘Is this person famous?’
level II: ‘Is this person primarily in the entertainment industry?’
level III: ‘Is this person an actor?’ or ‘Is this person a
sportsperson?’
level IV: ‘Is this person a comic actor (i.e. appeared in
comedy)?’ or ‘Is this person a tennis player?’
level V: ‘Is this person’s name _______?’
Knowledge at levels I–V was assessed progressively, in
descending order. The entire set of 40 face cues (both famous
and non-famous) were presented individually and randomly
at level I. From this level onwards, only the famous faces
were presented. At two knowledge levels (i.e. levels III and
IV), one of two questions was presented with faces. For half
of the items at level III, the participants were asked whether
the face was that of an actor and for the remainder whether
the face was that of a sports person. At level IV, the
participants were asked to judge whether actors were comic
actors and whether sports people were tennis players. Knowledge of face–name associations was examined at level V.
Knowledge at the first four levels was re-examined with
name cues using the same questions. As with the case for
faces, 40 name cues were presented at level I and 20 name
cues were presented at the remaining levels. All level I items
were presented before proceeding to level II and so on until
knowledge at the four levels had been examined.

Results and discussion
The face and name cue data for TG and patients with
anterograde amnesia only are provided in Table 2. Both hit
and false alarm rates are provided for each cue type.
Faces.
As expected, TG (the only retrograde amnesic) was the only
patient who had difficulty answering questions associated
with famous people. She was significantly better than chance
in discriminating famous from non-famous faces at level I

(P ⬍ 0.01), but her performance declined at the next three
levels. Thus, she had insufficient knowledge about famous
people to discriminate entertainers from non-entertainers,
actors from sports people and subtypes of actor as well as
subtypes of sports person.
TG’s performance at level V was unexpected. From this
result it appears that she can access sufficient information
about famous people to make accurate face–name associations
despite her inability to make less specific judgements at the
previous three levels. However, it is also possible that her
responses to these questions were based on familiarity with
names rather than knowledge of face–name associations.
This is because incorrect names would be sufficiently different
from studied names to be classified as unfamiliar. For
instance, TG may have decided that the name ‘Mel Gibbons’
was unfamiliar and on this basis judge that the association
is incorrect. To distinguish between these alternatives, the
famous faces were presented again but this time with two
famous names. The incorrect name in each pair was that of
another famous person included in the test sample (see
Appendix 1) from the same professional category (e.g. if the
target name belonged to an actor then the incorrect name
also belonged to an actor). TG’s task was to identify the
name that matched the face. Accurate performance on this
two-alternative forced choice task requires specific knowledge
of face–name associations. TG’s performance on this latter
task was no better than chance. She only matched 14 of the
20 face–name pairs correctly (χ2(1) ⫽ 2.5, ns; Yates’ correction
for continuity applied here), indicating that the result at level
V was not based on knowledge of particular face–name
associations but rather on a lack of familiarity with nonfamous or novel names. Overall, TG’s results with face cues
indicate that she was capable of accessing some knowledge
directly about famous people when cued with faces, but that
this was restricted to the most general level examined. In
contrast, the remaining patients (IP and GS) and matched
controls were significantly better than chance in their
responses to the questions at levels I–V (P ⬍ 0.01 at all
levels) and to items in the additional two-alternative forced
choice task (χ2(1) ranged between 14.45 and 18.05, all
Ps ⬍ 0.01).
Names.
The name cue data for patients are also presented in Table 2.
Again, TG was the only participant who showed any evidence
of impairment on this task, but only at the most specific level
of knowledge examined. She was significantly better than
chance in discriminating names on the basis of fame (P ⬍
0.01), general occupational category (P ⬍ 0.01) and basic
occupational category (P ⬍ 0.01). However, she did not
discriminate between subtypes of actor and subtypes of sports
person. The errors she made in the latter discrimination were
distributed fairly equally across the two occupational groups.
IP, GS and the matched controls responded accurately at all
four knowledge levels examined (P ⬍ 0.01 at all levels).
It was possible that TG’s performance with names was

