Brief Report: Developing Spatial Frequency Biases for Face Recognition in Autism and Williams Syndrome

Transcription

1 J Autism Dev Disord (2011) 41: DOI /s BRIEF REPORT Brief Report: Developing Spatial Frequency Biases for Face Recognition in Autism and Williams Syndrome Hayley C. Leonard Dagmara Annaz Annette Karmiloff-Smith Mark H. Johnson Published online: 14 October 2010 Ó Springer Science+Business Media, LLC 2010 Abstract The current study investigated whether contrasting face recognition abilities in autism and Williams syndrome could be explained by different spatial frequency biases over developmental time. Typically-developing children and groups with Williams syndrome and autism were asked to recognise faces in which low, middle and high spatial frequency bands were masked. All three groups demonstrated a gradual specialisation toward the mid-band. However, while the use of high spatial frequencies decreased in control and autism groups over development, the Williams syndrome group did not display a bias toward this band at any point. These data demonstrate that typical outcomes can be achieved through atypical developmental processes, and confirm the importance of cross-syndrome studies in the investigation of developmental disorders. Keywords Face recognition Spatial frequency Development Autism Williams syndrome Autism and Williams syndrome (WS) are two geneticallybased neuro-developmental disorders that display contrasting social phenotypes (Bailey et al. 1995; Osborne H. C. Leonard (&) A. Karmiloff-Smith M. H. Johnson Centre for Brain and Cognitive Development, School of Psychological Sciences, Birkbeck, University of London, Malet Street, London WC1E 7HX, UK h.leonard@psychology.bbk.ac.uk D. Annaz School of Health and Social Sciences, Middlesex University, London, UK 2006). While autism is defined by reduced and atypical social interactions (American Psychiatric Association 2000), individuals with WS are often hypersocial, with excessive interest in people (Bellugi et al. 2007). These differences are already apparent early in development, with infants with WS fixating for longer on faces (e.g., Mervis et al. 2003) and young children with autism displaying fewer and shorter fixations to faces (e.g., Swettenham et al. 1998; although see Klein-Tasman et al. 2007, and Lincoln et al. 2007, for examples of overlap in social-communication behaviour in the two groups). Furthermore, investigations of brain activity underlying face perception in the two disorders suggest different neural processes contribute to the atypical behavioural outcomes (e.g., Grice et al. 2001; Schultz et al. 2000; Meyer-Lindenberg et al. 2006). While face perception has received a great deal of attention in the autism and WS literature, there remain mixed results concerning how faces are processed in these populations. In typical development, it has been suggested that younger children may rely more on features for face perception, while older children and adults use a more configural strategy, in which the whole face and the relationship between different features is quickly assessed (e.g., Carey and Diamond 1977; Mondloch et al. 2002; Schwarzer 2000; although see Leonard et al. 2010, for further discussion of this topic). On the other hand, a featural bias resulting from a deficit in configural processing has been suggested in both autism (e.g., Hobson et al. 1988; Langdell 1978) and in WS (e.g., Deruelle et al. 1999; Karmiloff-Smith et al. 2004). However, face inversion effects have been found in both groups (e.g., Annaz et al (autism); Rose et al (WS)), implying that sensitivity to configural information may not be reduced in the two populations. One difficulty with interpretation of this research is that terms such as holistic, configural and featural face processing are

2 J Autism Dev Disord (2011) 41: often defined differently across research groups and a variety of tasks are used to tap into these distinct strategies. The current study focused on the use of low-level information, avoiding this confusion and investigating whether the two groups rely on different spatial frequencies to recognise faces throughout development. High spatial frequencies (HSFs) are usually proposed to convey the features of the face, and low spatial frequencies (LSFs) the more configural information (e.g., Boeschoten et al. 2005; Goffaux et al. 2005). However, adults tend to rely more on a mid-band of spatial frequencies (generally between 8 and 24 cycles per face) to recognise faces quickly and accurately (see Ruiz-Soler and Beltran 2006, for a review). In the first study of the development of this mid-band bias, Leonard et al. (2010) found that younger children were not reliant on the mid-band for face recognition, but tended to rely more on the HSFs than