Gillian Rhodes1, Linda Jeffery1 & Libby Taylor1, William G. Hayward1,2, & Louise Ewing1
1ARC Centre of Excellence in Cognition and its Disorders, School of Psychology, University of Western Australia
2Department of Psychology, University of Hong Kong, Hong Kong
WORD COUNT: 2503
Please address correspondence to:
Professor Gillian Rhodes
School of Psychology
University of Western Australia
35 Stirling Highway
Crawley, WA 6009
Running head: Individual differences in face adaptation
Keywords: face recognition, face identity aftereffects, face adaptation, individual differences
Abstract Despite their similarity as visual patterns, we can discriminate and recognize many thousands of faces. This expertise has been linked to two coding mechanisms: holistic integration of information across the face and adaptive coding of face identity using norms tuned by experience. Recently, individual differences in face recognition ability have been discovered and linked to differences in holistic coding. Here we show that they are also linked to individual differences in adaptive coding of face identity, measured using face identity aftereffects. Identity aftereffects correlated significantly with several measures of face-selective recognition ability. They also correlated marginally with own-race face recognition ability, suggesting a role for adaptive coding in the well-known other-race effect. More generally, these results highlight the important functional role of adaptive face-coding mechanisms in face expertise, taking us beyond the traditional focus on holistic coding mechanisms.
Our ability to discriminate and recognize many faces, despite their similarity as visual patterns, has traditionally been linked to holistic coding mechanisms. These integrate information across the face and represent spatial relations between component features as well as the features themselves (Farah, Wilson, Drain, & Tanaka, 1998; Maurer, Grand, & Mondloch, 2002; McKone, Kanwisher, & Duchaine, 2007; McKone & Robbins, 2011; Rhodes, 2013; Rossion, 2008; Tanaka & Gordon, 2011).
More recently, there has been growing interest in the role that adaptive face-coding mechanisms might play in face expertise (for reviews, see Rhodes, 2013; Rhodes & Leopold, 2011; Webster & MacLeod, 2011). The adaptive nature of face identity coding is highlighted by face identity aftereffects, where exposure to a face (e.g., antiDan) shifts the average (norm) towards that face, biasing perception selectively towards the opposite identity (e.g., Dan) (Leopold, O'Toole, Vetter, & Blanz, 2001; Rhodes & Jeffery, 2006; Tsao, Freiwald, Tootell, & Livingstone, 2006) (Figure 1). This selective bias towards the identity opposite (relative to the average) suggests that the average functions as a perceptual norm for coding identity. These aftereffects reflect adaptation of higher-level face-coding mechanisms and cannot be fully explained by adaptation of low-level or mid-level (generic) shape-coding mechanisms (e.g., Rhodes, Evangelista, & Jeffery, 2009; Susilo, McKone, & Edwards, 2010).
Figure 1. A hypothetical two-dimensional face-space with two target identities, reduced identity strength versions of those faces, and an average face. Each target identity (e.g., Dan) has a matching antiface (e.g., antiDan) with opposite attributes (e.g., Dan has thin lips, antiDan has thick lips). In the identity aftereffect, exposure to a face (e.g., antiDan) shifts the average (norm) towards that face, biasing perception selectively towards the opposite identity (e.g., Dan) and making it easier to identify low identity-strength versions of Dan.
Two lines of evidence suggest that adaptive coding of identity may contribute to face expertise. First, face identity aftereffects are reduced in populations with face recognition difficulties (Fiorentini, Gray, Rhodes, Jeffery, & Pellicano, 2012; Palermo, Rivolta, Wilson, & Jeffery, 2011; Pellicano, Jeffery, Burr, & Rhodes, 2007). Second, in typical populations, discrimination can be better around an average than a non-average face (Armann, Jeffery, Calder, Bülthoff, & Rhodes, 2011; Wilson, Loffler, & Wilkinson, 2002), suggesting that adaptive calibration of one’s norm to match the population average is useful (Rhodes, Watson, Jeffery, & Clifford, 2010). However, functional benefits are not always found, despite extensive testing (e.g., Nishimura, Doyle, Humphreys, & Behrmann, 2010; Rhodes, Maloney, Turner, & Ewing, 2007) and it is an open question whether adaptive coding contributes to face recognition ability.
