what hypothesis can you put together after reading this article

Superior visual search in adults with autism

MICHELLE A. O'RIORDANUniversityofCambridge,UK

ABSTRACTRecent studies have suggested that children with autism perform better than matched controls on visual search tasks and that this stems from a superior visual discrimination ability. This study assessed whether these findings generalize from children to adults with autism. Experiments 1 and 2 showed that, like children, adults with autism were superior to controls at searching for targets. Experiment 3 showed that increases in target-distractor similarity slowed the visual search performance of the control group significantly more than that of the autism group, suggesting that the adults with autism have a superior visual discrimination ability. Thus, these experiments replicate in adults previous findings in children with autism. Superior unique item detection in adults with autism, stemming from enhanced dis- crimination, is discussed in the light of the possible role of stimulus processing disturbances in the disorder in general.

Introduction

autism 2004 SAGE Publications and The National Autistic Society Vol 8(3) 229-248; 045219 1362-3613(200409)8:3

KEYWORDS

autism; developmental profile; visual dis- crimination; visual search

Autism is characterized by profound impairments in social and communi- cative behaviour together with the presence of repetitive behaviour (Kanner, 1943; Rutter, 1983; Wing and Gould, 1979). However, although these characteristics are necessary and sufficient to define autism, aspects of non-social stimulus processing also appear to be abnormal (Hayes, 1987; Kanner, 1943; NASC, 1978). A recent example of unusual stimulus pro- cessing is the finding of superior performance of children with autism on visual search tasks (O'Riordan et al., 2001; Plaisted et al., 1998a). In a visual search task, the participant is asked to indicate the presence or absence of a pre-specified target by pressing one of two buttons (Duncan and Humphreys, 1989; Treisman and Gelade, 1980; Wolfe et al., 1989). The target is present on 50 percent of the trials, and the number of distractors in each display varies from trial to trial. Thus, in each trial the participant knows what the target will be but knows neither whether the target will be

229

www.sagepublications.com DOI: 10.1177/1362361304045219

AUTISM8(3)

present, nor how many distractors will be presented. Visual search tasks are commonly distinguished into feature and conjunction search tasks. In feature search tasks the target differs from all simultaneously presented dis- tractors in terms of a single feature (e.g. a red F target hidden among green X and red T distractors). In conjunctive search tasks the target shares one feature with one set of distractors and another feature with the alternative distractor set and therefore is only uniquely defined by the combination of its component features. This may be a combination of features from different dimensions (e.g. a red X hidden among red T and green X dis- tractors) or a combination of features within a dimension (e.g. an R target hidden among P and Q distractors). The typical profile of performance is that individuals with autism are overall faster and show less increase in reaction time with increasing display size in feature compared with con- junctive tasks. Children with autism have been found to be superior at searching for a target regardless of whether it is defined by a single feature or a combination of features, as long as ceiling effects do not mask the difference (O'Riordan et al., 2001; Plaisted et al., 1998a).

There are several other reports of unusual stimulus processing in autism, which might be described as superior unique item detection. For example, children with autism perform better than normal on the Embedded Figures Task (Shah and Frith, 1983) and there are frequent reports of acute atten- tion to minor features or changes in the environment in autism (Hayes, 1987; Kanner, 1943; NASC, 1978). As considerably more research has been conducted on the mechanisms underlying normal visual search perform- ance than on the mechanisms involved in performing Embedded Figures and Block Design Tasks (Duncan and Humphreys, 1989; Treisman, 1988; Treisman and Gelade, 1980; Treisman and Gormican, 1988; Treisman and Sato, 1990; Wolfe, 1994; Wolfe et al., 1989), visual search is a more useful tool for investigating unique item detection in autism.

It is now well established that the critical factor determining search rate is the similarity of the target and the distractors (Duncan and Humphreys, 1989; Treisman, 1988; Wolfe et al., 1989). Under conditions of high simi- larity the target is difficult to find and hence performance is slow. Con- versely, under conditions of low similarity target detection is easy and rapid. This factor is thought to account for the relative ease of target detection in feature compared with conjunctive tasks, as target distractor similarity is usually much higher in the latter relative to the former.

