Content uploaded by Geoffrey Bird
Author content
All content in this area was uploaded by Geoffrey Bird on Sep 26, 2022
Content may be subject to copyright.
1
Individuals with developmental prosopagnosia show independent
impairments in face perception, face memory and face matching
Mirta Stantić1*, Zoë Pounder1*, Sarah Bate2, Tirta Susilo3, Caroline Catmur4, Geoffrey Bird1
* These authors contributed equally.
1 Department of Experimental Psychology, University of Oxford
2 Department of Psychology, Bournemouth University
3 School of Psychology, Victoria University of Wellington
4 Department of Psychology, King’s College London
Corresponding author**
Email address: z.pounder@psy.ox.ac.uk (Z. Pounder)
2
Abstract
Individuals with developmental prosopagnosia (DP) all exhibit impairments in face memory,
but the specificity of these face memory impairments is debated. One problem is that
standard behavioural tasks are not able to provide independent measurement of face
perception, face memory, and face matching (the decision process required to judge whether
two instances of a face are of the same individual or different individuals). The present study
utilised a new test of face matching, the Oxford Face Matching Test (OFMT), and a novel
analysis strategy to derive these independent indices. Twenty-nine individuals with DP and
the same number of matched neurotypical controls completed the OFMT, the Glasgow Face
Matching Test, and the Cambridge Face Memory Test. Results revealed individuals with DP
exhibit impairments in face perception, face memory and face matching. Collectively, these
results suggest that face processing impairments in DP are more comprehensive than has
previously been suggested.
Keywords: face perception; face matching, face memory, developmental prosopagnosia
3
Introduction
Developmental prosopagnosia (DP) is a neurodevelopmental condition characterised
by a lifelong inability to recognize faces (e.g., Behrmann & Avidan, 2005; Cook & Biotti,
2016; Susilo & Duchaine, 2013). While DP is, by definition, associated with problems
remembering faces, the particular aspect of face processing responsible for these memory
problems is unclear. Specifically, there is an ongoing debate as to whether individuals with
DP are able to form intact perceptual representations of faces but have difficulty
learning/recalling facial identities (‘memory hypothesis’; Jackson, Counter & Tree, 2017;
Stollhoff, Jost, Elze & Kennerknecht, 2011), or whether individuals with DP have difficulties
with both forming perceptual representations of faces and face memory (‘perceptual
hypothesis’; Biotti, Gray & Cook, 2019; Dalrymple, Garrido & Duchaine, 2014, see also
Towler, Fisher & Eimer, 2018). This debate is ongoing partly due to the heterogeneity of DP,
specifically the possibility that there may be subtypes of DP (e.g., Dalrymple, Corrow, Yonas
& Duchaine, 2012) or variants (e.g., apperceptive or associative) that differ in terms of the
nature of their face-processing impairment (Biotti et al., 2019; Corrow, Dalrymple & Barton,
2016). Nevertheless, little attention has been paid in DP to a third process, face matching,
which is also necessary for face recognition. It is worth noting that the term ‘Face Matching’
is used here in a very specific, and perhaps unusual, way. In contrast to the standard use of
the term face matching, which refers to a set of tasks in which participants are required to
judge whether two photographs of a face depict the same individual or different individuals,
face matching in this context relates to the decision-making process necessary to determine
whether two or more face images are of the same individual, or different individuals. Face
matching is the psychological process necessary for successful performance on face matching
tasks. Face matching is also required, however, when a participant is asked to determine
whether a face stimulus matches that stored in memory, such as when deciding whether a
4
photograph of a face matches a recently-learned identity. Although the former task is called a
face matching task, and the latter called a face memory or face recognition task, the decision-
making process (deciding whether two face instances are of the same individual or different
individuals, regardless of whether one of those instances is stored in memory) is the same,
and what we call face matching.
