**2. Perception**

106 Current Topics in Children's Learning and Cognition

over time (e.g., Nordin & Gillberg, 1998).

differences between typical development and ASD, no single difference appears to capture the disorder (for a review see Fein, 2011). And even though ASD appears to have a strong genetic component, genome variations appear diffuse (e.g., Devlin & Scherer, 2012). This raises the possibility that the disorder is not reducible to a static causal factor that can differentiate between typical development and ASD. Instead, the disorder might be the result of complex interdependence among multiple factors that change each other's effect as they interact over time. Take for example, the language impairments documented in ASD. Rather than being attributed to a stable factor (language-specific, neurological, genetic, or otherwise), these behaviors might have their origin in virtually undetectably minimal discrepancies in how the perceptual system combines information into higher-order patterns. The discrepancy from typical development might be minimal at first, but then get amplified by a variety of child-internal, environmental, and social factors (e.g., a difficulty detecting higher-order patterns, a low tolerance for over-stimulation, the hierarchical order inherent in a language, a disrupted communication synchrony between child and caregiver, etc). The coming-together of environmental factors further intensify the initially minimal difference in perceptual processes – which then in turn amplify environmental and social factors. In other words, what may start out as a barely noticeable difference in how information is integrated might enter a cycle of forces that amplify each other's effect over

time, an interdependence that heralds a major departure from typical development.

The view that ASD behavior is the result of interdependent factors that amplify each other's effects over time is a stark departure from the view that overt behavior is reducible to a stable factor that marks autism. And while there is no conclusive evidence to support the former view, there is nevertheless strong support in the developmental literature of ASD. First, it is difficult to predict the developmental trajectories of individual children (for a review, see Seltzer, Shattuck, Abbeduto & Greenberg, 2004). For example, while overall language abilities can improve over time (Sigman & McGovern, 2005), various atypicalities sometimes remain, including echolalia or fixations on various topics of interest (Lord, Rici, & Pickles, 2004). Some children may even exhibit an increase in general symptom severity

Second, developmental patterns tracked over time sometimes show a non-linear trajectory. For example, differences in social behavior (e.g. eye contact, visual tracking, visual disengagement, imitation, social interest, and sensory-motor behaviors), apparent at 12 months of age, are missing in younger children (Zwaigenbaum et al., 2005; Bryson , Brian, Roberts, Szatmari, Rombough, & McDermott 2007). And while 6-month-old infants with a high risk of autism show less frontal Gamma power than low-risk children, this difference is negligible when infants are 24 months old (Keehn, Luyster, Vogel-Farley, Tager-Flusberg & Nelson, 2012). Explaining such non-linear trajectories under the reductionist viewpoint would require an additional assumption, namely that the isolated causal factor comes online at a certain point in time. These trajectories imply instead that the disorder has to be

attributed to complex interactions among factors that change in nature over time.

The interdependence view on ASD has strong implications for how to go about studying the source of this disorder and its treatment. Rather than looking for black-and-white Perceiving meaningful configurations in the array of ever-changing information (visual or otherwise) requires the mind to combine separate bursts of sensation into an organized unit of perception. The mind has to detect or impose coherence (cf., Thagard, 1989) For example, in order to perceive a painting, the mind has to ignore the individual pixels of color and detect the higher-order organization of objects and scenes. The possible patterns of organization are nested hierarchically, ranging from a very local organization (e.g., an individual object in the painting) to a more global organization (e.g., the theme of the painting). The ability to detect patterns of organization at various levels of abstraction is commonly studied under the framework of local versus global perception, with the central question pertaining to the degree to which local and global perception interfere with one another (cf. e.g., Baylis & Driver, 1993; Herrmann & Bosch, 2001; Humphreys, Olson, Romani, & Riddoch, 1996; Kahneman & Henik, 1981; Kahneman & Treisman, 1984; Kimchi, 1992; Kramer & Jacobson, 1991; Maurer, Le Grand, & Mondloch, 2002; Moore & Egeth, 1997; Pellicano & Rhodes, 2003).

