Content uploaded by Brad Duchaine
Author content
All content in this area was uploaded by Brad Duchaine on Aug 25, 2016
Content may be subject to copyright.
The Cambridge Face Memory Test for Children (CFMT-C): A new tool
for measuring face recognition skills in childhood
Abigail Croydon
a
, Hannah Pimperton
b
, Louise Ewing
c,d
, Brad C. Duchaine
e
,
Elizabeth Pellicano
a,d,
n
a
Centre for Research in Autism and Education (CRAE), Institute of Education, University of London, London, UK
b
Institute of Cognitive Neuroscience, University College London, London, UK
c
Department of Psychological Science, Birkbeck, London, UK
d
School of Psychology, University of Western Australia, Perth, Australia
e
Department of Psychological and Brain Sciences, Dartmouth College, Hanover, NH, USA
article info
Article history:
Received 13 May 2014
Received in revised form
21 June 2014
Accepted 10 July 2014
Available online 20 July 2014
Keywords:
Face recognition
Face memory
Development
Children
Inversion effect
abstract
Face recognition ability follows a lengthy developmental course, not reaching maturity until well into
adulthood. Valid and reliable assessments of face recognition memory ability are necessary to examine
patterns of ability and disability in face processing, yet there is a dearth of such assessments for children.
We modified a well-known test of face memory in adults, the Cambridge Face Memory Test (Duchaine &
Nakayama, 2006, Neuropsychologia, 44, 576–585), to make it developmentally appropriate for children.
To establish its utility, we administered either the upright or inverted versions of the computerised
Cambridge Face Memory Test –Children (CFMT-C) to 401 children aged between 5 and 12 years. Our
results show that the CFMT-C is sufficiently sensitive to demonstrate age-related gains in the recognition
of unfamiliar upright and inverted faces, does not suffer from ceiling or floor effects, generates robust
inversion effects, and is capable of detecting difficulties in face memory in children diagnosed with
autism. Together, these findings indicate that the CFMT-C constitutes a new valid assessment tool for
children's face recognition skills.
&2014 Elsevier Ltd. All rights reserved.
1. Introduction
Faces are critical to social interaction. They provide a wealth of
information about an individual's gender, ethnicity, emotional state,
direction of attention and, crucially, they uniquely identify the owner.
The ability to identify persons from their facial appearance –face
identity recognition –begins early in development (e.g., Bushnell, Sai,
& Mullin, 1989; Pascalis, de Schonen, Morton, Deruelle, & Fabre-
Grenet, 1995). Despite this early facility, the emergence of adult face
expertise follows a protracted course of development, with perfor-
mance on tests of face recognition not approaching maturity until
well into adulthood (e.g., Germine,Duchaine,&Nakayama,2011;
Susilo,Germine,&Duchaine,2013).
Much current research is focused on understanding the per-
ceptual, cognitive, and neural mechanisms underlying this lengthy
developmental trajectory (e.g., Crookes & McKone, 2009) and
elucidating how such processes might develop differently in
children with neurodevelopmental conditions, such as autism
(e.g., Weigelt, Koldewyn, & Kanwisher 2012) and developmental
prosopagnosia (Wilson, Palermo, Schmalzl, & Brock, 2010). Such
research would be facilitated by standardized assessments of
unfamiliar face identity recognition, providing tools useful for
experimental and clinical settings and enabling direct comparison
between individuals with and without neurodevelopmental con-
ditions (see Dalrymple, Corrow, Yonas, and Duchaine (2012)).
Yet many of the existing standardized face recognition tests
for adults (e.g. the Benton Facial Recognition Test: Benton, Sivan,
Hamsher, Varney, & Spreen, 1983; the Recognition Memory Test
for Faces: Warrington, 1984) and children (e.g. the Identity
Matching Test: Bruce et al., 2000) suffer from significant short-
comings, where a score in the ‘normal range’does not necessarily
reflect typical face recognition skills (Duchaine & Nakayama, 2004;
Duchaine & Weidenfeld, 2003). For example, in the Benton Facial
Recognition Test both the target and the test faces are presented
simultaneously, which means that participants can derive the
correct responses by using a feature-matching strategy, while in
Warrington's Recognition Memory Test for Faces, participants can
use non-face information present in the stimuli to select the
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/neuropsychologia
Neuropsychologia
http://dx.doi.org/10.1016/j.neuropsychologia.2014.07.008
0028-3932/&2014 Elsevier Ltd. All rights reserved.
n
Corresponding author at: Centre for Research in Autism and Education,
Department of Psychology and Human Development, Institute of Education, 25
Woburn Square, London WC1H 0AA, UK. Tel.: þ44 207 331 5140.
E-mail address: l.pellicano@ioe.ac.uk (E. Pellicano).
Neuropsychologia 62 (2014) 60 –67
correct response. Both these limitations are also present in the
children's Identity Matching Test (Bruce et al., 2000).
The Cambridge Face Memory Test (CFMT) was developed both to
capitalise on the strengths of the previous adult assessments and to
overcome their limitations in order to provide researchers and
clinicians with a standardised test of face recognition that would
accurately and reliably measure face memory ability (Duchaine &
Nakayama, 2006). In the CFMT, adults study, and are subsequently
tested on, the facial identities of 6 men posing with neutral
expressions in three distinct experimental stages. Stage 1 requires
participants to identify the same learned face amongst two distractor
images, when the test image is identical to the study face. Stage
2 calls for recognition of the same learned faces in novel viewpoints
and/or lighting conditions. Stage 3 requires recognition of the learned
faces from novel images degraded by the presence of visual noise in
order to increase the difficulty and to force greater reliance on face-
specific mechanisms (McKone,Martini,&Nakayama,2001).