Problematically, existing tests are largely unable to isolate these distinct face
processes. For example, whether individuals with DP are impaired at face memory is
normally tested using the Cambridge Face Memory Test (CFMT; Duchaine & Nakayama,
2006), a task in which participants have to identify a pre-learned facial identity among
distractors. Although face memory is required in order to perform accurately on this task,
face perception and face matching are also required. Therefore, it is unclear whether poor
performance on the CFMT reflects impaired face memory, face perception or face matching,
or any combination of these processes. Similarly, face perception is often assessed using face
matching tasks such as the Glasgow Face Matching Test (GFMT; Burton, White & McNeill,
2010; see also Kent Facial Matching Test; Fysh & Bindemann, 2018), in which two facial
images are presented and participants are required to judge whether the facial images are of
the same person or different people. Matching tasks require intact face perception and face
matching, and so impaired performance on matching tasks in DP may reflect either one, or
both, of impaired face perception and face matching. When face matching tasks are not used
to assess face perception, the Cambridge Face Perception Task (CFPT; Duchaine, Germine &
Nakayama, 2007) is often used. This test requires participants to order a set of test face
stimuli in terms of their similarity to a target face, where those test stimuli are morphed faces
containing varying proportions of the target face and a foil face. The contribution of face
matching to performance on the CFPT is therefore hard to determine, as increasingly similar
faces are achieved by the test face becoming increasingly the same face as the target face, due
5
to the test face containing objectively more of the target face in the morph. Conversely,
decreasingly similarity is achieved by the test face containing objectively less of the target
face. Thus, responses of a participant basing their judgement on objective face perceptual
similarity only, and a participant basing their responses on the outcome of a face matching
decision process, would be perfectly correlated. Ideally, perceptual similarity would be
unconfounded from facial identity, so that facial images of the same person are sometimes
less similar (due to ageing, weight gain etc.), than facial images of different people. What has
been lacking thus far, therefore, is a means to obtain independent measures of face
perception, face matching, and face memory, in order to determine the degree to which these
are affected in DP.
The Oxford Face Matching Test (OFMT; Stantić et al., 2021) is a novel test designed
so that it can be used to assess individual differences in face processing abilities in clinical
and non-clinical populations in a non-biased manner (Stantić et al., 2021; Stantić, Brown,
Catmur & Bird, under review). Of relevance to the current study is that independent measures
of face perception and face matching can be obtained from the OFMT, and when these scores
are used to partition variance in CFMT scores, an independent measure of face memory can
also be derived. This approach has previously been used to show that, in neurotypical
individuals, face perception contributes to performance on face matching tests, and that face
perception and face matching make independent contributions to CFMT performance. In
addition, when this approach was used with volunteers with autism
1
, results suggested that
autism was associated with deficits in face perception and face memory, but not face
matching (Stantić et al., under review).
1
To respect the wishes of autistic individuals and report the study in line with scientific parlance, we
use language preferred by clinical professionals (e.g., 'individuals with autism’), as well as the term
‘autistic’, a term endorsed by many individuals with ASD (see Kenny et al., 2016).
6
Accordingly, the current study uses the OFMT and CFMT to derive independent
measures of face perception, face matching, and face memory in a group of adults with DP
and a matched neurotypical control group, such that the nature of face processing
impairment(s) in DP can be identified.
Methods
We report how we determined our sample size, all data exclusions, all
inclusion/exclusion criteria, whether inclusion/exclusion criteria were established prior to
data analysis, all manipulations, and all measures in the study.
Participants
Twenty-nine developmental prosopagnosics were recruited to participate in the study
(9 male; Mage = 42.31, SD = 11.31). Participants were selected as DPs from author databases
of individuals with DP. These participants met the criteria for impaired performance (defined
as 2 SDs below the neurotypical mean score) on at least two of three face processing
measures (CFMT, CFPT, and the Famous Face Test; (Bobak, Parris, Gregory, Bennetts &
Bate, 2017; Duchaine & Nakayama, 2006; Duchaine, Germine & Nakayama, 2007). No DP
participants were excluded from the current study for failing the attention check trials on the
OFMT (see below). On the PI-20 (see below), participants with DP scored a mean of 83.72
(SD = 5.80). An age- and gender-matched sample of 31 neurotypical participants were
recruited via Prolific.co and the authors’ database. Two participants were excluded for failing
to pass attention checks on the OFMT, providing a final sample of 29 neurotypical
participants (11 male; Mage = 42.41, SD = 9.81). On the PI-20, neurotypical controls scored a
mean of 45.69 (SD = 9.97). The DP and neurotypical groups did not differ significantly in
terms of age [t(56) = 0.04, p = .97, d = .01] or gender [X2(1) = 0.31, p = .58, w = .07], but, as
7
expected, the DP group self-reported more problems with face recognition on the PI-20 than
the neurotypical group (U = 0, p = < .001, r = .86). All participants reported having normal
or corrected-to-normal vision. Ethical approval was obtained from the Central University
Research Ethics Committee, University of Oxford. The procedures used in this study adhere
to the tenets of the Declaration of Helsinki.