Here we seek new evidence for a functional role of adaptive coding in face expertise using an individual differences approach. Despite a long-standing view that we are all face experts, it is now clear that there are strong and stable individual differences in face recognition ability (Wilhelm et al., 2010; Wilmer et al., 2010). These have been linked to variation in holistic coding, supporting a functional role of holistic coding in face expertise (DeGutis, Wilmer, Mercado, & Cohan, 2012; Richler, Cheung, & Gauthier, 2011; Wang, Li, Fang, Tian, & Liu, 2012) (but see Konar, Bennet, & Sekuler, 2010). Here, we asked whether they are also linked to individual differences in adaptive coding of identity, consistent with a functional role for adaptive coding mechanisms. Individual differences in face-selective recognition ability have recently been linked to figural eye-height aftereffects (Dennett, McKone, Edwards, & Susilo, 2012). However, although eye-height may be relevant to identity, these are distortion aftereffects, which transfer across identity and require judgments about normality, rather than identity. It remains to be seen, therefore, whether adaptive coding of identity itself is linked to face recognition ability.
Our first aim was to determine whether there are stable individual differences in face identity aftereffects that are related to face recognition ability. We measured adaptive coding of identity directly, using face identity aftereffects. We measured face recognition ability using the well-known Cambridge Face Memory Test (CFMT) (Duchaine & Nakayama, 2006) and an old-new recognition memory test developed for the present study. We also measured non-face memory, using the Cambridge Car Memory Test (CCMT) (Dennett et al., 2011), so that we could derive indices of face-selective recognition ability (DeGutis et al., 2012).
Our second aim was to determine whether adaptive coding of identity might be linked more specifically to own-race face expertise, i.e., to better recognition of own-race than other-race faces (Meissner & Brigham, 2001). Individual differences in own-race expertise have been linked to holistic coding differences (in Caucasian individuals) (DeGutis, Mercado, Wilmer, & Rosenblatt, 2013), but it is not known whether they are linked with adaptive coding differences. Previous studies have shown that Caucasian individuals maintain distinct norms for faces of different races (Caucasian and Asian) and have similar-sized face aftereffects for own- and other-race faces (Armann et al., 2011; Jaquet, Rhodes, & Hayward, 2008). Therefore, it is unlikely that greater adaptation to own- than other-race faces contributes to superior expertise for own-race faces. However, in a predominantly own-race environment, the coding mechanisms of people who adapt more (to all faces) might become more selectively tuned (calibrated) to own-race faces. If this calibration helps us recognize faces, as proposed, then face identity aftereffects, which index strength of adaptation, should be linked to own-race expertise.
To test whether adaptive coding of identity is linked specifically to own-race face expertise, we measured old-new recognition of other-race (Chinese) and own-race (Caucasian) faces in our Caucasian participants, and used residuals from a regression in which other-race recognition scores predicted own-race recognition scores to isolate own-race-selective expertise (following DeGutis et al., 2012). A positive correlation between these residuals and face identity aftereffects would link adaptability to own-race expertise.
Two-hundred-and-forty Caucasian adults (63 male) (M = 19.3 years, SD = 4.1 yrs, range = 17 – 46) participated for course credit.
Face Identity Aftereffect. This task measures adaptive coding of identity, and was adapted from previous studies (Jeffery et al., 2011; Rhodes et al., 2011) (details in Supplementary Materials). Briefly, on each trial participants view an adapting face, followed by a (low-identity-strength) target face, which they must identify (Figure 1). On match trials, the adapting antiface lies opposite the target identity (e.g., adapt antiDan, test Dan), facilitating its identification. On mismatch trials (e.g., adapt antiJim, test Dan), the adapting face is not opposite the target, impairing identification (because perception is biased towards the non-target identity that lies opposite the adaptor, e.g., Jim). The aftereffect is measured as accuracy on match trials minus accuracy on mismatch trials. Adapt and test faces were different sizes to minimize the contribution of low-level, retinotopic adaptation.
