Vision Research 126 (2016) 278–290

Contents lists available at ScienceDirect

Vision Research

journal homepage: www.elsevier .com/locate /v isres

A cognitive architecture account of the visual local advantagephenomenon in autism spectrum disorders

http://dx.doi.org/10.1016/j.visres.2015.04.0090042-6989/� 2015 Elsevier Ltd. All rights reserved.

E-mail address: peter.vanderhelm@ppw.kuleuven.be.URL: http://perswww.kuleuven.be/peter_van_der_helm.

Peter A. van der HelmLaboratory of Experimental Psychology, University of Leuven (K.U. Leuven), Tiensestraat 102 – Box 3711, Leuven B-3000, Belgium

a r t i c l e i n f o

Article history:Received 3 December 2014Received in revised form 17 February 2015Available online 24 April 2015

Keywords:Autism spectrum disordersCognitive architectureGlobal vs. local processingLocal advantage phenomenonNeuronal synchronizationPerceptual organization

a b s t r a c t

Ideally, a cognitive architecture is a neurally plausible model that unifies mental representations and cog-nitive processes. Here, I apply such a model to re-evaluate the local advantage phenomenon in autismspectrum disorders (ASD), that is, the better than typical performance on visual tasks in which localstimulus features are to be discerned. The model takes (a) perceptual organization as a predominantlystimulus-driven process yielding hierarchical stimulus organizations, and (b) attention as predominantlyscrutinizing the hierarchical structure of established percepts in a task-driven top-down fashion. Thisaccounts for a dominance of wholes over parts and implies that perceived global structures maskincompatible local features. The model also substantiates that impairments in neuronal synchronization– as found in ASD – reduce the emergence of global structures and, thereby, their masking effect onincompatible features. I argue that this explains the local advantage phenomenon and I discuss implica-tions and suggestions for future research.

� 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Autism spectrum disorders (ASD) are complex neurodevelop-mental disorders, the severity of which is based on social commu-nication impairments and restricted repetitive patterns of behavior(American Psychiatric Association, 2013). In addition to these diag-nostic features, ASD individuals also show atypical cognitive pro-cessing (for reviews, see Pellicano, 2011; Rajendran & Mitchell,2007), particularly in the visual domain (for reviews, see Dakin &Frith, 2005; Simmons et al., 2009). An intriguing example of atyp-ical visual processing in ASD is the local advantage phenomenon,that is, the better than typical performance on visual tasks in whichlocal stimulus features are to be discerned – such tasks being, forinstance, embedded figures tasks, block design tasks, and visualsearch tasks (Jolliffe & Baron-Cohen, 1997; Joseph et al., 2009;O’Riordan et al., 2001; Shah & Frith, 1983, 1993). Explanations ofthis phenomenon usually rely on either reduced global processing(promoted most prominently by Frith, 1989) or enhanced localprocessing (promoted most prominently by Mottron & Burack,2001). Thus far, however, empirical data on this phenomenonseemed inconclusive as to which explanation prevails.

In this theoretical study – which, of course, relies on empiricaldata too – I first integrate various ideas and findings on typical

perception and attention into a neurally plausible cognitive model,called PATVISH (acronym of Perception and ATtention in the VISualHierarchy). Then, I investigate what reduced global processing andenhanced local processing would yield according to this model,and I re-evaluate evidence on these alleged deviations in ASD. Iconclude that the local advantage phenomenon in ASD is primarilya side effect of reduced global processing caused by impaired neu-ronal synchronization, and I end with critical predictions. Forinstance, one of these predictions is that the local advantage phe-nomenon occurs only for local features that are incompatible withperceived global structures, that is, features that are not propersubstructures of perceived global structures. To be clear, incompat-ible features are not to be confused with incongruent features inthe sense of Navon (1977). This is illustrated in Fig. 1, in which per-ceived global structures are represented schematically above theirconstituent parts. It demonstrates that both congruent and incon-gruent features are compatible with perceived global structures. Inthe next section (with some material reproduced from van derHelm, 2012), I begin by discussing perception and attention in typ-ical individuals.

2. Typical perception and attention

Attention can mean many things. For instance, in cognitivescience, distinctions have been made between selective anddivided attention (i.e., concentrated on a specific thing vs. divided

Fig. 1. Incompatibility vs. incongruency. (a) A stimulus whose perceived organization is a triplet of triangles, and parts that are respectively compatible and incompatiblewith this perceived organization. (b) Stimuli composed of compatible local elements that are respectively congruent and incongruent with the global structure (after Navon,1977). The insets schematically represent the perceived hierarchical organizations, with global structures above their constituents parts.

P.A. van der Helm / Vision Research 126 (2016) 278–290 279

over several things); between overt and covert attention (i.e.,actively directed gaze vs. purely mental focus); and betweenexogenous bottom-up and endogenous top-down attention (i.e.,drawn by stimuli like a bright flash vs. directed to stimuli in func-tion of a task). These definitions overlap, but the form of attentionconsidered in this study is specified best by task-driven top-downattention. Be that as it may, notice that attention – of whateverform and involving whatever action – is basically the allocationof processing resources (Anderson, 2004). In other words, it maydecide what you focus on but not what you perceive before or afterfocusing. This is where perception comes in.

By perception, I mean visual perceptual organization. This is theneuro-cognitive process that enables us to perceive scenes asstructured wholes consisting of objects arranged in space. This pre-sumably automatic process may seem to occur effortlessly in dailylife, but by all accounts, it must be both complex and flexible. Gray(1999) gave the following gist of it. For a proximal stimulus, theperceptual organization process usually singles out one hypothesisabout the distal stimulus from among a myriad of hypotheses thatalso would fit the proximal stimulus (this is also called the inverseoptics problem). This means that multiple sets of features at mul-tiple, sometimes overlapping, locations in a stimulus must begrouped in parallel and that the process must cope with a largenumber of possible combinations simultaneously. This indicatesthat the combinatorial capacity of the perceptual organization pro-cess must be high, which, together with its high speed (it com-pletes in the range of 100–300 ms), reveals its truly impressivenature.

The exact nature of the interplay between perception and atten-tion is still unclear. Some hold that the perceptual organizationprocess is purely stimulus-driven (e.g., Gray, 1999; Pylyshyn,1999), while others hold that it is fully controlled by attention(e.g., Lamme & Roelfsema, 2000). As I specify next, my stance isclose to the former but leaves room for attention to modulatethe outcome of the perceptual organization process.

2.1. The cognitive architecture PATVISH

Cognitive architecture, or unified theory of cognition, is a con-cept from artificial intelligence (AI) research. It refers to a blueprintfor a system that acts like an intelligent system – taking into

account not only its resulting behavior but also physical or moreabstract properties implemented in it (Anderson, 1983; Newell,1990; for reviews, see Byrne, 2012; Langley, Laird, & Rogers,2009; Sun, 2004). Hence, it aims to capture not only competence(i.e., what is a system’s output?) but also performance (i.e., howdoes a system arrive at its output?). It therefore calls for a unifica-tion of dynamic processes (which include temporal factors) andstatic representations (which include structural factors).

Just as many recognized cognitive architectures – which comein various levels of detail and often focus on human language –PATVISH is not a full-blown model of human cognition as a whole.Yet, focusing on human perception and attention and takingbehavioral and neurophysiological data into account, it does aimto unify cognitive processes and representations. Furthermore, itmay not qualify as a typical AI model – as it is a verbal model (ablueprint) rather than an implemented model that can take actualinput – but it is sustained by a full-blown computational model ofa related issue (see below). More specifically, it incorporates thenext picture of processes and representations in the visual hierar-chy in the brain (for more details, see van der Helm, 2012, 2014,2015).

According to Felleman and van Essen (1991), the neural net-work in the visual hierarchy is organized with 10–14 distinguish-able hierarchical levels (with multiple distinguishable areaswithin each level), contains many short-range and long-range con-nections (both within and between areas and levels), and can besaid to perform a distributed hierarchical process. In line withLamme, Supèr, and Spekreijse (1998), PATVISH takes this processto consist of three neurally intertwined but functionally distin-guishable subprocesses. As illustrated in the left-hand panel inFig. 2, these subprocesses are responsible for, respectively, (a) feed-forward extraction of, or tuning to, features to which the visualsystem is sensitive, (b) horizontal binding of similar features, and(c) recurrent selection of different features. As illustrated in theright-hand panel in Fig. 2, these subprocesses together yield inte-grated percepts given by hierarchical stimulus organizations.

In PATVISH, the resulting hierarchical organizations are taken tobe the simplest ones, that is, organizations which – by exploitingvisual regularities such as repetition and symmetry – can be spec-ified using a minimum number of descriptive parameters. Thissimplicity principle (Hochberg & McAlister, 1953) is a descendant

Fig. 2. The PATVISH model of cognitive processing in the visual hierarchy in the brain. An exogenous, stimulus-driven, perceptual organization process (at the left) comprisesthree intertwined subprocesses, which, together, yield percepts in the form of hierarchical stimulus organizations (i.e., organizations in terms of wholes and their parts). Anendogenous, task-driven, attention process (at the right) may scrutinize such a hierarchical organization – starting at higher levels where relatively global structures arerepresented, and if required by task and allowed by time, descending to lower levels where relatively local features are represented.

280 P.A. van der Helm / Vision Research 126 (2016) 278–290

of the Gestalt law of Prägnanz and has been elaborated in struc-tural information theory (SIT; see Leeuwenberg & van der Helm,2013). To make quantitative predictions, SIT added a formal codinglanguage and a complexity metric for symbol strings representingstimulus interpretations – for a stimulus, the interpretation withthe simplest code then is predicted to be the one perceived byhumans. To this end, SIT also developed a minimal-coding algo-rithm, called PISA, which computes simplest codes of strings (vander Helm, 2004, 2015) and which in fact implements formal coun-terparts of the three subprocesses above (see Fig. 3). The algorithmis available on request or via a hyperlink in van der Helm (2015).

Among those three subprocesses – in PATVISH, and likewise, inPISA – the subprocess of feedforward extraction is reminiscent ofthe idea that, going up in the visual hierarchy, neural cells mediatedetection of increasingly complex features (Hubel & Wiesel, 1968).Furthermore, the subprocess of recurrent selection is reminiscentof the connectionist idea that, by parallel distributed processing(PDP), activation spreading in the brain’s neural network yieldspercepts represented by stable patterns of activation(Churchland, 1986). Dijkstra’s (1959) shortest path method, whichhas been implemented in PISA, is in fact a serial-processing simu-lation of PDP (see demos in van der Helm, 2004, 2012, 2014).

Currently most relevant is the subprocess of horizontal binding.This subprocess may be relatively underexposed in neuroscience,but in PISA – and presumably in the brain – it subserves a powerfulform of processing, called transparallel processing. A full technicalexposé on this would be beyond the scope of this article and can be

Fig. 3. (a) The three neurally intertwined perceptual subprocesses in PATVISH. (b) The talgorithm PISA.

found in van der Helm (2004, 2014, 2015); Appendix A gives thegist of it, and it basically boils down to the following. In PISA, setsof up to an exponential number of similar regularities in a stringgroup by nature in special distributed representations, calledhyperstrings. These are special in that, subsequently, those super-posed similar features can be recoded hierarchically in a transpar-allel fashion, that is, simultaneously as if only one feature wereconcerned. Conceptually, transparallel processing by hyperstringsis close to the idea of coactive architectures yielding supercapacity(Townsend & Nozawa, 1995). Furthermore, unlike parallel process-ing, it is feasible on single-processor classical computers (as PISAproves) and has the same extraordinary computing power as thatpromised by quantum computers (van der Helm, 2014, 2015).