That discriminability is the critical factor underlying search perform- ance raises the possibility that the reason for superior visual search in children with autism is an enhanced discrimination ability. A recent study tested this hypothesis by comparing the performance of children with and without autism on visual search tasks in which target-distractor similarity

230

O'RIORDAN:SUPERIOR VISUAL SEARCH

was systematically manipulated (O'Riordan and Plaisted, 2001). In line with the predictions, although all children were slowed by increases in target-distractor similarity, the typically developing children were affected significantly more than the autistic group. Thus, it was concluded that children with autism have a superior visual discrimination ability. Such ability might also underlie the superior performance of children with autism on the Embedded Figures Task and the unusual ability to notice minor features or changes in the environment.

A superior visual discrimination ability in autism could also have wider implications for the autistic disorder. More specifically, disturbances in such basic stimulus processes, which are integral to all cognition and behaviour, could at least contribute to the other behavioural manifestations of autism (O'Riordan et al., 2001). However, before we can determine the role of enhanced discrimination ability in the autistic disorder as a whole it is essential to first fully elucidate the precise nature of this abnormality.

Recent work has attempted to examine further the nature of enhanced discrimination in children with autism. For example, enhanced discrimi- nation ability appears to stem from differential low-level perceptual pro- cessing rather than from a differential attentional mechanism (O'Riordan, 2000), and it might also be present in the auditory modality (O'Riordan and Passetti, manuscript in preparation). However, one important issue, which remains to be addressed, is the developmental profile of enhanced discrimination and consequent superior unique item detection in autism. Most research looking at unique item detection and discrimination in autism has been conducted in children. Understanding the development of each feature of autism is essential if we are to determine the roles that they play in the natural history of the disorder. Further, the evolution of such features may provide clues about the fundamental psychological mechan- isms disrupted in autism. Finally, describing continuities and disconti- nuities of symptoms between childhood and later ages may have practical value in clarifying diagnosis and possibly identifying subgroups within the wider category (Burack, 1992; Burack et al., 2000; Piven et al., 1996). The present study focuses on the development of enhanced discrimination and subsequent unique item detection in autism. More specifically, this study investigates the performance of adults with and without autism on a series of visual search tasks. Experiment 1 compared performance of adults with and without autism on feature and conjunctive search tasks. In order to avoid the possible confound of ceiling effects, experiment 2 assessed performance on a difficult feature search task. In experiment 3 perform- ance was compared on two visual search tasks in which target-distractor similarity was systematically manipulated to test the enhanced discrimi- nation hypothesis.

231

AUTISM8(3)

Experiment 1

Experiment 1 compared the performance of adults with and without autism on within-dimension feature and conjunctive search tasks to see if the superiority of children with autism on visual search tasks generalizes to adults with autism.

Method

ParticipantsTwo groups of adults participated: a group of 10 adults with high-functioning autism, and a group of 10 developmentally normal adults. All adults in the group with autism had been diagnosed using the Autism Diagnostic Instrument-Revised (ADI-R: Lord et al., 1994). The ages of both groups ranged from 17 years to 27 years. All subjects were assessed for general ability using the Wechsler Abbreviated Scale of Intelligence (WASI: Wechsler, 1999). The two groups did not differ from one another in terms of age (t(18) = 0.08,p= 0.93) or general ability (t(18) = 0.06,p= 0.95). The mean age of both groups was 22.0 years (SD for both 3.6 years); the mean IQ of the control group was 111.1 (SD 15.7); the mean IQ of the autism group was 111.6 (SD 13.7).

ApparatusThe stimuli were generated by an Acorn Risc PC and displayed on a 14 inch colour monitor. Participants responded by pressing one of two keys on the keyboard (the '.' key with the right hand for target present responses, or the 'z' key with the left hand for target absent responses). In order to prevent irrelevant keys being pressed, the keyboard was covered by a hard black plastic cover that had two openings to allow access to just the response keys.

StimuliEach stimulus display consisted of 5, 15 or 25 elements (i.e. letters) arranged in an imaginary 16.8 cm by 16.8 cm square (approxi- mately 33 visual angle) centred around a central fixation point (a hash). Each element measured 0.5 cm by 0.5 cm, subtending approximately 1.0 of visual angle horizontally and 1.0 vertically. The minimum distances between elements in any display were 0.7 cm between positions in a row and 0.7 cm between positions in a column, and items were positioned randomly across the screen rather than in positions in an imaginary grid. Henceforth, the term 'display size' refers to the number of elements in the display, not to the physical boundaries of the display, which remained fixed throughout.

