These notes provide some further analysis of the processes involved when stimuli are selected for attention by individuals with ASD, with support for the view that anomalies are largely concerned with difficulties in switching or disengaging attention. Some challenge to the perceived over-arching effects of weak central coherence is presented.

The second theme concerns the pattern, in terms of reaction time and level of accuracy, by which presented items are placed into categories … with some difficulty (now supporting the significance of certain impairments linked to weak central coherence) observed among individuals with ASD in dealing with atypical exemplars.

While the long term focus in studies of ASD has been upon the triad issues of communication, social interaction, and imagination, it is noted by Fletcher-Watson et al (2006) that interest has increasingly explored the hypothesis that individuals with ASD can be differentiated from their non-ASD counterparts by the unusual style in which visual attention is organised.

However, attention is not some unitary skill or process but involves a number of components, and it is not clear where the anomalies or impairments are located.

For example, the authors cite evidence that people with ASD may show enhanced visual search capacities, but that they find difficulty in disengaging attention and in shifting the focus of attention from one stimulus or task or topic to another, and in visually processing whole figures (where the focus upon component detail may inhibit the gaining of an overview).

The problem is that such evidence is not consistent, and the authors are able to cite other research which has not identified such differences … although it does seem to be generally accepted that individuals with ASD are proficient at processing the detailed features of a stimulus array.

It is speculated that one distinctive and ASD-specific feature of visual attention concerns the use of semantic information (the meaning or role attributed to the item) or contextual information (the relationship of the item to the surroundings) by which to select particular features of a visual stimulus for attention.

Fletcher-Watson et al go on to describe a recently developed paradigm to assess the influence of these two factors … viz, "change blindness" … which refers to the difficulty experienced by certain individuals in detecting changes in scenes which have been made between the original presentation and a re-presentation after some interruption (which could involve natural shifts in eye movements, or a short-term removal of the stimulus such as a blanking of a monitor screen, etc.).

Change blindness may be observed after an interruption because there is no opportunity to note the activity within the stimulus (the motion cues) which would normally highlight the change taking place.

As set out in the work of Rensink et al (1997), change blindness occurs when there is some limitation in, or inhibition of, the attention required explicitly to perceive the visual stimulus.
It is further noted that attentional resources are finite, and the scope of attention is typically limited to around five items as one looks around the environment; and, if some change is introduced to an aspect of the stimulus to which attention is not being directed, there will be a failure to detect that change.

When it comes to the processes which govern the allocation of attention, one might refer again to Rensink (eg 2001) and his coherence theory (which may be related to but is not the same as the "weak central coherence" hypothesis by which to explain certain ASD symptoms and behaviours).

The Rensink view holds that one forms limited and unstable representations of many items in a scene, but selects for focused attention and more solid representations only a few of the items.

Such items represented in visual short term memory can be transferred into long term memory storage and explain good capacity for recalling a previously viewed scene, but difficulty in detecting small changes when the scene is presented again.

Fletcher-Watson et al raise questions about the influence of higher-order information

(such as context and meaning) on the selection of objects for attention within an overall array or scene. How does one choose those few items (5?) to which attention can be effectively focused ?

Existing evidence (such as that of Hollingworth and Henderson 2000) has supported the significance of semantic information in prioritising the most informative stimuli.

Individual differences will exist in terms of what is seen as important so that, for example, a knowledgeable football fan compared to someone with little interest or knowledge will more readily spot changes in a photograph of an incident in a game. In other words, a critical variable in capturing attention is salience … the extent to which a stimulus or an aspect thereof has some particular relevance to a given person (or is quite unimportant to another person). The speed of a person’s response to a change blindness task will reflect salience and the extent to which the area of change attracts attention.

The study completed by these present authors set out to explore how individuals with ASD, compared to non-ASD control individuals, select items for focused attention.

The first part of the study investigated whether an ASD sample respond in a typical or atypical way in change blindness trials.

The participants viewed alternating images of a tourist scene in which one (inanimate) object changes in one of three possible ways … a different colour, a horizontal or vertical shift in position, or simply in being present or absent.

In addition, via pilot studies involving the opinions of a sample of observers who were not aware of the overall scope of the study, it was possible to gain a measure of the significance of items of the stimulus scene to be changed. The items to which attention was consistently directed were deemed to be "central" and those not mentioned at all were deemed "marginal". (Centrality or marginality were, therefore, attributed according to their perceived importance or defining/semantic significance, not according to their location in the scene).