An essential difference between ASD and typical development is the degree to which the perception of global order interferes with the perception of local order. Rather than exhibiting a bias towards coherence, perception of individuals with ASD is typically characterized by what is known as weak central coherence. Best example of this difference was established with the classical Navon task, a task in which stimuli consist of many small letters configured in the arrangement of a large letter (cf., Navon, 1977). In typical development, results show a distinct interference of large letters on the perception of small letters, both in children (Ozonoff Strayer, McMahon & Filloux 1994; Plaisted, Swettenham,

and Rees, 1999) and in adults (Fagot & Deruelle, 1997; Navon, 1977). In particular, when participants are asked to focus on small letters, reaction time is longer for trials in which large and small letters differ than on trials in which large and small letters match. This global interference is non-detectable in participants with ASD: They perform equally fast in both letter-mismatch trials and letter-match trials – and that with high accuracy (e.g. Mottron, Burack, Stauder, & Robaey, 1999; Plaisted et al., 1999).

Beyond the Black-and-White of Autism: How Cognitive Performance Varies with Context 109

target, while the other one differed in a single feature. And on feature trials, the test items were individual features (e.g., eyes), one of which match the respective feature of the target. Critically, participants were sometimes provided with a cue indicating on which feature to focus. For example, they were told "look at the eyes." In cued configuration trials, the cue focused attention to the mismatched feature. On feature trials, it focused attention to the matching feature. The typically developing group demonstrated superior performance on configuration trials, compared to feature trials, regardless of cueing. The ASD group, in contrast, demonstrated a configural advantage only on cued trials, but not in un-cued trials. A similar effect of instruction was found when the task was to read sentences that contained homographs (words that have the same spelling are a pronounced differently, depending on the context of a sentence). As one would predict from a weak-central-coherence assumption, individuals with ASD are less likely than typically developing individuals to use the context of a sentence to disambiguate homographs (Frith & Snowling, 1983; Happé, 1997; Jolliffe & Baron-Cohen, 1999; Lopez & Leekam, 2003). For example, when asked to read aloud a sentence containing the homograph "tear", participants with ASD are less likely to take the sentence context into account when deciding on how to pronounce the word. However, this pattern of performance changes dramatically when attention is explicitly directed to the homographs (Snowling & Frith, 1986). That is, when explicitly told to look for homographs, their use of proper pronunciation approaches that of typically developing individuals.

Variations in instruction also affect ASD performance on tasks involving optical illusions, another area that attests to Gestalt interference in typical development. Typically, the task is to focus on a local piece of information and ignore the embedding context. For example, in the Muller-Lyer illusion, the task is to compare the length of the two lines, ignoring the arrows on each end of the lines. Findings show that typically developing participants are strongly affected by the embedded context, succumbing to the illusion to the expense of focusing on the local elements (for a review, see Changizi, Hsieh, Nijhawan, Kanai, & Shimojo, 2008). While individuals with ASD are far less affected by such visual configurations (e.g., Brosnan Scott, Fox, & Pye, 2004), important context effects are apparent. For example, participants with ASD were found to be more susceptible to Muller-Lyer illusions when asked, "which line *looks* longer," versus "which line *is* longer" (Scott, Brosnan, & Wheelwright, in preparation, as cited in Happé & Frith, 2006). It appears that individuals with ASD can see both that the lines are equally long, and that the lines look like

In sum, while individuals with ASD differ from typical development in the degree to which they focus on the higher-Gestalt of a pattern (vs. the local elements), this difference is susceptible to variations in task context. Findings reviewed here pertain to the domains of shifting attention from local elements to global patterns (and vice versa), face perception (which depends on detecting relations between facial parts), reading homographs (which require the entire sentence to be taken into account), and optical illusions (which depends on children failing to ignore the embedding aspects of the target). Such context effects on perception have led to some revisions in the ASD theory of weak central coherence (e.g. Happé & Booth, 2008). Rather than positing an all-or-none competence of Gestalt

they differ in length.

