SPECIAL ISSUE GAMMA ABNORMALITIES DURING PERCEPTION OF ILLUSORY FIGURES IN AUTISM Caroline Brown1, Thomas Gruber2, Jill Boucher3, Gina Rippon4 and Jon Brock5 (1School of Psychology, University of West of England, Bristol, UK; 2Institut für Allgemeine Psychologie, Universität Leipzig, Germany; 3Department of Psychology, University of Warwick, UK; 4Neurosciences Research Institute, School of Life and Health Sciences, Aston University, Birmingham, UK; 5Department of Psychology, University of Warwick, UK)

ABSTRACT This experiment was designed to test the hypothesis that perceptual abnormalities in autism might be associated with alteration of induced gamma activity patterns overlying visual cortical regions. EEG was recorded from six adolescents with autism and eight controls matched on chronological age, and verbal and nonverbal mental age, whilst identifying the presence or absence of an illusory Kanizsa shape. Although there were no reaction time or accuracy differences between the groups there were significant task-related differences in cortical activity. Control participants showed typical gammaband activity over parietal regions at around 350 msec post onset of shape trials, similar to gamma patterns found in previous studies with non-impaired adults. In contrast, autistic participants showed overall increased activity, including an early 100 msec gamma peak and a late induced peak, 50 to 70 msec earlier than that shown by the control group. We interpret the abnormal gamma activity to reflect decreased “signal to noise” due to decreased inhibitory processing. In this experiment we did not establish a link between altered perception and abnormal gamma, as the autistic participants’ behaviour did not differ from the controls. Future work should be designed to replicate this phenomenon and establish the perceptual consequences of altered gamma activity. Key words: EEG, 40 Hz, gamma EEG, binding, autism, development, attention, perception

INTRODUCTION Individuals with autism have specific deficits in social interaction and communication (American Psychiatric Association, 1994; World Health Organization, 1992). There is, however, considerable evidence that autism is also associated with certain visuo-perceptual and attentional abnormalities (e.g., Burack et al., 1997; Happé, 1999; Plaisted, 2001; Wing, 1996). The study reported in this paper investigated the neural correlates of atypical visual perception in autism by measuring EEG in adolescents with autism. Electrophysiological research has provided evidence that in non-impaired adults, visual perception is signalled at occipito-parietal regions by increased induced gamma activity (Keil et al., 2001; TallonBaudry and Bertrand, 1999). Such high frequency activity has been posited as a physiological marker of the co-activation of cortical cells engaged in processing the same stimuli (Singer and Gray, 1995), integrating different features of a stimulus in order to establish a cortical object representation (Müller, 2000), and over wide-ranging cortical regions in wider integration (Rodriguez et al., 1999). It has therefore been suggested that perceptual abnormalities in autism may stem from differences in cellular activity that may be measured as differences in induced gamma activity (Brock et al., 2002; Grice et al., 2001). Evidence for atypical visual-perception in autism comes from a wide range of paradigms. For Cortex, (2005) 41, 364-376

example, individuals with autism have consistently been found to outperform controls on visual search tasks (O’Riordan and Plaisted, 2001; O’Riordan et al., 2001; Plaisted et al., 1998); on the embedded figures task, in which a simple shape is identified within a complex picture (Shah and Frith, 1983; Tymchuk et al., 1977; Brian and Bryson, 1996); and on the block design test, in which a complex pattern has to be broken down and then reconstructed using coloured blocks (Shah and Frith, 1993; Happé, 1994; Jolliffe and Baron-Cohen, 1997; Pring et al., 1995). Furthermore, when required to attend to both the local and global levels of hierarchical figures, children with autism have been found to show an unusual advantage for processing local level detail (Plaisted et al., 1999). Finally, Happé (1996) reported that children with autism were less susceptible than controls to a number of visual illusions, although subsequent studies have failed to replicate this finding (Ropar and Mitchell, 1999, 2001). Various theoretical accounts have been proposed to explain these findings at the cognitive level. For example, it has been argued that people with autism have a tendency to focus on the local details of objects (e.g., Frith, 1989; Happé, 1999); that they fail to integrate local and global levels of stimuli (Mottron and Belleville, 1993); or that they pay more attention to the distinguishing features of objects than to their common features (e.g., O’Riordan and Plaisted, 2001; Plaisted, 2001). However, it is not clear from these accounts what