Accordingly, samples of adolescent and young adult participants with ASD and controls were compared for the accuracy of their responses to changes in central and marginal aspects of the scenes. Further, there was an analysis of reaction times (to determine whether or not the postulated advantage among individuals with ASD in visual search skills would be observed in this kind of task involving change blindness challenges).

The two groups of participants were matched as far as possible for age (ranges 17 to 26 and 17 to 32 respectively), gender, and measured cognitive ability (verbal, non-verbal, and full IQ).

The images were pairs of photographs of holiday scenes where differences of the kind described above were introduced. Half of the photographs involved a change to an item deemed central; half involved a change to a marginal item.

The participants were shown an image on a screen, and they were invited to press the space bar in order to reveal the paired (and altered) image. Participants could switch in this way as often as they wished until they had identified the change, and they were to press the enter key as soon as they had done so.

Scores were recorded in terms of mistakes (pressing the enter key by accident); incorrect responses about the changed item; giving up and choosing to move on to the next pair of images; or producing the correct responses.

The results showed that the two groups of participants responded is a similar way in that marginal changes to a scene took longer to detect than central changes. Thus, ASD individuals were like typically-developing individuals in respect of whether they perceived items as semantically central or marginal.

However, there was a difference between the two groups.

The ease with which the ASD group detected the central changes was similar to that of the controls, but they found the marginal changes much more difficult to spot.

Given that participants were matched for overall ability as well as for age and gender, it was held that this difference in responding was linked to the ASD status.

The ASD group were less ready and less quick to move their attentional focus towards marginal items, and it was further held that the group difference reflected a difficulty among the ASD group in disengaging their attention from the central items and in shifting their attention between items.

Response time and switching data led to the same conclusion … that the slower responding to marginal items indicated a poorer capacity among the ASD group to seek and find changes in marginal items.

In the second part of their study, Fletcher-Watson et al examined the effect of context upon change blindness.

Typically-developing individuals show an attentional bias towards items which are contextually inappropriate (incongruous in the particular setting); and this tendency would be assessed in various groups of participants by a task involving the detection of changes in items which either belong or are out of place in the room in which they are presented (such as a football in the kitchen).

The prediction would be that, for most people, the incongruous items, and changes thereto, would stand out and be detected rapidly.

The concept of weak central coherence would suggest that individuals with ASD have a bias towards local processing and a focus upon details, with corresponding problems in integrating information into a coherent whole.

This difficulty is likely to be observable in contextual processing tasks where an ASD sample, compared to controls, would be predicted to be less influenced by the contextual appropriateness of items within a setting.

However, recent studies cited (such as that of Lopez and Leekam 2003) have indicated that, contrary to the predictions associated with weak central coherence, samples of individuals with ASD have demonstrated an intact capacity to use context in (visual) processing of this kind.

In other words, the postulated effects of weak central coherence might not apply to all kinds of processing demands, or might be compensated by an ability to take advantage of contextual cues (of the kind available in change blindness tasks).

In this second study, Fletcher-Watson et al used pairs of photographs of rooms in ordinary homes, with the paired images differing in respect of the presence or absence of an object. In half of the pairs, the object in question was appropriate to the setting, and in the other half it was incongruous. (Categorization as appropriate or inappropriate was determined by pilot studies where individuals rated the likelihood of appearance in the setting of the target objects.)

The question under investigation concerned how context/setting affects the time taken by their participants with ASD and the controls to detect changes in items within a scene.

The results indicated that the two groups showed very similar responses to the two types of image presented. Response time and switching behaviour indicated that all the participants detected changes to inappropriate items more quickly, suggesting that both groups were influenced in their selective attention by contextual information.

This finding contradicted the effects that would have been predicted from weak central coherence … ie individuals with ASD appear able, in this kind of task at least, to make use of the overall context.

Meanwhile, the authors speculated that the results depended less upon the effects of context than upon the participants’ experience. However, even using existing experience to identify appropriate and inappropriate items in a scene was held to require the integration of different bits of information to identify the nature of the overall context (a kitchen) in the first place. This would suggest that individuals with ASD are able to identify and use contextual information in deciding where to focus their attention.

In their overall discussion and conclusions, the authors restated the finding that samples of ASD participants and controls seemed to select the same items in a scene for focused attention … central rather than marginal items, and inappropriate rather than appropriate items. This is a challenge to that view of ASD as being characterised by a difficulty in perceiving and using context.