did older children and adults, who demonstrated a clear bias toward the mid-band. While previous studies (Deruelle et al. 2004, 2008) have suggested that high-functioning children with autism between the ages of 4 and 16 are characterised by an HSF bias, no study has investigated whether spatial frequency biases for face recognition develop similarly in autism or WS, nor whether they eventually achieve the same mid-band bias found in typically-developing adults. The current study investigated the development of spatial frequency (SF) biases in autism and WS by building cross-syndrome developmental trajectories of performance on a face matching task. This approach promotes better understanding of the developmental processes involved in achieving specific behavioural outcomes (Thomas et al. 2009). Children with Autism Spectrum Disorder (ASD) and Williams syndrome (WS) were compared to typicallydeveloping controls on an identity matching task developed by Leonard et al. (2010) for use with young children. Participants were asked to match faces to a previously presented target when particular SF bands were masked in the stimulus. Reduced accuracy when a particular SF band was masked would suggest a greater bias toward this band during the task. In line with Leonard et al. (2010), it was expected that typically-developing children would demonstrate a greater bias toward HSFs early in development, and a mid-band bias by the age of 15. Based on the studies by Deruelle et al. (2004, 2008), it was hypothesised that children with ASD would demonstrate a bias toward HSFs into adolescence. Due to their different social profiles, it might also be expected that individuals with WS would differ from the ASD group in their use of spatial frequencies in this task. It is not clear, however, how they would differ from the control group. The use of trajectories allows further investigation of these questions, focusing on how spatial frequency biases emerge over development in the three groups. Method Participants Sixty-seven children (age range: 7 years, 2 months 15 years, 7 months) were divided into three groups: ASD (N = 17, males = 15), WS (N = 14, males = 5), and controls (N = 36, males = 17). This age range was chosen to replicate and extend the investigation by Leonard et al. (2010) in typically-developing children, which studied 7 10 year-olds. The age range was extended for the current study to increase the possibility that children in the ASD and WS groups would achieve the same level of performance as the youngest control children at the very least. Children in the WS group had a primary diagnosis of Williams syndrome from a qualified clinician and all had positive genetic tests. The children with ASD were recruited through specialist schools and clubs for individuals with autism on the basis that they had a British statement of special needs, with a primary diagnosis of autism, from a trained psychiatrist or pediatrician using established criteria for Autism Spectrum Disorders. Based on background data from schools and parents, it was ensured that the control group did not have any clinical diagnoses or learning difficulties before they were included in the study. Participant characteristics are provided in Table 1. All participants were tested for nonverbal Mental Age (using Raven s Progressive Matrices, Raven et al. 2000) and face recognition ability (Benton Facial Recognition Test, Benton et al. 1983). The two disorder groups were further assessed for verbal Mental Age using the British Picture Vocabulary Scale II (Dunn et al. 1997). The three groups did not differ in terms of mean chronological age, F(2,64) = 1.57, p =.22, but did differ from each other on their facial recognition scores, F(2,64) = 8.89, p \.001, and on measures of mental age, Fs [ 4.1, ps \.05, in line with the phenotypes of their disorders. However, these background measures did not explain a significant proportion of the variance in SF use in either disorder group. The focus of this paper is therefore on the change in SF use with increasing chronological age. In addition to these final groups, a further seventeen children were excluded from the analyses for failing to perform significantly above chance when matching the unmasked faces on the basis of identity (i.e., for not achieving at least 7 out of 8 correct). This included seven children with ASD (mean age = 10 years, 7 months), six with WS (mean age = 8 years, 11 months) and four TD children (mean age = 8 years, 9 months). This baseline measure identified both those children who could not perform the basic discrimination between the faces and those who lost concentration by the end of the task. The major