194 C. Haslam, M. Cook and M. Coltheart
Table 2. Hit rates (HR), false alarm rates (FAR) and P valuesa for face and name cues at knowledge levels investigated for famous people in patients with
retrograde (TG) and anterograde (IP and GS) amnesia
TG
Faces
Level

HR

FAR P

I
II
III
IV
V

0.7
0.6
0.5
0.6
0.9

0.1
0.3
0.2
0.5
0.1

⬍0.01
ns
ns
ns
⬍0.01

Anterograde amnesic (IP)

Anterograde amnesic (GS)

Names

Faces

Names

Faces

HR

FAR P

HR

FAR P

HR

FAR P

HR

FAR P

1.0
0.9
1.0
0.9
–

0.05
0
0.1
0.5
–

1.0
1.0
1.0
1.0
1.0

0.15
0
0
0.1
0

1.0
1.0
1.0
0.8
–

0.05
0
0
0
–

0.95
1.0
1.0
1.0
1.0

0.05
0
0
0.1
0

⬍0.001
⬍0.01
⬍0.01
ns
–

⬍0.001
⬍0.01
⬍0.01
⬍0.01
⬍0.01

⬍0.001
⬍0.01
⬍0.01
⬍0.01
–

Names

⬍0.001
⬍0.01
⬍0.01
⬍0.01
⬍0.01

HR

FAR P

1.0
1.0
1.0
0.8
–

0
0
0
0
–

⬍0.001
⬍0.01
⬍0.01
⬍0.01
–

aP value for Fisher’s exact test provided except when the number of items exceeded 30, in which case the P value for χ2 is provided.
ns, not significant.

facilitated by prior exposure to the same questions, as they
had been presented first with face cues. To examine this, the
experiment was repeated 3 months later, but this time the
order of cue presentation was reversed, i.e. knowledge was
examined first with name cues and then with face cues.
Despite the reversal of cues, the results on repeat testing
were consistent with those obtained on initial examination.
TG was significantly better than chance in discriminating
names at the first three levels (P ⬍ 0.01 at all three levels),
but could only discriminate faces at the most general level
(P ⬍ 0.05).
In summary, the results of this experiment show that TG
accessed considerably more information about famous people
from their names than she did from their faces. This pattern
emerged on initial testing and was confirmed on re-examination of TG’s knowledge several months later, indicating that
the original finding was not an isolated result. Given the
differentiation observed in TG’s knowledge about people in
the public domain, one might ask whether it also occurs in
the autobiographical domain. This question was investigated
in experiment 2.

Experiment 2
In experiment 2, TG’s knowledge about personally familiar
people was examined to determine whether a modality-based
difference emerged in access to autobiographical information.
As in experiment 1, a series of direct questions was asked
to elicit knowledge about people known personally to TG.
The questions probed knowledge at a range of levels, from
the very general to the very specific.

Design
The independent variables were cue type and knowledge
level, both of which were manipulated within subjects. Cue
type had two levels, faces and names. Five knowledge levels
were examined with faces, but only four were examined with
names. The dependent measure was memory performance

and this was determined on the basis of responses to yes/no
recognition.

Participants
One patient with retrograde amnesia, TG, took part in
this study.

Materials
Two sets of material were used—one set of face cards and
one set of name cards.
The set of face cards consisted of 66 faces. There were
two categories of faces. One category consisted of faces that
were familiar to TG (n ⫽ 46) and the other consisted of
novel or unknown faces (n ⫽ 20). The familiar faces
comprised: family members (n ⫽ 10), friends (n ⫽ 10),
friends’ children (n ⫽ 6) and famous people (n ⫽ 20).
Pictures of people were either collected from magazines or
supplied by TG’s husband. Both famous and unknown faces
were matched as closely as possible to those of family
members and friends on the basis of age, sex and attractiveness. A list of the famous people used in this experiment
is provided in Appendix 2. It consists of television, movie,
sporting and royal personalities. The pictures/photographs
were scanned into a computer and were reproduced using
the same procedure discussed for the previous experiment.
They were printed in black and white, had a neutral background and were 5 cm in width and 6–7 cm in height.
The name cue cards consisted of 68 names (first name and
surname pairs) and these belonged to familiar (n ⫽ 48) and
novel (n ⫽ 20) people. The familiar names comprised: family
members (n ⫽ 10), friends (n ⫽ 18) and famous people
(n ⫽ 20, also used with face cues, with the exception of
Queen Elizabeth II who was replaced with Elizabeth Taylor).
Each of these names was matched to those of family members
and friends on the basis of sex and syllable length (e.g. if
the first name of a family member had two syllables then
the novel first name also had two syllables). In addition,