The foregoing shows that, much like human perceptual organi-zation, PISA combines high combinatorial capacity and speed.Notice further that the hyperstrings in PISA are transient, that is,they are temporary distributed representations, which bind similarfeatures in the current input only (this contrast with standard con-nectionism, which assumes that one fixed network suffices formany different inputs). The transient nature of hyperstrings tooagrees with processing in the visual hierarchy in the brain, wherehorizontal binding of similar features is believed to be mediatedby transient neural assemblies, which signal their presence by syn-chronization of the neurons involved (see, e.g., Eckhorn, 1999;Edelman, 1987; Gilbert, 1992). These correspondences betweenhyperstrings and synchronized neural assemblies led me to postu-late – and to incorporate in PATVISH – that, in the brain, the

hree algorithmically intertwined counterparts implemented in the minimal-coding

P.A. van der Helm / Vision Research 126 (2016) 278–290 281

synchronization in those neural assemblies is the neural signatureof transparallel recoding of similar features (van der Helm, 2012,2014, 2015).

This idea is admittedly speculative but may be positioned fur-ther as follows. Both theoretically and empirically, neuronal syn-chronization has been associated with cognitive processing, andsynchronization in the gamma-band (30–70 Hz), in particular,has been associated with feature binding in perceptual organiza-tion (Eckhorn et al., 1988; Gray & Singer, 1989; Milner, 1974;von der Malsburg, 1981). Related ideas about its meaning are, forinstance, that it is a marker that an assembly has arrived at asteady state (Pollen, 1999), or that its strength is an index of thesalience of features (Finkel, Yen, & Menschik, 1998; Salinas &Sejnowski, 2001), or that more strongly synchronized assembliesin a visual area in the brain are locked on more easily by higherareas (Fries, 2005). These ideas sound plausible, that is, synchro-nization indeed might reflect a flexible and efficient mechanismsubserving the representation of information, the regulation ofthe flow of information, and the storage and retrieval of informa-tion (Sejnowski & Paulsen, 2006; Tallon-Baudry, 2009). To be clear,I do not challenge these ideas, but notice that they are about cog-nitive factors associated with synchronization rather than aboutthe nature of the underlying cognitive processes. Regarding the lat-ter, PATVISH now adds the idea that synchronization signalstransparallel recoding of similar features.

This transparallel recoding of similar features yields a hierarchyof feature constellations – that is, in PISA, a hierarchy of hyper-strings, and in PATVISH, a hierarchy of synchronized neural assem-blies. From this hierarchy of feature constellations, differentfeatures are selected to be integrated into percepts. In other words,transparallel processing underlies the perceptual integration capa-bility – as distinct from the feedforward extraction of visual fea-tures. This distinction between extraction and integration agreeswith that between base grouping and incremental grouping asput forward by Roelfsema (2006) (see also Roelfsema &Houtkamp, 2011) and may be clarified further as follows.

The way in which PISA implements feedforward extraction andrecurrent selection suggests that these subprocesses in the brainare like a fountain under increasing water pressure (for a similarpicture, see VanRullen & Thorpe, 2002). As the feedforward extrac-tion progresses along ascending connections (taking only about100 ms to reach the top end of the visual hierarchy, and thereforealso called the feedforward sweep), each passed level in the visualhierarchy forms the starting point of integrative recurrent process-ing along descending connections. This yields a gradual buildupfrom partial percepts at lower levels in the visual hierarchy to com-plete percepts near its top end.

Finally, PATVISH steers a middle course regarding the contro-versial issue of the effect of attention on perception. On the onehand, I think that the perceptual organization process alreadyhas done much of its integrative work before attention comes intoplay, that is, it is too fast to be controlled by attention. On the otherhand, the gradual buildup of percepts takes time, of course, so, itleaves room for attention to intrude and to modulate things beforea percept has completed. Be that as it may, for both partial andcomplete percepts, the role of attention may be specified furtheras follows.

Following Hochstein and Ahissar’s (2002) reverse hierarchy the-ory (RHT), PATVISH holds that, via recurrent connections fromhigher cognitive levels beyond the visual hierarchy, task-drivenattention can be deployed in a top-down fashion to any level inthe visual hierarchy (see also Ahissar & Hochstein, 2004; Wolfe,2007). This implies that it first captures things represented inhigher visual areas and, if required by task and allowed by time,may descend to capture things represented in lower areas. Justas RHT, PATVISH assumes that the things in higher areas are

relatively global structures and that the things in lower areas arerelatively local features (see Fig. 2). This is consistent with Hubeland Wiesel’s (1968) idea that, going up in the visual hierarchy,neural cells mediate increasingly complex features, and is sup-ported by recent neurophysiological data (de-Wit et al., 2012;Kubilius, Wagemans, & Op de Beeck, 2011). The above picture ofa top-down deployment of attention thus agrees with a top-downtraversal through the logical structure of a stimulus (Collard &Povel, 1982), with global structures represented at higher levelsand local features at lower levels (see the insets in Fig. 1). In spatialfrequency terms, it is similar to gradually switching off larger-scalefilters, thereby going from global to local information (vanRijsbergen & Schyns, 2009; Watt, 1987).

In sum, PATVISH specifies the combined action of perceptionand attention as follows. First, perceptual organization is a pre-dominantly stimulus-driven process that relies on transparallelprocessing mediated by temporarily synchronized neural assem-blies. Second, attention may modulate the outcome of this organi-zation process but predominantly subserves top-down scrutiny ofthe hierarchical structure of established percepts.

2.2. General implications

PATVISH focuses on visual form perception without ignoringbasic feature classes such as color, orientation, and spatial fre-quency. In my view, visual form perception builds on such basicfeatures to establish correlations that might give rise to perceptualgrouping. For instance, two squares of the same size but of differ-ent colors are not identical but are yet more similar to each otherthan to a rectangle of another color. Furthermore, PATVISH’s com-bined action of perception and attention calls for exercising duecaution when interpreting behavioral data in terms of cognitiveprocesses. In perception experiments, participants usually respondon the basis of what they have perceived, that is, on the basis oftheir mental representations of stimuli. PATVISH then suggeststhat, to respond, participants invoke task-driven attention to goin a top-down fashion through the hierarchical stimulus organiza-tions yielded by the perceptual organization process. The nextexample illustrates what this implies.

In the visual search for a target among distractors, a ‘‘pop-out’’is a target whose detection is fast and independent of the numberof distractors (e.g., a red item among blue items). Treisman andGelade (1980) proposed that pop-outs have features processed ear-lier than others during the perception process. However, a target isnot a pop-out by its own merits, but by the merits of the distrac-tors: The search for a target in a typical visual search display iseasier as the distractors are more similar to each other and moredifferent from the target (cf. Donderi, 2006; Duncan &Humphreys, 1989; Wolfe, 2007). In other words, a target may bea pop-out but only if allowed by the distractors. Hence, for a targetto become a pop-out, properties of all elements have to be pro-cessed first. This may well involve lateral inhibition among similarthings so that the target rises above the distractors (which hasbeen formulated also in terms of salience maps), but in any case,it seems plausible that the similarity of the distractors is repre-sented first in lower visual areas and that the representation ofthe target ends up in higher visual areas. This suggests that apop-out is a pop-out not because it is (nonconsciously) processedfirst during the bottom-up perception process but because its rep-resentation ends up in higher visual areas so that it is among thefirst things (consciously) encountered by the top-down attentionprocess. In other words, by this account, visual search in case ofpop-outs is fast and independent of the number of distractorsnot because the search has a presumed parallel nature but simplybecause pop-outs are the first things encountered.

282 P.A. van der Helm / Vision Research 126 (2016) 278–290

As indicated earlier, perceptual organization is characterizedmore general by a complex (i.e., nonlinear) interaction betweenparts. The founding fathers of Gestalt psychology – Wertheimer(1912, 1923), Köhler (1920), and Koffka (1935) – proposed that thisinteraction is governed by the law of Prägnanz. This law expressesthe idea that the brain, like any dynamic physical system, tends tosettle in relatively stable states. For vision, Koffka (1935, p. 138)formulated this by: ‘‘Of several geometrically possible organizationsthat one will actually occur which possesses the best, the most stableshape’’. The simplicity principle in PATVISH is in fact a moderninformation-theoretic elaboration of this idea (van der Helm,2000). The Gestaltists expressed this idea further in statementssuch as ‘‘the process of organization depends upon the properties ofits result’’ and ‘‘the response to a stimulus [. . .] is an organized patternin which each part depends on the organization of the whole’’ (Koffka,1935, pp. 151 and 601). Hence, rather than assuming a unidirec-tional process from parts to wholes, they postulated that wholesdominate parts – both during the perception process and in theresulting percepts.

Supported by empirical data on various behavioral tasks (for areview, see Wagemans et al., 2012), the dominance of wholes overparts has been specified further by notions such as global prece-dence (Navon, 1977), configural superiority (Pomerantz, Sager, &Stoever, 1977), primacy of holistic properties (Kimchi, 2003), andsuperstructure dominance (Leeuwenberg & van der Helm, 1991;Leeuwenberg, van der Helm, & van Lier, 1994; van Lier,Leeuwenberg, & van der Helm, 1997). By PATVISH, these behav-ioral notions do not apply to the perception process itself but,rather, to the resulting percepts. As said, to respond, participantsinvoke task-driven attention to go top-down through hierarchicalorganizations yielded by the perception process. This means thatglobal structures are experienced first and are therefore most deci-sive in participants’ responses (see also Stoll et al., 2015). As far as Ican tell, this explains most of the behavioral data on the domi-nance of wholes over parts.

In a similar vein, PATVISH might explain cases of change blind-ness in which local changes in a stimulus do not affect its perceivedglobal structure (cf. Poljac, de-Wit, & Wagemans, 2012). It alsoexplains that backward masking affects local features rather thanglobal structures. Next, I clarify the latter to specify a maskingeffect that, in Section 3, is argued to be relevant to the local advan-tage phenomenon in ASD.

2.3. Visual masking

Visual masking is the reduction in visibility of one stimulus orof a feature therein (the target) caused by another stimulus (themask) – typically under stimulus onset asynchronies (SOAs) wellbelow 500 ms (for reviews, see Breitmeyer, 2007; Breitmeyer &Ogmen, 2000, 2006). Usually, masking effects are defined in tem-poral terms. For instance, backward masking may occur whenthe mask is presented after the target and forward masking mayoccur when the mask is presented before the target. As said, theperception process completes in the range of 100–300 ms, so,masking experiments are among those that probe the perceptionprocess itself. The general idea is that the two stimuli trigger twosuccessive feedforward sweeps which, under those temporal con-ditions, interfere with each other (see, e.g., Bachmann, 1984;Lamme & Roelfsema, 2000).