Cambridge Face Memory Test (CFMT).The CFMT is a well-validated and widely used test of face recognition ability (Duchaine & Nakayama, 2006). Briefly, it tests memory for six male Caucasian faces, under three conditions: test faces (3AFC) that match the images studied, test faces that are different images from the study faces and test faces that are different images with visual noise added. We used the total score (maximum of 72) as the dependent measure.
Old-New Recognition Memory.We created an old-new recognition test to further measure memory for own-race (Caucasian) faces, and to measure memory for other-race (Asian) faces (details in Supplementary Materials). Briefly, participants saw study faces (male) in front view and had to identify these from a series of test faces shown in ¾ view. They saw six study-test blocks, each with 10 study faces (3000 ms each), followed by 20 test faces (5000 ms each). Half the blocks contained own-race faces and half contained other-race faces, with blocks alternating by race. The dependent measure, d, was calculated separately for own-race and other-race faces.
Cambridge Car Memory Test (CCMT).The CCMT is analogous to the CFMT, but uses cars instead of faces (Dennett et al., 2011).
Participants were tested individually in two sessions, lasting up to 40 minutes, one week apart. Participants completed two or more of the following tasks, in the order indicated: Face Identity Aftereffect (FIAE) (N = 129), Old-New Recognition (N = 149), CFMT (N = 240), CCMT (N = 65). The Face Identity Aftereffect task was split in half and completed over two sessions. These tasks were part of a larger battery that included tasks unrelated to the present study. All tasks were presented using SuperLab 4.0.6.
Table 1 shows task reliability. Reliability is well established for the CFMT (Bowles et al., 2009) and CCMT (Dennett et al., 2011). Importantly, reliability was reasonable for face identity aftereffects, indicating that there are stable individual differences in the adaptive coding of identity. Reliability was poorer for old-new recognition of other-race faces. There were no multivariate outliers, according to Mahalanobis distances. Univariate SPSS outliers were replaced by scores 2 SDs above/below the mean, as appropriate (CFMT, N = 1; Own-race d, N = 1, Other-race d, N = 1). All variables were normally distributed, except for the CFMT, but skew and kurtosis were within acceptable limits for parametric analysis (Table 1) (Stuart & Kendall, 1958). Descriptive statistics for the final distributions are shown in Table 1.
Table 1. Descriptive statistics and reliability for all tasks.
Face Identity AEs
Notes. Reliabilities for Face Identity Aftereffects (AEs), Own-race d and Other-race d are Spearman-Brown corrected split-half reliabilities (means from 50 random splits). Reliability for the face identity AE task was .60 when calculated using the two halves in which it was administered. CFMT reliability (from Bowles et al., 2009) and CCMT reliability (from Dennett et al., 2011) are Cronbach’s alphas. CFMT – Cambridge Face Memory Test. CCMT – Cambridge Car Memory Test. CFMT_residuals are residuals from a regression using CCMT scores to predict CFMT scores. Own-race_residuals are residuals from a regression using other- race d scores to predict own-race d scores.
Face identity aftereffects correlate with face-selective recognition ability
We found small-moderate, significant positive correlations of face identity aftereffects with the CFMT and own-race d scores (Figure 2, Table 2). Although these correlations weren’t large, they were quite substantial given the upper bounds imposed by reliability (square root of product of the two reliabilities) (CFMT upper bound r = .69; own-race d upper bound r = .60). In contrast, identity aftereffects correlated negatively (and non-significantly) with non-face (CCMT) recognition (Table 2).1 These results link adaptive coding of identity selectively with face recognition rather than visual recognition generally.
To more directly test whether adaptive coding of identity is linked to face-selective recognition ability, we used residuals from a regression that predicted CFMT scores from CCMT scores as an explicit measure of face-selective recognition ability. These residuals correlated moderately and significantly with face identity aftereffects, r = .288, p = .02, N = 65 (Figure 3).