‘I know your name but not your face’ 195

several of the novel surnames were repeated as there were
family members and friends who shared the same surname.

Procedure
There were two phases. In the first phase, knowledge about
people was examined using name cues and in the second
phase, knowledge was examined using face cues. A short
break of 10 min was provided between the two phases
of testing.
Four levels of knowledge were examined with name cues.
The materials used to test knowledge at these levels were
selected and prepared with the assistance of TG’s husband.
Names of people with the same surname and maiden name
as TG were not included because judgements involving these
names could be made on the basis of familiarity rather than
on the basis of knowledge about the particular person in
question. If the surname was the same as TG’s, then she may
have decided that the person is related to her without knowing
anything about them. Questions were presented with each
cue and these tapped a range of details from general to
increasingly more specific. Specificity was determined on the
basis of the amount of detailed knowledge required to answer
the questions. The number of items presented at each level
differed as did the category of name cue (i.e. whether the
person was family, friend, famous or unknown). The questions
asked at each level were as follows:
level I: ‘Is this person familiar?’
level II: ‘Is this person a family member/friend?’
level III: ‘Is this person family?’
level IV: ‘Is this person from (TG’s husband’s) side of the
family?’ ‘Is this an old (⬎ 10 years) friend?’
At the first level, TG was asked to make a judgement of
familiarity. At level II, knowledge of general relationships
was examined by asking TG to discriminate personally
familiar from famous people. The discrimination at level III
tested knowledge of basic relationships, and this involved
differentiating family members from friends. At level IV,
specific knowledge about family relationships and friendships
was examined. Knowledge of family relationships was examined by asking TG to discriminate between names belonging
to people from her side and her husband’s side of the family.
In the case of friendships, she was asked to discriminate old
friends from new. All items at level I were presented before
proceeding to level II and so on until knowledge at all levels
had been examined.
After a short break (about 10 min), knowledge was tested
with face cues. This time knowledge was examined at five
levels. The questions at each level were as follows:
level I: ‘Is this person familiar?’
level II: ‘Is this person a family member/friend?’
level III: ‘Is this person family?’
level IV: ‘Is this person from (TG’s husband’s) side of the
family?’ ‘Is this (X’s) child?’
level V: ‘Is this person’s name _______?’

The questions asked about people at levels I–III with name
cues were the same as those presented with face cues.
However, at level IV, a change in procedure was introduced
to address the problem of insufficient materials. The number
of items presented with the family relationship question was
reduced and this represented only a minor modification. The
most significant modification involved a change in the items
and question presented to test knowledge of friendships. To
test knowledge of specific friendships, TG was asked to
discriminate between children on the basis of whether they
belonged to a particular friend (i.e. X) or other friends. To
answer this question correctly, TG would need to have
knowledge of the parent–child relationships of her friends.
Thus, although the question was not the same as that used
with name cues, it nevertheless probed specific knowledge
about friends. A final question was asked to elicit knowledge
of face–name associations for family members and friends.
Only real names were included but half were correctly paired
with faces and half were incorrectly paired.