Duchaine and Nakayama (2006) established that the CFMT showed
excellent psychometric properties: the test showed a good range of
responses in typical adults (n¼50), did not suffer from ceiling or floor
effects, showed the classic face-inversion effect (a decrement in
performance when the face is turned upside-down) and could reliably
diagnose individuals with acquired prosopagnosia, who have profound
face-specificmemorydeficits (see Wilmer et al. (2010),forfurther
psychometric findings). Consequently, the test is now used widely to
test face identity recognition (e.g., Banissy et al., 2011; Bowles et al.,
2009; Di Simplicio, Massey-Chase, Cowen, & Harmer, 2009; Hedley,
Brewer, & Young, 2011; Kirchner, Hatri, Heekeren, & Dziobek, 2011;
O’Hearn, Schroer, Minshew, & Luna, 2010; Richler, Cheung, & Gauthier,
2011; Wilmer et al., 2010)–at least in adults. A study using the CFMT
with typically developing children aged between 9 and 17 years
showed that performance in the younger children (9- to 12-year-olds)
was poor and that the test was not sufficiently sensitive to distinguish
between children with autism and typically developing children at this
age (O’Hearn et al., 2010).
In the current study, we adapted Duchaine and Nakayama's
CFMT to create a developmentally sensitive test of face recognition
for children aged between 5 and 12 years. Here, we report on the
performance of a large sample of primary and secondary school
children on the upright and inverted versions of this new test, the
CFMT for children (CFMT-C). We also examined the validity of the
CFMT-C by administering it to a group of children diagnosed with
autism, who have marked difficulties in social interaction and for
whom problems with face identity recognition have been consis-
tently implicated (Weigelt et al., 2012).
2. Methods
2.1. Participants
Four hundred and one participants (202 females) aged between 5 and 12 years
were recruited from primary and secondary schools in England, UK, to take part in this
study. Two hundred and eighty-two children completed the upright version of the test,
including 37 5-year-olds (13 females), 38 6-year-olds (20 females), 37 7-year-olds
(21 females), 40 8-year-olds (20 females), 33 9-year-olds (17 females), 41 10-year-olds
(24 females), 29 11-year-olds (13 females) and 27 12-year-olds (13 females). One
hundred and nineteen children completed the inverted version, including 12 5-year-olds
(3 females), 10 6-year-olds (4 females), 12 7-year-olds (6 females), 18 8-year-olds
(9 females), 20 9-year-olds (12 females), 23 10-year-olds (12 females), 12 11-year-olds
(7 fem ale s) and 12 12 -year-olds (8 fem ales ).
2.2. Stimuli
The face stimuli were selected from those used in the adult version of the CFMT
(Duchaine & Nakayama, 2006). The faces were greyscale photographs of men in early
adulthood posing with neutral expressions. Each face was photographed in the same
three poses and lighting conditions and cropped to remove the hairline and any
facial blemishes (see Fig. 1 for example stimuli). Similar to the CFMT, the same faces
were used in the upright and inverted versions of the CFMT-C, the difference being
that in the inverted version all images were presented upside-down (rotated 1801).
Based on pilot testing with 16 children aged between 5 and 12 years, several
modifications were made to the original CFMT to create the CFMT-C. First, the
number of target faces that the children viewed was reduced from six to five.
Asking children to remember six target faces resulted in a floor effect during the
most difficult ‘noise’section of the test, while four target faces induced a ceiling
effect for older children. The face selected for elimination from the test was the one
that produced the lowest percentage of correct responses out of the six target faces
in the original test (Duchaine & Nakayama, 2006). Second, test items were altered
from a three- to a two-alternative forced choice design, comprising one target face
and one distractor face (see Fig. 1A). Based on error data from adult participants,
the least distinct distractor (i.e., the one most similar to the target and therefore
incorrectly selected most often by adult participants) was removed from each test
item. Third, the exposure time of the target faces was increased from 300 0 ms to
5000 ms in stage 1 of the test to give children more time to encode the faces.
Finally, the wording of the on-screen instructions was made appropriate for
children, including practice trials showing a popular cartoon face, and reinforce-
ment slides (e.g., “Keep up the good work!”) to provide encouragement and help
maintain children's attention.
2.3. Procedure: upright version
As in the original CFMT, the CFMT-C was comprised of a short practice stage
plus three progressively more difficult stages (same images, novel images and novel
images with noise).
2.3.1. Practice
This stage was used to familiarise children with the task structure. In the study
phase, children saw the face of a popular cartoon character presented three times
sequentially for 5000 ms each shown from different viewpoints (left-facing, front
view, right-facing). Children were instructed to look at the images very carefully
because they would need to remember them later. In the test phase, children saw
the face of the same cartoon character alongside another character. They were
asked to select which of the two faces they had just seen and to make the
corresponding keypress (‘1’or ‘2’). There were 3 test trials, one for each of the three
different viewpoints. If an error was made on one or more practice trials, the
practice stage was repeated until the child achieved perfect performance.
2.3.2. Stage 1: same images
Children were told that they would now need to memorise the faces of five
different (real) people. During the study phase (see Fig. 1A), three different images
(left-facing, front view, right-facing) of each face were presented sequentially for
5000 ms –just like in the practice stage. Next, in the test phase (see Fig. 1B), each of
the three images was presented alongside a distractor face and children were
required to choose which face they had just seen by making the corresponding
keypress (‘1’or ‘2’) themselves (older children) or by informing the experimenter of
their answer, who pressed the corresponding key on their behalf (young children:
5- to 6-year-olds). No feedback was given. One point was given for each correct
response (maximum¼15).