Visual perception and gamma activity in autism

the underlying neurophysiological mechanisms might be. In a recent paper (Brock et al., 2002), we suggested that many of these behavioural phenomena in autism could be understood in terms of impairment in the integration or ‘binding’ of functionally discrete information processing from across the brain. Evidence from intra-cellular recording (Singer and Gray, 1995) and scalp recorded EEG has linked this binding process with gamma-band activity (Müller et al., 1996; 1997; Tallon-Baudry and Bertrand, 1999; Keil et al., 2001). In unimpaired participants, the onset of a visual stimulus gives rise to a burst of gamma activity over occipital and parieto-occipital sites at around 250-300 msec. This also occurs when illusory objects are perceived, for instance when Kanizsa stimuli are presented (Tallon-Baudry et al., 1996). These types of tasks may best be considered as “bottom-up” or perceptual tasks. When more complex tasks are undertaken, discrete bursts of gamma activity have been identified overlying cortical regions thought to be engaged in those tasks. For instance modulation of attention gives rise to occipito-parietal bursts of gamma activity (Gruber et al., 1999; Müller et al., 2000; Müller and Gruber, 2001), and word and non-word tasks show differential gamma activity over auditory cortical regions (Pulvermüller et al., 1999). Additionally, tasks that involve attention or topdown integration of features have given rise to simultaneous bursts of gamma activity over frontal and occipito-parietal regions (Rodriguez et al., 1999). These findings led us to predict that, where integration of cortically-distant information was impaired, variations in gamma-band activity might be found. Correspondingly, we predicted that perceptual abnormalities in autism might be associated with alteration of induced gamma activity patterns overlying visual cortical regions. EEG recording and analysis has the advantage of very high temporal resolution, compared to say functional Magnetic Resonance Imaging (fMRI). In EEG neural activity changes occurring within a few milliseconds can be shown, whereas in fMRI recordings of cerebral blood flow change require several minutes. In conventional EEG analysis all the correct, artifact-free trials for a given condition are initially baselined around the mean voltage of a period preceding onset of the trial, and then summed and an average signal calculated for each channel recorded. This mean value is the Evoked Potential1, and depending upon both scalp location and task being investigated a set of recognized components may be identified. Of particular interest to this study are visually evoked components called N1, P1, N2, P2, and P3. The letter indicates 1 This description simplifies the many variations that may be made in EEG recording, for instance ongoing changes or response-related recordings may be required, depending on the experimental paradigm.

365

polarity, and the number the period after onset of a visual trial, i.e. 1 is in the 100-200 msec period, 2, the 200-300 msec and 3 at (variable) times after 300 msec (Luck et al., 1990; O’Donnell et al., 1997). Most conventional analysis removes high frequency activity by filtering it out, but as it has such relatively low power compared to the lower frequency activity, it is in fact quite acceptable to calculate evoked potentials with the high frequency component maintained. Evoked potentials are considered to be the summed activity of highly co-ordinated (synchronized) neural activity underlying the scalp (Hillyard and Picton, 1987), and it is this property that led electrophysiologists to investigate high frequency (gamma) activity, as it has been suggested that gamma activity is the signal between assemblies of neurons that they are working on the same object (percept, idea, cognition). Consequently, early components (N1, P1) may be thought of as the overall increase in generalized activity, and early gamma components are measured at the same time over the same cortical regions during the attentional control of perception (Herrmann et al., 1999). Later components (N2, P2) may correspond with the binding of inter-areal information, and corresponding gamma components have also been identified (Tomberg, 1999). Because the P3 represents a large positive extra-cellular charge and is variable in both onset and duration, it could signal that processing has been completed. The method used to identify gamma activity at later time periods (induced gamma) identifies a gamma burst that occurs at varying times in each trial, in the period preceding the P3 latency. Evoked gamma band activity has been identified at a latency of around 100 msec post stimulus onset, over visual and auditory regions (Herrmann and Mecklinger, 2001; Yordanova et al., 1997; Bertrand and Tallon-Baudry, 2000 review). Evoked gamma-band response is highly phaselocked to stimulus onset, and is measured from the average (mean) of all summed trials, whereas induced gamma band activity occurring later has variable onset latency and therefore can only be identified on a trial-by-trial analysis. Hermann and Mecklinger argue that evoked gamma band activity reflects the effect of attention on early visual processing. The separation of gamma activity into early and late components might be thought of as indirectly reflecting the different levels at which binding of neural response is required. The early component may be interpreted as the binding of perceptual information within the same cortical area (i.e. intra-areal), whereas the late component is assumed to reflect the binding of feed-forward and feed-back processing information in a whole network of cortical areas (corticocortical) (Müller et al., 2000; Müller, 2000; Shibata et al., 1999).

366

Caroline Brown and Others

Developmentally, onset of gamma band activity is critically related to the development of the perceptual ability to discriminate shapes. This aptitude develops in infants between the ages of 6 and 9 months (Csibra, 2001), and Csibra et al. (2000) recently showed that gamma band activity over frontal regions became apparent as the perceptual ability to discriminate Kanizsa stimuli developed. Research has shown that the locus of gamma band activity shifts during development, which may be related to the development of mature attention systems. Gamma activity measured in children during auditory attention tasks (Yordanova et al., 2000, 2002) showed that earlier in development, (9-12 years) frontal gamma power was greater and more discriminative of stimulus type whereas later in development (13-16 years) parietal gamma power increased during attention demanding tasks. Grice et al. (2001) compared gamma band activity over frontal regions in adults with autism, and unimpaired adults during a face discrimination paradigm. Unimpaired adults showed clear discriminative increases in frontal gamma activity during presentation of upright faces compared with inverted faces. However, although adults with autism showed similar patterns of activity, the extent of gamma activity did not differ significantly between upright and inverted faces. This suggests that in autism the gamma response does not allow discrimination between stimuli. In the current study, we further investigated the neural correlates of visual perception in autism by measuring gamma EEG during perception of illusory Kanisza figures. The Kanizsa illusion has been widely used to investigate processes underlying perception, because the percept of the shape is overwhelming despite its illusory nature. Its arrangement allows the same objects to form both a target-present and a target-absent condition, with inducers that can be oriented towards or away from each other. Conventional EEG research has shown that perception of Kanizsa shapes gives rise to a large negative peak (N2) at around 250 msec after onset, followed by a positive peak (P300) at around 390 msec (Proverbio and Zani, 2002). In spectral EEG research perception of Kanizsa has given rise to early evoked gamma band activity (Herrmann and Mecklinger, 1999, 2001; Böttger, et al., 2002) and to the later induced gamma band component (Tallon et al., 1995; Tallon-Baudry et al., 1996). The induced component occurs at around the same latency as the N2 component reported in ERP studies, that is 250 msec. It is unclear from previous behavioural studies whether or not individuals with autism should be less susceptible to the Kanisza illusion than nonautistic individuals. Happé (1996) reported that children with autism were less likely than controls to report seeing an illusory shape. However, she