Results and discussion
Accuracy of memory performance at each level with each
cue type is provided in Table 3. With name cues, TG’s
responses were significantly better than chance at all levels
examined. That is, she was significantly better than chance
in discriminating known from unknown names (P ⬍ 0.001),
personally familiar from famous names (P ⬍ 0.001), and the
names of family members from friends (P ⬍ 0.01). The
responses to the two questions at level IV were combined
and the results indicated that TG could identify the names
of people from her husband’s side of the family and the
names of old friends (P ⬍ 0.01). Errors at the latter level
were only made in response to questions eliciting knowledge
about friends. Thus, TG’s performance with name cues
indicated that she had access to a range of knowledge about
people known to her personally.
In contrast, performance with face cues was poor. TG was
only accurate in discriminating faces at the first two levels.
That is, she was significantly better than chance in discrimin-

Table 3. Hit rates (HR), false alarm rates (FAR) and P valuesa for face and
name cues at knowledge levels investigated for people known personally to TG
Faces

Names

Level

HR

FAR

P

HR

FAR

P

I
II
III
IV
V

0.95
0.95
1.0
0.8
0.6

0.05
0.1
0.8
0.3
0.2

⬍0.001
⬍0.001
ns
ns
ns

1.0
1.0
1.0
1.0
–

0
0
0.2
0.4
–

⬍0.001
⬍0.001
⬍0.01
⬍0.01
–

aP value for Fisher’s exact test provided except when the number of items

exceeded 30, in which case the P value for χ2 is provided.
ns, not significant.

196 C. Haslam, M. Cook and M. Coltheart

ating known from unknown faces (P ⬍ 0.001) and personally
familiar from famous faces (P ⬍ 0.001). When making
the latter judgement in response to famous faces, TG
frequently stated that while she ‘would like the person to
be a family member/friend, they were not’, suggesting that
she had access to more detailed information about these
people. The same statement was never provided in response
to personally familiar faces. While this evidence is only
anecdotal, it suggests that her level II judgements were
based on more than mere familiarity. TG’s performance
with face cues broke down at level III and this result can
be directly compared with that for name cues. Knowledge
about the same people was examined at level III with
name cues, and TG’s performance in this case was
significantly better than chance. This indicates that information available in one modality was not available in another.
Performance at the remaining levels, examining knowledge
about specific family relationships and friendships as well
as face–name associations, was also no better than chance.
These results confirm the existence of a differentiation
in TG’s memory performance between face and name
cues, but this time in the autobiographical domain.
Performance with name cues was again clearly superior
to that with face cues. In the context of understanding the
breakdown in amnesia, one might ask whether TG’s deficit
in the case of faces is one of storage or retrieval. An
answer to this question was provided, at least in the
autobiographical domain by a supplementary experiment.
In the case of face cues in experiment 2, TG failed to
access sufficient knowledge on direct questioning to
differentiate between family members and friends (i.e. at
level III) and then to make more specific judgements.
Using an indirect procedure (see Haslam et al., 1998 for
a description of the procedure) as a means of facilitating
access to knowledge, we found that TG could make this
discrimination with a high degree of accuracy. Specifically,
she was asked to identify faces according to whether they
were from her side of the family, from her husband’s side
of the family, an old friend or a new friend. This
categorization enabled examination of knowledge at two
levels—general and specific. The four-alternative forced
choice judgement examined specific knowledge of these
relationships. These four categories can also be classified
at a more general, or higher level, of knowledge involving
family members and friends. This latter discrimination was
one that TG failed to make in experiment 2 when
given direct instructions. However, under these conditions,
knowledge of family and friends can be revealed indirectly
in response confusions when categorizing faces in the
lower level task. While TG failed to make the specific
discrimination she rarely misclassified a family member as
a friend or vice versa (χ2(1) ⫽ 22.3, P ⬍ 0.001). In order
to make this higher level discrimination, information
concerning family members and friends must be available
in storage. In other words, TG’s knowledge about these
people at least at the level investigated was not lost,

merely inaccessible (see Haslam, 1998, Chapter 8 for a
more thorough description of this experiment).