Notice, however, that those temporal conditions are not suffi-cient for masking to occur. For instance, the same temporal condi-tions also may yield priming, that is, the increase in visibility of atarget, or of a feature therein, when preceded by a prime (Beller,1971). Originally, priming was thought to rely on physical corre-spondences between prime and target, but later, it became clearthat the perceived structure of the prime – and thereby attention

– is relevant too (see, e.g., Gerbino & Salmaso, 1987; Sekuler &Palmer, 1992; van Lier, van der Helm, & Leeuwenberg, 1994).

By the same token, perceived structure and attention are rele-vant to masking (see, e.g., Breitmeyer & Ogmen, 2006; Enns & DiLollo, 1997; Herzog, 2007; van der Vloed, Csathó, & van derHelm, 2007). For instance, Hermens and Herzog (2007) found that,in backward masking, target processing is affected by the mask’sspatial layout. Furthermore, Leeuwenberg, Mens, and Calis (1985)found that perceived global structures and incompatible local fea-tures have a backward masking effect on each other. In view ofPATVISH, the situation in standard backward masking can in factbe understood as follows. A structured target and a subsequentunstructured mask trigger successive feedforward sweeps, andthe second sweep (by the mask) then may perturb the neural traceof the first sweep (by the target). Because the mask is unstructured,its trace will not penetrate high up in the visual hierarchy, so, itwill perturb predominantly local target features represented inlower visual areas and top-down attention may still capture therelatively intact global target structure represented in higher visualareas. This explains that backward masking affects local featuresrather than global structures.

Hence, in both priming and masking, temporal conditions setthe context within which structural factors determine whicheffects occur. Of course, the influence of structural factors is smal-ler for some stimuli and SOAs than for others. However, structuralfactors take over completely in case of simultaneous priming ormasking, that is, when target and prime or mask are presentedsimultaneously in one stimulus. One of many examples of simulta-neous priming is van der Helm and Treder’s (2009) finding thatdetection of mirror symmetry or repetition in designated contourparts of two shapes is easier when the entire stimulus is a mirrorsymmetry or a repetition than when it is not. A currently relevantexample of simultaneous masking is that the perceived globalstructure of a stimulus masks local stimulus features that areincompatible with this global structure (see Fig. 1, also for the dif-ference between incompatibility and incongruency). By PATVISH,this can be understood as follows. The perceptual organization pro-cess yields perceived hierarchical organizations in terms of globalstructures and their constituent local features. This means that itpreserves the representations of the (compatible) constituents,and suppresses or eliminates those of other (incompatible) parts.Thus, if a to-be-detected local feature is compatible, the top-downattention process may exploit the perceived hierarchical organiza-tion to descend easily from its global structure to this local feature.If it is not compatible – as is typical in embedded figures tasks, forinstance – the top-down attention process is first misled by theperceived global structure and then has to find a way around it.

2.4. Summary

In this section, I discussed how the cognitive architecturePATVISH incorporates existing ideas about typical perception andattention to specify their combined action. In PATVISH, perceptualorganization is a predominantly autonomous stimulus-driven pro-cess, in which transparallel processing of similar features – medi-ated by temporarily synchronized neural assemblies – underliesthe perceptual integration capability. The latter yields hierarchicalstimulus organizations, with local features represented in lowerareas in the visual hierarchy in the brain and global structures inhigher areas. Attention may modulate the outcome of this organi-zation process but predominantly subserves top-down scrutiny ofthe hierarchical structure of established percepts. This explains thedominance of wholes over parts and various priming and maskingeffects – in particular, and relevant to ASD, that perceived globalstructures mask incompatible local features.

P.A. van der Helm / Vision Research 126 (2016) 278–290 283

3. The local advantage phenomenon in ASD

Here, I do not discuss theories on ASD in general, but I focus onthe local advantage phenomenon, which, as said, is the phe-nomenon that ASD individuals tend to exhibit a better than typicalperformance on tasks in which local features are to be discerned.Insofar as theories on ASD included considerations specific to thisphenomenon, these considerations tend to fall into two categories.One category reflects the idea that the phenomenon is to be attrib-uted indirectly to inferior processing of global structures – this ideahas been advocated most prominently in the weak central coher-ence theory (WCC) proposed by Frith (1989) (see also Happé &Booth, 2008). The other category reflects the idea that the phe-nomenon is to be attributed directly to superior processing of localfeatures – this idea has been advocated most prominently in theenhanced perceptual functioning theory (EPF) proposed byMottron and Burack (2001) (see also Mottron et al., 2006). To shedmore light on this issue, I next re-evaluate the local advantage phe-nomenon starting from the cognitive architecture PATVISH.

3.1. Global vs. local processing

Mottron et al. (2006, p. 40) rightly expressed concerns about thepossibly misleading use of the terms low-level and high-level pro-cessing as, among others, in their EPF principle ‘‘higher-order pro-cessing is optional in autism and mandatory in non-autistics’’.Indeed, it is not always clear how these terms map onto perceptualand attentional processes, for instance. By PATVISH, perception is apredominantly nonconscious stimulus-driven process, while atten-tion predominantly subserves top-down scrutiny of the hierarchicalstructure of established percepts. Hence, PATVISH takes perceptionto be an autonomous and therefore ‘‘mandatory’’ process, whereas ittakes attention to be ‘‘optional’’ in that it can be directed consciouslyin function of a task. Therefore, here, I prefer to use the terms low-level and high-level processing to refer to perceptual processing oflocal features and global structures, respectively. In my view, bothare ‘‘mandatory’’ in typical perception and the current question iswhether, in ASD, one or the other deviates.

EPF advocates that local processing deviates in ASD, and to thisend, it also claims to provide evidence that global processing isintact in ASD. However, I have reservations regarding this evi-dence, because it mostly relies on the usage of Navon’s (1977) typeof stimuli (see, e.g., Iarocci et al., 2006; Mottron et al., 2003; Wanget al., 2007). On the one hand, Navon-type stimuli perfectly illus-trate the asymmetrical hierarchical relationship between globalstructures and local constituents (Fig. 1b gives two examples).On the other hand, the global structures in such stimuli are usuallyso simple that they can be discerned easily via conscious patternrecognition. For instance, when explicitly asked for, high-function-ing ASD individuals seem well able to correctly report such globalstructures (Koldewyn et al., 2013; Mottron et al., 1999; Plaisted,Swettenham, & Rees, 1999; Rinehart et al., 2000). Such instructionsin fact appeal to (conscious) attention rather than to (noncon-scious) perception, and I do not exclude that, by way of an ‘‘op-tional’’ long-term attentional strategy, high-functioning ASDindividuals train themselves in conscious pattern recognition tocompensate for the fact that – as WCC suggests – they do not per-ceive global structures as instantaneously as typical individuals do.

Notice further that, unlike to-be-discerned parts in embeddedfigures tasks, the local constituents in a Navon-type stimulus arealways compatible with the typically perceived global structure.That is, they may or may not be congruent with it but they arealways proper substructures of it (see Fig. 1). By PATVISH, thismeans that typical individuals can easily deploy attention via theperceived global structures to the local elements, that is, probably

as easily as – according to WCC – ASD individuals can do moredirectly because, in their case, attention hardly encounters per-ceived global structures. Besides, data based on Navon-type stimuliare known to depend heavily on the employed empirical design(Kimchi, 1992), so, all in all, it is no surprise that they can go oneway or the other regarding differences between typical and ASDindividuals (see, e.g., Bernardino et al., 2012; Scherf et al., 2008).In other words, the usage of such stimuli can lead to a conflationof local and global processing, so that it can hardly yield conclusiveevidence about whether or not global processing is intact (cf.Happé & Booth, 2008).

To be clear, the foregoing does not go against enhanced localprocessing, but it does question the alleged evidence for intact glo-bal processing. In fact, the idea of intact global processing is prob-lematic in view of the following two observations.

First, categorization in typical individuals relies on perceivedglobal structures, which, by definition, means that minor differ-ences tend to be ignored (see the earlier-given references on thedominance of wholes over parts). So, if global processing in ASDindividuals would be intact, they would also show typical catego-rization, but if it is not, they would naturally put more value onminor differences – leading to atypical categorization involving,more specifically, smaller categories. That categorization in ASDis indeed atypical is sustained by studies that focused on catego-rization tasks in which the to-be-categorized exemplars showeddeviations from prototypes (Klinger & Dawson, 2001; see alsoChurch et al., 2010; Gastgeb et al., 2009; Johnson & Rakison,2006; Molesworth, Bowler, & Hampton, 2008; Plaisted, 2000).Such categorization tasks require the formation of prototypes,and according to Newell et al. (2010), these findings indeed suggestthat, unlike typical exemplars, atypical exemplars are processedmore like members of a (relatively small) subordinate categorythan as members of a (relatively large) basic level category.

Second, even when local processing is enhanced, typical globalprocessing would preclude the local advantage phenomenon,because it would still yield global structures that tend to masklocal features. Notice that, by PATVISH, this masking effect appliesto features that are incompatible with perceived global structures,that is, not to compatible features. To my knowledge, this differen-tiation has not been made in WCC. It does not, however, alter thefact that – while enhanced local processing would be helpful –reduced global processing seems necessary for the local advantagephenomenon to arise.

3.2. Neurophysiological evidence

There is ample evidence of different neuro-cognitive processingin ASD individuals as compared to typical individuals. As for theWCC and EPF accounts, much of this evidence has been taken tobe in favor of one or against the other (see Happé & Frith, 1996;Mottron et al., 2006). To date, however, more specific data havebecome available.

For instance, Peters et al.’s (2013) data suggest that, in ASD,there is a trade-off between local and global processing. Theyfound that, as compared to typical individuals, ASD individualsexhibit decreased long-range and increased short-range neuralconnectivity (see also Barttfeld et al., 2011). By nature, short-rangeconnections are thought to be involved in the processing of rela-tively local things and long-range connections in the processingof relatively global things. Peters et al. (2013) measured connectiv-ity in terms of EEG synchronization in the theta and alpha bands(4–12 Hz), that is, not in the gamma band (30–70 Hz) which is usu-ally associated with visual processes. Nevertheless, by proxy, theirfinding suggests that the local advantage phenomenon is due to acombination of enhanced local processing and reduced globalprocessing.

284 P.A. van der Helm / Vision Research 126 (2016) 278–290

To qualify the foregoing, the evidence for decreased long-rangeconnectivity seems considerably stronger than that for increasedshort-range connectivity (see, e.g., Hoppenbrouwers,Vandermosten, & Boets, 2014; Vissers, Cohen, & Geurts, 2012).This agrees with the observation above that – while enhanced localprocessing would be helpful – reduced global processing seemsnecessary to trigger the local advantage phenomenon.

Possibly related to the atypical connectivity, ASD individualsalso exhibit atypical, or impaired, oscillations in the 30–70 Hzgamma band (Grice et al., 2001; Maxwell et al., 2015; Milneet al., 2009; Sun et al., 2012; Wright et al., 2012). Gamma-bandoscillations are particularly suited for establishing precise neuronalsynchronization in the visual hierarchy (Fries, 2009), which, as dis-cussed, generally is taken to subserve perceptual integration.Hence, like the atypical connectivity, also the atypical gammaactivity suggests that global processing in ASD is reduced. To inves-tigate this further, I next take a closer look at what the role ofgamma synchronization might be.