2.3.3. Stage 2: novel images
In the study phase, children initially inspected a single screenshot showing
front views of all the five target faces for 20 s (see Fig. 1C). They were instructed to
look carefully at the faces and to try to memorise them. During the test phase,
children completed 25 trials, each consisting of one of the target faces and a
distractor face, which remained onscreen until a response (keypress: ‘1’or ‘2’).
Children selected which of the two faces they thought was one of the five target
faces they had been asked to memorise. Each target face was shown 5 times each in
afixed, randomized order. Children were given one point for each correct response
(maximum¼25). Each distractor face appeared several times throughout the test
phase to avoid the possibility that participants might select the correct answer
simply by choosing faces that looked familiar. All target faces presented during the
test phase were novel images, that is, either the lighting or viewpoint was different
from those used in stage 1 (same images) (see Fig. 1C; for further detail, see
Duchaine & Nakayama, 2006).
2.3.4. Stage 3: novel images with noise
The final stage began with another study phase –a single presentation of all
five target faces, which children were again instructed to review for 20 s (see
Fig. 1D). The subsequent test phase consisted of 20 trials in which a novel image of
each target face was presented together with a distractor face 4 times each.
Critically, both target and distractor images were masked with a pre-specified
degree of Gaussian noise (see Fig. 1D), making the judgment more difficult than in
stage 2. Trials were presented in a fixed, random order and the faces remained
onscreen until children indicated which of the two faces most closely resembled
one of the five target faces. One point was given for each correct response
(maximum¼20).
A. Croydon et al. / Neuropsychologia 62 (2014) 60–67 61
Like in the adult CFMT, scores across all three stages were summed to yield a
total recognition accuracy score for each child (out of 60).
2.4. Procedure: inverted version
The procedure for the inverted version of the CFMT-C was identical to the
upright version. The exception to this was that children were told that there was
“something a bit strange about the faces –they are going to be all upside-down!”
They were encouraged to look at them carefully, just like in the upright version.
2.5. General procedure
Written informed consent was provided by the parents of all children prior to
participation. All children were seen individually in a quiet space within their
Fig. 1. Stimuli and procedure for the Cambridge Face Memory Test for Children (CFMT-C). A: Stage 1 (same images study phase). During the study phase, participants viewed
5 target faces, one at a time from three different viewpoints. B: Stage 1 (same images test phase). In the test phase, participants were then required to judge which of two
faces was the one they had just seen. C: Stage 2 (novel images). During the study phase, participants were presented with all five target faces and asked to review them
carefully. In the test phase, one of these faces was presented in a different viewpoint and children were asked to identify which of two faces (a novel image of the target face
and a distractor face) was one of the 5 target faces. D: Stage 3 (novel images with noise). Similar to the study phase of stage 2, participants were asked to review the same
5 target faces. During the test phase, they were shown two faces comprising a novel image of target face and a distractor face, both masked with Gaussian noise.
A. Croydon et al. / Neuropsychologia 62 (2014) 60–6762
school. We used a 15-in. Macintosh Powerbook G4 running OSX, and children were
seated at a distance of approximately 50 cm from the computer screen. The test
took approximately 8–10 min to complete.
3. Results
To begin, we determine the presence of any floor or ceiling
effects in the upright dataset. We then examine the relationship
between age and total performance on the upright and inverted
versions of the CFMT-C and also, using ANOVA, we examine age-
related differences on the three stages of each test version
(upright, inverted) and provide normative data for children of
different ages. Finally, we establish the reliability and validity of
the CFMT-C. Note that children's scores were converted to per-
centages to facilitate comparison across the three stages.
3.1. Floor/ceiling effects
One sample ttests (with Bonferroni correction for multiple
comparisons; corrected po.004) on children's upright perfor-
mance showed that there were no floor and ceiling effects (see
Table 1 for scores). Five- and 6-year-olds scored significantly above
chance (50%) in all three stages (all pso.001). Eleven- and 12-
year-olds scored significantly below ceiling (100%) under all
conditions (all pso.002). The mean performance of the oldest
children (12-year-olds; n¼27) was 82.6% (SD ¼9.4%). This perfor-
mance is comparable to Duchaine and Nakayama's (2006) adult
participants (M¼80.42%, SD¼11.0), suggesting that test difficulty
for the older children in the CFMT-C is closely matched to test
difficulty for adults in the original test.
3.2. Age-related gains in overall CFMT-C performance
To examine the relationship between age and children's total
CFMT-C performance, we conducted two separate regression
analyses on children's total percentage correct for each version
(upright, inverted) separately. Fig. 2 shows that there is much
individual variation in performance on both tasks but also clear
age-related improvements in face identity recognition. For the
upright task, age accounted for 25% of the variance in children's
performance, F(1, 280)¼94.9, po.001. Each additional birthday
resulted in a 2.8% increase in children's total score. Similarly, for
the inverted task, age accounted for a significant amount of
variance (R
2
¼.31) in the model, F(1,117)¼52.10, po.001. Children
improved in their performance on the inverted task by approxi-
mately 2.8% with each birthday.
3.3. Age-related differences on stages of the CMFT-C
Table 1 shows children's performance on each stage of the upright
and inverted versions of the CFMT-C at each age. To examine age-
related differences on the CFMT-C, we performed a repeated-measures
ANOVAwithagegroup(5years,6years,7years,8years,9years,10
years, 11 years, 12 years) and version (upright, inverted) as the
between-participants factors and stage (same images, novel images,
novel images with noise) as the within-participants factor. An addi-
tional ANOVA with gender as a factor (male, female) showed no main
effect of gender or any interaction involving gender (all pso.08).