3 970 J Autism Dev Disord (2011) 41: Table 1 Group means, ranges and (standard deviations) of chronological ages, mental ages and facial recognition scores Group Chronological age (years;months) Verbal mental age a (years;months) Nonverbal mental age b (years;months) Benton facial recognition score Control 11;5 (2;7) N/A 10;2 (2;11) (2.33) 7;6 15;5 N/A 6;6 15; WS 11;10 (2;8) 7;6 (1;7) 5;9 (1;4) (2.74) 7;7 15;7 4;7 10;2 3;0 7; ASD 10;4 (2;5) 9;3 (2;8) 9;6 (2;9) (3.23) 7;2 15;3 6;4 16;8 6;6 15; a Age equivalence scores from the BPVS b Age equivalence scores from the Ravens Progressive Matrices reason for exclusion was therefore a difficulty maintaining concentration to the end of the experiment in the two disorder groups. In addition, previous piloting in typicallydeveloping children found that decreasing testing age below 7 years resulted in a drastically increased drop-out rate. Children with developmental delay are therefore more likely to fail to reach baseline than their chronological agematched controls. Materials and Procedure All materials were taken from Leonard et al. (2010). Only two face stimuli (identified as Bob and Jimmy to participants) were chosen from the original set of images for the current study. These were judged to be the two faces that differed most from each other in terms of their configural properties from the stimuli used in the previous study, making it easier to discriminate between them. Specifically, the distances from the pupil to the tip of the nose and from the tip of the nose to the mouth were measured. As Leonard et al. (2010) found no effect of stimulus pair in their original study, the current face images were not expected to produce stimulus-specific responses that would not generalise to other stimuli, and so only the two faces were utilised for the current study (see Leonard et al. 2010, for further discussion of the stimuli). Three altered versions of these faces were presented during the computer task, with each of the two identities covered by noise masks at 8, 16 and 32 cycles per image (LSF, MSF and HSF respectively). These corresponded to 1.1, 2.2 and 4.4 cycles per degree during presentation (see Näsänen 1999, for further details concerning stimulus production). Following the child procedure from Leonard et al. (2010), participants undertook a training phase in which they learned the identities of the two faces and practised naming them through a number of different games before beginning the computer task (see Fig. 1). During the test trials, each of the SF masks was presented a total of eight times, with eight unmasked faces randomly presented throughout these trials to provide the baseline measure (producing a total of 32 trials). Trials were initiated by the experimenter and began when participants were attending to the fixation point. One of the two faces then appeared with a spatial frequency mask in the center of the screen for 500 ms (for those aged over 10 years in the control group) or 2 s (younger control children and all individuals in the disorder groups), before being replaced by the two unmasked faces in the same orientation in randomised positions (e.g., left or right) until a response was given. The exposure durations in the control group were identical to those in Leonard et al. (2010), as older children were found to need less time for the face discrimination than younger children, and no differences were found between the two exposure times in the responses of a group of 10-year-olds. The 2-second exposure was adopted for the two disorder groups, irrespective of age, in order to give individuals the same amount of time as younger control children to process the masked face. For further details concerning the main procedure, see Leonard et al. (2010). Once the task was finished, children completed the standardised tests appropriate for their groups. Results Figure 2 illustrates the linear trajectories modelling the use of (or bias toward) each SF band in relation to chronological age for upright trials (see Thomas et al. 2009, for details concerning trajectory analyses). These scores were calculated by subtracting task accuracy from 100% (i.e., achieving 100% accuracy for the LSF mask would demonstrate that the LSF band was not being used in the task, resulting in a use score of 0%). A fully factorial repeated-measures Analysis of Covariance (ANCOVA) was first conducted on the upright data in each group to compare the changing use of spatial frequencies (within-subjects factor) with chronological