The temporal correlation hypothesis (Milner, 1974; von derMalsburg, 1981; for a review, see Gray, 1999) applies to the inte-gration of different features into percepts. It holds that gammasynchronization binds those neurons which, together, representone perceptual entity, say, one object (see also Eckhorn et al.,2001). Building on this hypothesis, Brock et al. (2002) proposedthe temporal binding deficit hypothesis of ASD. The latter hypoth-esis holds that, due to impaired gamma synchronization, the tem-poral binding between local networks in the brain is impaired,while the temporal binding within local networks is intact or pos-sibly even enhanced. In an update (Rippon et al., 2007), they main-tained the idea of a temporal binding deficit due to impairedgamma synchronization but – probably because gamma activityis usually associated with local computations – they seemed todownplay the within vs. between local networks differentiation.

It is, of course, relevant to distinguish processes within andbetween local networks, but a difficulty with the line of reasoningabove is that it relies on the idea that synchronization is a forcethat binds neurons. The temporal correlation hypothesis is indeednot without questions and concerns (e.g., Thiele & Stoner, 2003). Inparticular, taking synchronization as a binding force begs the ques-tion of which neurons are to be bound (Shadlen & Movshon, 1999).I do not speculate about the mechanisms of neuron binding, but asI discussed earlier, PATVISH assigns a different role tosynchronization.

To recall, in PATVISH, percepts are assumed to be given by sim-plest hierarchical organizations obtained by extracting a maximumamount of visual regularity. In PISA, PATVISH’s computationalmodel for strings, the extracted regularities (e.g., repetitions andsymmetries) are such that similar regularities group naturally intodistributed representations called hyperstrings, which are specialin that, subsequently, those similar regularities can be hierarchi-cally recoded in a transparallel fashion (i.e., simultaneously as ifonly one regularity were concerned; see Appendix A). Thistransparallel recoding then yields a hierarchy of feature constella-tions from which different features are selected to be integratedinto percepts. PATVISH takes hyperstrings to correspond to tem-porarily synchronized neural assemblies, whose synchronizationis postulated to be the neural signature of transparallel recodingof similar features represented in such assemblies.

The foregoing retains the idea that synchronization subservesperceptual integration, but instead of taking synchronization as aforce that binds features, it takes synchronization as a manifesta-tion of the further processing of bound features. This suggests thatimpairments in synchronization do not so much affect the repre-sentation of local features but rather hamper the further process-ing of those features – including the coupling of different neuralassemblies and the integration of different features into global

structures. Hence, for ASD, this alternative picture of synchroniza-tion does not exclude enhanced local processing but it does sustainthe idea that the atypical gamma activity implies reduced globalprocessing. (Note: this does not interfere with the idea that atypi-cal gamma activity reflects decreased signal-to-noise ratios due todecreased inhibitory processing; Brown et al., 2005; Dinstein et al.,2012; Milne, 2011; Rippon et al., 2007; Rubenstein & Merzenich,2003; Simmons et al., 2009; Snijders, Milivojevic, & Kemner, 2013.)

Hence, all in all, I think that empirical (both behavioral and neu-rophysiological) and theoretical (both logical and computational)evidence converges towards the idea that the local advantage phe-nomenon in ASD arises primarily due to reduced global processingcaused by impaired gamma synchronization. That is, it convergestowards the idea that the latter hampers the integration of localfeatures into global structures, which, in turn, implies that themasking effect of global structures on incompatible features isweaker, so that, in general, task-driven top-down attention hasbetter access to local features.

4. Discussion

Based on the neurally plausible cognitive architecture PATVISH– which models the combined action of perceptual organizationand attention – I argued that gamma synchronization is not somuch a force that binds features but rather a manifestation ofthe further processing of bound features. This strengthens the ideathat impaired gamma activity – as found in ASD – implies reducedglobal processing. For ASD, this does not exclude enhanced localprocessing but it does suggest that the local advantage phe-nomenon is to be attributed primarily to reduced global process-ing. In this section, I discuss several implications (includingcritical predictions) and suggestions for future research.

By PATVISH, the human perceptual organization process isautonomous, so that typical individuals can do nothing but inte-grate incoming pieces of visual information into perceived wholes.Due to an impaired integration capability, however, ASD individu-als seem to be left with something in between (think of an unfin-ished jigsaw puzzle). Then, top-down attention hardly hasanything global to focus on, so that it naturally exhibits a focuson local things. In other words, reduced global processing naturallyresults in a narrowed spatial focus of attention (as Baron-Cohen,2004, proposed) and in putting relatively more value on smallerrors (as Van de Cruys et al., 2013, 2014, proposed within a pre-dictive coding approach). To be clear, ideas like the latter may leadto powerful models, but my point is that they comply just as wellwith reduced global processing as with enhanced local processing.

More indicative are neurophysiological data (see above) andbehavioral data (for a comprehensive meta-analysis, see Van derHallen et al., 2014). For instance, as discussed, categorization relieson perceived global structures, and ASD individuals tend to catego-rize deviations from prototypes into smaller categories than typicalindividuals do (for a similar finding in amodal completion, see deWit et al., 2007). This suggests that they perceptually integrate fea-tures to a lesser degree than typical individuals do.

This could be investigated further by means of triadic compar-isons of stimuli like those in Fig. 4a. Such stimuli depict three-dimensional (3-D) objects that could be characterized formally interms of generalized cylinders (Binford, 1981; Biederman, 1987).Here, however, I focus on their perceptual characterization interms of superstructures that determine the positions of identicalsubordinate structures – as if one could produce such an objectby moving a fixed shape along the superstructure (Leeuwenberg& van der Helm, 1991; Leeuwenberg et al., 1994; van Lier et al.,1997). In general, objects are typically categorized by their percep-tually dominant superstructures (i.e., same superstructure, same

Fig. 4. (a) In the top row, objects for which, as indicated in the bottom row, a superstructure (solid line) determines the positions of identical subordinate structures (dashedline) – as if one could produce such an object by moving a fixed shape along the superstructure. The central object thus has a reversed hierarchy as compared to the one at theleft, and has the same superstructure as the one at the right. (b) At the top, a stimulus (after Kastens & Ishikawa, 2006) with a typically perceived organization given by twotriangular shapes, plus one of these compatible parts (bottom left) and an incompatible part (bottom right).

P.A. van der Helm / Vision Research 126 (2016) 278–290 285

category) but notice that, while the superstructures of two-dimen-sional (2-D) objects are always larger than their subordinate struc-tures, the superstructures of 3-D objects can be smaller (seeFig. 4a). The latter means that the asymmetrical hierarchical rela-tionship between superstructures and subordinate structures isnot determined by individual properties like size, but is purelythe result of their integration into organized wholes. I thereforepredict that, due to reduced global processing, ASD individuals cat-egorize such 3-D objects less on the basis of their typically per-ceived superstructures.

Furthermore, the local advantage phenomenon in ASD is mostprominent in visual search tasks, block design tasks, and embed-ded figures tasks (Jolliffe & Baron-Cohen, 1997; Joseph et al.,2009; O’Riordan et al., 2001; Shah & Frith, 1983). The availablebehavioral data on these tasks might be multi-interpretable butare at the least consistent with reduced global processing. Thatis, as argued, the detectability of a target in a visual search displaydepends typically on the perceived whole of target plus distractors– hence, less so if global processing is impaired. In block designtasks, the typically perceived organization of a design may conflictwith the blocks from which it is to be constructed – but, again, lessso if global processing is impaired. By PATVISH, the (in)compatibil-ity of typically perceived wholes and to-be-discerned parts is rele-vant in both these tasks, but even more so in embedded figurestasks. In fact, for embedded figures tasks, I predict that ASD indi-viduals exhibit the local advantage phenomenon for incompatibleparts but not, or hardly, for compatible parts – because, in typicalindividuals, perceived global structures mask incompatible partsbut not compatible parts (see Fig. 4b).

Both predictions above have, to my knowledge, not yet beentested, but notice that they are critical in that they would not fol-low from enhanced local processing plus intact global processing.Both predictions in fact illustrate that, in empirical studies on thelocal advantage phenomenon in ASD, it is imperative not only todistinguish between typically perceived global structures and localfeatures but also to dissociate global processing and local process-ing. For instance, as argued, Navon-type stimuli allow for the for-mer but not for the latter, because the local constituents arealways compatible with the perceived global structures. The deli-cateness of empirical designs also shows itself in ASD researchon geometrical visual illusions, which rely on discrepanciesbetween properties of 2-D images and properties of their 3-D per-cepts. Happé (1996) found that ASD individuals are not susceptible

to such illusions (see also Bölte et al., 2007), which is what onewould expect if they have a reduced perceptual integration capa-bility. However, Ropar and Mitchell (1999) found that they are assusceptible as typical individuals. As usual, the truth probably liessomewhere between the two. Mitchell et al. (2010), for instance,found that they are susceptible – but less than typical individuals– to Shepard’s (1981) table-top illusion (surfaces that are identicalin a 2-D image but not in its 3-D percept). As I discuss next, a sim-ilar situation holds for research on symmetry perception, which,however, also gives rise to a quantitative tool to investigate indi-vidual differences.

Csathó, van der Vloed, and van der Helm (2003) conducted asymmetry detection experiment using stimuli composed of onemirror symmetrical or random part surrounding another with asame or different spatial scale. They aimed at typical individualsand, afterwards, they therefore excluded a participant who wassuspected to be a high-functioning ASD individual. Whereas theother participants’ correct scores varied considerably with theempirical conditions, this participant not only responded muchfaster than the others but also scored nearly 100% correct in allconditions – so, no speed–accuracy trade-off. Csathó et al. didnot pursue this point further, but later, Perreault et al. (2011) alsofound an increased sensitivity to mirror symmetry in ASD.However, using another paradigm, Falter and Bailey (2012) founda decreased sensitivity to mirror symmetry in ASD – characterized,moreover, by a speed–accuracy trade-off.

I concur with Falter (2013) that these opposite findings mightbe due to the fact that Perreault et al.’s (2011) paradigm allowedfor a local attentional strategy, whereas Falter and Bailey’s (2012)paradigm required a global attentional strategy. In this respect,notice not only that symmetry detection requires the integrationof correlations between stimulus elements (see Fig. 5), but alsothat symmetry is a stimulus attribute that precedes object percep-tion (Perreault et al., 2011; van der Helm & Treder, 2009). Hence,symmetry can be said to be a mid-level feature. In fact, typical indi-viduals integrate symmetry and noise into a percept in which thesymmetry-to-noise ratio determines how well the symmetry isdetected (Csathó, van der Vloed, & van der Helm, 2004).

More specifically, for a mirror symmetrical dot pattern consist-ing of R symmetry pairs perturbed by N noise elements, thedetectability (d0) of the symmetry typically follows the psy-

chophysical law d0 ¼ g=ð2þ N=RÞ, where the proportionality

Fig. 5. An imperfect mirror symmetry with 40 symmetry pairs (at the left) and an imperfect Glass pattern with 40 dipoles (at the right). Despite the noise (20 unpaired dots),each regularity remains detectable due to the intact correlations that give rise to the regularity. In symmetry, these correlations are given by parallel and midpoint-collinearvirtual lines connecting elements in symmetry pairs, and in Glass patterns, by the coherent orientation of dot dipoles.