Another example of weak central coherences in ASD comes from face-perception tasks. The identity of a face is defined not only by its individual parts (e.g., nose, eyes, mouth), but also by the holistic configuration of these parts, something that appears to be disrupted when faces are presented upside down. For typically developing children, recognition accuracy decreases when faces are presented upside down, compared to trials in which faces are presented upright (Mondloch, Le Grand, & Maurer, 2002). In contrast, children with ASD do not perform differently as a function of face orientation (Langdell, 1978; Tantam, Monaghan, Nicholson, & Stirling, 1989). Along the same lines, participants with ASD could classify faces better when local rather than global features were exaggerated (through the use of a high-pass vs. low-pass filter; Deruelle, Rondan, Salle-Collemiche, Bastard-Rosset, & Da Fonséca, 2008). The inverse pattern of results was obtained for typically developing children.

A final example of preferential local focus comes from research involving auditory perception (Foxton Stewart, Barnard, Rodgers, Young, O'Brien, & Griffiths 2003). Stimuli were 5-tone sequences that varied in specific tones, pattern of switch in pitch direction (e.g., a down sequence was followed by up sequence), and timing of the switch. In the crucial task (a global-interference condition), participants had to focus on only one of these features, ignoring changes in the other features. In particular, they had to decide on whether two sequences match in the patterns of switch, ignoring differences in specific tones or differences in timing. Result show superior performance for participants with ASD than matched controls. Vice versa, when sequences differed only in specific tones (the patterns and timing of the sequences being identical), ASD performance matched that of control participants (see also Mottron, Peretz & Menard, 2000). Further evidence for enhanced local processing of auditory information comes from the finding that individuals with ASD can label isolated tones better than TD controls and are more likely to have perfect pitch, meaning that they can replicate or identify individual musical tones without assistance (Bonnel, Mottron, Peretz, Trudel, Gallun, & Bonnel 2003; Heaton, Hermelin, & Pring, 1998).

In broad strokes, while typical development is characterized by a bias towards perceiving higher-order Gestalts over perceiving an isolated detail, this bias is thought to be missing or at least less prevalent in ASD (for reviews, see Happé, 2000; Happé & Booth, 2008; Happé & Frith, 2006). However, even though research generally supports the idea of a weak bias towards higher-order Gestalt in ASD, there are some interesting exceptions. For example, when individuals with ASD are told to look at relevant information in face processing tasks, they perform in a similar way to typically developing children (Lopez, Donnelly, Hadwin, & Leekam, 2004). The task was to determine which of two test items matched with a target face. In configuration trials, the test items were faces, one of which always matched the target, while the other one differed in a single feature. And on feature trials, the test items were individual features (e.g., eyes), one of which match the respective feature of the target. Critically, participants were sometimes provided with a cue indicating on which feature to focus. For example, they were told "look at the eyes." In cued configuration trials, the cue focused attention to the mismatched feature. On feature trials, it focused attention to the matching feature. The typically developing group demonstrated superior performance on configuration trials, compared to feature trials, regardless of cueing. The ASD group, in contrast, demonstrated a configural advantage only on cued trials, but not in un-cued trials.

108 Current Topics in Children's Learning and Cognition

children.

Mottron, Burack, Stauder, & Robaey, 1999; Plaisted et al., 1999).

and Rees, 1999) and in adults (Fagot & Deruelle, 1997; Navon, 1977). In particular, when participants are asked to focus on small letters, reaction time is longer for trials in which large and small letters differ than on trials in which large and small letters match. This global interference is non-detectable in participants with ASD: They perform equally fast in both letter-mismatch trials and letter-match trials – and that with high accuracy (e.g.