There was a main effect of version, F(1, 385)¼136.06, po.001,
η
2
p
¼.26. As expected, there was a significant inversion effect.
Performance was significantly better in the upright (M¼76.3%;
SD¼12.2) than the inverted (M¼65.6%; SD ¼10.2) version of the
CFMT-C. On average, the effect of inversion led to a 10.7%
reduction in performance. There was no interaction between
version and age group, Fo1, indicating no developmental increase
in the size of the inversion effect.
There was also a main effect of stage, F(2, 770) ¼285.94,
po.001, η
2
p
¼.43. These main effects were qualified by a stage-
version interaction, F(2, 770)¼8.61, po.001, η
2
p
¼.02. In the
upright version, performance in stage 1 (same images:
M¼90.5%; SD¼12.3) was significantly better than in stage 2 (novel
images: M¼76.2%; SD¼15.6), t(281)¼17.68, po.001, which in
turn was significantly better than performance in stage 3 (novel
images with noise: M¼67.2%; SD¼14.7), t(281)¼10.44, po.001.
In the inverted version, performance in stage 1 (same images:
M¼77.1%; SD¼16.6) was significantly better than in stage 2 (novel
images: M¼63.2%; SD¼13.8), t(118)¼9.48, po.001, which in turn
was significantly better than inverted performance in stage 3
(novel images with noise: M¼59.6%; SD¼11.9), t(118)¼2.39,
po.05. The source of the interaction came from comparison of
the difference between performance in the inverted stages. The
difference in performance between stages 1 and 2 was signifi-
cantly greater (M¼13.92; SD¼16.02) than the difference between
stages 2 and 3 (M¼3.58; SD¼16.36), t(118)¼4.30, po.001, most
likely because performance in stage 3 was approaching floor.
Table 1
Children's mean performance (% correct) on the different stages of the upright and inverted versions of the CFMT-C.
CFMT-C stage
Version Age group Same images Novel images Novel images with noise Total % correct
M (SD) M (SD) M (SD) M (SD)
Upright 5-year-olds (n¼37) 80.0 (12.1) 62.3 (14.3) 60.5 (13.0) 66.1 (11.0)
6-year-olds (n¼38) 83.7 (13.3) 67.3 (15.6) 58.4 (13.0) 66.9 (11.3)
7-year-olds (n¼37) 88.4 (11.8) 73.5 (14.4) 67.3 (14.3) 75.2 (10.6)
8-year-olds (n¼40) 90.8 (10.5) 76.8 (13.1) 67.6 (12.3) 76.1 (9.8)
9-year-olds (n¼33) 94.5 (8.2) 82.2 (11.9) 71.3 (13.9) 81.6 (9.0)
10-year-olds (n¼41) 95.6 (7.1) 81.3 (15.6) 69.6 (15.3) 80.1 (12.0)
11-year-olds (n¼29) 97.0 (4.6) 85.1 (12.0) 70.5 (14.0) 83.2 (9.2)
12-year-olds (n¼27) 97.3 (7.4) 86.2 (9.0) 75.6 (14.4) 85.6 (8.2)
Inverted 5-year-olds (n¼12) 66.1 (12.8) 47.3 (8.8) 52.9 (11.2) 53.4 (5.3)
6-year-olds (n¼10) 64.0 (12.3) 51.6 (13.1) 58.0 (8.9) 56.5 (7.2)
7-year-olds (n¼12) 73.9 (16.7) 59.3 (13.6) 54.2 (14.9) 64.0 (11.8)
8-year-olds (n¼18) 77.0 (16.4) 62.0 (9.4) 60.8 (14.3) 65.4 (7.3)
9-year-olds (n¼20) 81.7 (11.4) 67.6 (9.9) 58.8 (10.4) 67.9 (6.1)
10-year-olds (n¼23) 76.5 (18.9) 65.6 (13.5) 60.9 (9.6) 66.8 (9.9)
11-year-olds (n¼12) 86.7 (14.2) 74.7 (11.5) 64.2 (13.4) 74.2 (9.1)
12-year-olds (n¼12) 86.1 (17.2) 70.7 (12.8) 65.4 (9.4) 72.8 (10.1)
Note that chance performance on each stage of the CFMT-C is 50%.
A. Croydon et al. / Neuropsychologia 62 (2014) 60–67 63
There was a significant main effect of age group F(7, 385)¼18 .47,
po.001, η
2
p
¼.25. Post-hoc comparisons with Bonferroni correction
confirmed a general pattern of age-related improvements in face
identity recognition. The performance difference between adjacent
ages, however, was not always significant (see Table 1 for scores).
Five-year-olds performed significantly worse than all other age
groups (all pso.001) apart from 6-year-olds, 6-year-olds performed
worse than all other groups (all pso.005) apart from 5- and 7-year-
olds, 7-year-olds performed significantly better than 5-year-olds
(po.001) and significantly worse than 11- and 12-year-olds
(pso.001), 8-year-olds performed better than 5- and 6-year-olds
(pso.005) but worse than 11- and 12-year-olds (pso.02), both
9- and 10-year-olds performed better than 5- and 6-year-olds
(pso.001) and similar to all other age groups, and 11- and 12-
year-olds obtained higher scores than all age groups (pso.002)
with the exception of 9-year-olds. There were no significant
interactions involving age group (ps4.09). These results mirror
the regression analyses showing a gradual age-related increase in
face identity recognition.