We also derived a second measure of face-selective recognition ability from a PCA conducted on the three face recognition scores (CFMT, own-race d, other-race d) and car recognition (CCMT) scores. As expected, the PCA yielded two factors, with face scores loading on a “face PC” (explaining 40.2% of variance), and car scores loading on a “car PC” (explaining 27.3% of variance). Face identity aftereffects correlated positively and significantly with the face PC, but negatively (and non-significantly) with the car PC (Table 2), further indicating that adaptive coding is linked selectively to face recognition ability, and not to visual memory generally.
Figure 2. Scatterplot with best-fitting regression line illustrating the relationship between face identity aftereffects and two measures of face recognition ability: CFMT and own-race d.
Table 2. Pearson correlations between face identity aftereffects (AEs), face memory variables and non-face (car) memory variables. Face PC and Car PC are principal components from a Principal Components Analysis conducted on face recognition scores (CFMT, own-race d, other-race d) and car recognition scores (CCMT).
Face Identity AEs
Note. Ns are lower than in Table 1 because not all participants did all tasks.
* p < .05, ** p < .01
Figure 3. Scatterplot with best-fitting regression line illustrating the relationship between face identity aftereffects and face-selective recognition ability (residuals from regression predicting CFMT scores from CCMT scores).
Face identity aftereffects correlate with own-race expertise As expected, recognition (d) was better for own-race than other-race faces, t(148) = 7.97, p < .0001 (Cohen’s d = .71), confirming that our test was sensitive to own-race expertise (Table 1). We used residuals from a regression that predicted own-race recognition from other-race recognition as our measure of own-race expertise. Face identity aftereffects showed a small-to-moderate correlation with own-race expertise, r = .194, p = .076, N = 85 (but none with other-race expertise, i.e., residuals from regression predicting other-race from own-race recognition, r = .068, p = .534, N = 85). Although only marginally significant, the effect size indicates a modest link between adaptive coding of identity and own-race expertise.
Our results provide direct evidence that individual differences in face recognition ability are linked to differences in adaptive coding of identity. Face identity aftereffects correlated positively with several measures of face recognition, including measures of face-selective recognition, but negatively (non-significantly) with non-face (car) recognition ability. This link between adaptive coding of identity and face-selective recognition ability supports a functional role for adaptive coding in face expertise.
Our results are consistent with evidence that face adaptation is reduced in some clinical populations with face-processing difficulties (Ewing, Pellicano, & Rhodes, 2012; Palermo et al., 2011; Pellicano et al., 2007; Pellicano, Rhodes, & Calder, 2013). Importantly, they show that this link between face adaptation and recognition performance extends across the full range of neuro-typical performance. One previous study has linked face aftereffects to face recognition ability in the neuro-typical population (Dennett et al., 2012). However, that study measured how adaptation to a single face dimension, eye-height, affected perceptions of normality. Here we measured identity aftereffects, which directly assess how adaptation on all identity-related dimensions affects the perception of identity.
We also found a small-moderate, marginally significant, correlation between face identity aftereffects (for own-race faces) and own-race face expertise in Caucasian participants. This result is expected if face adaptation calibrates coding mechanisms to the diet of faces, because people who adapt more strongly should have mechanisms that are tuned more selectively to own-race faces than their peers (when own-race faces predominate). However, caution is needed. The correlation was only marginally significant, possibly because our measure of other-race recognition had relatively low reliability, and only Caucasian participants were tested. Therefore, it will be important to confirm this link between adaptability of face-coding mechanisms and own-race-selective expertise in future studies.