4 J Autism Dev Disord (2011) 41: Fig. 1 An example of a test trial presenting an HSF-masked face. A fixation stimulus was followed by a masked face. Participants judged which of the two choice stimuli had been the target. Note that the relative importance of different spatial frequencies depends on the size and viewing distance of the stimuli, and the images illustrated are not the same size as the actual experimental stimuli Fixation Masked target Choice stimuli age (covariate). 1 Separate analyses were conducted in this way to avoid masking any effects found in single groups by the difference in variability between groups (e.g., Annaz et al. 2009). Control Data In line with the results of Leonard et al. (2010), the control group displayed a greater use of HSFs at 7 years than of MSF and LSFs, F(2,70) = 16.14, p \.001, g p 2 =.3, and a change in the use of SF bands with age, F(1,34) = 6.40, p =.02, g p 2 =.2. A significant interaction between SF use and age demonstrates that only HSF use decreased reliably during this period, F(2,68) = 6.53, p \.01, g p 2 =.2. Disorder Group Data Both disorder groups demonstrated significant effects of SF band, F(2,32) = 10.78, p \.001, g p 2 =.4 (ASD), and F(2,26) = 6.12, p \.01, g p 2 =.3 (WS). However, as can be seen in Fig. 2, this was caused by a greater use of HSFs at 7 years in the ASD group, while the WS group displayed a greater use of LSFs at this age. Age was not a reliable predictor of spatial frequency use in either group, F(1,15) =.84, p =.38, g p 2 =.05 (ASD) and F(1,12) =.54, p =.5, g p 2 =.04 (WS). However, the WS group did show a significant interaction between age and spatial frequency use, with the use of LSFs decreasing more rapidly than the use of MSFs or HSFs, F(2,24) = 3.84, p =.04, g p 2 =.2. This interaction did not reach significance in the ASD group, F(2,30) = 2.34, p =.11, g p 2 =.1. A mixed-design ANCOVA, with SF use and disorder group as factors and Chronological Age as a covariate, revealed 1 Note that the main effect of spatial frequency is independent of the covariate, and is reported from an analysis excluding age as a factor; degrees of freedom may therefore differ between main effects and interactions and within- or between-subjects factors (see Annaz et al for further explanation). that the two groups differed significantly in their changing use of SF bands between the ages of 7 and 15, F(2,54) = 3.21, p =.05, g p 2 =.1. Two further planned contrasts revealed that while the WS group also differed from the control group, F(2,92) = 3.21, p =.05, g p 2 =.1, the control and ASD groups did not differ in their development of spatial frequency biases, F(2,98) =.33, p =.72, g p 2 =.01. Discussion The current study replicated previous findings in TD children of a gradual specialisation toward the mid-band, with a greater reliance on HSFs earlier in development (Leonard et al. 2010). In line with the current hypotheses, children with WS displayed a very different development of SF biases from the ASD group, as well as the control group, demonstrating a greater reliance on LSFs and very little use of HSFs at 7 years of age. Surprisingly, the ASD and control groups appear to develop similarly to each other, although the smaller ASD sample meant that the overall pattern did not reach significance. The most interesting finding, however, is that all three groups ultimately achieved the same pattern of spatial frequency biases by the age of 15, with a greater reliance on the mid-band than other SFs. This supports previous research suggesting that typical face processing outcomes can result from atypical developmental trajectories (Karmiloff-Smith et al. 2004). It is interesting to note that the ASD and control groups demonstrate a highly similar level of performance for each SF band, while the WS group demonstrate generally poorer performance overall (see Fig. 2). It could be that that the reduced flexibility in the pattern of biases in WS is a less efficient way of dealing with face stimuli, perhaps reducing the overall level of performance in this group. On the other hand it could be due to the masking technique used and the inability to resolve face stimuli through noise in this group. Future studies could resolve this issue by comparing spatial

5 972 J Autism Dev Disord (2011) 41: Fig. 2 The use of LSFs, MSFs and HSFs in upright face recognition, plotted against chronological age, for a the control group, b the ASD group, and c the WS group. Individual data are not presented for reasons of clarity. R 2 values indicate the proportion of variance explained by each trajectory frequency masks to filtered stimuli and assessing the resulting pattern of biases within WS. The fact that the ASD group performed so similarly to the control group was unexpected and seems to contradict findings by Deruelle et al. (2004, 2008) pointing to different SF biases in the two populations. However, the LSF faces used by these researchers were based on different SF cut-offs to the present stimuli, making it difficult to compare across studies (Leonard et al. 2010). In addition, younger children with ASD did display an HSF bias similar to the previous findings, but