Procedure
In a randomized order, participants completed the PI-20 and three measures of face
processing ability: two matching paradigms, the OFMT (Stantić et al., 2021) and the GFMT
(Burton et al., 2010), as well as a face memory paradigm, the CFMT (Duchaine &
Nakayama, 2006). Participants for whom the CFMT scores were available from previous
testing (specifically, DPs who completed the CFMT as part of their prosopagnosia screening)
did not complete the test again and instead the existing score was used to avoid practice
effects. All tasks were undertaken using the online behavioural platform Gorilla
(www.gorilla.sc).
The 20-item Prosopagnosia Index (PI-20; Shah, Gaule, Sowden, Bird & Cook, 2015)
The PI-20 is a self-report measure of face recognition ability and comprises 20 items
whereby participants are asked to rate on a Likert scale (1 = strongly disagree to 5 = strongly
agree) their face recognition difficulties in everyday life (e.g., ‘I have always had a bad
memory for faces’). Five questions are reverse scored, and the total calculated by summing
the scores from all items. The PI-20 can be used in conjunction with the CFMT to identify
individuals with DP (Gray, Bird & Cook, 2017); note that the PI-20 was not used to identify
DPs in this study). The maximum possible score is 100. The PI-20 is publically available
8
from:https://royalsocietypublishing.org/action/downloadSupplement?doi=10.1098%2Frsos.1
40343&file=rsos140343supp1.pdf.
Oxford Face Matching Test (OFMT; Stantić et al., 2021)
The OFMT (Figure 1A) is a novel face matching task that contains 200 trials (100
match (same) and 100 mismatch (different) face pairs). As an attention check, the OFMT
contains an additional 12 trials that are designed to be answered correctly even by individuals
with severe face processing impairments. Participants were excluded from all analyses if they
answered two or more of these trials incorrectly. Participants are presented with a face pair
for 1600ms and asked to determine whether faces are of the same person or different people.
The maximum possible matching score is 200. In addition, participants provide a perceptual
similarity judgment for each pair of faces on a scale from 0-100 (from very dissimilar to very
similar). The OFMT is deliberately constructed such that faces in match and mismatch trials
contain overlapping similarity distributions – two images of the same person can be
perceptually markedly different, while images of two different individuals can be
perceptually very similar. Thus, perceptual similarity can be dissociated from the outcome of
a face matching process. The OFMT is available to researchers on the Gorilla Open Materials
repository (https://gorilla.sc/openmaterials/134286) for non-commercial use upon request.
Glasgow Face Matching Test (GFMT; Burton et al., 2010)
The GFMT (Figure 1B) is an established face matching task that contains 40 trials (20
match and 20 mismatch). Participants are presented with face pairs and can view them for an
unlimited amount of time before making a decision about whether the faces are the same or
different. The maximum possible score is 40.
9
Cambridge Face Memory Task (CFMT; Duchaine & Nakayama, 2006)
The CFMT (Figure 1C) is an established face memory task that contains 72 trials in
three stages of increasing difficulty. Participants initially learn six faces and are afterwards
tested on three-alternative-forced-choice trials with two distractors and one image of a
previously learnt identity. There are 18 trials with no changes to viewpoint or lighting, 30
trials with changes to viewpoint and lighting, and 24 trials with changes to viewpoint and
lighting as well as the addition of visual noise. The maximum possible score is 72.
Figure 1. A sample trial of three face processing tasks: A – the Oxford Face Matching Test
(OFMT); B – the Glasgow Face Matching Test, GFMT, and C – the Cambridge Face
Memory Test, CFMT.
10
Analysis strategy
Independent measures of face perception, face matching and face memory are
required to address the aims of the study. To derive a measure of face perception,
participants’ ratings of the similarity of face pairs on the OFMT were compared to similarity
ratings derived from the average of three leading facial recognition algorithms (AWS
Rekognition (https://aws.amazon.com/rekognition/), FaceSoft (retrieved from
http://facesoft.io/) and Azure Face Recognition (https://azure.microsoft.com/en-
us/services/cognitive-services/face/), see Stantić et al, 2021). Each of these algorithms
provide a similarity index from 0 to 100, from which a mean index of similarity can be
calculated. For each participant, an average absolute deviation from algorithmically-provided
similarity was calculated. This value represents the difference between a participant’s
similarity score and the average similarity score provided by the algorithms (i.e., the higher
the deviation score, the greater the difference between the average similarity value provided
by the algorithms and the value provided by the participant). Participant similarity ratings
were compared with algorithmic similarity ratings for two reasons: the first is that such
algorithms regularly outperform human observers (Phillips & O’Toole, 2014; Phillips et al.,
2018) implying that their similarity ratings are valid, and the second is that the use of
algorithms (rather than large groups of human raters) to determine similarity avoids a
systemic bias towards whichever group rates the stimuli.