286 P.A. van der Helm / Vision Research 126 (2016) 278–290

constant g is a psychophysical parameter which allows for moredetailed data fits than just rank orders and which is to be estimatedusing experimental data (van der Helm, 2010). Taking the symme-try-to-noise ratio R=N as signal, this law improves on Weber’s law,which requires more parameters to fit data and which misses floorand ceiling effects. Interestingly, it holds also for Glass patternsconsisting of R randomly positioned but coherently oriented dotdipoles perturbed by N noise elements (Maloney, Mitchison, &Barlow, 1987). For such Glass patterns (see Fig. 5), Koldewyn,Whitney, and Rivera (2010) found no difference in detectabilitybetween typical and ASD individuals. Spencer and O’Brien (2006),on the other hand, used perfect Glass patterns surrounded by noiseelements. They distinguished an autism group and an Aspergergroup and found a heightened detection threshold for the formerbut not for the latter (but see also Simmons et al., 2009).

Hence, for symmetry and Glass pattern detection, it seems hardto arrive at a consensus about group differences. However, noticethat the psychophysical law above provides a quantitative tool toinvestigate individual differences. To fit data, this law involves onlyone parameter. Such a psychophysical parameter is to be deter-mined experimentally because it depends on stimulus type andobserver. Thus, not only this law’s goodness-of-fit but also itsobserver-dependent parameter is predicted to be informativeabout individual differences – in general, so, also in ASD.Perreault et al. (2011) and Koldewyn et al. (2010) did not provideenough data to test this here but their paradigms would be appro-priate (Note: such paradigms might also be relevant to the ideathat atypical gamma activity in ASD reflects decreased signal-to-noise ratios; Simmons et al., 2009.)

Finally, a focus on individual differences is appropriate also con-sidering that perceptual integration is a nonlinear and thereforefragile process. That is, neuronal synchronization – which, byPATVISH, underlies perceptual integration – is a dynamic processthat relies on the precise timing of the firing rates of neurons (cf.van Leeuwen, Steyvers, & Nooter, 1997). This means that a minorchange in neural parameters may have a large impact on cognitivebehavior. It is therefore understandable that impairments manifestthemselves in various shapes and forms. This agrees with the factthat the autism spectrum is a continuum with many individual dif-ferences, which reflect on performance in perceptual tasks (see,e.g., Dakin & Frith, 2005; Scherf et al., 2008; Simmons et al.,2009). More general, weak central coherence seems even morepronounced in Williams Syndrome (Bernardino et al., 2012) andimpaired gamma synchronization also has been found inWilliams Syndrome and schizophrenia (Grice et al., 2001;Uhlhaas, Silverstein, & Phillips, 2005). This suggests that these dis-orders share a reduced perceptual integration capability. Becauseone’s perception is the primary source of knowledge about one’ssurrounding world, it guides one’s interaction with that world.The idea of a reduced perceptual integration capability mighttherefore be pivotal to a better understanding of the neuro-cogni-tive side of such disorders in general and of individual differencesin particular.

5. Conclusion

In this theoretical study, I strengthened the idea that the localadvantage phenomenon in ASD is primarily due to reduced globalprocessing caused by impaired neuronal synchronization in thegamma band. To investigate this phenomenon, I started from theneurally plausible cognitive architecture PATVISH, which models(a) perception as a predominantly stimulus-driven process yieldinghierarchical stimulus organizations, and (b) attention as predomi-nantly scrutinizing the hierarchical structure of established per-cepts in a task-driven top-down fashion. As I discussed, thisexplains a range of phenomena in typical vision and has interestingimplications for ASD research.

At the cognitive side of the local advantage phenomenon, Iargued that it is imperative to dissociate global processing andlocal processing. I therefore proposed a categorization task inwhich the asymmetrical hierarchical relationship between globalstructures and local features is put to a critical test. By the sametoken, I predicted that, in embedded figures tasks, ASD individualsexhibit the local advantage phenomenon for parts that are incom-patible with typically perceived global structures but not, orhardly, for compatible parts. Furthermore, I argued that individualdifferences in ASD can be probed quantitatively by means of a psy-chophysical law, which, for typical individuals, specifies thedetectability of mirror symmetries and Glass patterns in the pres-ence of noise.

At the neuronal side, I argued that gamma synchronization is amanifestation of transparallel processing of similar features. Likeother interpretations, this interpretation is admittedly speculative,but unlike the others, it finds support in a powerful minimal-cod-ing algorithm. This provides a firmer theoretical basis for the ideathat impaired gamma activity – as found in ASD – implies reducedglobal processing.

Acknowledgments

I thank Árpád Csathó, Emanuel Leeuwenberg, Bilge Sayim,Sander Van de Cruys, Ruth Van der Hallen, Gert van der Vloed,and Johan Wagemans for valuable discussions on various aspectsof this study. This research was supported by Methusalem grantMETH/08/02 awarded to Johan Wagemans (www.gestaltrevision.be).

Appendix A

The key property of the minimal-coding algorithm PISA is that itemploys transparallel processing by hyperstrings to hierarchicallyrecoded exponential numbers of similar features simultaneously asif only one feature were concerned. As mentioned earlier, full tech-nical exposés on PISA can be found elsewhere. Here, I give only thegist of it.

Fig. A.1. Hyperstrings. The graph represents the arguments of all symmetries into which the string ababfababbabafbaba can be encoded. The graph is a hyperstring and cantherefore be hierarchically recoded as if it were a single normal string h1h2h3h4h5h6h7h8h9, whose substrings correspond one-to-one to hypersubstrings in the graph.

P.A. van der Helm / Vision Research 126 (2016) 278–290 287

To compute simplest codes of strings, PISA applies the codinglanguage and the complexity metric developed in structural infor-mation theory (SIT). One of the coding rules in SIT’s coding lan-guage is the S-rule, which captures bilateral symmetries. Forinstance, by the S-rule, the string ababfababbabafbaba can beencoded into S[(aba)(b)(f)(aba)(b)], whose argument (aba)(b)(f)(a-ba)(b) is represented in the graph in Fig. A.1 by the path along ver-tices 1, 4, 5, 6, 9, and 10. In fact, this graph represents, in adistributed fashion, the arguments of all symmetries into whichthe string can be encoded.

To assess which of these symmetries is the simplest one, theirarguments have to be hierarchically recoded first. For instance,the argument (aba)(b)(f)(aba)(b) above can be hierarchicallyrecoded into S[((aba)(b)),((f))] which gives a further reduction incomplexity. The problem now is that there may be up to an expo-nential symmetries into which a string can be encoded, and that itwould take a superexponential amount of work and time to recodeeach of their arguments separately. Provably, however, graphs likethe one in Fig. A.1 are hyperstrings, whose formal graph-theoreti-cal definition is:

Definition A.1. A hyperstring is a simple semi-Hamiltoniandirected acyclic graph ðV ; EÞ with a labeling of the edges in E suchthat, for all vertices i; j; p; q 2 V:

either pði; jÞ ¼ pðp; qÞ or pði; jÞ \ pðp; qÞ ¼£

where substring set pðv1;v2Þ is the set of label strings representedby the paths between vertices v1 and v2; the subgraph on the ver-tices and edges in these paths is a hypersubstring.

Definition A.1 holds that, in a hyperstring, substring sets repre-sented by hypersubstring are either completely identical or com-pletely disjoint – never something in between. This implies thatthe hyperstring in Fig. A.1 can be treated as if it were a single nor-mal string H ¼ h1h2h3h4h5h6h7h8h9, whose substrings correspondone-to-one to hypersubstrings. For instance, substrings h1h2h3h4

and h6h7h8h9 are identical, because they both represent the sub-strings ðaÞðbÞðaÞðbÞ; ðabaÞðbÞ, and ðaÞðbabÞ in candidate symmetryarguments. In other words, the single identity relationshipbetween substrings in string H corresponds, in one go, to threeidentity relationships between substrings in candidate symmetryarguments. The identity relationship in string H means that Hcan be encoded into the symmetry S½ðh1h2h3h4Þ; ðh5Þ�, which thusrepresents, in one go, three symmetries in different candidate sym-metry arguments, namely:

S½ððaÞðbÞðaÞðbÞÞ;ððf ÞÞ� in the argument ðaÞðbÞðaÞðbÞðf ÞðaÞðbÞðaÞðbÞS½ððabaÞðbÞÞ;ððf ÞÞ� in the argument ðabaÞðbÞðf ÞðabaÞðbÞS½ððaÞðbabÞÞ;ððf ÞÞ� in the argument ðaÞðbabÞðf ÞðaÞðbabÞ

Hence, by encoding the hyperstring, one in fact hierarchicallyrecodes all candidate symmetry arguments in one go, without hav-ing to distinguish explicitly between them.

There is, of course, much more one has to reckon with to get afull-blown minimal-coding algorithm (for that, see the full techni-cal exposés). However, the foregoing shows that the candidatesymmetry arguments do not have to recoded in a serial fashion(i.e., one after the other by one processor) or in a parallel fashion(i.e., simultaneously by many processors). Instead, they can berecoded simultaneously by one processor (e.g., a single-processorclassical computer) as if only one symmetry argument were con-cerned. This also holds for the other coding rules in SIT’s codinglanguage, and this is the extraordinary form of processing I dubbedtransparallel processing.

References

Ahissar, M., & Hochstein, S. (2004). The reverse hierarchy theory of visualperceptual learning. Trends in Cognitive Science, 8, 457–464. http://dx.doi.org/10.1016/j.tics.2004.08.011.

American Psychiatric Association (2013). Diagnostic and statistical manual of mentaldisorders (5th ed.). Arlington, VA: American Psychiatric Publishing.

Anderson, J. R. (1983). The architecture of cognition. Cambridge, MA: HarvardUniversity Press.

Anderson, J. R. (2004). Cognitive psychology and its implications. Richmond, UK:Worth Publishers.

Bachmann, T. (1984). The process of perceptual retouch: Nonspecific afferentactivation dynamics in explaining visual masking. Perception & Psychophysics,35, 69–84. http://dx.doi.org/10.3758/BF03205926.

Baron-Cohen, S. (2004). The cognitive neuroscience of autism. Journal of NeurologyNeurosurgery & Psychiatry, 75, 945–948. http://dx.doi.org/10.1136/jnnp.2003.018713.

Barttfeld, P., Wicker, B., Cukier, S., Navarta, S., Lew, S., & Sigman, M. (2011). A big-world network in ASD: Dynamical connectivity analysis reflects a deficit inlong-range connections and an excess of short-range connections.Neuropsychologia, 49, 254–263. http://dx.doi.org/10.1016/j.neuropsychologia.2010.11.024.

Beller, H. (1971). Priming: Effects of advance information on matching. Journal ofExperimental Psychology, 87, 176–182. http://dx.doi.org/10.1037/h0030553.

Bernardino, I., Mouga, S., Almeida, J., van Asselen, M., Oliveira, G., & Castelo-Branco,M. (2012). A direct comparison of local-global integration in autism and otherdevelopmental disorders: Implications for the central coherence hypothesis.PLoS One, 7, e39351. http://dx.doi.org/10.1371/journal.pone.0039351.

Biederman, I. (1987). Recognition-by-components: A theory of human imageunderstanding. Psychological Review, 94, 115–147. http://dx.doi.org/10.1037/0033-295X.94.2.115.

