Reduced face identity aftereffects in relatives of children with autism
Highlights
► Adaptive mechanisms are fundamental for coding a wide variety of attributes, including faces. ► We demonstrate atypicalities in adaptive face-coding in relatives of children with autism. ► Face identity aftereffects were significantly reduced in relatives of children with autism. ► Atypicalities in adaptive face-coding could be a candidate endophenotype for autism.
Introduction
Autism spectrum disorders (ASD) are pervasive developmental conditions characterised by often striking difficulties in social communication and social interaction in addition to repetitive and unusually focused behaviours and interests (American Psychiatric Association, 2000). Early twin and family history studies firmly established that genetic factors play a crucial role in the aetiology of autism (Bailey et al., 1995, Bolton et al., 1994, Folstein and Rutter, 1977). Yet, despite numerous studies using linkage or candidate gene approaches, the discovery of a single genetic locus of major effect has not been forthcoming. Instead, there is now clear consensus among geneticists that autism is both oligogenic – resulting from the action of multiple interacting genes – and multifactorial – resulting from interactions between genes and environmental factors, which have yet to be fully identified (see Geschwind, 2011, for review).
A significant minority of parents and siblings of individuals with ASD show behavioural traits that are qualitatively similar to the defining features of ASD, albeit in more subtle form (see Bailey, Palferman, Heavey, & Le Couteur, 1998, for review). Many studies have attempted to identify the various components of this so-called “broad autism phenotype”, which can include rigid or aloof personality traits, difficulties initiating and maintaining friendships, limited communicative use of language, and overly focused and unusual interests and activities (Bishop et al., 2006, Losh et al., 2008, Losh and Piven, 2007, Piven et al., 1997).
Focusing on the behavioural level alone, however, is not an ideal basis for identifying genetic mechanisms: the same genotype can give rise to different behavioural phenotypes, and the same phenotype can arise from a range of genotypes (Gottesman & Gould, 2003). Researchers have therefore turned their attention towards discovering neurobiological or cognitive markers that are initially unobservable but which are more proximal to the underlying aetiology of autism than overt behaviours themselves. Such “endophenotypes” are thought to index genetic liability to autism in otherwise apparently “unaffected” individuals (Flint & Munafò, 2007).
Atypical face-processing mechanisms have been proposed to be one such candidate endophenotype for autism (Dawson et al. 2002). Difficulties in perceiving and discriminating faces have been well documented in individuals with ASD (e.g., see Dawson, Webb, & McPartland, 2005, for review). Children and adults with ASD show poorer performance on a variety of face processing tasks compared with non-autistic individuals, including face recognition (Boucher et al., 1998, Ewing et al., 2011a), face discrimination (Ewing et al., 2011a, Wallace et al., 2008a), expression recognition (Rump et al., 2009, Wallace et al., 2008b), and eye-gaze perception (Wallace, Coleman, Pascalis, & Bailey, 2006). Even when their performance is similar to that of non-autistic individuals, individuals with ASD appear to use atypical strategies, such as paying more attention to the mouth than the eyes (Klin et al., 2002, Neumann et al., 2006; although see Falck-Ytter & von Hofsten, 2011, for a critique), and applying a local rather than holistic processing style (Joseph & Tanaka, 2002).
A similar range of face processing atypicalities has also been reported in parents and siblings of individuals with ASD. Such atypicalities include less time spent looking at the eyes during a face processing task (Dalton, Nacewicz, Alexander & Davidson, 2005), difficulties discriminating subtle differences between faces, identifying facial expressions of fear and disgust, and judging direct eye contact (Wallace, Sebastian, Pellicano, Parr, & Bailey, 2010) and problems on a standardized test of facial identity recognition (Wilson, Freeman, Brock, Burton, & Palermo, 2010), compared to parents and siblings of typically developing individuals. These studies clearly show that atypicalities in various behavioural aspects of face processing are shared by individuals with autism and their relatives.