An index of face matching independent of face perception was derived by regressing
OFMT accuracy scores on average algorithmic deviations across participants. The residuals
from this regression constitute face matching scores as they represent the variance in OFMT
accuracy (i.e., in same/different judgements) that cannot be explained by perceptual face
similarity judgements. These face matching scores (the residuals from the above regression)
are thus statistically independent from face perceptual similarity judgements due to the way
11
they are derived. Face matching scores can also be derived from GFMT scores, if GFMT
scores are regressed on average algorithmic deviations obtained from the OFMT. This double
coding of face matching ability provides a data robustness check – allowing the replicability
of results to be ascertained when using matching scores from the GFMT after controlling for
OFMT perceptual similarity judgements. Note that this robustness check stems from the
independence of the two judgments made on each trial of the OFMT. The binary
same/different decision about two faces, akin to the one made on each trial of the GFMT,
constitutes an independent data point from the judgment of perceptual similarity of the two
presented faces. The latter can therefore be used as an index of participants’ perceptual ability
independent of their face matching ability. In the current study, we use these residuals (i.e.,
the measure of face matching independent of face perception accuracy) in univariate
analyses, e.g., in post-hoc tests. In multivariate analyses, entering OFMT or GFMT test
scores and algorithmic deviations into the same analysis achieves the same aim (the variance
explained by face matching performance can be identified independently of that accounted
for by face perception - i.e., when face perception ability is held constant).
Finally, face memory scores that are independent of both face perception and face
matching can be obtained by regressing CFMT accuracy scores on average algorithmic
deviations from the OFMT and face matching test scores from either the OFMT or GFMT.
In the current study, we determined whether DP impacts face perception, face
matching (independent of face perception) and face memory (independent of face perception
and face matching) using a series of regression analyses in which group (DP vs Control),
predictors (e.g., face perception), and their interaction, were used to predict test scores (e.g.,
CFMT test scores). Including the interaction term in regression models allows for the
relationship between, e.g., face perception (i.e., algorithmic deviation) and CFMT scores, to
vary across groups. Data was analysed using SPSS Statistics Version 28. For between group
12
comparisons (DP vs Control), participant demographics and task performance were analysed
with independent t-tests or the non-parametric equivalent, the Mann Whitney test, when
normality assumptions were violated. All statistical analyses were performed with a
significance level of p < .05, and all p values are two-tailed. All data are available at
https://osf.io/vzsqd and no part of the study procedures or analysis was formally preregistered
(though these are the same as used in Stantić et al., under review).
Results
Group Comparisons – Standard Test Scores
OFMT: Matching Performance
Nine participants with DP (31%) performed two standard deviations below the
Control group’s mean score, and an additional 13 participants with DP (45%) performed one
standard deviation below the Control group mean. One DP had an OFMT score that was
greater than the neurotypical mean score (but less than one standard deviation above the
control mean). At the group level, an independent t-test showed a significant difference in
OFMT accuracy [t(56) = 5.90, p <.001, d = 1.55] with DPs less accurate (M = 66%, SD =
6%, range: 55%-78%) than control participants (M = 74%, SD = 5%, range: 65% - 84%).
Signal Detection Theory (Green & Swets, 1966) was used to characterise performance
on the matching component of the OFMT, providing a measure of sensitivity (d') and bias
(criterion). An independent t-test showed a significant difference in d' [t(56) = 5.40, p <.001,
d = 1.42] and criterion [t(56) = 3.21, p = .002, d = 0.84] values between the DP and Control
groups, with the DP group showing less sensitivity and an increased bias (DP d': M = 0.91,
SD = 0.31; criterion: M = -0.36, SD = 0.33; Control d': M = 1.38, SD = 0.35; criterion: M = -
0.06, SD =. 0.38). Thus, individuals with DP exhibited less sensitivity to signals relevant to
13
face matching, and needed more evidence that the faces were the same before they made this
response.
OFMT: Algorithmic Deviation
Nine participants with DP (31%) performed two standard deviations above (indicating
worse performance) the Control mean deviation score, while an additional 11 participants
with DP (38%) performed one standard deviation above the Control mean. Eight participants
with DP (28%) had a deviation score that was greater than the neurotypical deviation mean
(but less than one standard deviation). At group level, a Mann-Whitney test revealed a
significant difference in deviation scores (U = 105, p = < .001, r = .64) with the DP group
having worse deviation scores, i.e., higher deviation from algorithmic judgements (median =
26.69, range: 22.81 – 48.28), than the Control group (median = 23.57, range: 16.66 – 27.38).