General discussion
The two studies reported in this paper investigated residual
biographical knowledge in a patient with retrograde
amnesia. The distinctive feature in this patient’s memory
performance was the difference in the amount of biographical information she could access in response to face and
name cues—a difference that cannot be explained by
existing models of person recognition. This difference
emerged in experiment 1, examining knowledge in the
public domain, and was replicated in experiment 2,
examining knowledge in the autobiographical domain.
In experiment 1, TG only accessed sufficient information
to support fame judgements when cued with famous faces.
Yet when presented with famous names, she accessed
additional information to allow her to discriminate between
people on the basis of their status as an entertainer (i.e.
at level II) and their basic occupational category (i.e. at
level III involving the actor/sports person discrimination).
In each modality, TG’s knowledge breaks down when
attempting to access semantic information (i.e. at the level
of semantic information units), but the level of biographical
detail available differs. In the case of faces, accurate fame
judgements require more than mere familiarity with a
person’s face and, hence, can only be supported by access
to basic semantic information. In the case of name cues,
TG accessed more detailed knowledge, sufficient to
discriminate between people at the level of their basic
occupational category. While this judgement also requires
semantic knowledge, it represents a level of detail that
arguably exceeds that available with face cues. If semantic
information is common to all modalities, then TG’s ability
to access additional biographical details in response to
name cues cannot be explained.
A similar difficulty was encountered in experiment 2 in
which autobiographical knowledge was investigated. In this
experiment, TG responded accurately to all questions posed
with name cues (i.e. involving four levels of knowledge),
but was only accurate at the first two levels with face
cues. In other words, detailed semantic information was
available in response to name cues, whereas superficial
information only was available in response to face cues.
Again the breakdown, at least in the case of faces, appears
to involve semantic knowledge. TG was capable of
accessing sufficient information to discriminate personally
familiar from novel people at level I and personally
familiar from famous people at level II. Clearly, the former
judgement requires nothing more than access to the face
recognition units. While the latter judgement may also
have been based on familiarity (i.e. given the greater
strength of familiarity with people known personally to
TG), we would argue that in TG’s case semantic knowledge
was involved. For many of the famous items, TG indicated

‘I know your name but not your face’ 197

that although she would like to be associated with the
person in question, she knew they were neither family nor
friend. This suggests that her judgement was based on
more than familiarity with the person’s face. Furthermore,
if she based her judgement on some other superficial
criterion, such as attractiveness, her performance at this
level would have been far from perfect—many of the
famous faces used in experiment 2 would not be considered
attractive. In fact, TG made only three errors (misses and
false alarms combined) at this level, where 40 items were
presented. Of course it remains possible that TG did not
utilize semantic knowledge when judging whether the
person was a family member/friend. However, this seems
unlikely when the accuracy of her performance is considered
together with the comments she provided when making
these judgements. If we accept the latter argument, that
TG had access to some biographical information in relation
to faces, then we encounter the same difficulty in explaining
why semantic information available in one modality cannot
be accessed by another. This time, though, the difficulty
arises in the context of autobiographical knowledge.
TG’s failure to access some biographical details in
response to face cues is best explained by dysfunction at
the level of the semantic information units and this is
where the problem arises in relation to models of person
recognition. In its present form, there is no distinction
between modalities in access to biographical information—
information available in one modality is available in all
modalities. TG’s superior performance with name cues
relative to face cues challenges this assumption. On
superficial analysis, it might be argued that the discrepancy
arises because information associated with names is more
readily accessible or less vulnerable than that associated
with faces. If this was the case, then differences in
processing complexity should be represented somehow in
existing models. However, the suggestion is incompatible
with evidence of the reverse pattern where knowledge
associated with faces is superior to that associated with
names. This occurred in Eslinger et al.’s (1996) patient,
EK. Her evidence alone is sufficient to question the
relevance of processing complexity arguments in the
context of person recognition.
Another explanation for the presence of modality-based
differences is offered by Eslinger et al. (1996). As noted
in the introduction, these researchers raised the possibility
that semantic information may be modality specific.
Biographical information associated with names may be
stored and processed separately from that associated with
faces. Taken at face value, TG’s data appear to support
Eslinger et al.’s proposal. Nonetheless, there is limited
appeal in this adaptation of the model, primarily due to
the storage of redundant information. Information specific
to faces would be repeated with names and this is not an
efficient means of storage. Although the present findings
do not provide an explanation for the differentiation in
TG’s performance, they raise questions about the accuracy