Studies that go one step further to pinpoint the underlying mechanisms responsible for such atypicalities in relatives of individuals with autism should therefore bring us closer for isolating a potential endophenotype for autism at the neurocognitive level. One study has demonstrated distinct face-processing strategies during emotion recognition in relatives of individuals with ASD (Adolphs, Spezio, Parlier, & Piven, 2008). These authors showed that parents of typical children showed substantial use of the eyes when judging emotions like fear or happiness. Yet parents of autistic children, especially those with an aloof personality, made much less use of the eyes when making these judgments, using more cues from the mouth (Adolphs et al., 2008), a strategy that closely mirrors the behaviour of individuals diagnosed with ASD (Spezio, Adolphs, Hurley, & Piven, 2007).
The present study extends the search for candidate endophenotypes for autism by focusing on mechanisms underlying another important aspect of face processing the recognition of facial identity.
Typical children and adults code faces relative to an implicitly-stored internal average or norm, which is continuously updated by experience to represent the central tendency of the population of faces experienced (for review see Rhodes & Leopold, 2011). The strongest evidence for this adaptive norm-based coding mechanism comes from aftereffect paradigms. Aftereffects occur throughout perceptual systems and are illustrative of how perceptual attributes, such as colour and motion, are coded by these systems. For example, the motion aftereffect occurs when adaptation (prolonged exposure) to a stimulus moving in a particular direction causes a subsequently viewed stationary stimulus to be perceived as moving in the opposite direction. Similarly, in face identity aftereffects, adapting to a face biases us to see a subsequent face as having opposite properties (e.g., Leopold, O'Toole, Vetter, & Blanz, 2001).
In a typical face identity aftereffect task, participants learn some target identities (e.g., Ted and Rob; see Fig. 1), and are then tested on their recognition of faces with weaker identity strengths of Ted (e.g., 30%, 60%, etc., including 0% average face) both before and after adaptation to anti-Ted. In typical adults (e.g., Leopold et al., 2001, Rhodes and Jeffery, 2006) and children (e.g., Nishimura, Maurer, Jeffery, Pellicano, & Rhodes, 2008), identification of the target identity (e.g., Ted) is facilitated following adaptation to its antiface (antiTed), that is, a face with opposite properties. For example, if Ted has smaller lips than average, antiTed will have larger lips than average, and so on for many other facial attributes. In terms of face space, anti-faces lie along the same vector as the target identity, but are situated on the other side of the average face. After adaptation to antiTed, the previously neutral 0% (average) face will be perceived as Ted. Adaptation to antiTed “shifts” the observer’s internal average toward anti-Ted, causing the actual average (identity neutral) to look more like Ted. Furthermore, the identity aftereffect is selective for opposite face pairs (e.g., Ted/anti-Ted but not Ted/anti-Rob), suggesting that facial identity is coded opponently, with pairs of neural populations coding for above- and below-average values along particular dimensions in face space (Rhodes and Jeffery, 2006, Rhodes et al., 2005, Robbins et al., 2007).
Pellicano, Jeffery, Burr and Rhodes (2007) investigated adaptation to facial identity in children with and without ASD aged between 8 and 13 years. They found that the extent to which the children shifted their perception following adaptation was significantly attenuated in those with autism. Furthermore, the degree of adaptation correlated significantly and negatively with children’s current levels of autistic symptoms, such that the children with the smallest aftereffects exhibited greater levels of symptoms. These findings suggest that the mechanisms responsible for coding facial identity might be less flexible or adaptable in children with autism, so that, relative to typical children, they might be less capable of generating and updating face norms with experience (Pellicano et al., 2007).
The primary aim of the current study was to investigate whether diminished face adaptation might be a candidate endophenotype for autism. To address this aim, we administered a face identity aftereffect task to parents and siblings of children with and without autism. If atypicalities in adaptive face-coding mechanisms are a potential neurocognitive endophenotype for ASD, then we should expect attenuated face identity aftereffects in relatives of individuals with autism compared with relatives of typically developing children.