3.4. Reliability
Following Duchaine and Nakayama (2006) and Bowles et al.
(2009), we calculated the reliability of the upright version of the
CFMT-C across all participants in two ways. First, we correlated
children's performance on stage 2 of the test (novel images) with
their performance on stage 3 (novel images with noise). Perfor-
mance was significantly correlated across these conditions, r(279)
¼.54, po.001. This moderately-sized correlation was not as strong
as that reported by Duchaine and Nakayama (2006);r¼.74; n¼50
adults; r¼.75 or Bowles et al. (2009),n¼124 young adults and
may be due to the possibility that children's performance is less
stable with development than adults’.
Second, we examined the internal consistency of the upright
version of the CFMT-C. The estimate of Cronbach's alpha was high
(α¼.88) and comparable to that reported with the adult CFMT
(Bowles et al., 2009;α¼.89). Like the CFMT, the CFMT-C therefore
meets the standard reliability requirements for clinical tests
(Cronbach's alpha 4.85; Aiken, 2003).
3.5. Validity
To determine the validity of the CMFT-C, we administered the
upright version on a population with known difficulties in face
identity recognition (see Weigelt et a l. (2012),forreview).Forty-four
children (9 girls) diagnosed with an autism spectrum condition
aged between 7 and 12 years (M age¼10.7 years; SD¼1.6) took
part in this validation study. All children had received indepen-
dent clinical diagnoses of autism (n¼36) or Asperger syndrome
(n¼8) and obtained a score of at least 15 or above (the cut-off for
autism) on the Social Communication Questionnaire (SCQ; Rutter
et al., 2003). Children with autism were compared to a subsample
of typical children (n¼44; 11 females) who had completed the
upright version of the CFMT-C (M age ¼10.6 years; SD¼1. 7 ). Al l
typical children fell well below the cutoff score of 15 on the SCQ,
suggesting that they showed few behavioural features of autism
(see Table 2).
The groups were of similar chronological age, F(1,86)¼.29,
p¼.87, verbal IQ, F(1,86)¼.71, p¼.40, and performance IQ, F
(1,86)¼.02, p¼.89, as measured by the Wechsler Abbreviated
Scales of Intelligence (WASI; Wechsler, 1999) (see Table 2 for
scores). All children completed the upright version of the CFMT-C
in a single, individual session alongside the measure of intellectual
functioning. Parents gave informed written consent for their child
to take part.
To examine potential group differences in CFMT-C perfor-
mance, an ANOVA with group (autism, typical) as the between-
participants factor and stage (same images, novel images, novel
images with noise) as the within-participants factor was per-
formed on children's scores (percentage correct). There was a
Fig. 2. Total percentage correct on the CFMT-C plotted against age for children who completed the upright (open blue circles) and inverted (open green circles) versions. The
regression lines are shown for each relationship (upright: solid blue, y¼51.89þ(2.78) age; inverted: solid green, y¼40.09þ(2.79) age) and 95% confidence intervals. Dotted
line represents chance performance (50%). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
A. Croydon et al. / Neuropsychologia 62 (2014) 60–6764
significant main effect of stage, F(2, 172)¼153.76, po.001, η
2
p
¼.64.
Overall, participants performed better in stage 1 (same images:
M¼89.8%; SD¼12.5) than in stage 2 (novel images phase:
M¼77.5%; SD¼13.3), t(87)¼9.40, po.001, and stage 3 (novel
images with noise: M¼65.2%; SD¼15.1), t(87) ¼8.46, po.001.
There was also a main effect of group, F(1,86)¼15.63, po.001,
η
2
p
¼.15. Children with autism (M¼71.7%; SD¼12.8) obtained
significantly lower scores than typical children (M¼81.3%;
SD¼10.2). There was no interaction between group and condition,
Fo1.
Autistic children's total performance on the CFMT-C was
unrelated to their age, verbal IQ or performance IQ (all ps4.11).
Yet there was a significant negative correlation between children's
total % correct and their SCQ scores, r(43)¼.40, p¼.007. Greater
degrees of autistic symptomatology were related to worse face
memory performance on the CFMT-C. Overall, these results sug-
gest that the CFMT-C is sensitive for detecting atypicalities in face
identity recognition in children with autism.
4. Discussion
Face identity recognition skills follow a lengthy trajectory and
are at risk of developing atypically in individuals with neurode-
velopmental conditions, such as autism and developmental pro-
sopagnosia. There are, however, remarkably few tests that are
appropriate for assessing these skills in children and those that do
exist (e.g., Bruce et al., 2000) are limited in various ways. In this
paper, we describe the development and application of a child-
friendly version of the CFMT (Duchaine & Nakayama, 2006), a test
that is increasingly being used by researchers as a valid and
reliable assessment of face memory skills in adulthood. Here, we
show in a large group of typically developing children aged
between 5 and 12 years that the CFMT-C is sensitive to develop-
mental differences in the recognition of unfamiliar upright and
inverted faces, is sufficiently reliable to provide an accurate
indication of a child's performance and is capable of detecting
difficulties in face memory in children diagnosed with autism.
Together, these findings suggest that the CFMT-C is a valid and
reliable tool for assessing face recognition in middle childhood.
