These results indicate abnormal modulation not only at short SOAs as predicted by the hypothesis of nonspecific sensory gating, but also at longer SOAs. This latter finding is remarkable given the relatively intact behavioural performance of persons with autism at these longer SOAs. Physiological abnormality at longer SOAs is consistent, however, with electrophysiological evidence showing lack of frontal negativities and diminished P3b response even where behavioural performance is unimpaired, and suggests that the autistic brain implements attentional processing via a mechanism different than that of the normal brain.
When persons with autism had only a short amount of time in which to re-orient attention (short SOAs), they performed less accurately by behavioural measures, and electrophysiological measures of SSVEP and background EEG amplitudes indicated a failure of the two hemispheres to operate independently: in both hemispheres the SSVEP was much more responsive to rightward shifts than to leftward shifts, and in both hemispheres the background EEG decreased during rightward shifts and increased during leftward shifts. These electrophysiological results are consistent with other reports of loss of hemispheric specialisation in autism [Dawson & Lewy 1989; Chiron & al. 1995], and suggest that in autism a generalised arousal substitutes for the normal brain's more fine-grained, spatiotopically specific pattern of activation. This high level of generalised arousal may account for the autistic subjects' combination of low accuracy and especially quick response times at short SOAs: if a person with autism is able to distinguish the target from the rest of the flood of sensory data, the response to that target occurs with a very low threshold.
Activation patterns at longer SOAs suggest a lack of functional specificity of attention, in addition to the aforementioned failure of spatial specificity. At longer SOAs persons with autism had more time to re-orient, and performed more accurately but still not quite at the level of the controls. The lack of modulation of background EEG and the high variability in the SSVEP under these conditions suggest that irrelevant processing is not appropriately supressed and that relevant stimuli are not reliably selected. It thus becomes difficult for the person with autism to pick out and to act upon task-relevant stimuli. The variability in the SSVEP data is interesting in light of suggestions that autism may be viewed functionally as a sort of dysmetria of thought [Courchesne & Allen 1997], in which reciprocal connections with subcortical structures fail to tune cortical excitability appropriately for the current sensory inputs and behavioural set.
Group social activities present many simultaneous signals, between which attention must be switched effortlessly and rapidly. In a conversation, information is transferred not only in the text of the dialogue but also in tone of voice (auditory), facial expression (visual), gesture (peripheral visual), and reference to third parties or to external objects or events (peripheral auditory and visual). Without the ability to integrate rapidly stimuli in separate modalities and spatial locations, the everyday task of constructing a coherent mental narrative of the external world becomes nearly impossible.
Many persons with autism have reported episodes of sensory overload that prevent their processing any inputs at all. Sounds can become painfully intense, and visual stimuli can become `whited out'. The electrophysiological findings reported here reveal that the autistic brain attempts to compensate for impaired attention by applying a much coarser, ersatz mechanism of sensory gating: instead of the normal spatially and functionally specific enhancement of sensory processing, there is a generalised, nonspecific arousal that heightens response to all stimuli. It is quite remarkable that persons with autism are able most of the time to sort through the resulting cacophany. When such generalised hyper-arousal is uncontrolled, sensory overload may be the result.
The use of steady-state potentials to quantify attentional modulation of sensory input is a technique previously untried in clinical populations, and its application in this study was restricted to a narrowly defined, largely high-functioning group of adult men with autism. This study should therefore be regarded as a preliminary report, and general conclusions should be approached with caution. Although there were no detectable differences in attentional processing between male and female controls, the same is not necessarily true of people with autism: autistic males as a group are more severely impaired in early social development than are autistic females [McLennan & al. 1993], and this behavioural distinction may reflect a quantitative or even qualitative difference in the process of attentional control. A similar distinction may exist between the high-functioning subjects studied here and persons whose autism is complicated by severe mental retardation. Furthermore, even within our diagnostically homogeneous group of high-functioning male subjects, physiological subgroups may exist, and the presence of such subgroups may have been masked by high variability in the data combined with a small sample size.
Although this study sheds light on the physiological basis of attention in autistic adults, it does not address the developmental processes of which this adult physiology is the end result. It is known that the neurochemistry of the the autistic child changes during development [Chugani & al. 1999]. As corresponding differences in neurophysiology are to be expected, the extension of this electrophysiological work to autistic children would be of interest.
In light of physiological studies, the autistic obsession with replicable, patterned sensory stimulation becomes explicable as a consequence of attentional dysfunction: any human being confronted with such a profound difficulty in integrating the many separate elements of a complex stimulus will naturally gravitate to simple stimuli that can be repeated or replayed over and over again. It is only by imposing such limits that persons with profound impairments in the shifting of attention can make sense of the world. Thus, many autistic behaviours can be understood as the products of a normal human mind confronted with a highly abnormal sensory experience.
Historically in the field of autism, biological studies and development of behavioural interventions have proceeded independently of each other. This has been a loss, since physiology supplies a unique window onto the ways in which persons with autism process information. Educational and social approaches to people with autism must take these biologically based strengths and weaknesses into account. In particular, they must endeavour to avoid rapid presentation of information in disjoint spatial locations or in different sensory modalities, since the integration of such complex sensory stimuli demands a level of attentional control physiologically impossible for people with autism.
This study adds a new type of electrophysiological evidence to the growing body of findings indicating abnormal shifting and distribution of attention in autism. In the future, it will be important to localise these shift-related activations and thus to ascertain what anatomical regions and functional networks are responsible for the spatiotopically and functionally nonspecific pattern of activation seen in autism. In the long term, this information will inform behavioural and pharmacological interventions, and point the way to an understanding of the abnormal neurological development that leads to autistic perception and behaviour.
ACKNOWLEDGEMENTS: I thank Eric Courchesne for access to his laboratory, Alan Lincoln for diagnostic data, Brian Egaas and Greg Allen for helpful discussions, Robert Ringrose for computing facilities, and Marissa Westerfield for invaluable assistance in conducting the experiments. Most importantly, thanks are due to the people with autism who volunteered to participate in this study.