A secondary aim was to determine whether the face coding mechanisms shown by the parents and siblings of children with autism differ qualitatively from those seen in typical children and adults. In typical individuals, more physically extreme adaptors (e.g., faces that lie further away from the average) produce a larger change or bias in perception of the average face than less extreme adaptors (Jeffery et al., 2010, Jeffery et al., 2011, Robbins et al., 2007). This result is consistent with an opponent (norm-based) coding model in which face dimensions (e.g., the size of the lips) are coded by the relative output of only two oppositely tuned pools of neurons: one pool responds maximally to one extreme of the value range (e.g., very small lips) and minimally to the opposite extreme (e.g., very large lips), while the other shows the reverse tuning. A larger aftereffect for more extreme adaptors is not consistent with a multi-channel, exemplar-based coding system, in which values along a dimension are coded by multiple pools of neurons each with bell-shaped tuning centred over a different value. Indeed, that model predicts a larger aftereffect for less extreme adaptors, which will have a greater effect on neurons tuned to the average face. Only the opponent coding model predicts that a more physically extreme adaptor (e.g., a face with lips much smaller than the average) will produce a larger change in the perception of the average face than a less extreme adaptor because adaptation reduces neural response in proportion to the adapted firing rate (e.g., Maddess et al., 1988, Movshon and Lennie, 1979).
By including adaptors both near and far from the average face in our task, we sought to determine whether potential differences in face coding between relatives of individuals with ASD and relatives of typical children are purely quantitative or also qualitative. If both groups show stronger aftereffects for far relative to near adaptors, this would suggest that the face space of relatives of children with ASD is qualitatively similar to that of relatives of typical children, in that it relies on norm-based rather than exemplar-based coding mechanisms.
Section snippets
Participants
Relatives of both children with autism (hereafter, “ASD group”) and typically developing children (hereafter, “TD group”) were recruited for this study. The ASD group (n=28) included 20 parents (13 mothers, 7 fathers) and 8 siblings1 (6 females) of individuals with ASD (14 families in total; see Table 1), who were recruited via an existing database of families with children with autism, whose children had received independent diagnoses according to
Results
An initial ANOVA on the proportion of correct responses for the 80% target faces with group (relatives of children with ASD, relatives of TD children) and relative status (parent, sibling) showed no main effects of group, relative status or any interactions involving these factors (all Fs<2.95, all ps>0.16), indicating that relatives of children with ASD (M=0.93, SD=0.08) could recognise strong versions of the target faces just as well as relatives of TD children (M=0.96, SD=0.04).2
Discussion
This study investigated whether parents and siblings of children with autism show diminished aftereffects to facial identity as children with autism do (Pellicano et al., 2007). Using a face identity adaptation task, we demonstrated that, for relatives of typically developing children, adaptation to antiRob caused the average face to appear more Rob-like, and the magnitude of this bias in perception (the aftereffect) was stronger when the adapting face was further from the average face than
Acknowledgements
We are extremely grateful to all of the families that generously took part in this research. C.F. was supported by a Swiss National Science Foundation (FNS) Fellowship for Young Researchers during the period of this research. Research at the Centre for Research in Autism and Education (CRAE) is supported by The Clothworkers’ Foundation and Pears Foundation (L.G. & E.P.). G.R., L.J. and E.P. were supported by the Australian Research Council Centre of Excellence in Cognition and its Disorders
References (51)
- et al.
Distinct face-processing strategies in parents of autistic children
Current Biology
(2008) - et al.
How special is social looking in ASD: A review
Progress in Brain Research
(2011) Genetics of autism spectrum disorders
Trends in Cognitive Sciences
(2011)- et al.
Abnormal adaptive face-coding mechanisms in children with autism spectrum disorder
Current Biology
(2007) Autism: Face-processing clues to inheritance
Current Biology
(2008)- et al.
The role of higher level adaptive coding mechanisms in the development of face recognition
Journal of Experimental Child Psychology
(2009) - et al.
Adaptive norm-based coding of facial identity
Vision Research
(2006) - et al.
The size-tuning of the face-distortion after-effect
Vision Research
(2001) - American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., Text...
- et al.
Autism as a strongly genetic disorder: Evidence from a British twin study
Psychological Medicine
(1995)