did not include a control condition in which illusory shapes were not present. This concern was addressed by Ropar and Mitchell (1999, 2001), who failed to find any differences between children with autism and controls in their susceptibility to a number of visual illusions, but these studies did not include any Kanizsa stimuli. Visual illusions may be subject to different cognitive processes and neurobiological pathways (Gregory, 2001), so it remains possible that individuals with autism may be relatively immune to Kanisza illusions. The current study offered the opportunity to investigate both the behavioural response to Kanisza illusions and the underlying neural mechanisms in autism. As part of a larger experiment, EEG recordings were made of children with autism whilst they made decisions regarding the presence or absence of Kanizsa stimuli. It was predicted that the perceptual disturbances associated with autism would be accompanied by differences in gamma band power over occipital, occipito-parietal or parietal regions (cf. Brock et al., 2002). A crucial consideration was the control group to employ in this exploratory study. Behavioural studies of autism often include younger typically developing children matched on a measure of mental age. However, this approach offers the potential confound that group differences in EEG patterns may be explained simply in terms of differences in cortical processing related to chronological age differences. To ensure that differences in performance or neurophysiological correlates of performance of the Kanizsa task were not confounded by differences in intelligence or developmental stage, it was decided to test a group of children matched for both chronological age and IQ. In the current study, therefore, children with autism were compared with a group of mentally retarded children whose age and cognitive profiles matched those of the children with autism. METHOD Participants Twenty-four children between the ages of 11 and 17 were recruited through special schools in and around Coventry, and as respondents to adverts placed in the West Midlands Autistic Society Newsletter. All participants in the autism group had a clinical diagnosis of autism or Asperger syndrome. The control participants were recruited on the basis that they showed no other developmental disorder (e.g., Down Syndrome or Fragile X Syndrome) to control only for mental ability with the group with autism. In the UK, children with non-syndrome related mental retardation with reasonable language skills are collectively assigned to a group described as

Visual perception and gamma activity in autism

367

TABLE I

Matching and Grouping Variables for Participant Groups ASD AGE (years) TROG (score) TROG (age) RAVENS (score) RAVENS (age) CARS (score)

MLD

M

SD

M

SD

t (12)

p

14.7 68.5 7.9 26.8 ~ 8.9 34.5

0.8 8.5 2.7 7.3

14.0 69.0 7.9 24.6 ~ 8.4 19.8

1.9 4.3 1.7 7.6

0.82 – 0.14 – 0.55 – 9.47

0.43 0.89 – 0.59 – 0.00

3.2

having “moderate learning difficulties” (MLD), which will be used to describe the control group in the methods and results sections, whilst the children in the experimental group will be described using the acronym ASD, (autistic spectrum disorder). All participating children had sufficiently high language and cognitive skills to enable their understanding of the task and the experimental procedure, and all were free from epilepsy, had suffered no encephalic trauma and were not taking neuroleptic or psychotropic medication. Participants each received a £10 gift voucher in recognition of their time and participation. As might be expected given the difficulties involved in collecting EEG data, it was found that EEG data from many of these children were not useable because of movement and eye-movement generated artifacts. Data are therefore reported from sub-groups of six children with autism (the ASD group) and eight mentally retarded children (the MLD group). Verbal and nonverbal mental ages were established using the Test of Reception of Grammar [TROG] (Bishop, 1989) and the Ravens Coloured Progressive Matrices [RAVENS] (Raven, 1993) respectively. Parents were also interviewed using the Childhood Autism Rating Scale [CARS] (Schopler et al., 1988). Independentsamples t-tests showed that the two groups were closely matched on chronological age, verbal mental age and non-verbal mental age, but that the children in the autism group had significantly higher scores on the CARS, as shown in Table I.

2.6

Procedure The volunteer children were initially tested for verbal age ability using the TROG in an interview either at their home or in an initial visit to the laboratory. During this session the procedure was explained and outlined to both the child and their carers using photographs of children being recorded using the SensorNet system. In accordance with British Psychological Society Ethical Guidelines participants and carers were reassured that they could withdraw from the study at any time, both in this and the later session. The remaining cognitive measures were administered when the child attended the laboratory for EEG testing. In addition to the Ravens Matrices, participants were tested on the Block Design subtest (BDT) of the Wechsler Intelligence Scale for Children (Wechsler, 1974), and on the Children’s Embedded Figures Test (Witkin et al., 1971). The CEFT was scored using the length of time (sec) it took participants to identify the embedded figure, up to a maximum of 60 seconds (a time of 60 seconds was recorded when participants failed to identify the target shape), and reaction times were log transformed (cf. Brian and Bryson, 1996). Training was given in two stages. Initially participants were shown Kanizsa-shape and Kanizsa-random stimuli in a book, and were instructed to point to a blue or green square to

Stimuli Stimuli for the EEG task were presented as black and white images on a white screen within a black border subtending 0.5° visual angle, the whole image subtending 6.3° by 4.0° (h × w) visual angle in size. Examples of Kanizsa stimuli are shown in Figure 1. Stimuli were created in rectangles of 5 × 5 circles within the black border. This gave a set of 16 Kanizsa shape-stimuli each of which were presented twice. Four exemplars of Kanizsa-random stimuli were created and used as ‘target-absent’ control stimuli. Stimuli were presented in a block of sixty-four trials, thirty-two each of Kanizsa-shape and Kanizsa-random exemplars in randomised order.