Display elements were constructed from combinations of the three components O, \ and P. For example, an R was constructed from the combi- nation of P and the diagonal line and the Q was constructed from the

232

O'RIORDAN:SUPERIOR VISUAL SEARCH

combination of O and the diagonal line. All items were red in colour. In the feature search task, the target was uniquely defined by the shape dimen- sion (i.e. an N target among P and Q distractors). In the conjunction search task, each distractor shared one feature with the target (i.e. an R target among P and Q distractors). Here the conjunctive target was uniquely defined by the combination of component shapes (i.e. a within-dimension conjunction) rather than a combination of separable dimensions such as colour and form (i.e. a between-dimension conjunction). Superior visual search in children with autism has been found on both within- and between-conjunctive search tasks (O'Riordan, 1998) and the performance of children with and without autism on a series of search tasks revealed this within-dimension conjunction search to be difficult. Thus, the within- dimension conjunction task was selected so that any potential autistic superiority in these high-functioning adults might be revealed. (See Figure 1 for a diagrammatic representation of the within-dimension conjunctive target and how it shared features with the distractor items.)

DesignThe experiment comprised two different search tasks (feature or conjunction task). Each task contained two fully crossed factors: display size (5, 15 or 25 items) and probe (target present or target absent), yielding six possible display types. There were 10 trials for each of these display types, giving 60 trials per task. Trials were randomized within blocks of 30 for each search task, with equal representation of all experimental factors in each block.

,EIJH=?JHI

6=HCAJ

PP

2=HJI B JDA @EIJH=?JHI ?J=EA@ E JDA J=HCAJ

+PP

Example of how the component parts of the target item are contained within each of the distractor items

233

Figure 1

AUTISM8(3)

The order of positive and negative trials and of different display sizes were randomized within each session; thus the participant knew what the target was, but did not know whether a target would be present or what the display size would be on any trial. The participants performed a binary choice reaction time task indicating present or absent for the single pre- specified target, via button presses on each trial.

ProcedureEach participant was tested on both the conjunctive and feature search tasks in a single session and the tasks were separated by a 5 minute rest period. The order in which tasks were presented was counterbalanced across participants within each group. The participant was informed of the target to search for and that certain keys were to be pressed depending on whether the target was present or absent. Prior to each task participants were given a block of 12 practice trials involving the stimuli for that task, with the experimenter's instruction and assistance. Following these practice trials (immediately prior to the test trials) participants were instructed to respond as quickly as possible and with as few errors as possible.

On each trial the sequence of events was as follows: a fixation hash mark was presented on an otherwise blank screen for 500 ms. The search display was then presented, at which point the timing was initiated. The search display remained on for 10 seconds or until the participant responded, whichever was the sooner. If the former occurred, the phrase 'You were too slow' appeared in the centre of the screen for 500 ms, followed by the presentation of the central hash for 500 ms indicating the onset of the next trial. If the correct response was made, the next trial was initiated. If an incorrect response was made, a tone sounded as an indication of the error. An incorrect trial was followed by a dummy trial, the response to which was not recorded. This was to allow the participant to recover from an error. On the rare occasion that a button press occurred before the search display appeared on the screen the phrase 'You pressed too soon!' was displayed at the centre of the screen for 500 ms. Then the task resumed, as before starting with the trial that had been interrupted by the premature response.

Results

Except where otherwise stated, a significance level ofp< 0.05 was adopted for all statistical comparisons in this experiment and likewise for those that follow. Performance of the two groups of participants was compared on the feature search task and the conjunctive task. For each participant, reaction time (RT) data (for correct trials) and error data were averaged for the 10 trials for each particular combination of task, display size and probe. The mean RT data and the error data were initially analysed using a mixed ANOVA, with one between-subject factor of group (control or autistic), and

234

O'RIORDAN:SUPERIOR VISUAL SEARCH

three within-subject factors of task (feature or conjunctive), probe (present or absent) and display size (5, 15 or 25).

Figure 2 displays the mean RT data as a function of group, task, probe and display size. The left-hand panel shows the pattern of results for target present trials in both search tasks, for the normal and autistic participants. This graph suggests that in control participants, conjunctive target detec- tion time increased more substantially with display size than for feature search; furthermore, RT appeared overall much slower in the conjunctive than the feature task. This pattern of results in the normal group replicates the standard pattern for easy feature versus hard conjunctive search (e.g. Treisman and Gelade, 1980).