developed a typical mid-band bias later. This demonstrates the importance of studying task-specific developmental trajectories (Karmiloff-Smith 1998) and not treating childhood as a static point in time. However, it is possible that the stringent selection process for inclusion in the analyses (i.e., the baseline measure) removed those lower-functioning children with ASD who might have displayed a different pattern of biases. This problem could be overcome by using a visual preference paradigm, in which participants are not required to make an overt response (i.e., pointing or naming), thereby including younger or lower-functioning children. Measurement of brain activity could also provide some insight into whether atypical neural processes underlie the seemingly typical outcome in this task. The current study is the first cross-syndrome developmental comparison of spatial frequency biases in faceidentity processing. This approach supports suggestions that biases toward different facial information during development could contribute to the differences in face recognition found in WS in previous research, and delineates how the use of these different types of information interact and change with chronological age. In particular, the reduced use of HSFs compared to the other groups suggests that children with WS are using alternative strategies for face-identity processing. One possibility is that the WS group are missing out this step of typical development, i.e., the mid-band specialisation occurs earlier, thus reducing the flexibility of the SF processing system. It is not clear how the mid-band of spatial frequencies might relate to the configural properties of a face, making it difficult to equate the MSF bias found here to the configural deficit reported previously in WS (e.g., Deruelle et al. 1999; Karmiloff-Smith et al. 2004). However, it clear that achieving a mid-band bias in WS and typical development occurs through different developmental processes, which is in line with these previous reports. Our results are also consistent with the idea that a midband spatial frequency bias for face processing occurs because this frequency band contains the optimal diagnostic features for face identification (Schyns and Oliva 1999). This pressure to utilise mid-band frequencies may drive different developmental trajectories to reach the same end attractor state, i.e., a state in which optimal information is provided for face recognition. As with the ASD group, it will be useful in future research to study younger and lower-functioning children with WS. Comparisons could also be made with the pattern of biases used for object recognition over time to determine the face-specificity of the mid-band bias. In sum, while some questions remain open, the current study s novel approach to investigating face-identity processing in atypical populations provides insight into the cascading effects of early visual biases over developmental time.

6 J Autism Dev Disord (2011) 41: Acknowledgments The current research was supported by a UK Medical Research Council Grant G ID: to the last author and by a UK Medical Research Council studentship to the first author. We would like to thank Michael Thomas for his advice and support and the charities and schools who helped with recruitment. Our thanks go to all the parents and children who participated in the study. References Annaz, D., Karmiloff-Smith, A., Johnson, M. H., & Thomas, M. S. C. (2009). A cross- syndrome study of the development of holistic face recognition in children with autism, Down syndrome and Williams syndrome. Journal of Experimental Child Psychology, 102, American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed. ed.). Washington D.C.: Author. Bailey, A., Le Couteur, A., Gottesman, I., Bolton, P., Simonoff, E., Yuzda, E., et al. (1995). Autism as a strongly genetic disorder: Evidence from a British twin study. Psychological Medicine, 25, Bellugi, U., Järvinen-Pasley, A., Doyle, T. F., Reilly, J., Reiss, A. L., & Korenberg, J. R. (2007). Affect, social behaviour, and the brain in Williams syndrome. Current Directions in Psychological Science, 6(2), Benton, A., Hamsher, K., Varney, N. R., & Spreen, O. (1983). Benton test of facial recognition. New York: Oxford University Press. Boeschoten, M. A., Kemner, C., Kenemans, J. L., & van Engeland, H. (2005). The relationship between local and global processing and the processing of high and low spatial frequencies studied by event-related potentials and source modelling. Cognitive Brain Research, 24, Carey, S., & Diamond, R. (1977). From piecemeal to configurational representation of faces. Science, 195, Deruelle, C., Mancini, J., Livet, M. O., Cassé-Perrot, C., & de Schonen, S. (1999). Configural and local processing of faces in children