Binford, T. (1981). Inferring surfaces from images. Artificial Intelligence, 17, 205–244.http://dx.doi.org/10.1016/0004-3702(81)90025-4.

Bölte, S., Holtmann, M., Poustka, F., Scheurich, A., & Schmidt, L. (2007). Gestaltperception and local–global processing in high-functioning autism. Journal ofAutism and Developmental Disorders, 37, 1493–1504. http://dx.doi.org/10.1007/s10803-006-0231-x.

Breitmeyer, B. G. (2007). Visual masking: Past accomplishments, present status,future developments. Advances in Cognitive Psychology, 3, 9–20. http://dx.doi.org/10.2478/v10053-008-0010-7.

Breitmeyer, B. G., & Ogmen, H. (2000). Recent models and findings in visualbackward masking: A comparison, review, and update. Perception &Psychophysics, 62, 1572–1595. http://dx.doi.org/10.3758/BF03212157.

Breitmeyer, B. G., & Ogmen, H. (2006). Visual masking: Time slices through consciousand unconscious vision. Oxford, UK: Oxford University Press.

Brock, J., Brown, C. C., Boucher, J., & Rippon, G. (2002). The temporal binding deficithypothesis of autism. Development and Psychopathology, 14, 209–224. http://dx.doi.org/10.1017/S0954579402002018.

Brown, C., Gruber, T., Rippon, G., Brock, J., & Boucher, J. (2005). Gammaabnormalities during perception of illusory figures in autism. Cortex, 41,264–376. http://dx.doi.org/10.1016/S0010-9452(08)70273-9.

Byrne, M. D. (2012). Unified theories of cognition. WIREs Cognitive Science, 3,431–438. http://dx.doi.org/10.1002/wcs.1180.

288 P.A. van der Helm / Vision Research 126 (2016) 278–290

Church, B. A., Krauss, M. S., Lopata, C., Toomey, J. A., Thomeer, M. L., Coutinho, M. V.,et al. (2010). Atypical categorization in children with high functioning autismspectrum disorder. Psychonomic Bulletin & Review, 17, 862–868. http://dx.doi.org/10.3758/PBR.17.6.862.

Churchland, P. S. (1986). Neurophilosophy. Cambridge, MA: MIT Press.Collard, R. F., & Povel, D. J. (1982). Theory of serial pattern production: Tree

traversals. Psychological Review, 89, 693–707. http://dx.doi.org/10.1037/0033-295X.89.6.693.

Csathó, Á., van der Vloed, G., & van der Helm, P. A. (2003). Blobs strengthenrepetition but weaken symmetry. Vision Research, 43, 993–1007. http://dx.doi.org/10.1016/S0042-6989(03)00073-7.

Csathó, Á., van der Vloed, G., & van der Helm, P. A. (2004). The force of symmetryrevisited: Symmetry-to-noise ratios regulate (a)symmetry effects. ActaPsychologica, 117, 233–250. http://dx.doi.org/10.1016/j.actpsy.2004.06.004.

Dakin, S., & Frith, U. (2005). Vagaries of visual perception in autism. Neuron, 48,497–507. http://dx.doi.org/10.1016/j.neuron.2005.10.018.

de-Wit, L., Kubilius, J., Wagemans, J., & Op de Beeck, H. P. (2012). Bi-stable Gestaltsreduce activity in the whole of V1 not just the retinotopically predicted parts.Journal of Vision, 12, 1–14. http://dx.doi.org/10.1167/12.11.12.

de Wit, T. C. J., Schlooz, W., Hulstijn, W., & Van Lier, R. (2007). Visual completion andcomplexity of visual shape in children with pervasive developmental disorder.European Child & Adolescent Psychology, 16, 168–177. http://dx.doi.org/10.1007/s00787-006-0585-9.

Dinstein, I., Heeger, D. J., Lorenzi, L., Minshew, N. J., Malach, R., & Behrmann, M.(2012). Unreliable evoked responses in autism. Neuron, 75, 981–991. http://dx.doi.org/10.1016/j.neuron.2012.07.026.

Donderi, D. C. (2006). Visual complexity: A review. Psychological Bulletin, 132,73–97. http://dx.doi.org/10.1037/0033-2909.132.1.73.

Duncan, J., & Humphreys, G. W. (1989). Visual search and stimulus similarity.Psychological Review, 96, 433–458. http://dx.doi.org/10.1037/0033-295X.96.3.433.

Eckhorn, R. (1999). Neural mechanisms of visual feature binding investigated withmicroelectrodes and models. Visual Cognition, 3&4, 231–265. http://dx.doi.org/10.1080/135062899394975.

Eckhorn, R., Bauer, R., Jordan, W., Brosch, M., Kruse, W., Munk, M., et al. (1988).Coherent oscillations: A mechanisms of feature linking in the visual cortex?Biological Cybernetics, 60, 121–130. http://dx.doi.org/10.1007/BF00202899.

Eckhorn, R., Bruns, A., Saam, M., Gail, A., Gabriel, A., & Brinksmeyer, H. J. (2001).Flexible cortical gamma-band correlations suggest neural principles of visualprocessing. Visual Cognition, 8, 519–530. http://dx.doi.org/10.1080/13506280143000098.

Edelman, G. M. (1987). Neural Darwinism: The theory of neuronal group selection.New York: Basic Books.

Enns, J. T., & Di Lollo, V. (1997). Object substitution: A new form of masking inunattended visual locations. Psychological Science, 8, 135–139. http://dx.doi.org/10.1111/j.1467-9280.1997.tb00696.x.

Falter, C. M. (2013). Impaired and superior mirror symmetry perception in autism.Autism, 17, 117–118. http://dx.doi.org/10.1177/1362361312438.

Falter, C. M., & Bailey, A. J. (2012). Perception of mirror symmetry in autismspectrum disorder. Autism, 16, 622–626. http://dx.doi.org/10.1177/1362361311407353.

Felleman, D. J., & van Essen, D. C. (1991). Distributed hierarchical processing in theprimate cerebral cortex. Cerebral Cortex, 1, 1–47. http://dx.doi.org/10.1093/cercor/1.1.1.

Finkel, L. H., Yen, S.-C., & Menschik, E. D. (1998). Synchronization: Thecomputational currency of cognition. In L. Niklasson, M. Boden, & T. Ziemke(Eds.), ICANN 98, Proceedings of the 8th International Conference on ArtificialNeural Networks (Skövde, Sweden, 2–4 September, 1998). New York, NY:Springer-Verlag.

Fries, P. (2005). A mechanism for cognitive dynamics: Neuronal communicationthrough neuronal coherence. Trends in Cognitive Sciences, 9, 474–480. http://dx.doi.org/10.1016/j.tics.2005.08.011.

Fries, P. (2009). Neuronal gamma-band synchronization as a fundamental processin cortical computation. Annual Review of Neuroscience, 32, 209–224. http://dx.doi.org/10.1146/annurev.neuro.051508.135603.

Frith, U. (1989). Autism: Explaining the enigma. Oxford, UK: Basil Blackwell.Gastgeb, H., Rump, K. M., Best, C. A., Minshew, N. J., & Strauss, M. S. (2009).

Prototype formation in autism: Can individuals with autism abstract facialprototypes? Autism Research, 2, 279–284. http://dx.doi.org/10.1002/aur.93.

Gerbino, W., & Salmaso, D. (1987). The effect of a modal completion on visualmatching. Acta Psychologica, 65, 25–46. http://dx.doi.org/10.1016/0001-6918(87)90045-X.

Gilbert, C. D. (1992). Horizontal integration and cortical dynamics. Neuron, 9, 1–13.http://dx.doi.org/10.1016/0896-6273(92)90215-Y.

Gray, C. M. (1999). The temporal correlation hypothesis of visual featureintegration: Still alive and well. Neuron, 24, 31–47. http://dx.doi.org/10.1016/S0896-6273(00)80820-X.

Gray, C. M., & Singer, W. (1989). Stimulus-specific neuronal oscillations inorientation columns of cat visual cortex. Proceedings of the National Academyof Sciences USA, 86, 1698–1702. http://dx.doi.org/10.1073/pnas.86.5.1698.

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 inautism and Williams Syndrome. NeuroReport, 12, 2697–2700. http://dx.doi.org/10.1097/00001756-200108280-00021.

Happé, F. G. E. (1996). Studying weak central coherence at low levels: Children withautism do not succumb to visual illusions. A research note. Journal of Child

Psychology and Psychiatry, 37, 873–877. http://dx.doi.org/10.1111/j.1469-7610.1996.tb01483.x.

Happé, F. G. E., & Booth, R. D. L. (2008). The power of the positive: Revisiting weakcoherence in autism spectrum disorders. The Quarterly Journal of ExperimentalPsychology, 61, 50–63. http://dx.doi.org/10.1080/17470210701508731.

Happé, F. G. E., & Frith, U. (1996). The neuropsychology of autism. Brain, 119,1377–1400. http://dx.doi.org/10.1093/brain/119.4.1377.

Hermens, F., & Herzog, M. H. (2007). The effects of the global structure of the maskin visual backward masking. Vision Research, 47, 1790–1797. http://dx.doi.org/10.1016/j.visres.2007.02.020.

Herzog, M. H. (2007). Spatial processing and visual backward masking. Advances inCognitive Psychology, 3, 85–92. http://dx.doi.org/10.2478/v10053-008-0016-1.

Hochberg, J. E., & McAlister, E. (1953). A quantitative approach to figural goodness.Journal of Experimental Psychology, 46, 361–364. http://dx.doi.org/10.1037/h0055809.

Hochstein, S., & Ahissar, M. (2002). View from the top: Hierarchies and reversehierarchies in the visual system. Neuron, 36, 791–804. http://dx.doi.org/10.1016/S0896-6273(02)01091-7.

Hoppenbrouwers, M., Vandermosten, M., & Boets, B. (2014). Autism as adisconnection syndrome: A qualitative and quantitative review of diffusiontensor imaging studies. Research in Autism Spectrum Disorders, 8, 387–412.http://dx.doi.org/10.1016/j.rasd.2013.12.018.

Hubel, D. H., & Wiesel, T. N. (1968). Receptive fields and functional architecture ofmonkey striate cortex. Journal of Physiology (London), 195, 215–243. http://dx.doi.org/10.1113/jphysiol.1968.sp008455.

Iarocci, G., Burack, J. A., Shore, D. I., Mottron, L., & Enns, J. T. (2006). Global-localvisual processing in high functioning children with autism: Structural versusimplicit task biases. Journal of Autism and Developmental Disorders, 36, 117–129.http://dx.doi.org/10.1007/s10803-005-0045-2.

Johnson, C. R., & Rakison, D. H. (2006). Early categorization of animate/inanimateconcepts in young children with autism. Journal of Developmental and PhysicalDisabilities, 18, 73–89. http://dx.doi.org/10.1007/s10882-006-9007-7.

Jolliffe, T., & Baron-Cohen, S. J. (1997). Are people with autism and Aspergersyndrome faster than normal on the Embedded Figures Test? Journal of ChildPsychology and Psychiatry, 38, 527–534. http://dx.doi.org/10.1111/j.1469-7610.1997.tb01539.x.