Fig. 1 – Kanisza stimuli a) shape condition and b) random condition.

368

Caroline Brown and Others

indicate whether they could see a shape or not. Only participants whose responses were consistently correct (100%) in this task proceeded onto the second training phase. Three participants were excluded at this point. The second phase of training was given once the SensorNet was in place and recording conditions achieved within the recording booth. Training for the EEG task was given by presenting stimuli on a computer screen, participants responding by means of a response box, pressing the blue button when they could see a rectangle, and pressing the green button when it was absent. Collection of EEG data was undertaken using the Geodesics Net system, which is much simpler to use than traditional caps because it consists of a lightweight net into which sponge covered electrodes are embedded. No pain or discomfort is reported by users, and this system is regularly used in the collection of EEG data from infants (de Haan et al., 2002). Recording sessions took around 25 minutes in total with the number of trials being limited to ensure that participants were able to maintain necessary attention and alertness levels. The procedure was exactly the same as that described for the second training phase above. All participants met the experimenters prior to taking part in the EEG recording, all were sufficiently language-able to understand both the task and the recording procedures. Pace of the experiment was set by the individual child and follow-up comment from parents (and some of the children) suggested that the experience was one the participants had enjoyed. EEG Recording and Analysis Continuous EEG data were recorded with a 128-channel Electrical Geodesics Inc. Net Station amplifier, via a dense array 128 electrode Geodesics Sensor Net. The Vertex (Cz) was chosen as the reference, and impedances were kept below 50 kW as recommended for the EGI high input impedance amplifier. Sampling rate was 250 Hz (any effect of aliasing would be expected to generate introduction of lower frequency power, rather than high frequency gamma band effects), with an on-line bandpass filter of 0.1 to 200Hz, vertical and horizontal eye movements were monitored using four of the 128 electrodes. Specific analysis was performed off-line on data from trials in which a correct answer was given. Further analysis was performed on a subset of electrodes corresponding to the sites of the extended 10-20 system (Jasper, 1958), as shown in Figure 2. Electrode sites used for the TimeFrequency plot were selected on the basis of previous findings regarding visual information processing (Tallon-Baudry and Bertrand, 1999) and findings demonstrating the establishment of cortical object representations (Gruber et al., 2002).

Data Reduction and Analysis EEG was initially segmented into epochs of 2048 msec, consisting of 768 msec prior to onset and 1024 msec post onset. Artifact correction was performed using the EGI NetStation correction algorithms; rejecting channels exceeding both a voltage threshold of 200 µV and a transition threshold of 100 µV. Channels were also rejected where 80% or more of segments were identified as not acceptable. Individual segments were rejected where more than 20 channels were rejected; eye-blink threshold was set at 70 µV. Artifact correction was implemented using spherical-spline interpolation (Ferree, 2000). In the data reported here, all participants provided at least seven accurate, artifact free trials in both Kanizsa shape and Kanizsa-random conditions. A descriptive table of trials used in the analysis is shown in Table II, no group differences in either accuracy or number of artifact-free trials were found. Significantly fewer trials were available for analysis in the Kanizsa-random condition than the Kanizsa-shape condition [F (1, 12) = 5.090, p < 0.05]. Analysis of Spectral EEG This analysis focused on non-phase locked oscillatory activity, which was identified by subtracting the mean evoked response (i.e. the ERP) from each trial (Müller et al., 1996)a. Spectral changes in bands ranging from delta to gamma band (4.9-97.7 Hz) activity were analysed using wavelet analysis of artifact-free or corrected single epochs. Spectral changes in oscillatory activity during the experiment were analysed by means of a Morlet wavelet analysis. This method provides a good compromise between time and frequency a

A given EEG-epoch can be modelled by the sum of the evoked response plus the trial-by-trial fluctuation around the mean (Priestley, 1988). This analysis focused on non-phase locked oscillatory activity, which was identified by subtracting the mean evoked response (i.e., the ERP) from each trial (Müller et al., 1996). Spectral changes in bands ranging from delta to gamma band activity were analysed using wavelet analysis of artifact-free or corrected single epochs. Morlet wavelets were used, thus giving reasonable time and frequency resolution (Sinnkonen et al., 1995), yielding information about higher frequencies over shorter time periods. Complex Morlet wavelets g may be generated for different frequencies f0 within the time domain by g (t, f0 ) = A' e

–

t2 2 σ t2 2 i π f02

e

(1)

Where A’ depends upon sf , the width of the wavelet in the frequency domain, the frequency band being analysed f0 and the constant m: m A' = σf 2 π3 (2) f0 π with m=

f0 σf

(3)

Therefore, using the constant ratio m, the width of the wavelet within the frequency domain sf changes as a function of the frequency band f0. The wavelet family used in this analysis was defined by m = f0/sf = 7; f0 range 4.88 to 96.04 Hz in 0.49 Hz steps. This procedure has been proposed by Bertrand and Pantev (1994) and is described in detail in Tallon-Baudry et al., 1998; Tallon-Baudry et al., 1997).