with Williams syndrome. Brain and Cognition, 41, Deruelle, C., Rondan, C., Gepner, B., & Tardif, C. (2004). Spatial frequency and face processing in children with autism and Asperger syndrome. Journal of Autism and Developmental Disorders, 34, Deruelle, C., Rondan, C., Salle-Collemiche, X., Bastard-Rosset, D., & Da Fonseca, D. (2008). Attention to low- and high-spatial frequencies in categorizing facial identities, emotions and gender in children with autism. Brain and Cognition, 66, 115. Dunn, L. M., Whetton, C., & Pintilie, D. (1997). British picture vocabulary scale. Windsor, UK: NFER Nelson. Goffaux, V., Hault, B., Michel, C., Vuong, Q. C., & Rossion, B. (2005). The respective role of low and high spatial frequencies in supporting configural and featural processing of faces. Perception, 34, Grice, S. J., Spratling, M. W., Karmiloff-Smith, A., Halit, H., Csibra, G., de Haan, M., et al. (2001). Disordered visual processing and oscillatory brain activity in autism and Williams syndrome. NeuroReport, 12, Hobson, P., Ouston, J., & Lee, A. (1988). What s in a face? The case of autism. British Journal of Psychology, 79, Karmiloff-Smith, A. (1998). Development itself is the key to understanding developmental disorders. Trends in Cognitive Sciences, 2, Karmiloff-Smith, A., Thomas, M., Annaz, D., Humphreys, K., Ewing, S., Brace, N., et al. (2004). Exploring the Williams syndrome face processing debate: The importance of building developmental trajectories. Journal of Child Psychology and Psychiatry, 45, Klein-Tasman, B. P., Mervis, C. B., Lord, C., & Phillips, K. D. (2007). Socio- communicative deficits in young children with Williams syndrome: Performance on the autism diagnostic observation schedule. Child Neuropsychology, 13, Langdell, T. (1978). Recognition of faces: An approach to the study of autism. Journal of Child Psychology and Psychiatry, 19, Leonard, H. C., Karmiloff-Smith, A., & Johnson, M. H. (2010). The development of spatial frequency biases in face recognition. Journal of Experimental Child Psychology, 106, Lincoln, A. J., Searcy, Y. M., Jones, W., & Lord, C. (2007). Social interaction behaviors discriminate young children with autism and Williams syndrome. Journal of the American Academy of Child and Adolescent Psychiatry, 46, Mervis, C. B., Morris, C. A., Klein-Tasman, B. P., Kwitny, S., Appelbaum, L. G., & Rice, C. E. (2003). Attentional characteristics of infants and toddlers with Williams syndrome during triadic interactions. Developmental Neuropsychology, 23, Meyer-Lindenberg, A., Mervis, C. B., & Berman, K. F. (2006). Neural mechanisms in Williams syndrome: A unique window to genetic influences on cognition and behaviour. Nature Reviews Neuroscience, 7, Mondloch, C. J., Le Grand, R., & Maurer, D. (2002). Configural face processing develops more slowly than featural face processing. Perception, 31, Näsänen, R. (1999). Spatial frequency bandwidth used in the recognition of facial images. Vision Research, 39, Osborne, L. (2006) The molecular basis of a multisystem disorder. In Williams- Beuren Syndrome: Research and Clinical Perspectives (pp ). Baltimore: Johns Hopkins University Press. Raven, J., Raven, J. C., & Court, J. H. (2000, updated 2004). Manual for Raven s Progressive Matrices and Vocabulary Scales. Section 3: The Standard Progressive Matrices. San Antonio, TX: Harcourt Assessment. Rose, F. E., Lincoln, A. J., Lai, Z., Ene, M., Searcy, Y. M., & Bellugi, U. (2007). Orientation and affective expression effects on face recognition in Williams syndrome and autism. Journal of Autism and Developmental Disorders, 37, Ruiz-Soler, M., & Beltran, F. S. (2006). Face perception: An integrative review of the role of spatial frequencies. Psychological Research, 70, Schultz, R. T., Gauthier, I., Klin, A., Fulbright, R., Anderson, A., Volkmar, F., et al. (2000). Abnormal ventral temporal cortical activity during face discrimination among individuals with autism and Asperger syndrome. Archives of General Psychiatry, 57, Schwarzer, G. (2000). Development of face processing: The effect of face inversion. Child Development, 71, Schyns, P. G., & Oliva, A. (1999). Dr. Angry and Mr. Smile: When categorization flexibly modifies the perception of faces in rapid visual presentations. Cognition, 69, Swettenham, J., Baron-Cohen, S., Charman, T., Cox, A., Baird, G., Drew, A., et al. (1998). The frequency and distribution of spontaneous attention shifts between social and nonsocial stimuli in autistic, typically developing, and nonautistic developmentally delayed infants. Journal of Child Psychology and Psychiatry, 39, Thomas, M. S., Annaz, D., Ansari, D., Scerif, G., Jarrold, C., & Karmiloff-Smith, A. (2009). Using developmental trajectories to understand developmental disorders. Journal of Speech, Language and Hearing Research, 52,