GFMT
Nine participants with DP (31%) performed two standard deviations below the
Control mean score, while an additional 11 DPs (38%) performed one standard deviation
below the Control mean. Five participants with DP (17%) had a GFMT score that was greater
than the neurotypical mean score (but less than one standard deviation above the control
mean). At group level, an independent t-test showed a significant difference in GFMT
accuracy [t(56) = 5.14, p <.001, d = 1.35], with DPs less accurate (M = 26.59, SD = 4.78,
range: 19 – 35) than Control participants (M = 32.83, SD = 4.46, range: 23 – 39).
As with the OFMT, Signal Detection Theory (Green & Swets, 1966) was used to
characterise performance on the matching component of the GFMT. An independent t-test
showed a significant difference in d' [t(56) = 4.69, p <.001, d = 1.23] with individuals with
14
DP less sensitive to face matching signals (M = 0.98, SD = 0.83) than Control participants (M
= 2.10, SD = 0.98). A Mann-Whitney test revealed a significant difference in criterion scores
(U = 254, p =.01, r = .34; DP: median: -0.14, range: -1.46 – 2.49; Control: median: 0, range: -
0.49 – 1.24). Thus, as on the OFMT, individuals were less sensitive to signals suggesting the
faces were the same, and needed more evidence to make this response, on the GFMT.
CFMT
Thirteen participants with DP (45%) performed 2 SDs below the neurotypical mean
score, with the remaining members of the DP group (55%) performing between one and two
standard deviations below the Control mean. As expected (given it was used to select
participants), a Mann-Whitney test revealed a significant difference in CFMT performance
between groups (U = 40.50, p = < .001, r = .78), with DPs less accurate (median = 36, range:
23 - 43) than the Control group (median = 56, range: 33 - 70).
Face Matching, Controlling for Face Perception
Group (DP vs Control), Deviation scores, and their interaction were entered into two
regressions, the first predicting OFMT matching accuracy and the second GFMT matching
accuracy. For the OFMT analysis, Deviation scores were a significant predictor (β = -.70, t = -
6.58, p = <.001) of OFMT matching accuracy, suggesting that face perception abilities are
related to OFMT performance. Group was also a significant predictor of OFMT matching
accuracy (β = -.24, t = -2.46, p = .02), indicating that face matching was worse in the DP group
even after accounting for face perception ability. The interaction between Group and Deviation
15
scores was not a significant predictor (β = .09, t = 1, p = .32), suggesting that the relationship
between face perception and OFMT matching performance did not vary as a function of
group.
The same pattern of significance was observed in the analysis predicting GFMT
matching accuracy - both Deviation scores (β = -.53, t = -3.95, p = <.001) and Group (β = -
.28, t = -2.27, p = .03) were significant predictors of GFMT matching accuracy, while the
interaction between Group and Deviation was not (β = .08, t = .72, p = .48).
Face Memory, Controlling for Face Perception and Face Matching
Group (DP vs control), Deviation scores, Face Matching scores (OFMT, and separately
GFMT matching accuracy) and the interactions between Deviation scores and Group, and Face
Matching and Group, were entered into regressions predicting CFMT scores. For the OFMT
analysis, results showed significant independent contributions of face perception (as measured by
Deviation scores; β = -.28, t = -2.08, p = .04) and face matching (β = .32, t = 2.52, p = .02).
Group was a significant predictor (β = -.39, t = -4.20, p = <.001), indicating that face memory
was worse in DP even after controlling for face perception and face matching. The two
interactions were not significant predictors, indicating the relationships between face memory
and face perception, and face memory and face matching, did not vary as a function of group
(Group x Deviation scores: β = .16, t = 1.38, p = .17; Group x Face Matching: β = -.10, t = -
1.03, p = .31).
Similarly, for the GFMT analysis, results showed significant independent contributions of
face perception (as measured by Deviation scores; β = -.36, t = -3.38, p = .001) and face
matching (β = .26, t = 2.70, p = .009). Group was also a significant predictor (β = -.40, t = -
4.46, p = <.001). The interactions between Group and Deviation scores and Group and Face
Matching scores were not significant predictors (Group x Deviation scores: β = .16, t = 1.72,
16
p = .09; Group x Face Matching: β = -.16, t = -2.00, p = .050). Given that the Group x Face
Matching interaction approached significance, exploratory analyses assessed the correlation
between GFMT residuals (which represent GFMT face matching independent of face
perception, see Analysis Strategy section) and CFMT scores in each group. These revealed a
significant positive correlation in the Control group (rs = .46, p = .01) and a weaker, non-
significant correlation (rs = .02, p = .91) in the DP group, suggesting that the contribution of
face matching to face memory performance, independent of face perception, was stronger for
the control compared to the DP group.