As expected, we observed gradual gains in face memory skills
between 5 and 12 years of age. These findings are consistent with
existing research showing that children's memory for faces follows
a protracted developmental course both at the behavioural (e.g.,
Carey & Diamond, 1977;Johnston & Ellis, 1995) and neural levels
(e.g., Golarai, Liberman, Yoon, & Grill-Spector 2009; Haist, Adamo,
Han Wazny, Lee, & Stiles, 2013). We also found a significant effect
of face inversion: upright faces were recognised more accurately
than were inverted faces across all age groups.
The face inversion effect, that inversion disproportionately
impairs the recognition of faces to a greater degree than the
recognition of other classes of objects, is one of the most robust
findings in the face processing literature and reflects the purport-
edly special status of faces (Diamond & Carey, 1986;Yin, 1969). In a
seminal study, Carey and Diamond (1977) found that while 8- and
10-year-olds were much better at recognising the faces when they
were presented upright than upside-down, 6-year-olds showed no
such inversion effect. On this basis, Carey and Diamond (1977)
proposed that young children process faces in terms of individual
facial features until the age of 10 years, when there is a qualitative
shift to a more adult-like processing style involving representa-
tions of the overall facial configuration. This proposal has since
been vigorously debated in the literature (e.g., see Crookes and
McKone (2009),Mondloch, Le Grand, and Maurer (2002),Pellicano
and Rhodes (2003), and Pellicano, Rhodes, and Peters (2006)).
Crookes and McKone (2009) have highlighted that when studies
have reported differences in the magnitude of child and adult
inversion effects (e.g., Brace et al., 2001; Carey & Diamond, 1977),
these effects may be attributable to the presence of ceiling and
floor effects. Indeed, when efforts are made to avoid such effects,
children as young as 3 years show the classic inversion effect
(Sangrigoli & de Schonen, 2004, Experiment 3; see also Pascalis,
Demont, de Haan, and Campbell (2001)).
The absence of a significant interaction between version
(upright, inverted) and age group in the current study provides
further evidence of no age-related changes in the size of the
inversion effect –at least between the ages of 5 and 12 years –and
accords with other findings challenging qualitative differences in
the way that younger and older children recognise faces (e.g., see
Pellicano et al. (2006) and Crookes and McKone (2009)). Instead,
the mechanisms responsible for face processing appear to be
mature in early childhood, with developmental improvements in
face memory skills potentially arising from more general gains in
memory, attention and processing speed (Crookes & McKone,
2009; though see Mondloch, Le Grand, and Maurer (2010), for an
alternative view).
We also assessed the validity of the CFMT-C by administering it
to children with autism, a neurodevelopmental condition that
affects the way an individual interacts with and experiences the
world around them (American Psychiatric Association, 2013).
Some key early indicators, including limited eye contact, poor
social orienting, and reduced social responsiveness (Dawson,
Webb, & McPartland, 2005; Zwaigenbaum et al., 2005), have led
some researchers to suggest that face processing difficulties –and
sociocognitive impairments more broadly –might be at the core of
autism (Dawson et al., 2005; Schultz, 2005). A recent review of 90
experiments investigating face processing in autism suggested
that, on average, autistic individuals perform significantly worse
than typical individuals on tasks tapping face recognition (Weigelt
Table 2
Descriptive statistics for chronological age, measures of intellectual functioning,
autistic symptomatology and upright CFMT-C scores (% correct) in each group
separately.
Group
Children with
autism (n¼44)
Children without
autism (n¼44)
Age (in months)
M (SD) 128.52 (19.42) 127.80 (20.80)
Range 90–155 91 –155
Verbal IQ
a
M (SD) 99.27 (14.75) 101.45 (8.90)
Range 61–131 82–12 0
Performance IQ
a
M (SD) 98.36 (11.50) 98.77 (11.50)
Range 73–129 66–123
SCQ
b
M (SD) 25.23 (5.48) 3.73 (3.04)
Range 15–36 0–11
CFMT-C stage 1 (same images)
M (SD) 85.15 (14.08) 94.54 (8.54)
Range 46.67–100 66.67–100
CFMT-C stage 2 (novel images)
M (SD) 71.82 (16.25) 83.27 (12.02)
Range 32–100 5 2–10 0
CFMT-C stage 3 (novel images with noise)
M (SD) 61.36 (15.19) 68.98 (14.08)
Range 35–90 35–95
CFMT-C total % correct
M (SD) 71.67 (12.78) 81.32 (10.18)
Range 43.44–91.67 51.67–98.33
a
Children's intellectual functioning was measured using the Wechsler Abbre-
viated Scales of Intelligence (WASI; Wechsler, 1999), standard scores reported here.
b
SCQ: Social Communication Questionnaire (Rutter et al., 2003).
A. Croydon et al. / Neuropsychologia 62 (2014) 60–67 65
et al., 2012). Our findings are consistent with Weigelt and
colleagues’conclusions. We showed that 7- to 12-year-olds with
autism performed significantly worse than typical children of
similar age and ability –in fact, they performed on average one
standard deviation below the mean of typical children. Further-
more, the absence of an interaction between CFMT-C stage and
group supports the view that face identity recognition might be
qualitatively similar in these children with and without autism
(Weigelt et al., 2012).
Notably, our results are in contrast with the results of one study
using the adult version of the CFMT. O’Hearn et al. (2010) reported
face-processing difficulties in adolescents, but not in children,
with autism and concluded that face memory difficulties emerge
in autism only after adolescence. It is plausible, however, that
these authors failed to identify such difficulties at earlier ages
because the adult CFMT was not sufficiently sensitive to detect
difficulties in children (with or without autism), many of whom
seemed to perform at floor. The discrepancy between our findings
using the modified CFMT-C and those of O’Hearn et al. serves to
reinforce the importance of creating valid and developmentally
appropriate measures of face identity recognition.