A plausible reason for the observed link between adaptive coding of identity and recognition ability is that over time adaptation calibrates coding mechanisms to the population of faces that we experience and this calibration facilitates recognition (for reviews see Clifford & Rhodes, 2005; Rhodes & Leopold, 2011; Webster & MacLeod, 2011). Alternatively, Dennett et al. (2012) have proposed that the link between adaptation and face recognition could stem from the slope of the tuning functions of the neural populations that code face dimensions. They argue that steeper tuning curves (over a fixed range) produce both larger aftereffects (because adaptors will produce more activation and hence more adaptation) and better recognition (because subtler discriminations can be made). A similar account might apply to all identity-related dimensions. This account avoids any chicken-and-egg problem of whether more adaptation causes better recognition or vice versa, because both reflect the shape of the underlying neural tuning functions. Of course, the question then becomes, what long-term factors determine the slope of the tuning functions? A good candidate is adaptation (Georgeson, 2004).
Individual differences in face recognition ability have a genetic basis (McKone & Palermo, 2010; Wilmer et al., 2010; Zhu et al., 2010) (but see DeGutis, Wilmer, Mercado, & Cohan, 2013 for a critique of subtraction-based measures used by Zhu et al.). It is possible, therefore, that the individual differences in adaptive coding of identity that are linked to those differences also have a genetic basis. Certainly, individual differences in holistic coding that have been linked to recognition ability appear to have a genetic basis (McKone & Palermo, 2010). Future studies using twins could determine whether the variation in adaptive face-coding mechanisms observed here also has a genetic basis.
Our expertise in recognizing faces has been linked to two coding mechanisms: holistic integration of information across the face and adaptive norm-based coding of identity using norms tuned by experience. The effects obtained here between adaptive coding and face recognition ability were similar in size (small-moderate) to those reported previously between holistic coding and face recognition ability (DeGutis et al., 2012; Richler et al., 2011; Wang et al., 2012). Therefore, holistic and adaptive coding of identity may make similar contributions to our face expertise. It will be interesting in future studies to directly compare the relative contributions of these two coding mechanisms, and try to determine how much of the individual variation in face expertise they can explain together.
This research was supported by the Australian Research Council Centre of Excellence in Cognition and its Disorders (CE110001021), an ARC Professorial Fellowship to Rhodes (DP0877379) and an ARC Discovery Outstanding Researcher Award to Rhodes (DP130102300). We thank Christian Meissner for supplying Caucasian faces used in the old/new recognition tests, Rachell Barker for assistance with stimulus preparation, Ainsley Read for assistance with testing, and Mayu Nishimura and Daphne Maurer for co-creating the Robbers Game used in the Identity Aftereffects task. Ethical approval was granted by the Human Research Ethics Committee of the University of Western Australia.
References Armann, R., Jeffery, L., Calder, A. J., Bülthoff, I., & Rhodes, G. (2011). Race-specific norms for coding face identity and a functional role for norms. Journal of Vision, 11(13:9), 1-14.
Bowles, D. C., McKone, E., Dawel, A., Duchaine, B., Palermo, R., Schmalzl, L., . . . Yovel, G. (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(5), 423-455.
Clifford, C. W. G., & Rhodes, G. (Eds.). (2005). Fitting the Mind to the World: Adaptation and After-Effects in High-Level Vision. Oxford: Oxford University Press.
DeGutis, J., Mercado, Rogelio J, Wilmer, Jeremy, & Rosenblatt, Andrew. (2013). Individual Differences in Holistic Processing Predict the Own-Race Advantage in Recognition Memory. PloS one, 8(4), e58253.
DeGutis, J., Wilmer, J. B., Mercado, R. J., & Cohan, S. (2012). Using regression to measure holistic face processing reveals a strong link with face recognition ability. Cognition, 126, 87-100.
DeGutis, J., Wilmer, J. B., Mercado, R. J., & Cohan, S. (2013). Using regression to measure holistic face processing reveals a strong link with face recognition ability. Cognition, 126, 87-100.
Dennett, H. W., McKone, E., Edwards, M., & Susilo, T. (2012). Face aftereffect predict individual differences in face recognition ability. Psychological Science, 23(11), 1279-1287.