Joseph, R. M., Keehn, B., Connolly, C., Wolfe, J. M., & Horowitz, T. S. (2009). Why isvisual search superior in autism spectrum disorder? Developmental Science, 12,1083–1096. http://dx.doi.org/10.1111/j.1467-7687.2009.00855.x.

Kastens, K. A., & Ishikawa, T. (2006). Spatial thinking in the geosciences andcognitive sciences: A cross-disciplinary look at the intersection of the two fields.Geological Society of America Special Papers, 413, 53–76. http://dx.doi.org/10.1130/2006.2413(05).

Kimchi, R. (1992). Primacy of holistic processing and global/local paradigm: acritical review. Psychological Bulletin, 112, 24–38. http://dx.doi.org/10.1037/0033-2909.112.1.24.

Kimchi, R. (2003). Relative dominance of holistic and component properties in theperceptual organization of visual objects. In M. A. Peterson & G. Rhodes (Eds.),Perception of faces, objects, and scenes: Analytic and holistic processes(pp. 235–263). New York, NY: Oxford University Press.

Klinger, L., & Dawson, G. (2001). Prototype formation in autism. Developmentand Psychopathology, 13, 111–124. http://dx.doi.org/10.1017/S0954579401001080.

Koffka, K. (1935). Principles of Gestalt psychology. London: Routledge & Kegan Paul.Köhler, W. (1920). Die physischen Gestalten in Ruhe und im stationären Zustand [Static

and stationary physical shapes]. Braunschweig, Germany: Vieweg.Koldewyn, K., Jiang, Y., Weigelt, S., & Kanwisher, N. (2013). Global/local processing

in autism: Not a disability, but a disinclination. Journal of Autism andDevelopmental Disorders, 43, 2329–2340. http://dx.doi.org/10.1007/s10803-013-1777-z.

Koldewyn, K., Whitney, D., & Rivera, S. M. (2010). The psychophysics of visualmotion and global form processing in autism. Brain, 133, 599–610. http://dx.doi.org/10.1093/brain/awp272.

Kubilius, J., Wagemans, J., & Op de Beeck, H. P. (2011). Emergence of perceptualGestalts in the human visual cortex: The case of the configural superiorityeffect. Psychological Science, 22, 1296–1303. http://dx.doi.org/10.1177/0956797611417000.

Lamme, V. A. F., & Roelfsema, P. R. (2000). The distinct modes of vision offered byfeedforward and recurrent processing. Trends in Neuroscience, 23, 571–579.http://dx.doi.org/10.1016/S0166-2236(00)01657-X.

Lamme, V. A. F., Supèr, H., & Spekreijse, H. (1998). Feedforward, horizontal, andfeedback processing in the visual cortex. Current Opinion in Neurobiology, 8,529–535. http://dx.doi.org/10.1016/S0959-4388(98)80042-1.

Langley, P., Laird, J. E., & Rogers, S. (2009). Cognitive architectures: Research issuesand challenges. Cognitive Systems Research, 10, 141–160. http://dx.doi.org/10.1016/j.cogsys.2006.07.004.

Leeuwenberg, E. L. J., Mens, L., & Calis, G. (1985). Knowledge within perception:Masking caused by incompatible interpretation. Acta Psychologica, 59, 91–102.http://dx.doi.org/10.1016/0001-6918(85)90043-5.

Leeuwenberg, E. L. J., & van der Helm, P. A. (1991). Unity and variety in visual form.Perception, 20, 595–622. http://dx.doi.org/10.1068/p200595.

Leeuwenberg, E. L. J., & van der Helm, P. A. (2013). Structural information theory: Thesimplicity of visual form. Cambridge, UK: Cambridge University Press.

Leeuwenberg, E. L. J., van der Helm, P. A., & van Lier, R. J. (1994). From geons tostructure: A note on object classification. Perception, 23, 505–515. http://dx.doi.org/10.1068/p230505.

P.A. van der Helm / Vision Research 126 (2016) 278–290 289

Maloney, R. K., Mitchison, G. J., & Barlow, H. B. (1987). Limit to the detection of Glasspatterns in the presence of noise. Journal of the Optical Society of America A, 4,2336–2341. http://dx.doi.org/10.1364/JOSAA.4.002336.

Maxwell, C. R., Villalobos, M. E., Schultz, R. T., Herpertz-Dahlmann, B., Konrad, K., &Kohls, G. (2015). Atypical laterality of resting gamma oscillations in autismspectrum disorders. Journal of Autism and Developmental Disorders, 45, 292–297.http://dx.doi.org/10.1007/s10803-013-1842-7.

Milne, E. (2011). Increased intra-participant variability in children with autisticspectrum disorder: Evidence from single trial analyses of evoked EEG. Frontiersin Psychology, 2, 51. http://dx.doi.org/10.3389/fpsyg.2011.00051.

Milner, P. (1974). A model for visual shape recognition. Psychological Review, 81,521–535. http://dx.doi.org/10.1037/h0037149.

Milne, E., Scope, A., Pascalis, O., Buckley, D., & Makeig, S. (2009). Independentcomponent analysis reveals atypical electroencephalographic activity duringvisual perception in individuals with autism. Biological Psychiatry, 65, 22–30.http://dx.doi.org/10.1016/j.biopsych.2008.07.017.

Mitchell, P., Mottron, L., Soulières, I., & Ropar, D. (2010). Susceptibility to theShepard illusion in participants with autism: Reduced top-down influenceswithin perception? Autism Research, 3, 113–119. http://dx.doi.org/10.1002/aur.130.

Molesworth, C. J., Bowler, D. M., & Hampton, J. A. (2008). When prototypes are notbest: Judgments made by children with autism. Journal of Autism andDevelopmental Disorders, 38, 1721–1730. http://dx.doi.org/10.1007/s10803-008-0557-7.

Mottron, L., & Burack, J. (2001). Enhanced perceptual functioning in thedevelopment of autism. In J. Burack, T. Charman, N. Yirmiya, & P. Zelazo(Eds.), The development of autism: Perspectives from theory and research(pp. 131–148). Mahwah, NJ: Erlbaum.

Mottron, L., Burack, J., Iarocci, G., Belleville, S., & Enns, J. (2003). Locally orientedperception with intact global processing among adolescents with highfunctioning autism: Evidence from Multiple Paradigms. Journal of ChildPsychology and Psychiatry, 44, 906–913. http://dx.doi.org/10.1111/1469-7610.00174.

Mottron, L., Burack, J., Stauder, J., & Robaey, P. (1999). Perceptual processing amonghigh-functioning persons with autism. Journal of Child Psychology and Psychiatry,40, 203–212. http://dx.doi.org/10.1111/1469-7610.00433.

Mottron, L., Dawson, M., Soulières, I., Hubert, B., & Burack, J. (2006). Enhancedperceptual functioning in autism: An update, and eight principles of autisticperception. Journal of Autism and Developmental Disorders, 36, 27–43. http://dx.doi.org/10.1007/s10803-005-0040-7.

Navon, D. (1977). Forest before trees: The precedence of global features in visualperception. Cognitive Psychology, 9, 353–383. http://dx.doi.org/10.1016/0010-0285(77)90012-3.

Newell, A. (1990). Unified theories of cognition. Cambridge, MA: Harvard UniversityPress.

Newell, L. C., Best, C. A., Gastgeb, H., Rump, K. M., & Strauss, M. S. (2010). Thedevelopment of categorization and facial knowledge: Implications for the studyof autism. In L. M. Oakes, C. H. Cashon, M. Casasola, & R. H. Rakison (Eds.), Infantperception and cognition: Recent advances, emerging theories, and futuredirections. http://dx.doi.org/10.1093/acprof:oso/9780195366709.003.0011.

O’Riordan, M. A., Plaisted, K. C., Driver, J., & Baron-Cohen, S. (2001). Superior visualsearch in autism. Journal of Experimental Psychology: Human Perception andPerformance, 27, 719–730. http://dx.doi.org/10.1037/0096-1523.27.3.719.

Pellicano, E. (2011). Psychological models of autism: An overview. In I. Roth & P.Rezaie (Eds.), Researching the Autism Spectrum: Contemporary Perspectives(pp. 219–265). Cambridge, UK: Cambridge University Press.

Perreault, A., Gurnsey, R., Dawson, M., Mottron, L., & Bertone, A. (2011). Increasedsensitivity to mirror symmetry in autism. PLoS One, 6, e19519. http://dx.doi.org/10.1371/journal.pone.0019519.

Peters, J. M., Taquet, M., Jeste, S. S., Fernández, I. S., Tan, J., Nelson, C. A., et al. (2013).Brain functional networks in syndromic and non-syndromic autism: A graphtheoretical study of EEG connectivity. BioMed Central Medicine, 11, 54–69.http://dx.doi.org/10.1186/1741-7015-11-54.

Plaisted, K. C. (2000). Aspects of autism that theory of mind cannot explain. In S.Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other minds:Perspectives from developmental cognitive neuroscience (pp. 222–250). New York:Oxford University Press.

Plaisted, K., Swettenham, J., & Rees, L. (1999). Children with autism show localprecedence in a divided attention task and global precedence in a selectiveattention task. Journal of Child Psychology and Psychiatry, 40, 733–742. http://dx.doi.org/10.1111/1469-7610.00489.

Poljac, E., de-Wit, L., & Wagemans, J. (2012). Perceptual wholes can reduce theconscious accessibility of their parts. Cognition, 123, 308–312. http://dx.doi.org/10.1016/j.cognition.2012.01.001.

Pollen, D. A. (1999). On the neural correlates of visual perception. Cerebral Cortex, 9,4–19. http://dx.doi.org/10.1093/cercor/9.1.4.

Pomerantz, J. R., Sager, L. C., & Stoever, R. J. (1977). Perception of wholes and theircomponent parts: Some configural superiority effects. Journal of ExperimentalPsychology: Human Perception and Performance, 3, 422–435. http://dx.doi.org/10.1037/0096-1523.3.3.422.

Pylyshyn, Z. W. (1999). Is vision continuous with cognition? The case ofimpenetrability of visual perception. Behavioral and Brain Sciences, 22,341–423. http://dx.doi.org/10.1017/S0140525X99002022.

Rajendran, G., & Mitchell, P. (2007). Cognitive theories of autism. DevelopmentalReview, 27, 224–260. http://dx.doi.org/10.1016/j.dr.2007.02.001.

Rinehart, N. J., Bradshaw, J. L., Moss, S. A., Brereton, A. V., & Tonge, B. J. (2000).Atypical interference of local detail on global processing in high-functioningautism and Asperger’s disorder. Journal of Child Psychology and Psychiatry, 41,769–778. http://dx.doi.org/10.1017/S002196309900596X.

Rippon, G., Brock, J., Brown, C., & Boucher, J. (2007). Disordered connectivity in theautistic brain: Challenges for the new psychophysiology. International Journal ofPsychophysiology, 63, 164–172. http://dx.doi.org/10.1016/j.ijpsycho.2006.03.012.

Roelfsema, P. R. (2006). Cortical algorithms for perceptual grouping. Annual Reviewof Neuroscience, 29, 203–227. http://dx.doi.org/10.1146/annurev.neuro.29.051605.112939.