Another example of weak central coherences in ASD comes from face-perception tasks. The identity of a face is defined not only by its individual parts (e.g., nose, eyes, mouth), but also by the holistic configuration of these parts, something that appears to be disrupted when faces are presented upside down. For typically developing children, recognition accuracy decreases when faces are presented upside down, compared to trials in which faces are presented upright (Mondloch, Le Grand, & Maurer, 2002). In contrast, children with ASD do not perform differently as a function of face orientation (Langdell, 1978; Tantam, Monaghan, Nicholson, & Stirling, 1989). Along the same lines, participants with ASD could classify faces better when local rather than global features were exaggerated (through the use of a high-pass vs. low-pass filter; Deruelle, Rondan, Salle-Collemiche, Bastard-Rosset, & Da Fonséca, 2008). The inverse pattern of results was obtained for typically developing

A final example of preferential local focus comes from research involving auditory perception (Foxton Stewart, Barnard, Rodgers, Young, O'Brien, & Griffiths 2003). Stimuli were 5-tone sequences that varied in specific tones, pattern of switch in pitch direction (e.g., a down sequence was followed by up sequence), and timing of the switch. In the crucial task (a global-interference condition), participants had to focus on only one of these features, ignoring changes in the other features. In particular, they had to decide on whether two sequences match in the patterns of switch, ignoring differences in specific tones or differences in timing. Result show superior performance for participants with ASD than matched controls. Vice versa, when sequences differed only in specific tones (the patterns and timing of the sequences being identical), ASD performance matched that of control participants (see also Mottron, Peretz & Menard, 2000). Further evidence for enhanced local processing of auditory information comes from the finding that individuals with ASD can label isolated tones better than TD controls and are more likely to have perfect pitch, meaning that they can replicate or identify individual musical tones without assistance (Bonnel, Mottron, Peretz, Trudel, Gallun, & Bonnel 2003; Heaton, Hermelin, & Pring, 1998). In broad strokes, while typical development is characterized by a bias towards perceiving higher-order Gestalts over perceiving an isolated detail, this bias is thought to be missing or at least less prevalent in ASD (for reviews, see Happé, 2000; Happé & Booth, 2008; Happé & Frith, 2006). However, even though research generally supports the idea of a weak bias towards higher-order Gestalt in ASD, there are some interesting exceptions. For example, when individuals with ASD are told to look at relevant information in face processing tasks, they perform in a similar way to typically developing children (Lopez, Donnelly, Hadwin, & Leekam, 2004). The task was to determine which of two test items matched with a target face. In configuration trials, the test items were faces, one of which always matched the A similar effect of instruction was found when the task was to read sentences that contained homographs (words that have the same spelling are a pronounced differently, depending on the context of a sentence). As one would predict from a weak-central-coherence assumption, individuals with ASD are less likely than typically developing individuals to use the context of a sentence to disambiguate homographs (Frith & Snowling, 1983; Happé, 1997; Jolliffe & Baron-Cohen, 1999; Lopez & Leekam, 2003). For example, when asked to read aloud a sentence containing the homograph "tear", participants with ASD are less likely to take the sentence context into account when deciding on how to pronounce the word. However, this pattern of performance changes dramatically when attention is explicitly directed to the homographs (Snowling & Frith, 1986). That is, when explicitly told to look for homographs, their use of proper pronunciation approaches that of typically developing individuals.

Variations in instruction also affect ASD performance on tasks involving optical illusions, another area that attests to Gestalt interference in typical development. Typically, the task is to focus on a local piece of information and ignore the embedding context. For example, in the Muller-Lyer illusion, the task is to compare the length of the two lines, ignoring the arrows on each end of the lines. Findings show that typically developing participants are strongly affected by the embedded context, succumbing to the illusion to the expense of focusing on the local elements (for a review, see Changizi, Hsieh, Nijhawan, Kanai, & Shimojo, 2008). While individuals with ASD are far less affected by such visual configurations (e.g., Brosnan Scott, Fox, & Pye, 2004), important context effects are apparent. For example, participants with ASD were found to be more susceptible to Muller-Lyer illusions when asked, "which line *looks* longer," versus "which line *is* longer" (Scott, Brosnan, & Wheelwright, in preparation, as cited in Happé & Frith, 2006). It appears that individuals with ASD can see both that the lines are equally long, and that the lines look like they differ in length.