Dennett, H. W., McKone, E., Tavashmi, R., Hall, A., Pidcock, M., Edwards, M., & Duchaine, B. (2011). The Cambridge Car Memory Test: A task matched in format to the Cambridge Face Memory Test, with norms, reliability, sex differences, dissociations from face memory, and expertise effects. Behavior research methods, 44, 587-605.
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.
Ewing, L., Pellicano, E., & Rhodes, G. (2012). Atypical updating of face representations with experience in children with autism. Submitted for publication.
Farah, M. J., Wilson, K. D., Drain, M., & Tanaka, J. N. (1998). What is "special" about face perception? Psychological Review, 105(482-498).
Fiorentini, C., Gray, L., Rhodes, G., Jeffery, L., & Pellicano, E. (2012). Reduced face identity aftereffects in relatives of children with autism. Neuropsychologia, 50(12), 2926-2932.
Georgeson, Mark. (2004). Visual aftereffects: Cortical neurons change their tune. Current Biology, 14(18), R751-R753.
Jaquet, E., Rhodes, G., & Hayward, W. G. (2008). Race-contingent aftereffects suggest distinct perceptual norms for different race faces. Visual Cognition, 16(6), 734-753.
Jeffery, L., Rhodes, G., McKone, E., Pellicano, E., Crookes, K., & Taylor, E. (2011). Distinguishing norm-based from exemplar-based coding of identity in children: Evidence from face identity aftereffects. Journal of Experimental Psychology: Human Perception and Performance, 37(6), 1824-1840. doi: 10.1037/a0025643
Konar, Y., Bennet, P. J., & Sekuler, A. B. (2010). Holistic processing is not correlated with face-identification accuracy. Psychological Science, 21, 38-43.
Leopold, D. A., O'Toole, A. J., Vetter, T., & Blanz, V. (2001). Prototype-referenced shape encoding revealed by high-level aftereffects. nature neuroscience, 4(1), 89-94.
Maurer, D., Grand, R. L., & Mondloch, C. J. (2002). The many faces of configural processing. Trends in Cognitive Sciences, 6(6), 255-260.
McKone, E., Kanwisher, N., & Duchaine, B. (2007). Can generic expertise explain special processing for faces? Trends in Cognitive Sciences, 11(1), 8-15.
McKone, E., & Palermo, R. (2010). A strong role for nature in face recognition. Proceedings of the National Academy of Sciences, 107(11), 4795-4796.
McKone, E., & Robbins, R. (2011). Are faces special? In A. J. Calder, G. Rhodes, M. H. Johnson & J. V. Haxby (Eds.), The Oxford Handbook of Face Perception (pp. 149-176). Oxford: Oxford University Press.
Meissner, Christian A, & Brigham, John C. (2001). Thirty years of investigating the own-race bias in memory for faces: A meta-analytic review. Psychology, Public Policy, and Law, 7(1), 3.
Nishimura, M., Doyle, J., Humphreys, K., & Behrmann, M. (2010). Probing the face-space of individuals with prosopagnosia. Neuropsychologia, 48(6), 1828-1841.
Palermo, R., Rivolta, D., Wilson, C. E., & Jeffery, L. (2011). Adaptive face space coding in congenital prosopagnosia: Typical figural aftereffects but abnormal identity aftereffects. Neuropsychologia, 49(14), 3801-3812.
Pellicano, E., Jeffery, L., Burr, D., & Rhodes, G. (2007). Abnormal adaptive face-coding mechanisms in children with autism spectrum disorder. Current Biology, 17(17), 1508-1512.
Pellicano, E., Rhodes, Gillian, & Calder, Andrew J. (2013). Reduced gaze aftereffects are related to difficulties categorizing gaze direction in autism. Neuropsychologia.
Rhodes, G. (2013). Face Recognition. In D. Resiberg (Ed.), Oxford Handbook of Cognitive Psychology (pp. 46-68). New York: Oxford University Press.
Rhodes, G., Evangelista, E., & Jeffery, L. (2009). Orientation-sensitivity of face identity aftereffects. Vision research, 49, 2379-2385.