Roelfsema, P. R., & Houtkamp, R. (2011). Incremental grouping of image elements invision. Attention, Perception & Psychophysics, 73, 2542–2572. http://dx.doi.org/10.3758/s13414-011-0200-0.

Ropar, D., & Mitchell, P. (1999). Are individual’s with autism and Asperger’sSyndrome susceptible to visual illusions? Journal of Child Psychiatry andPsychology, 40, 1283–1293. http://dx.doi.org/10.1111/1469-7610.00544.

Rubenstein, J. L., & Merzenich, M. M. (2003). Model of autism: Increased ratio ofexcitation/inhibition in key neural systems. Genes, Brain and Behavior, 2,255–267. http://dx.doi.org/10.1034/j.1601-183X.2003.00037.x.

Salinas, E., & Sejnowski, T. J. (2001). Correlated neuronal activity and the flow ofneural information. Nature Reviews Neuroscience, 2, 539–550. http://dx.doi.org/10.1038/35086012.

Scherf, K. S., Luna, B., Kimchi, R., Minshew, N., & Behrmann, M. (2008). Missing thebig picture: Impaired development of global shape processing in autism. AutismResearch, 1, 114–129. http://dx.doi.org/10.1002/aur.17.

Sejnowski, T. J., & Paulsen, O. (2006). Network oscillations: Emerging computationalprinciples. Journal of Neuroscience, 26, 1673–1676. http://dx.doi.org/10.1523/JNEUROSCI.3737-05d.2006.

Sekuler, A. B., & Palmer, S. E. (1992). Perception of partly occluded objects: Amicrogenetic analysis. Journal of Experimental Psychology: General, 121, 95–111.http://dx.doi.org/10.1037/0096-3445.121.1.95.

Shadlen, M. N., & Movshon, J. A. (1999). Synchrony unbound: A critical evaluation ofthe temporal binding hypothesis. Neuron, 24, 67–77. http://dx.doi.org/10.1016/S0896-6273(00)80822-3.

Shah, A., & Frith, U. (1983). An islet of ability in autistic children: A research note.Journal of Child Psychology and Psychiatry, 24, 613–620. http://dx.doi.org/10.1111/j.1469-7610.1983.tb00137.x.

Shah, A., & Frith, U. (1993). Why do autistic individuals show superior performanceon the block design task? Journal of Child Psychology and Psychiatry, 34,1351–1364. http://dx.doi.org/10.1111/j.1469-7610.1993.tb02095.x.

Shepard, R. N. (1981). Psychological complementarity. In M. Kubovy & J. R.Pomerantz (Eds.), Perceptual organization (pp. 279–342). Hillsdale, NJ: Erlbaum.

Simmons, D. R., Robertson, A. E., McKay, L. S., Toal, E., McAleer, P., & Pollick, F. E.(2009). Vision in autism spectrum disorders. Vision Research, 49, 2705–2739.http://dx.doi.org/10.1016/j.visres.2009.08.005.

Snijders, T. M., Milivojevic, B., & Kemner, C. (2013). Atypical excitation–inhibitionbalance in autism captured by the gamma response to contextualmodulation. Neuroimage: Clinical, 3, 65–72. http://dx.doi.org/10.1016/j.nicl.2013.06.015.

Spencer, J. V., & O’Brien, J. M. D. (2006). Visual form-processing deficits in autism.Perception, 35, 1047–1055. http://dx.doi.org/10.1068/p5328.

Stoll, J., Thrun, M., Nuthmann, A., & Einhäuser, W. (2015). Overt attention in naturalscenes: Objects dominate features. Vision Research, 107, 36–48. http://dx.doi.org/10.1016/j.visres.2014.11.006.

Sun, R. (2004). Desiderata for cognitive architectures. Philosophical Psychology, 3,341–373. http://dx.doi.org/10.1080/0951508042000286721.

Sun, L., Grützner, C., Bölte, S., Wibral, M., Tozman, T., Schlitt, S., et al. (2012).Impaired gamma-band activity during perceptual organization in adults withautism spectrum disorders: Evidence for dysfunctional network activity infrontal-posterior cortices. Journal of Neuroscience, 32, 9563–9573. http://dx.doi.org/10.1523/JNEUROSCI.1073-12.2012.

Tallon-Baudry, C. (2009). The roles of gamma-band oscillatory synchrony in humanvisual cognition. Frontiers in Bioscience, 14, 321–332. http://dx.doi.org/10.2741/3246.

Thiele, A., & Stoner, G. (2003). Neural synchrony does not correlate with motioncoherence in cortical area MT. Nature, 421, 366–370. http://dx.doi.org/10.1038/nature01285.

Townsend, J. T., & Nozawa, G. (1995). Spatio-temporal properties of elementaryperception: An investigation of parallel, serial, and coactive theories. Journal ofMathematical Psychology, 39, 321–359. http://dx.doi.org/10.1006/jmps.1995.1033.

Treisman, A., & Gelade, G. (1980). A feature integration theory of attention. CognitivePsychology, 12, 97–136. http://dx.doi.org/10.1016/0010-0285(80)90005-5.

Uhlhaas, P. J., Silverstein, S. M., & Phillips, W. A. (2005). The course and clinicalcorrelates of dysfunctions in visual perceptual organization in schizophreniaduring the remission of psychotic symptoms. Schizophrenia Research, 75,183–192. http://dx.doi.org/10.1016/j.schres.2004.11.005.

Van de Cruys, S., de-Wit, L., Evers, K., Boets, B., & Wagemans, J. (2013). Weak priorsversus overfitting of predictions in autism: Reply to Pellicano and Burr (TICS,2012). i-Perception, 4, 95–97. doi:10.1068/i0580ic.

Van de Cruys, S., Evers, K., Van der Hallen, R., Van Eylen, L., Boets, B., de-Wit, L., et al.(2014). Precise minds in uncertain worlds: Predictive coding in autism.Psychological Review, 121, 649–675. http://dx.doi.org/10.1037/a0037665.

Van der Hallen, R., Evers, K., Brewaeys, K., Van den Noortgate, W., & Wagemans, J.(2014). Global processing takes time: A meta-analysis on local-global visual

290 P.A. van der Helm / Vision Research 126 (2016) 278–290

processing in ASD. Psychological Bulletin. http://dx.doi.org/10.1037/bul0000004.Online First.

van der Helm, P. A. (2000). Simplicity versus likelihood in visual perception: Fromsurprisals to precisals. Psychological Bulletin, 126, 770–800. http://dx.doi.org/10.1037//0033-2909.126.5.770.

van der Helm, P. A. (2004). Transparallel processing by hyperstrings. Proceedings ofthe National Academy of Sciences USA, 101(30), 10862–10867. http://dx.doi.org/10.1073/pnas.0403402101.

van der Helm, P. A. (2010). Weber-Fechner behaviour in symmetry perception?Attention, Perception & Psychophysics, 72, 1854–1864. http://dx.doi.org/10.3758/APP.72.7.1854.

van der Helm, P. A. (2012). Cognitive architecture of perceptual organization: Fromneurons to gnosons. Cognitive Processing, 13, 13–40. http://dx.doi.org/10.1007/s10339-011-0425-9.

van der Helm, P. A. (2014). Simplicity in vision: A multidisciplinary account ofperceptual organization. Cambridge, UK: Cambridge University Press.

van der Helm, P.A. (2015). Transparallel mind: Classical computing with quantumpower. Artificial Intelligence Review, http://dx.doi.org/10.1007/s10462-015-9429-7, Online First.

van der Helm, P. A., & Treder, M. S. (2009). Detection of (anti)symmetry and(anti)repetition: Perceptual mechanisms versus cognitive strategies. VisionResearch, 49, 2754–2763. http://dx.doi.org/10.1016/j.visres.2009.08.015.

van der Vloed, G., Csathó, Á., & van der Helm, P. A. (2007). Effects of asynchrony onsymmetry perception. Psychological Research, 71, 170–177. http://dx.doi.org/10.1007/s00426-005-0002-x.

van Leeuwen, C., Steyvers, M., & Nooter, M. (1997). Stability and intermittency inlarge-scale coupled oscillator models for perceptual segmentation. Journal ofMathematical Psychology, 41, 319–344. http://dx.doi.org/10.1006/jmps.1997.1177.

van Lier, R. J., Leeuwenberg, E. L. J., & van der Helm, P. A. (1997). In support ofhierarchy in object representations. Psychological Research, 60, 134–143. http://dx.doi.org/10.1007/BF00419761.

van Lier, R. J., van der Helm, P. A., & Leeuwenberg, E. L. J. (1994). Integrating globaland local aspects of visual occlusion. Perception, 23, 883–903. http://dx.doi.org/10.1068/p230883.

van Rijsbergen, N. J., & Schyns, P. G. (2009). Dynamics of trimming the content offace representations for categorization in the brain. PLoS Computational Biology,11, e1000561. http://dx.doi.org/10.1371/journal.pcbi.1000561.

VanRullen, R., & Thorpe, S. J. (2002). Surfing a spike wave down the ventral stream.Vision Research, 42, 2593–2615. http://dx.doi.org/10.1016/S0042-6989(02)00298-5.

Vissers, M. E., Cohen, M. X., & Geurts, H. M. (2012). Brain connectivity and highfunctioning autism: A promising path of research that needs refined models,methodological convergence, and stronger behavioral links. Neuroscience &Biobehavioral Reviews, 36, 604–625. http://dx.doi.org/10.1016/j.neubiorev.2011.09.003.

von der Malsburg, C. (1981). The correlation theory of brain function. Internal Report81–2, Max-Planck-Institute for Biophysical Chemistry, Göttingen, Germany.

Wagemans, J., Feldman, J., Gepshtein, S., Kimchi, R., Pomerantz, J. R., van der Helm, P.A., et al. (2012). A century of Gestalt psychology in visual perception: II.Conceptual and theoretical foundations. Psychological Bulletin, 138, 1218–1252.http://dx.doi.org/10.1037/a0029334.

Wang, L., Mottron, L., Peng, D., Berthiaume, C., & Dawson, M. (2007). Local bias andlocal-to-global interference without global deficit: A robust finding in autismunder various conditions of attention, exposure time, and visual angle. CognitiveNeuropsychology, 24, 550–574. http://dx.doi.org/10.1080/13546800701417096.

Watt, R. J. (1987). Scanning from coarse to fine spatial scales in the human visualsystem after the onset of a stimulus. Journal of the Optical Society of America A,10, 2006–2021. http://dx.doi.org/10.1364/JOSAA.4.002006.

Wertheimer, M. (1912). Experimentelle Studien über das Sehen von Bewegung.Zeitschrift für Psychologie, 12, 161–265.

Wertheimer, M. (1923). Untersuchungen zur Lehre von der Gestalt [On Gestalttheory]. Psychologische Forschung, 4, 301–350.

Wolfe, J. M. (2007). Guided search 4.0: Current progress with a model of visualsearch. In W. Gray (Ed.), Integrated models of cognitive systems (pp. 99–119).New York: Oxford.

Wright, B., Alderson-Day, B., Prendergast, G., Bennett, S., Jordan, J., Whitton, C., et al.(2012). Gamma activation in young people with autism spectrum disorders andtypically-developing controls when viewing emotions on faces. PLoS One, 7,e41326. http://dx.doi.org/10.1371/journal.pone.0041326.