of current functional models of person recognition. The
findings also bear on the broader question of the existence
of modality-specific subsystems in semantic memory (see
special issue on category-specific deficits, Neurocase, 1998,
Volume 4, issue 4/5). Although this proposal appears to
have fallen out of favour in the general semantic memory
literature (see Caramazza, 1998), our findings seem to
support a modality-specific structure in the case of
biographical knowledge.
Despite TG’s superior performance with name cues, her
knowledge about people in this modality was not entirely
preserved and this was demonstrated most clearly in
our examination of knowledge about famous people in
experiment 1. This is important to note for two reasons.
First, it indicates that TG’s pathology is best characterized
by amnesia and not prosopagnosia. Were she prosopagnosic,
then her knowledge about people in all other modalities
apart from faces should be preserved. Second, her failure
at particular knowledge levels with name cues highlights
a pattern in memory performance—namely, that TG’s
remote public memory is characterized by preservation of
general level knowledge and impairment of specific
knowledge. TG was able to access more general facts
about famous people in response to name cues, but was
no better than chance at the most specific level examined
(i.e. differentiating between comic and non-comic actors
as well as between tennis players and all other sports
people). Not only did this pattern emerge when testing
knowledge about famous people in response to name cues,
it also emerged with face cues in both public and
autobiographical domains. A good example of the latter
is provided in TG’s performance in experiment 2 where
she was only accurate in providing information at the first
two levels of knowledge examined. This pattern of
preserved high level relative to low level knowledge, or
preservation of memory for generalities, is one that we
have reported previously, but in the context of new learning
(Haslam et al., 1997 and unpublished thesis, 1998).
Preservation of memory for generalities was found in the
public domain with both cue types and replicated in the
autobiographical domain with face cues. This indicates
that preservation of memory for generalities occurs in
retrograde amnesia as well as anterograde amnesia.
In conclusion, this paper adds to the existing evidence
of modality-based differences in access to person-related
knowledge in the public domain and elaborates upon this
by providing evidence of the same phenomenon in the
autobiographical domain. In doing so, it raises our
awareness of a possible faulty assumption in models of
person recognition—namely, that semantic information is
accessed equally in all modalities. While an explanation
for TG’s results can be provided, it is important to note
that it is not particularly appealing when considering
efficiency of storage. This provides a challenge either to
provide a reasonable explanation for storage of redundant

198 C. Haslam, M. Cook and M. Coltheart

information or to develop alternative explanations for
modality-based differences in accessing biographical details.

Appendix 1

References

Actors
Kim Basinger
Jodie Foster
Nicole Kidman
Michelle Pfeiffer
Meg Ryan
Bruce Willis
Mel Gibson
Tom Hanks
Hugh Grant
Kevin Costner

Modified name (if used)
Kim Basinford
–
–
–
Meg Ryder
Bruce Williams
Mel Gibbons
–
–
Kevin Costin

Sports people
Kathy Freeman
Steffi Graf
Conchita Martinez
Melinda Gainsford
Monica Seles
André Agassi
Kieren Perkins
Pete Sampras
Laurie Daley
Greg Norman

Modified name (if used)
–
Steffi Grant
–
Melinda Gainey
–
–
Kieren Perley
Pete Sampson
–
Greg Norden