Visual perception and gamma activity in autism

369

Fig. 2 – EGI Sensor Net Diagram, showing channels approximating to P1; P3; P2 and P4.

resolution (Sinkkonen et al., 1995). In order to achieve good time and frequency resolution in the gamma frequency range the wavelet family used is defined by a constant m = f0/σf =7 (Tallon-Baudry and Bertrand, 1999) with wavelets ranging from 4.88 to 97.65 Hz in 0.49 Hz steps. The analysis reported here was applied to wavelet coefficients generated separately for all channels from a time period from 32 msec prior to onset until 512 msec after onset. Frequency powers identified during the segment 32 msec prior to onset were used as baseline values, which were subtracted from post-

onset power values. The short baseline period reported here was chosen to avoid interference from electro-oculogram artifacts in order to maximize the number of trials available for this analysis. Mean baseline power values for the 100 msec prior to trial onset are shown in Figure 3a. Baseline values were statistically compared and no difference between Kanizsa-condition or between groups was found. Time-frequency plots (– 32 msec to 512 msec) show negative as well as positive changes in power across the power spectrum, where power levels were below those measured in

TABLE II

Number of Artifact-free Trials used in Data Analysis ASD

Kanizsa-shape Kanizsa-random

MLD

M

SD

Range

M

SD

Range

22.5 22.7

5.8 8.52

8-30 8-31

16.5 22.5

5.6 8.3

9-26 7-31

370

Caroline Brown and Others

Fig. 3 – Mean 30-40 Hz gamma activity summed over channels P3; P1; P2: P4. a) Line plot showing changes in gamma activity in the 100 msec prior to stimulus onset, including the 32 msec baseline period. b) Line plot showing changes in gamma activity from 32 msec prior to, until 512 msec after trial onset. Both plots show the mean power values of the frequency band 30-40 Hz as they change over time summed across all four channels. The upper plot, shows the level of 30-40 Hz power measured during the 100ms prior to baseline, with the baseline period used in this analysis indicated as a line from the x-axis at – 32 ms. The data from both the Kanizsa-shape and Kanizsa-random conditions are represented by thick and thin lines respectively, and from the group with autism (ASD) in red and the mentally retarded group (MLD) in blue. Because baseline values are subtracted from each trial prior to the averaging process, changes in gamma power can be positive – that is increased compared to baseline, or negative – where levels are reduced compared to baseline.

the baseline period. Theoretical considerations suggested that earlier (< 150 msec) peaks should be treated as a separate component from later (> 250 msec) gamma band activity (Tallon-Baudry and Bertrand, 1999; Bertrand and Tallon-Baudry, 2000). Mean power values for channels corresponding to P1; P3; P2 and P4, were calculated for frequencies between 29.3 and 41.5 Hz, and graphed in the timefrequency plot shown in Figure 3b. Separate statistical comparisons were made for: 0-400 msec

(the time period as a whole), 80-120 msec (the early peak) and from 250-400 msec (the late peak). RESULTS Cognitive Battery The ASD group (M = 25.8; SD = 14.8) scored higher than controls (M = 20.7; SD = 18.6) on the

Visual perception and gamma activity in autism

371

TABLE III

Reaction Time and Accuracy Value Means and Standard Deviations for Kanizsa Task ASD

MLD

% Acc

Kanizsa-shape Kanizsa-random

RT (ms)

% Acc

RT (ms)

M

SD

M

SD

M

SD

M

SD

93 90

6.6 13.5

1828 3005

1170 1785

88 89

5.0 15.1

1640 2949

397 829

block design test, although independent samples ttests showed this was not significant [t (11) = 0.54, p = 0.60]. In the CEFT log-transformed scores of reaction time and accuracy were lower for the ASD group (M = 2.397 SD = 0.176) than the MLD group (M = 2.569 SD = 0.188) indicating superior performance, but again an independent t-test of these scores failed to reach significance [t (12) = – 1.74, p = 0.11]. When the same comparison was made for all 24 participants, the ASD group (n = 13, M = 2.045 SD = 0.258) performed significantly better than MLD group (n = 10, M = 2.268 SD = 0.189) [t (21) = 2.301, p < 0.05]. Behavioural Task The reaction time and accuracy data for detection of the Kanisza shapes (mean and standard deviations shown in Table III) were subjected to ANOVA with group and condition (Kanizsa-shape vs. Kanizsa-random) as factors. For accuracy data there were no significant main effects or interactions (all F-values < 1). Reaction times were significantly longer in the Kanizsa-random condition than in the Kanizsa-shape condition [F (1, 12) = 8.813, p < 0.05]. However, the main effect of group and the interaction between group and condition were not significant (F-values < 1). Spectral Analysis of EEG Three observations can be made from inspection of the time-frequency plots: overall the ASD group showed higher gamma power than the MLD group (0-400 msec); an early peak observed in the ASD group was not found in the MLD group (80-140 msec); the onset of the later gamma component was earlier in the ASD group than in the MLD group (250-400 msec). i) 0-400 msec The averaged 30-40 Hz power across the first 400 msec after onset was subjected to ANOVA with group and stimulus-condition as factors. The main effect of stimulus condition approached significance [F (1, 12) = 3.404, p = 0.09], as the Kanizsa-shape condition was associated with greater gamma power than Kanizsa-random condition. The interaction between stimuluscondition and group was highly significant [F (1, 12) = 13.282, p < 0.01]. The interaction reflects the overall increase in power manifested by the ASD