In sum, while individuals with ASD differ from typical development in the degree to which they focus on the higher-Gestalt of a pattern (vs. the local elements), this difference is susceptible to variations in task context. Findings reviewed here pertain to the domains of shifting attention from local elements to global patterns (and vice versa), face perception (which depends on detecting relations between facial parts), reading homographs (which require the entire sentence to be taken into account), and optical illusions (which depends on children failing to ignore the embedding aspects of the target). Such context effects on perception have led to some revisions in the ASD theory of weak central coherence (e.g. Happé & Booth, 2008). Rather than positing an all-or-none competence of Gestalt processing, weak central coherence is now seen as a tendency, a preference of some sort that could be changed under ideal task contexts.

Beyond the Black-and-White of Autism: How Cognitive Performance Varies with Context 111

ASD. Overall, these findings have undermined a claim that ASD is characterized by a general difficulty with implicit learning, in turn undermining an effort to explain social

There are many ways in which context effects on implicit learning could be explained. For example, one could address the differences in findings by looking for differences in the groups of participants, whether in age, symtomatology, or co-morbidity. It is possible that the findings fail to univocally address the question of implicit-learning competence in ASD because participants differ across different tasks. Or one could look for differences in other internal processes that could explain the pattern of performance. Tasks might differ in the degree to which they tap a participant's working memory. Or they differ in the extent to which they require the integration of gross-motor movements. Or they differ in whether they afford or undermine the use of explicit (i.e., hypothesis-testing) strategies. Indeed, ASD performance is comparable to that of typically developing children when the prototype learning task required a rule-based approach (Klinger & Dawson, 2001). And an exceedingly short inter-stimulus interval might have forced the minds of participants with ASD to abandon their bias to use a hypothesis-testing strategy and therefore make room for an implicit-learning process. There are multiple problems with this kind of reasoning, a major one being that it fails to address the

Executive functioning (EF) is an umbrella term to describe various cognitive abilities assumed to be involved in conscious problem-solving. They pertain, for example, to inhibiting incorrect but dominant actions, planning a future action, and flexibly switching attention when instructed to do so (e.g., Zelazo & Mϋeller, 2002). EF plays an important role in cognitive development, as it leads to an improved ability to override automatic responses (Garon, Bryson, & Smith, 2008). A classical EF task – but by far not the only one – is the Stroop task, a task in which participants are asked to name the color of the ink used for a printed word, the word spelling a particular color (Stroop, 1935). The central finding is a slowing in reaction time when the ink color differs from the spelled-out color (compared to trials in which the ink color matches the spelled-out color), demonstrating the difficulty of

EF is thought to be associated with typical ASD attributes, including the need for sameness, difficulty with switching attention, a tendency to perseverate, and a lack of impulse control. Indeed, there are tasks in which participants with ASD show difficulty with inhibition (for a review see Rajendran & Mitchell, 2007). Consider, for example, an inhibition task in which participants have to point to an empty window in order to receive the reward shown in a non-empty window (Hala, & Russel 2001). Unlike control participants, a majority of participants with ASD have difficulty inhibiting their natural response of pointing to the reward they desire, compared to controls matched on mental age. Other examples of EF difficulties consist of difficulties with planning (e.g., (Ozonoff & Jensen, 1999, Ozonoff, Pennington, & Rogers, 1991), mental flexibility (e.g., Hughes, Russell, & Robbins, 1994;

deficits, motor abnormalities, and language deficits associated with the disorder.

entire list of context effects – beyond a comparison of a few studies.

inhibiting the automatic tendency to read the word.

**4. Executive functioning** 