Discussion
This study examined the performance of individuals with DP on several face
processing tasks, with the aim of deriving independent measures of face matching, face
perception and face memory. Results showed that, as expected when standard test scores
were compared, the group of individuals with DP performed worse on the three face
processing tasks (OFMT, GFMT and CFMT) compared to the matched control group. It
should be noted that previous studies have not always found that prosopagnosic individuals
perform poorly on the GFMT (e.g. Fysh & Ramon, 2021; White, Rivolta, Burton, Al-Janabi
& Palermo, 2017), which may be due to issues with the sensitivity of the GFMT itself
(White, Guilbert, Varela, Jenkins & Burton, 2022), as it tends to be the case that studies with
larger sample sizes are more likely to find an impairment in DP. More interestingly, analysis
of OFMT facial similarity judgements revealed that individuals with DP were worse at
judging the perceptual similarity of two faces compared to the matched control group.
Furthermore, even after controlling for face perception, individuals with DP exhibited worse
face matching performance (i.e., they were worse at deciding whether two face images were
from the same person, or different people). This result was robust: it was observed when
17
matching was assessed using both the OFMT and the GFMT. This novel finding indicates
that individuals with DP have difficulties with both face perception and face matching.
Interestingly, this pattern is different from that seen in autistic individuals (Stantić et al.,
under review), whereby autistic individuals exhibit difficulties in face perception but not in
face matching. Although it is usually claimed that there are high rates of prosopagnosia in the
autistic population (e.g., Cook, Shah, Gaule, Brewer & Bird, 2015; Wilson, Palermo,
Schmalzl & Brock, 2010), these data indicate that there may be subtle differences in face
processing between autistic individuals and non-autistic individuals with DP.
Comparison of the results of autistic and prosopagnosic individuals reinforces the
distinction between the psychological processes of face perception and face matching.
Furthermore, it suggests that although face perception is likely necessary for accurate face
matching it is not sufficient. In addition to being able to form accurate perceptual
representations of faces from memory or from a pictorial representation, one must have an
accurate model of how, and how much, faces are allowed to vary before deciding they belong
to different people. Results of the Signal Detection Theory analysis of the OFMT and GFMT
matching task may be informative as to this point. In addition to a lower d prime, participants
with DP exhibited a more extreme bias towards ‘different’ responses. That is, they needed
more evidence that the faces were the same before they responded that they were the same.
This is consistent with personal reports from prosopagnosic individuals who report that they
fail to recognise individuals – i.e., they fail to recognise instances of faces (whether stored in
memory or available for visual inspection) are of the same facial identity, and therefore faces
are more likely to be judged as different. It is also consistent with claims that individuals with
DP show impaired performance on matching tasks specifically for trials in which the two
faces depict the same person, compared to trials in which different individuals are depicted
(White et al., 2017).
18
Finally, individuals with DP exhibited impaired face memory even after accounting
for their difficulties with face perception and face matching, indicating problems with all
three facial identity processes tested. Despite the clear pattern of impairments seen in those
with DP, it should be acknowledged that there were two main limitations of the current study
which raise questions about the degree of generalisability that can be assumed from the
current results. The first is the lack of any control stimulus class, meaning that it is not clear
whether the perceptual, matching and memory impairments seen for faces would extend to
other stimulus classes. This problem is a general one for the field, and reflects the fact that
appropriate control stimuli are difficult to identify (e.g., Fry, Wilmer, Xie, Verfaellie &
DeGutis, 2020; Susilo et al., 2010). The second limitation is that prosopagnosic individuals
were selected on the basis that they showed impaired performance on two out of three tests,
one of which was the Cambridge Face Perception Task. As such, the sample of DP
individuals identified may have been biased towards those that have perceptual difficulties in
addition to problems with face memory. This possibility will only be able to be tested with
further testing of the DP population with varying recruitment criteria, though it is worth
noting that a fairly large degree of variation in test scores was observed within the group of
individuals with DP. Unfortunately, the sample size was too low to allow formal testing for
the presence of sub-groups (Dalmaijer, Nord & Astle, 2022) but if future work adopts the
same testing and analysis procedure then samples could be combined to enable this approach.
Collectively, these results suggest that individuals with DP exhibit impaired face
perception, face memory and face matching. Thus, we suggest that neither the ‘memory’ nor
the ‘perceptual’ hypothesis is a sufficiently comprehensive account of face processing
difficulties in DP, and that decision-making processes involved in matching perceptual
stimuli with stored face representations (e.g., Bruce & Young, 1986), and how they may be
impaired, are also of importance in understanding DP.