group compared to that in the control group, where a very significant reduction in power compared to baseline values in the Kanizsa-shape condition was found. Power values were approximately equivalent for both groups in the Kanizsa-random condition. ii) 80-140 msec The early component (80-140 msec) was subjected to ANOVA with group (ASD and MLD) and stimulus-condition (Kanizsa-shape and Kanizsa-random) as factors. No main effect of stimulus-condition was found [F (1, 12) = 2.554, p > 0.1]. A significant group by stimulus-condition interaction was found [F (1, 12) = 5.032, p < 0.05]. Post hoc Tukeys HSD tests were performed on the differences between mean power values at this time period. The power in both Kanizsa-shape and Kanizsa-random conditions for the ASD group were not significantly different from one another, whereas the power levels shown by the MLD group in the Kanizsa-shape condition were significantly different from Kanizsa-random condition (p < 0.01) and from the ASD group (p < 0.01) in both conditions. iii) 250-450 msec The later component was subjected to ANOVA with group, stimulus-condition, and time period (250-300 msec vs. 300-350 msec vs. 350-450 msec) as factors. No main effects of time or stimuluscondition were found. There were significant interactions between time and group [F (2, 24) = 5.83, p < 0.01] and between stimulus–condition and group [F (1, 12) = 7.796, p < 0.05]. The interaction between group and time represents the earlier peak power value shown by the ASD group, in the 250300 msec period, compared to the MLD group peak power value in the 350-400 msec period [t (12) = 3.685, p < 0.01]. The stimulus–condition by group interaction reflects the equivalent levels of power in the Kanizsa-random condition in both ASD and MLD groups, compared to the increased power identified in the ASD group during Kanizsa-shape condition and the decrease in power in the MLD group. The difference in power between ASD and MLD groups in the Kanizsa-shape condition approached significance [t (12) = 2.025, p = 0.07]. DISCUSSION The hypothesis guiding this study was that abnormal gamma power would accompany

372

Caroline Brown and Others

perceptual disturbances in autism. This hypothesis was only partially supported. There were significant group differences in gamma activity but no group differences in behaviour. We found that the control group of participants, mentally retarded adolescents, evinced the pattern of gamma found in typical, non-impaired adults, that is a distinct peak of gamma activity in the period 300-400 msec after onset of the stimulus at parietal sites (TallonBaudry et al., 1996), embedded within an overall reduction in power. By contrast, the autistic group demonstrated overall increased activity (compared to baseline), and the peaks occurred at earlier time points than in the mentally retarded group. Group differences in the timing of gamma activity were only found in Kanizsa-shape trials, with both groups showing comparable gamma activity during Kanizsa-random trials. Whereas the control group showed reduced power levels in the first 300 msec of the Kanizsa-shape trials, the group with autism showed increased gamma power. The early differences in gamma activity suggest that there may be an early recruitment of cells within visual areas of individuals with autism. Although, it is widely believed that visual perception in ASD is atypical, it has been particularly difficult to define experimentally, with some researchers reporting better performance in perceptual tasks such as embedded figures (e.g. Shah and Frith, 1983) and others not (e.g. Brian and Bryson, 1996). Equally the finding that people with autism are less susceptible to visual illusions (Happé, 1996) is not always replicated (Ropar and Mitchell, 1999). Behaviourally, the autistic group did not perform any better than the control group, which indicates that, even if the group with autism did perceive the Kanizsa shapes more rapidly than the control group, the combination mechanism that allows a judgement to be made and a motor control decision to be planned and instigated was not enabled. The disorder of autism is not homogenous, and with such a small sample no final conclusions may be drawn from the results reported here. Nevertheless there are clear gamma activity differences between both groups as predicted, despite the lack of behavioural difference. As outlined in the introduction, gamma activity is ubiquitous and continuously measured in EEG. In non-autistic individuals it has been found to increase during ‘target-present’ compared to ‘target-absent’ type trials (Tallon-Baudry et al., 1996; Müller et al., 1996), here we found no difference between shape and random conditions, suggesting that gamma activity in autism is not discriminating between stimulus conditions, a result which is in line with that reported by Grice and colleagues (Grice et al., 2001). The difference between groups in gamma activity at later stages in the epoch occurs in the difference of onset of the late induced gamma peak at around 350 msec. In a non-impaired adult

population, such an increase is typically embedded within an overall reduction in gamma-band activity prior to and after the peak at a slightly reduced offset of around 300 msec, similar to that shown by the control group (Tallon-Baudry et al., 1996). The time-frequency line plots shown in Figure 3 and the spectrogram in Figure 4 show that the control group gamma peak post-300 msec only reaches approximately baseline values in response to Kanizsa-shape stimuli. However, gamma levels prior to and subsequent to the peak are markedly lower than baseline values throughout Kanizsashape trials and compared to overall levels throughout the Kanizsa-random trials. This implies that it is not the increase in gamma power to a specific level, but an improvement in the signal to noise ratio that is important in signalling relevance. Improving the signal to noise ratio requires either that the relevant signal is increased in power or that irrelevant ‘noise’ is reduced and of course ideally both solutions work in tandem. It is possible that the gamma signal itself increases, hence the early increase in power identified in other research (e.g. Herrmann et al., 1996) as ‘evoked gamma’, and that surrounding noise (other non-essential neurons firing at high frequency) is reduced so that those cells that are binding information can do so unambiguously and without extraneous input from neural cells whose output is not relevant. We suggest therefore that the induced gamma component reflects an inhibitive process, which attenuates high frequency activity in non-participating neurons within an area during processing, and that it is this inhibitive process which is defective in the group with autism. The overall increase in control group gamma power measured during Kanizsa-random trials is puzzling, but remarkably similar to activity patterns during the same trials in the group with autism. Statistical comparison of absolute power values in the control group late component peaks shows that the Kanizsa-random condition is indeed greater than that in the Kanizsa-shape condition, and this is not what is usually found. However, in this task participants were shown images in which they effectively had to search for a Kanizsa shape, rather like a serial search task. In ‘shape’ conditions, the shape did indeed “pop-out”, and the signal to noise ratio in gamma power was increased markedly for control group participants. In the ‘Kanizsa-random’ condition trials therefore the increased overall gamma power may reflect increased search activity across the image, each possible “shape” candidate eliciting a burst of gamma activity. One of the most pertinent criticisms of the gamma solution to the binding problem has been that an assembly of cells raising its activity level to around 40 Hz does not in itself provide an unequivocal marker of stimulus processing. If other neural groupings processing different stimuli also