Rhodes, G., & Jeffery, L. (2006). Adaptive norm-based coding of facial identity. Vision research, 46, 2977-2987.
Rhodes, G., Jeffery, L., Evangelista, E., Ewing, L., Peters, M, & Taylor, L. (2011). Enhanced attention amplifies face adaptation. Vision research, 51, 1811-1819.
Rhodes, G., & Leopold, D. A. (2011). Adaptive norm-based coding of face identity. In A. J. Calder, G. Rhodes, M. H. Johnson & J. V. Haxby (Eds.), The Oxford Handbook of Face Perception (pp. 263-286). Oxford: Oxford University Press.
Rhodes, G., Maloney, L. T., Turner, J., & Ewing, L. (2007). Adaptive face coding and discrimination around the average face. Vision research, 47, 974-989.
Rhodes, G., Watson, T. L., Jeffery, L., & Clifford, C. W. G. (2010). Perceptual adaptation helps us identify faces. Vision research, 50, 963-968.
Richler, J. J., Cheung, O. S., & Gauthier, I. (2011). Holistic processing predicts face recognition. Psychological Science, 22(4), 464-471.
Rossion, B. (2008). Picture-plane inversion leads to qualitative changes of face perception. Acta Psychologica, 128, 274-289.
Susilo, T., McKone, E., & Edwards, M. (2010). Solving the upside-down puzzle: Why do upright and inverted face aftereffects look alike? Journal of Vision, 10(13), 1-16.
Tanaka, J. W., & Gordon, I. (2011). Features, configuration, and holistic face processing. In A. J. Calder, G. Rhodes, M. H. Johnson & J. V. Haxby (Eds.), The Oxford Handbook of Face Perception (pp. 177-194). Oxford: Oxford University Press.
Tsao, D. Y., Freiwald, W. A., Tootell, R. B. H., & Livingstone, M. S. (2006). A cortical region consisting entirely of face-selective cells. Science, 311(5761), 670.
Wang, R., Li, J., Fang, H., Tian, M., & Liu, J. (2012). Individual differences in holistic processing predict face recognition ability. Psychological Science, 23, 169-177.
Webster, M. A., & MacLeod, D. I. A. (2011). Visual adaptation and face perception. Philosophical Transactions of the Royal Society: Biological Sciences, 366, 1702-1725.
Wilhelm, Oliver, Herzmann, Grit, Kunina, Olga, Danthiir, Vanessa, Schacht, Annekathrin, & Sommer, Werner. (2010). Individual differences in perceiving and recognizing faces-One element of social cognition. Journal of personality and social psychology, 99(3), 530-548.
Wilmer, J. B., Germine, L., Chabris, C. F., Chatterjee, G., Williams, M., Loken, E., . . . Duchaine, B. C. (2010). Human face recognition ability is specific and highly heritable. Proceedings of the National Academy of Sciences, 107, 5238-5241.
Wilson, Hugh R, Loffler, Gunter, & Wilkinson, Frances. (2002). Synthetic faces, face cubes, and the geometry of face space. Vision research, 42(27), 2909-2923.
Zhu, Q., Song, Y., Hu, S., Li, X., Tian, M., Zhen, Z., . . . Liu, J. (2010). Heritability of the specific cognitive ability of face perception. Current Biology, 20, 137-142.
Supplementary Materials - Task Details
Face identity aftereffects task
Stimuli. Grayscale photographs of four male faces comprised the targets (“Dan”, “Jim”, “Rob”, and “Ted”) (taken from Jeffery et al., 2011). Reduced strength versions were made by morphing each face toward an average face (constructed from 20 adult male faces) using Gryphon Morph 2.5 (Maxwell, 1994), in varying increments to produce 15%, 30% and 90% versions of each target. The 100% and 30% stimuli were used only in training and the 90% and 15% stimuli were used as test faces (see Figure 1). Adapting antifaces were made by caricaturing the form of the average male away from each target face by 80% (after Leopold, O'Toole, Vetter, & Blanz, 2001) using Gryphon Morph 2.5 (Maxwell, 1994). To minimize non-facial cues stimuli were masked so that the top of the hairline was obscured. Adapting faces subtended an average visual angle of 5.5° (v) x 5.6° (h), viewed from approximately 65 cm. Test faces were 20% smaller, to minimise the impact of low-level adaptation, measuring 4.8° (v) X 4.5° (h).