Brédart S, Valentine T, Caldor A, Gassi L. An interactive activation model of
face naming. The Quarterly Journal of Experimental Psychology 1995; 48:
466–86.
Bruce V, Young AW. Understanding face recognition. British Journal of
Psychology 1986; 77: 305–27.
Burton AM, Bruce V. I recognize your face but I can’t remember your name:
A simple explanation? British Journal of Psychology 1992; 83: 45–60.
Burton AM, Bruce V. Naming faces and naming names: Exploring an
interactive activation model of person recognition. Memory 1993; 1: 457–80.
Burton AM, Bruce V, Johnston RA. Understanding face recognition with an
interactive activation model. British Journal of Psychology 1990; 81:
361–80.
Caramazza A. The interpretation of semantic category-specific deficits: What
do they reveal about the organization of conceptual knowledge in the brain?
Neurocase 1998; 4: 265–72.
Eslinger PJ, Easton A, Grattan LM, Van Hoesen GW. Distinctive forms of
partial retrograde amnesia after asymmetric temporal lobe lesions: Possible
role of the occipitotemporal gyri in memory. Cerebral Cortex 1996; 6: 530–9.
Evans JJ, Heggs AJ, Antoun N, Hodges JR. Progressive prosopagnosia
associated with selective right temporal lobe atrophy: A new syndrome?
Brain 1995; 118: 1–13.
Haslam C. Memory for generalities in amnesia. Canberra: The Australian
National University. Thesis, 1998.
Haslam C, Coltheart M, Cook M. Preserved category learning in amnesia.
Neurocase 1997; 3: 337–47.
Haslam C, Cook M, McKone E. Memory for generalities: Access to higherlevel categorical relationships in amnesia. Cognitive Neuropsychology 1998;
15: 401–37.
McCarthy RA, Warrington EK. Actors but not scripts: The dissociation of
people and events in retrograde amnesia. Neuropsychologia 1992; 30:
633–44.
Valentine T, Brédart S, Lawson R, Ward G. What’s in a name? Access to
information from people’s names. European Journal of Cognitive Psychology
1991; 3: 147–76.

Received on 15 March, 1999; resubmitted on 12 April, 2000; accepted
on 5 September, 2000

List of famous people used in experiment 1, together with modified surnames

Appendix 2
List of famous people used in experiment 2
Flo Bjelke Peterson
Jean Kitson
Queen Elizabeth II/Elizabeth Taylor
Barbara Stanwyck
Alicia Silverstone
Tony Pearen
Kylie Minogue
Susan Olsen
Anna Paquin
Denise Roberts

Rowena Wallace
Andrew Lloyd-Webber
Rob de Castella
Grant Kenny
Gene Hackman
Clint Eastwood
James Richter
Derryn Hinch
Lloyd Bridges
Macauley Culkin

‘I know your name but not your face’ 199

‘I know your name but not your face’:
explaining modality-based differences in
access to biographical knowledge in a
patient with retrograde amnesia
C. Haslam, M. Cook and M. Coltheart
Abstract
In current functional models of person recognition it is proposed that there
exists a single common body of semantic information concerning people
accessible from all modalities (e.g. names, faces). In this paper, we present
evidence that challenges this view. Two experiments are discussed investigating
residual autobiographical and public knowledge in a patient suffering from
retrograde amnesia. Knowledge about people was investigated in each case
by asking a series of questions, ranging from the very general to the very
specific. Experiment 1 examined knowledge about famous people. The results
showed that the patient accessed more information about famous people when
cued with names than when cued with faces. Experiment 2 examined
knowledge about people known personally to the patient. Again, the same
pattern of results emerged. While the patient responded accurately to all
questions posed with name cues, her ability to recall the same information
when prompted with face cues was clearly inferior. This modality-based
difference in accessing biographical knowledge is discussed in relation to
models of person recognition.

Journal
Neurocase 2001; 7: 189–99

Neurocase Reference Number:
O217

Primary diagnosis of interest
Herpes simplex encephalitis resulting in retrograde amnesia

Author’s designation of case
TG

Key theoretical issue
d Impact of modality differences in accessing biographical knowledge on

existing models of person recognition
Key words: person recognition; biographical knowledge; modality differences

Scan, EEG and related measures
MRI, EEG

Standardized assessment
Wechsler Adult Intelligence Scale-Revised (WAIS-R), Wechsler Memory
Scale-Revised (WMS-R), Rey Auditory Verbal Learning Test (RAVLT), Rey
figure, Autobiographical Memory Interview (AMI), modified Crovitz, letter
and category fluency, picture naming, Token test, Benton’s facial recognition,
judgement of line orientation, visual retention tests

Other assessment
Autobiographical and public knowledge tests

Lesion location
d Bilateral temporal lobes, more extensive in right hemisphere

Lesion type
Consequence of herpes simplex encephalitis

Language
English