Visual perception and gamma activity in autism

373

Fig. 4 – Spectrogram showing power change in frequency band 30-40 Hz over first 500 msec of trial onset: Upper panels show recordings from children with autism (ASD group); Lower panels show recordings from IQ matched controls (MLD group); Left hand panels show recordings during Kanizsa shape condition; Right hand panels show recordings during Kanizsa random condition. Spectral activity (in this case in the 30-40Hz bandwidth) is ascertained from every correct behavioural trial in which there are no EEG artifacts, and the spectral activity found in the preceding baseline period is subtracted from that found in the trial. This gives rise to recorded changes in spectral power that can be both positive, where an increase in power is found (shown as red on colour bars) and negative, where a decrease in power is identified (shown as blue on colour bars). The spectrogram is a contour plot that enables the observation of changes in three dimensions: power, frequency and time. The data shown here are the average changes recorded over parietal sites corresponding to P1, P2, P3 and P4.

raise their activity levels then the question remains as to how independent assemblies signal relatedness of their own stimulus whilst remaining distinguishable from one another (Treisman, 1999; Shadlen and Movshon, 1999). We suggest here therefore that peaks of gamma activity are cognitively salient only if they occur embedded within a general inhibition of gamma activity in surrounding cells. The necessity of gamma peaks being embedded in more generalised gamma inhibition has a parallel in the very techniques employed here. EEG recordings are particularly macroscopic in measurement, typically picking up the activity of millions of underlying cortical cells, which leads to much of the theoretical testing of theories of binding being drawn from many different research techniques (Casanova et al., 2002). Increases in gamma power are very small compared to increases in power at lower frequencies when measured at such a macroscopic level. Cortical responses are likely to require some sort of dynamic modulation to enable small increases or decreases to be detected. For the signal to power ratio of a response to be sufficiently useful

therefore it is likely that widespread inhibition needs to be part of the dynamic processing system. The population samples used in this preliminary study were necessarily very small, with some inevitable loss of viable EEG data occurring with more learning disabled individuals and excluded from the data reported here. It was important that the control group used in this experiment were matched not only on intelligence but also in chronological age, so that maturational differences in cognition and cognitive related gamma activity (Yordanova et al., 2000) were not confounded. Future research should be undertaken, replicating and increasing the depth of this study, using more trials and more participants. Additionally the use of a task like the embedded figures task (but greatly increased in scope as twelve exemplars would be insufficient for EEG recordings to be meaningful) should be undertaken. If the perceptual anomalies in autism are directly measurable as very early bursts of uninhibited gamma over visual areas, such a task would be predicted to show clear differences over visual areas. Embedded figures tasks have been used as visuo-spatial tasks in non-patient populations, and have shown that alpha power is related to search

374

Caroline Brown and Others

Fig. 5 – Modified Kanizsa task. A better control for identifying “pop-out” of perceived shape as illusory contours are bound together would be for all the inducers to be faced in the same quadrant. In the control condition this gives four sets of control stimulus (a) and in the shape condition in this arrangement would give a possible 64 (16 locations by 4 inducer direction) stimuli (b). To further test whether children with autism do not see the shape but instead look for a particular arrangement of inducers a third condition (c) may be used, in which a non-shape composition of inducers may be cued.

strategy (Lowenhard, 1972), but to the knowledge of the authors have not been used in similar research into gamma activity. Two issues arise from this study with respect to the difficulty in identifying perceptual differences in autism. Firstly, it is particularly difficult to find behavioural tasks in which all people with autism respond differently from those without autism. Secondly, therefore, it is difficult to identify which element or process in perception is missing or impaired. It would be possible however to compare whether autistic children ‘bind’ the Kanizsa visual illusion in the same way as non-autistic individuals using an adapted form of the stimuli reported here. In future experiments both the control (see Figure 5a) and experimental conditions (see Figure 5b) utilized here would be best changed so that all the inducers are aligned to face in the same quadrant – excepting those inducing the shape in the ‘shape’ condition. An additional third condition could be tested in which an alternative arrangement of inducers, not arranged to form a shape, (see Figure 5c) could be included. If children with autism do not see the shape, but instead look for a specific arrangement of inducers, they should perform the third condition equally as well as the shape condition. Conversely, non-autistic populations would see the shape condition faster and more accurately than either the control or ‘arrangement’ condition. To summarise, the current study showed that the perception of illusory objects is associated with increased overall levels of gamma activity in children with autism, and further, we showed that