Importantly, the degree of face memory difficulties in children
with autism is not so profound that many of them would be
considered prosopagnosic. Rather, their face-memory difficulties
are reasonably subtle. The CFMT-C has been used with a handful of
young children with developmental prosopagnosia, who often
show severe face recognition problems (Wilson et al., 2010).
Future research should seek to validate further the CFMT-C with
this group of children.
In sum, these results indicate that the Cambridge Face Memory
Test –Children is a valid and reliable measure of unfamiliar face
recognition ability that is sensitive to a wide range of abilities. The
large number of typically developing children in this study
warrants confidence in the results. We note here that adult face
stimuli were used in the CFMT-C to allow comparison with the
adult version of the CFMT (Duchaine & Nakayama, 2006). While
evidence exists for an ‘other age effect’, where participants are
better at recognising faces from their own rather than another age
group, the impact of such an effect in children appears to be small
(Hedges g¼.24; Rhodes & Anastasi, 2012) and, in any case, such an
effect is not an issue for this particular test because children's
scores can be compared to the normative data presented here.
Furthermore, the development of a version of the CFMT-C with
children's faces (see Dalrymple, Gomez, & Duchaine, 2012) will
allow us to test the degree of impact, if any, of the other age effect.
One limitation of the CFMT-C, just like its adult counterpart, is that
the same faces are used for the upright and inverted tasks,
precluding the possibility of using the task to examine the degree
of the inversion effect in individual participants. Nevertheless, the
test, which is freely available for research purposes, will unques-
tionably prove useful for those wanting to examine patterns of
ability and disability in face processing in children and to compare
face identity recognition performance across laboratory sites and
across populations.
Acknowledgements
This paper is in memory of Dr Andy Calder, a mentor and
friend, whose dedication to the highest quality scientific research
in face processing was unsurpassed. We are very grateful to the
children, families and school staff who kindly took part in this
research. Thanks also to Laura Dixon, Louise Edgington, Marlene
Flögel, Emma Jaquet, Lydia King, Elena Klaric, Catherine Manning,
Romina Palermo, Gill Rhodes, Erica Salomone, Martin Thirkettle
and Ellie Wilson for help with data collection and to Marc Stears
for comments on a previous version of this manuscript. Research
at the Centre for Research in Autism and Education (CRAE) is also
supported by The Clothworkers' Foundation and Pears Foundation.
References
American Psychiatric Association. (2013). Diagnostic and statistical manual of mental
disorders (5th ed.). Arlington, VA: American Psychiatric Publishing.
Aiken, L. R. (2003). Psychological testing and assessment (11th ed.). Boston, MA: Allyn
and Bacon.
Banissy, M. J., Garrido, L., Kusnir, F., Duchaine, B., Walsh, V., & Ward, J. (2011).
Superior facial expression, but not identity recognition, in mirror-touch
synesthesia. The Journal of Neuroscience,31, 1820–1824.
Benton, A., Sivan, A., Hamsher, K., Varney, N., & Spreen, O. (1983). The Benton Facial
Recognition TestNew York: Oxford University Press.
Bowles, D. C., McKone, E., Dawel, A., Duchaine, B., Palermo, R., Schmalzl, L., et al.
(2009). Diagnosing prosopagnosia: effects of ageing, sex, and participant–
stimulus ethnic match on the Cambridge Face Memory Test and Cambridge
Face Perception Test. Cognitive Neuropsychology,26, 423–455.
Brace, N., Hole, G., Kemp, R., Pike, G., Van Duuren, M., & Norgate, L. (2001).
Developmental changes in the effect of inversion: using a picture book to
investigate face recognition. Perception,30,85–94.
Bruce, V., Campbell, R., Doherty Sneddon, G., Import, A., Langton, S., McAuley, S.,
et al. (2000). Testing face processing skills in children. British Journal of
Developmental Psychology,18,319–333.
Bushnell, I. W. R., Sai, F., & Mullin, J. T. (1989). Neonatal recognition of the mother's
face. British Journal of Developmental Psychology,7,3–15.
Carey, S., & Diamond, R. (1977). From piecemeal to configurational representation
of faces. Science,195 , 312–314.
Crookes, K., & McKone, E. (2009). Early maturity of face recognition: No childhood
development of holistic processing, novel face encoding, or face-space. Cogni-
tion,111,219–247.
Dalrymple, K. A., Corrow, S., Yonas, A., & Duchaine, B. (2012a). Developmental
prosopagnosia in childhood. Cognitive Neuropsychology,29, 393–418.
Dalrymple, K. A., Gomez, J., & Duchaine, B. (2012b). CFMT-Kids: a new test of face
memory for children. Journal of Vision,9, 492.
Dawson, G., Webb, S., & McPartland, J. (2005). Understanding the nature of face
processing impairment in autism: insights from behavioral and electrophysio-
logical studies. Developmental Neuropsychology,27, 403–424.
Diamond, R., & Carey, S. (1986). Why faces are and are not special: An effect of
expertise. Journal of Experimental Psychology: General,115,107–117 .
Di Simplicio, M., Massey-Chase, R., Cowen, P., & Harmer, C. (2009). Oxytocin
enhances processing of positive versus negative emotional information in
healthy male volunteers. Journal of Psychopharmacology,23,241–24 8.
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,
576–585.
Duchaine, B. C., & Nakayama, K. (2004). Developmental prosopagnosia and the
Benton Facial Recognition Test. Neurology,62,1219–1220.