Similarly, there is a refutation of the prediction that visual search skills would elicit faster responding from the ASD group.

Nevertheless, the individuals with ASD could be differentiated from controls in respect of their greater difficulty in responding to marginal items and in switching behaviours.
The explanation offered concerned a problem in disengaging attention, and in moving the focus of attention between items.

The lack of group effects in the second part of the study could be taken as support for the disengagement hypothesis in that no differences were observed because there was no need to disengage attention when the item in question was always a prominent feature of the scene, whether appropriate or not.

Some limitations of the study were acknowledged, such as the issue of cueing. The fact that the participants were asked to look for changes might have altered their "normal" attentional focus, so that the results reflect what individuals are able to do rather than what they typically choose to do.

Nevertheless, the results are seen as suggestive of an apparently normal attentional processing of visual input (at least in the kinds of task currently used).

Where atypicalities were noted, the underlying impairment could be less concerned with selective attention per se and more with the disengaging and shifting of attention once a pattern has been established.

(The present writer – MJC – would comment that such findings appear to match converging evidence, from studies of social skill or theory of mind development, that children can benefit from direct instruction in the use of certain skills, or demonstrate a greater prowess than they spontaneously choose to exercise and which would result from any reliance upon incidental learning rather than explicit teaching.)

The paper by Gastgeb et al (2006) is prefaced by the comment that the emphasis in autism research has been upon the social deficits (given that they are critical for a diagnosis), but increasing attention has been directed to cognitive anomalies or impairments with the emergence of hypotheses concerning deficits in executive functioning, central coherence, or aspects of attention.

The current authors suggest that one aspect of cognitive processing which has been subject to very little evaluation is categorisation despite its importance in reducing demands upon memory and the opportunity afforded to focus upon what is salient among stimuli and to ignore irrelevant details.

The hypothesis is that, if individuals with ASD have some impairment in their capacity to make categorisations, there is at least a partial basis for the social, communicative, and behavioural impairments which are characteristic of this condition. A limitation in the means of organising input and of applying some order to the environment could underlie the experience of being overwhelmed by incoming stimuli or of failing to attune to the communications of other people, resulting in a retreat from the social milieu into one’s own private (and safe ?) world.

Such a hypothesis does not belittle the likely deficits in theory of mind or in language, but the suggestion is that what abilities do exist can become rapidly over-stretched by the failure automatically to categorize the incoming information.

Some (natural) categories like " birds" are not organised according to strict criteria but have somewhat fuzzy boundaries. "Typicality" is also a relevant construct in that some members of a category are more representative, and better examples, of a category than others. A robin could be cited as a good exemplar and more typical of the "bird" category, but a penguin could be cited as a poor exemplar of the category and, thus, less typical.

Individuals tend to agree about which members of a category are good and poor exemplars of a category; and reaction times in respect of assigning items to a given category are faster for good exemplars (more typical of the category) than poor exemplars. This pattern reflects memory storage in that the more typical exemplars are easier to retrieve from the memory store than less typical.

(Further, evidence shows [Barrett 1995] that children learn the names of typical examples more readily and quickly than they learn the names of atypical ones .)

While there is considerable information about this topic relating to typically-developing individuals, there is little evidence relating to those with autism; and inconsistent findings have emerged from what research has been completed.

Some early studies suggested that the ability to form categories is intact in ASD, but were criticised in terms of using simple and unarguable categories (such as colour) and leaving unanswered any questions about the capacity of individuals with autism to deal with more complex categories or those whose features are less perceptually obvious.

Further, inconsistency of findings could be linked to a failure to take account of typicality with the hypothesized possibility that individuals with ASD will find greater difficulty as the exemplars of a category become less typical.

This hypothesis has received support from some recent studies (such as Minshew et al 2002) who found that high functioning autism seems to be linked with adequate capacity to organise information according to basic rules, but difficulty is experienced when the task demands the eliciting of concepts (categories) from complex information.

The current authors noted also that there is no evidence available concerning the development of processing/categorising skills over time from childhood into adulthood, so that it is not clear whether these categorisation anomalies apply to both children and older people, and whether improvements are noted with age and experience.

Accordingly, the study by Gastgeb et al themselves examined both categorisation processes across development from childhood to adulthood and the significance of typicality or task difficulty in seeking to categorise stimulus information.