The task was presented as the “Robbers Game”, following Jeffery et al (2011). The game context was devised for use with children but has also proved an engaging way to present the task to adults (Jeffery et al, 2011, 2013). Participants began with training in recognizing the targets (100% versions) and familiarization with lower strength versions of the targets (30% versions), who were introduced as the brothers of each target. Participants were told to respond with the team leader’s name (target) whenever they saw the target or a member of his team (his brothers). Participants responded using labeled keyboard keys. Auditory feedback (beeps) was provided for correct and incorrect responses throughout training.
Figure 1. Face stimuli used in the identity aftereffect task. The first row shows the four target identities “Dan”, “Jim”, “Rob”, and “Ted” (l-r). Rows 2, 3, and 4 show weaker versions of the targets used in training and test. Row 5 shows the adapting antifaces.
Participants then completed the adaptation task. The experimenter explained that a robber’s face (adapting antiface) would appear on the screen, “while he was stealing things”, and then it would disappear and be followed by a very brief presentation of the face of the team member (test stimulus) who caught the robber. The participant was asked to identify the catcher’s team. The task comprised 96 trials. The 15% stimuli were the test faces on 64 trials, which comprised 32 match trials, in which the adapting antiface and test face lay on the same identity trajectory and 32 mismatch trials in which the adapt and test face came from different identity trajectories (see Figure 2). Each target was shown equally often in each condition (i.e. 8 trials per target). The remaining 32 trials featured 90% targets as test faces (16 match and 16 mismatch trials). These easy-to-identify stimuli were included to ensure participants remained motivated throughout the task and data from these trials were not analyzed.
Each trial began with blank grey screen (300 ms) then the adapting faces was shown (5000 ms), followed by a 150 ms inter-stimulus interval, then the test face (400ms). A blank screen followed and remained until the participant responded. Following response participants initiated the next trial by pressing the spacebar. Trial sequence is shown in Figure 2. Trials were presented in one of two pseudorandom orders, constrained so that the same adapting face did not appear more than four times consecutively, to avoid building up adaptation to one face. Participants were randomly allocated to one order. The 96 trials were divided into eight blocks of 12 trials. Half the trials were completed in the first test session and the remainder in the week following (second session). Training was completed prior to the adaptation task on both days. The experimenter observed each participant to ensure they looked at the adapting faces for the full presentation.
Figure 2. Example match and mismatch trial sequence and timings in the Face Identity Aftereffect Task (adapted from Jeffery et al, 2011).
Old-new recognition of own-race and other-race faces
This was a standard old-new recognition task using Asian and Caucasian faces. It took approximately 30 minutes to complete.
Stimuli. Study faces were front-views of 30 Caucasian and 30 Asian faces. Test faces consisted of the study faces photographed at 3/4 view, as well as an 30 3/4 view distractors for each race. All faces were male and displaying a neutral expression. Caucasian face stimuli were obtained from Christian Meissner and Asian face stimuli were obtained from William Hayward. To minimize the use of non-face cues, the hair (including forehead hairline) was masked. Study faces measured approximately 5.9° x 5.3° (Asian), and 6.0° x 5.3 (Caucasian), and test faces measured 6.5° x 5.5° (h) (Asian), and 6.8° x 5.7° (h) (Caucasian), when viewed from 60 cm.
Procedure. Participants were verbally informed that they would see 10 faces consecutively, which they must try to remember, and that once they have seen all 10 faces, they would see the faces a second time to enhance their memory. Participants were instructed that once this study phase was over, they would then be shown a bigger group of faces, one at a time, to which they must indicate whether the face was