19
Acknowledgements: We would like to greatly thank all our participants who took the time to
undertake this study. MS is funded by an ESRC DTP studentship and a Wilfrid Knapp
Science Scholarship. TS was supported by the Royal Society of New Zealand Marsden Fund
16-VUW-175. This publication was made possible through the support of a grant from the
John Templeton Foundation. The opinions expressed in this publication are those of the
author(s) and do not necessarily reflect the views of the John Templeton Foundation.
CRediT authorship contribution statement: M. Stantić: Study conceptualisation,
Methodology, Investigation, Project administration, Software, Data curation, Writing – review
& editing; Z. Pounder: Formal analysis, Writing – original draft, Writing – review & editing,
Visualisation; S. Bate: Methodology, Resources; Writing – review & editing; T. Susilo:
Methodology, Resources; Writing – review & editing; C. Catmur: Study conceptualisation,
Methodology, Writing – review & editing; G. Bird: Study conceptualisation, Methodology,
Funding acquisition, Supervision, Writing – review & editing.
20
References
Behrmann, M., & Avidan, G. (2005). Congenital prosopagnosia: face-blind from birth.
Trends Cogn Sci, 9(4), 180-187. doi:10.1016/j.tics.2005.02.011
Biotti, F., Gray, K. L. H., & Cook, R. (2019). Is developmental prosopagnosia best
characterised as an apperceptive or mnemonic condition? Neuropsychologia, 124, 285-298.
doi:10.1016/j.neuropsychologia.2018.11.014
Bobak, A. K., Parris, B. A., Gregory, N. J., Bennetts, R. J., & Bate, S. (2017). Eye-
Movement Strategies in Developmental Prosopagnosia and “Super” Face Recognition.
Quarterly Journal of Experimental Psychology, 70(2), 201-217.
doi:10.1080/17470218.2016.1161059
Bruce, V., & Young, A. (1986). Understanding face recognition. British Journal of
Psychology, 77(3), 305-327. doi:https://doi.org/10.1111/j.2044-8295.1986.tb02199.x
Burton, A. M., White, D., & McNeill, A. (2010). The Glasgow Face Matching Test.
Behavior Research Methods, 42(1), 286-291. doi:10.3758/BRM.42.1.286
Cook, R., & Biotti, F. (2016). Developmental prosopagnosia. Curr Biol, 26(8), R312-
313. doi:10.1016/j.cub.2016.01.008
Cook, R., Shah, P., Gaule, A., Brewer, R., & Bird, G. (2015). Autism and
Developmental Prosopagnosia: A Cross-Disorder Study. Journal of Vision, 15(12), 1211-
1211. doi:10.1167/15.12.1211
Corrow, S. L., Dalrymple, K. A., & Barton, J. J. (2016). Prosopagnosia: current
perspectives. Eye Brain, 8, 165-175. doi:10.2147/eb.S92838
Dalmaijer, E. S., Nord, C. L., Astle, D. E. (2022). Statistical power for cluster
analysis. BMC Bioinformatics, 23 (1) https://doi.org/10.1186/s12859-022-04675-1
21
Dalrymple, K. A., Corrow, S., Yonas, A., & Duchaine, B. (2012). Developmental
prosopagnosia in childhood. Cogn Neuropsychol, 29(5-6), 393-418.
doi:10.1080/02643294.2012.722547
Dalrymple, K. A., Garrido, L., & Duchaine, B. (2014). Dissociation between face
perception and face memory in adults, but not children, with developmental prosopagnosia.
Dev Cogn Neurosci, 10, 10-20. doi:10.1016/j.dcn.2014.07.003
Duchaine, B., Germine, L., & Nakayama, K. (2007). Family resemblance: Ten family
members with prosopagnosia and within-class object agnosia. Cognitive Neuropsychology,
24(4), 419-430. doi:10.1080/02643290701380491
Duchaine, B., & Nakayama, K. (2006). The Cambridge Face Memory Test: results for
neurologically intact individuals and an investigation of its validity using inverted face
stimuli and prosopagnosic participants. Neuropsychologia, 44(4), 576-585.
doi:10.1016/j.neuropsychologia.2005.07.001
Fry, R., Wilmer, J., Xie, I., Verfaellie, M., & DeGutis, J. (2020). Evidence for normal
novel object recognition abilities in developmental prosopagnosia. R Soc Open Sci, 7(9),
200988. doi:10.1098/rsos.200988
Fysh, M. C., & Bindemann, M. (2018). The Kent Face Matching Test. British Journal
of Psychology, 109(2), 219-231. doi:https://doi.org/10.1111/bjop.12260
Fysh, M. C., & Ramon, M. (2022). Accurate but inefficient: Standard face identity
matching tests fail to identify prosopagnosia. Neuropsychologia, 165, 108119.