In the first part of the study, children and adolescents were compared. The child sample covered the age range from 9 to 12 years; and the adolescent sample covered the range 13 to 16 years. In each case, the target sample of individuals with high functioning autism was matched with a typically developing control group on age, verbal ability, non-verbal ability, and full-scale ability.

Gender matching was completed as far as possible. In the younger samples, the M : F ratios were 27 : 1 for the target group and 19 : 5 for the controls; the respective ratios in the older samples were 15 : 5 and 16 : 3.

The task required the participants to examine a monitor and to utilise just two keys on a keyboard marked "true" and "false". All other keys were concealed.

The stimulus categories consisted of 2 "natural" and animate items, dogs and cats; and 2 inanimate items, couches and chairs.

(A pilot study involving a large sample of students had produced mean ratings of typicality of the items using a 7-point scale.)

The participants were given an auditory stimulus (the name of the item to follow … dog or cat; couch or chair) then shown a picture of an object. The task was to decide whether or not the pictured object belonged to the category cited, and to register their decision by pressing either the true or false key as quickly as possible.

In 25% of the trials, the category and the picture did not match. In the rest of the trials, the picture and category did match, and there were equal numbers of typical, somewhat typical, and atypical exemplars.

The results showed that the reaction times of both the younger groups were slower for the somewhat typical and the atypical items, with the effect more evident among the children with autism. Thus, typicality had a greater influence over the reaction times of children with autism than over those of the control children.

Among the older group (the adolescent samples), the reaction times of the group with autism were not significantly slower than those of controls for the somewhat typical items. The implication was that, with experience, individuals with autism can achieve as accurate a categorisation of somewhat challenging exemplars as that achieved by control adolescents.

However, when it came to atypical items, the reaction times of the adolescents with autism were significantly slower than those of the control group (in all but the "dog" category of items). It was speculated that children and adolescents with autism could be differentiated in terms of the strategies used to make categorisations.

The second part of the study concerned whether or not the differences observed would apply to adult samples, with the possibility to explore that greater experience would increase the accuracy and fluency of categorisation among individuals with autism even with atypical (complex) items.

This time, the participants were adults with a mean age of 24 years for the target group and 23 years for the control group. The two groups were matched according to the same criteria as in the first part of the study.

The task, procedures, and dependent measures of reaction time and accuracy were also similar.

The results indicated that adults with autism, like their child and adolescent counterparts, responded significantly more slowly than controls to atypical exemplars, and were a little less accurate.

In other words, experience did improve the processing of somewhat typical items, but adults with autism did not reach the level of accuracy achieved by control adults across the whole range of typicalities.

In their overall discussion, the authors describe the degree of improvement that individuals do show with regard to categorisation over time; but repeat that individuals with autism continue to be differentiable from controls across the time span from childhood to adulthood in terms of poor or atypical exemplars of a category. In other words, high functioning autism does not limit the capacity to process simple and typical objects, but does involve a weaknesses when the material is more complex or atypical.

The authors return to the anomaly whereby the observed pattern applied to three of the categories of items, but not to the category of "dogs".

They put forward the possibility that the lack of an effect here was a function of the particular set of exemplars used … ie… the supposedly less typical or atypical examples were still clearly and rapidly identifiable as dogs, and none appeared truly different from the general perceptual characteristics that define dogs. Thus, no additional and time-consuming processing was required for allegedly atypical exemplars.

Alternatively, it was considered possible that the exemplars used were sufficiently familiar to all participants so that the task was not about categorisation but about simple recognition.

A more significant question was seen to lie in the finding that improvements were shown with age in dealing with somewhat typical items by the individuals with autism, but the processing of atypical items remained relatively inefficient even in adulthood.

One hypothesis is that the efficiency of categorisation is a matter of a person’s familiarity with the category, and ASD-control differences may reflect differences in experiences.

On the other hand, it is possible to cite evidence that typicality effects are relatively independent of familiarity … rare but typical-looking exemplars do not require longer processing time. Further, in this particular study, there is no reason to suppose that the participants with autism and controls differed in respect of their familiarity with cats, chairs and couches … one might even be able to argue that individuals with autism spend less time attending to social stimuli and correspondingly more time attending to non-social stimulus objects.

A further hypothesis is that atypical exemplars are subject to a different type of processing.

They require more processing time because they do not have the same basic and observable characteristics as simple exemplars. The additional processing time could be used for more perceptual analysis, or for additional semantic analysis in determining the category in which the items should be placed.

The authors refer to these additional analyses as " subordinate perceptual processes " and describe three subdivisions.