Duchaine, B. C., & Weidenfeld, A. (2003). An evaluation of two commonly used tests
of unfamiliar face recognition. Neuropsychologia,41,713–720.
Germine, L., Duchaine, B., & Nakayama, K. (2011). Where cognitive development
and aging meet: Face learning ability peaks after age 30. Cognition,11 8,
201–210.
Golarai, G., Liberman, A., Yoon, J. M., & Grill-Spector, K. (2009). Differential
development of the ventral visual cortex extends through adolescence.Article 80,3
http://dx.doi.org/10.3389/neuro.09.080.2009.
Haist, F., Adamo, M., Han Wazny, J., Lee, K., & Stiles, J. (2013). The functional
architecture for face-processing expertise: fMRI evidence of the developmental
trajectory of the core and the extended face systems. Neuropsychologia,51 ,
2893–2908.
Hedley, D., Brewer, N., & Young, R. (2011). Face recognition performance of
individuals with Asperger syndrome on the Cambridge Face Memory Test.
Autism Research,4,449–455.
Johnston, R. A., & Ellis, H. (1995). The development of face recognition. In:
T. Valentine (Ed.), Cognitive and computational aspects of face recognition:
Explorations in face space (pp. 1–23). London: Routledge.
Kirchner, J. C., Hatri, A., Heekeren, H. R., & Dziobek, I. (2011). Autistic symptomatol-
ogy, face processing abilities, and eye fixation patterns. Journal of Autism and
Developmental Disorders,41,158–167.
McKone, E., Martini, P., & Nakayama, K. (2001). Categorical perception of face
identity in noise isolates configural processing. Journal of Experimental Psychol-
ogy: Human Perception and Performance,27,573–599.
Mondloch, C. J., Le Grand, R., & Maurer, D. (2002). Configural face processing
develops more slowly than featural face processing. Perception,31, 553–566.
Mondloch, C. J., Le Grand, R., & Maurer, D. (2010). Development of expertise in face
recognition. In: I. Gauthier, M. J. Tarr, & D. Bub (Eds.), Perceptual expertise:
bridging brain and behavior (pp. 67–106). Oxford, UK: Oxford University Press.
O’Hearn, K., Schroer, E., Minshew, N., & Luna, B. (2010). Lack of developmental
improvement on a face memory task during adolescence in autism. Neuropsy-
chologia,48, 3955–3960.
A. Croydon et al. / Neuropsychologia 62 (2014) 60–6766
Pascalis, O., de Schonen, S., Morton, J., Deruelle, C., & Fabre-Grenet, M. (1995).
Mother’s face recognition by neonates: a replication and an extension. Infant
Behavior &Development,18,79–85.
Pascalis, O., Demont, E., de Haan, M., & Campbell, R. (20 01). Recognition of faces of
different species: A developmental study between 5 and 8 years of age. Infant
and Child Development,10,39–45.
Pellicano, E., & Rhodes, G. (2003). Holistic processing of faces in preschool children
and adults. Psychological Science,14,618–622.
Pellicano, E., Rhodes, G., & Peters, M. (2006). Are preschoolers sensitive to
configural information in faces? Developmental Science,9, 270–277.
Rhodes, M. G., & Anastasi, J. S. (2012). The own-age bias in face recognition: a meta-
analytic and theoretical review. Psychological Bulletin,138 ,146–174.
Richler, J. J., Cheung, O. S., & Gauthier, I. (2011). Holistic processing predicts face
recognition. Psychological Science,22, 464–471.
Rutter, M., Bailey, A., & Lord, C. (2003). Social Communication Questionnaire. Los
Angeles, CA: Western Psychological Services.
Sangrigoli, S., & de Schonen, S. (2004). Recognition of own-race and other-race
faces by three-month-old infants. Journal of Child Psychology and Psychiatry,45,
219–1227 .
Schultz, R. T. (2005). Developmental deficits in social perception in autism: the role
of the amygdala and fusiform face area. International Journal of Developmental
Neuroscience,23,125–141.
Susilo, T., Germine, L., & Duchaine, B. (2013). Face recognition ability matures late:
evidence from individual differences in young adults. Journal of Experimental
Psychology: Human Perception &Performance,39, 1212–1217.
Warrington, E. (1984). Manual for the recognition memory test for words and
facesWindsor,UK: NFER-Nelson.
Wechsler, D. (1999). Wechsler Abbreviated Scale of Intelligence. San Antonio, TX:
Psychological Corporation.
Weigelt, S., Koldewyn, K., & Kanwisher, N. (2012). Face identity recognition in
autism spectrum disorders: a review of behavioural studies. Neuroscience and
Biobehavioral Reviews,36, 1060–1084.
Wilmer, J. B., Germine, L., Chabris, C. F., Chatterjee, G., Williams, M., Loken, E., et al.
(2010). Human face recognition ability is specific and highly heritable. Proceed-
ings of the National Academy of Sciences of the United States of America,107,
5238–5241.
Wilson, C. E., Palermo, R., Schmalzl, L., & Brock, J. (2010). Specificity of impaired
facial identity recognition in children with suspected developmental proso-
pagnosia. Cognitive Neuropsychology,27,30–45.
Yin, R. K. (1969). Looking at upside-down faces. Journal of Experimental Psychology,
81,141–145.
Zwaigenbaum, L., Bryson, S., Rogers, T., Roberts, W., Brian, J., & Szatmari, P. (2005).
Behavioral manifestations of autism in the first year of life. International Journal
of Developmental Neuroscience,23,143–152.
A. Croydon et al. / Neuropsychologia 62 (2014) 60–67 67