doi:10.1016/j.neuropsychologia.2021.108119
Gray, K. L. H., Bird, G., & Cook, R. (2017). Robust associations between the 20-item
prosopagnosia index and the Cambridge Face Memory Test in the general population. Royal
Society Open Science, 4(3), 160923. doi:doi:10.1098/rsos.160923
22
Green, D. M., & Swets, J. A. (1966). Signal detection theory and psychophysics.
Oxford, England: John Wiley.
Jackson, M. C., Counter, P., & Tree, J. J. (2017). Face working memory deficits in
developmental prosopagnosia: Tests of encoding limits and updating processes.
Neuropsychologia, 106, 60-70. doi:10.1016/j.neuropsychologia.2017.09.003
Kenny, L., Hattersley, C., Molins, B., Buckley, C., Povey, C., & Pellicano, E. (2016).
Which terms should be used to describe autism? Perspectives from the UK autism
community. Autism, 20(4), 442-462. doi:10.1177/1362361315588200
Phillips, P. J., & O'Toole, A. J. (2014). Comparison of human and computer
performance across face recognition experiments. Image and Vision Computing, 32(1), 74-
85. doi:https://doi.org/10.1016/j.imavis.2013.12.002
Phillips, P. J., Yates, A. N., Hu, Y., Hahn, C. A., Noyes, E., Jackson, K., Cavazos, J.
G., Jeckeln, G., Ranjan, R., Sankaranarayanan, S., Chen, JC., Castillo, C. D., Chellappa, R.,
White, D., O’Toole, A. J. (2018). Face recognition accuracy of forensic examiners,
superrecognizers, and face recognition algorithms. Proceedings of the National Academy of
Sciences, 115(24), 6171-6176. doi:doi:10.1073/pnas.1721355115
Shah, P., Gaule, A., Sowden, S., Bird, G., & Cook, R. (2015). The 20-item
prosopagnosia index (PI20): a self-report instrument for identifying developmental
prosopagnosia. Royal Society Open Science, 2(6), 140343. doi:doi:10.1098/rsos.140343
Stantic, M., Brewer, R., Duchaine, B., Banissy, M. J., Bate, S., Susilo, T., Catmur, C.,
Bird, G. (2021). The Oxford Face Matching Test: A non-biased test of the full range of
individual differences in face perception. Behav Res Methods, 54(1), 158-173.
Stollhoff, R., Jost, J., Elze, T., & Kennerknecht, I. (2011). Deficits in long-term
recognition memory reveal dissociated subtypes in congenital prosopagnosia. PLoS ONE,
6(1), e15702. doi:10.1371/journal.pone.0015702
23
Susilo, T., & Duchaine, B. (2013). Advances in developmental prosopagnosia
research. Curr Opin Neurobiol, 23(3), 423-429. doi:10.1016/j.conb.2012.12.011
Susilo, T., McKone, E., Dennett, H., Darke, H., Palermo, R., Hall, A., Pidcock, M.,
Dawel, A., Jeffrey, L., Wilson, C. E., Rhodes, G. (2010). Face recognition impairments
despite normal holistic processing and face space coding: Evidence from a case of
developmental prosopagnosia. Cognitive Neuropsychology, 27(8), 636-664.
doi:10.1080/02643294.2011.613372
Towler, J., Fisher, K., & Eimer, M. (2018). Holistic face perception is impaired in
developmental prosopagnosia. Cortex, 108, 112-126. doi:10.1016/j.cortex.2018.07.019
White, D., Guilbert, D., Varela, V. P. L., Jenkins, R., & Burton, A. M. (2022).
GFMT2: A psychometric measure of face matching ability. Behav Res Methods, 54(1), 252-
260. doi:10.3758/s13428-021-01638-x
White, D., Rivolta, D., Burton, A. M., Al-Janabi, S., & Palermo, R. (2017). Face
matching impairment in developmental prosopagnosia. Q J Exp Psychol (Hove), 70(2), 287-
297. doi:10.1080/17470218.2016.1173076
Wilson, C. E., Palermo, R., Schmalzl, L., & Brock, J. (2010). Specificity of impaired
facial identity recognition in children with suspected developmental prosopagnosia. Cogn
Neuropsychol, 27(1), 30-45. doi:10.1080/02643294.2010.490207