Firstly, one might examine an item that does not seem readily to fall into a given category (eg neither quite chair nor couch), so that it is necessary to use further and increasingly subtle quantitative information

Secondly, simple quantitative features like "short" vs "not short" may not suffice so that one has to go on to a comparison between the presented item and stored memories of examples of the categories.

Thirdly, it may prove necessary to go to a further "level" which involves the examination of multiple features and to give weight to such features in coming to a decision about categorisation.

The authors cite evidence from existing research findings that individuals with autism have difficulty with subordinate perceptual processes …. although such findings are usually discussed in the context of global/holistic perceptual processing which can prove a differentiating factor between individuals with autism and typically-developing individuals.

It is held that ASD is marked by a lack of capacity to derive prototypes or average representations of the features of a category; and that processing (of faces, for example) on the part of individuals with ASD commonly involves an emphasis upon single feature or parts rather than more subtle comparisons.

In other words, the results produced in this present study could be attributable to actual differences/deficits in processing capacities, notably in these subordinate perceptual processes, among the individuals with high functioning autism.

A further question concerns what underlies the improvement that is noted in respect of categorising poor exemplars.

One relevant finding in existing research on face perception is that normally developing individuals gradually shift from a predominant reliance on part-processing to a more holistic processing. As older children or adults, individuals increasingly have available both types of processing capacity and utilise each according to the task and their developmental level.

So, some comparisons can be very simple and involve comparing basic features (one man with a beard, the other man without); but comparisons of relatively similar faces require more holistic analysis.

With time, individuals gain greater proficiency and can use subordinate perceptual processing more effectively in examining atypical stimulus items. However, individuals with autism never reach the same level of efficiency as that shown by their non-autistic counterparts.

This kind of thinking provides a challenge to the view that an early lack of interest and attention to faces that underlies the lack of subordinate or holistic face processing capacities. Instead of a social deficit, the underlying issue may be an actual processing deficit with the ASD weakness more evident as the task requires greater involvement of subordinate perceptual processing.

The final point made by Gastgeb et al is that there are similarities between their current thoughts based upon the findings in this two-part study and those theories which suggest that individuals with autism and ASD are more attentive to local details at the expense of attending to more global information …. weak central coherence.

This study has added to the thinking by highlighting those underlying processes with respect to perceptual recognition and categorisation that could well be different in individuals with autism and ASD.

The difference/difficulty would cover social and non-social domains, and may be more profound in individuals younger than the age groups involved in this study or in those with a lower overall level of functioning …. all of which could prove reflective of a pervasive cognitive deficit rather than of a more specific social-perceptual deficit.

(One further point occurring to the present writer – MJC – [ as it usually does ] is that the use of labels, such as ASD, can be helpful in defining the general category of need under discussion, but can also be misleading if one tends to attribute all the [negative or positive] attributes relating to that label to all the individuals to whom the label can legitimately be applied. Individual strengths and weaknesses may show wide variation, and the capacities that do exist may be differentially observable according to the precise nature of the task and the context, including the wording and specificity of directions and availability of cues. Accordingly, it remains preferable to avoid statements which begin …. " The autistic child does, is, has, ……" and instead to use statements which begin … " This autistic child does, is, has …. " )

* * * * *

Barrett M. 1995 Early lexical development. In P. Fletcher and B. Macwhinney (Eds) The Handbook of Child Language. Oxford : Blackwell

Fletcher-Watson S., Leekam S., Turner M., and Moxon L. 2006 Do people with autistic spectrum disorder show normal selection for attention ? British Journal of Psychology 97 537-554

Gastgeb H., Strauss M., and Minshew N. 2006 Do individuals with autism process categories differently ? The effect of typicality and development. Child Development 77(6) 1717-1729

Hollingworth A. and Henderson J. 2000 Eye movements, visual memory, and scene representation. Michigan State University : Eye Movement Laboratory. Technical Report 5 1-17

Lopez B. and Leekam S. 2003 Do children with autism fail to process information in context ? Journal of Child Psychology and Psychiatry 44(2) 285-300

Minshew N., Meyer J., and Goldstein G. 2002 Abstract reasoning in autism. Neuropsychology 16 327-334

Rensink R., O’Regan K., and Clark J. 1997 To see or not to see. Psychological Science 8(5) 368-373

Rensink R. 2001 Change blindness : implications for the nature of visual attention.

In M. Jenkin and L. Harris (Eds) Vision and Attention. New York : Springer