While the participants with autism showed a similar tendency in the feature task, they appeared to performbetterthan normal in the conjunctive task. In particular, the increase in RT with display size in the conjunctive task was not as dramatic for the participants with autism as it was for the control participants, and likewise the overall mean RT was not as high in this task for the autistic group. The right-hand panel shows the pattern of results for target absent trials in both search tasks, for the normal partici- pants and those with autism. The same pattern is apparent as for the target present trials.

4000

3000

RT (ms)2000

1000

4000

3000

2000 RT (ms)

1000

Target present

Target absent

Control-Feature Control-Conjunctive Autistic-Feature Autistic-Conjunctive

00

5 15 25 5 15 25 Display size Display size

Figure 2Reaction time (RT) data from experiment 1: each data point shows mean RT SEM

235

AUTISM8(3)

Reaction time analysisANOVA revealed that the group by task inter- action was significant (F(1, 18) = 12.52). Simple effects revealed that while there was no difference between the performance of the two groups in the feature task (F< 1.0), the control group was significantly slower in the con- junctive task (F(1, 25) = 10.79).

The group by display size interaction was also significant (F(2, 36) = 4.64). Simple effects revealed that the group with autism was significantly faster than the controls when searching the largest display of 25 items but not when searching 15-item or 5-item displays (F(1, 28) = 8.56,F(1, 28) = 3.52,p= 0.071, andF< 1.0 for display sizes of 25 items, 15 items and 5 items respectively).

There was also a three-way interaction between group, task and display size (F(2, 36) = 4.09). Separate analysis of the data from each task revealed that the source of this interaction was the presence of a group by display size interaction in the data from the conjunctive (F(2, 36) = 5.41) but not the feature task (F< 1). In the conjunctive task individuals with autism were slowed less than controls by increases in display size, but the effect of increasing display size on each group was comparable in the feature task. Thus, the superiority of individuals with autism at larger display sizes was confined to the conjunctive target task.

Other findings applied across groups. These main effects and inter- actions all replicate standard visual search results (e.g. Duncan and Humphreys, 1989; Treisman and Gelade, 1980). There were significant main effects of task (F(1, 18) = 157.31), probe (F(1, 18) = 198.15) and display size (F(2, 36) = 169.26). There were also significant interactions between task and display size (F(2, 36) = 53.03); between probe and display size (F(2, 36) = 61.15); between task and probe (F(1, 18) = 24.49); and between task, probe and display size (F(2, 36) = 4.56).

Accuracy analysisFigure 3 shows the pattern of accuracy as a function of group, task, probe and display size. There was no effect of group in the analysis of accuracy (F< 1.0); the control group made 4.4 percent errors while the autistic group made 3.7 percent errors. The only interaction was between group and display size (F(2, 36) = 3.27). Simple effects revealed that the source of this interaction was that the autistic group was more accurate than controls at display sizes of 15 items (F(1, 50) = 4.22) but there was no difference between the accuracy of the groups at other display sizes (F< 1.0, andF(1, 50) = 1.38,p= 0.246, for display sizes of 5 and 25 items respectively).

Therefore, although the accuracy of the two groups was generally similar, the autistic group was overall more accurate than the control group and significantly more accurate at display sizes of 15 items. Importantly, the

236

Target present

Target absent

O'RIORDAN:SUPERIOR VISUAL SEARCH

Control-Feature Control-Conjunctive Autistic-Feature Autistic-Conjunctive

Accuracy (%)

100

95

90

85

80

75

70

100

95

90

85

80

75

70

Accuracy (%)

5 15 25 5 15 25 Display size Display size

Figure 3Accuracy data from experiment 1: each data point shows mean percentage correct SEM

faster reaction time of the autistic group was not accompanied by poorer accuracy rates and thus can be taken to indicate search differences rather than merely detection criterion differences.

Once again in accordance with standard findings (e.g. Duncan and Humphreys, 1989; Treisman and Gelade, 1980) this analysis also revealed significant main effects of task (F(1, 18) = 33.00), probe (F(1, 18) = 37.32) and display size (F(2, 36) = 5.73); and interactions between task and probe (F(1, 18) = 20.69) and probe and display size (F(2, 36) = 7.98).

In summary, this experiment showed that, like children with autism, adults with autism are superior to matched controls at conjunctive visual search. However, there was no difference between the performance of the groups in the feature search task. One possibility is that this feature search task was too easy and that ceiling effects masked any possible group differ- ences here. Alternatively, superior visual search in adults with autism might be confined to search for conjunctive targets.