increased induced gamma power occurs earlier in the group with autism than in age and IQ matched controls performing the same task. The control group of mentally retarded children showed a similar pattern of gamma activity to that reported in unimpaired adults (Tallon-Baudry et al., 1996), but at lower amplitude and at a longer latency. These results tentatively provide interesting and useful information, both about the specific neurological processes occurring in autistic brains and giving an insight into the large body of gamma research in typically functioning populations, which would certainly benefit from replication with a larger group of participants and including a typically developing control group. Acknowledgments: This work was carried out as part of PPP Research Grant reference: 445/828, BrainBehaviour Links in Autism. Grateful acknowledgement is made to Gaynor Evans and Sarah Crispin for their assistance in data collection. REFERENCES AMERICAN PSYCHOLOGICAL ASSOCIATION. Diagnostic and Statistical Manual of Mental Disorders (4th Edition). Washington DC: Author, 1994. BERTRAND O and PANTEV C. Stimulus frequency dependence of the transient oscillatory auditory evoked response (40Hz) studied by electric and magnetic recordings in humans. In C Pantev, T Elbert and B Lutekenhoner (Eds), Oscillatory Event-related Brain Dynamics. New York: Plenum, 1994. BERTRAND O and TALLON-BAUDRY C. Oscillatory gamma activity in humans: A possible role for object representation. International Journal of Psychophysiology, 38: 211-223, 2000. BISHOP DVM. TROG (Test for Reception of Grammar). Manchester University: Age and Cognitive Performance Research Centre. 2nd Edition, 1989.

Visual perception and gamma activity in autism BÖTTGER D, HERRMANN CS and VON CRAMON DY. Amplitude differences of evoked alpha and gamma oscillations in two different age groups. International Journal of Psychophysiology, 45: 245-251, 2002. BRIAN JA and BRYSON SE. Disembedding performance and recognition memory in autism/PDD. Journal of Child Psychiatry, 37: 865-872, 1996. BROCK J, BROWN C, BOUCHER J and RIPPON G. The temporal binding deficit hypothesis of autism. Development and Psychopathology, 14: 209-224, 2002. BURACK JA, ENNS JT, STAUDER JEA, MOTTRON L and RANDOLPH B. Attention and autism: Behavioral and electrophysiological evidence. In DJ Cohen and FR Volkmar (Eds), Handbook of Autism and Pervasive Developmental Disorders. New York: Wiley, 1997. BULLOCK TH. Signals and signs in the nervous system: The dynamic anatomy of electrical activity is probably information-rich. Proceedings of the National Academy of Sciences of the United States of America, 94: 1-6, 1997. CASANOVA MF, BUXHOEVEDEN DP and BROWN C. Clinical and macroscopic correlates of minicolumnar pathology in autism. Journal of Child Neurology, 17: 692-695 2002. CSIBRA G. Fast-track report: Illusory contour figures are perceived as occluding surfaces by 8 month old infants. Developmental Science, 4: F7-F11, 2001. CSIBRA G, DAVIS G, SPRATLING MW and JOHNSON MH. Gamma oscillations and object processing in the infant brain. Science, 290: 1582-1585, 2000. DE HAAN M, PASCALIS O and JOHNSON MH. Specialization of neural mechanisms underlying face recognition in human infants. Journal of Cognitive Neuroscience, 14: 199-209, 2002. FERREE TC. Spline interpolation of the scalp EEG. Secondary Titl. EGI, 2000. http://www.egi.com/Technotes/SplineInterpolation.pdf FRITH U. Autism: Explaining the Enigma. Oxford: Blackwell Scientific Publications, 1989. GREGORY RL. Action potentials to potential actions. In TE Parks (Ed), Looking at Looking: An Introduction to the Intelligence of Vision. Thousand Oaks: Sage Publications, 2001, pp. 5171. GRICE SJ, SPRATLING MW, KARMILOFF-SMITH A, HALIT H, CSIBRA G, DE HAAN M and JOHNSON MH. Disordered visual processing and oscillatory brain activity in autism and Williams syndrome. Neuroreport, 12: 2697-2700, 2001. GRUBER T, MÜLLER MM and KEIL A. Modulation of induced gamma band responses in a perceptual learning task in the human EEG. Journal of Cognitive Neuroscience, 14: 732744, 2002. GRUBER T, MULLER MM, KEIL A and ELBERT T. Selective visualspatial attention alters induced gamma band responses in the human EEG. Clinical Neurophysiology, 110: 2074-2085, 1999. HAPPÉ F. Autism: Cognitive deficit or cognitive style? Trends in Cognitive Sciences, 3: 216-222, 1999. HAPPÉ FGE. Wechsler IQ profile and theory of mind in autism: A research note. Journal of Child Psychology and Psychiatry, 35: 1461-1471, 1994. HAPPÉ FGE. Studying weak central coherence at low levels: Children with autism do not succumb to visual illusions: A research note. Journal of Child Psychology and Psychiatry and Allied Disciplines, 37: 873-877, 1996. HERRMANN CS and MECKLINGER A. Magnetoencephalographic responses to illusory figures: Early evoked gamma is affected by processing of stimulus features. International Journal of Psychophysiology, 38: 265-281, 1999. HERRMANN CS and MECKLINGER A. Gamma activity in human EEG is related to high-speed memory comparisons during object selective attention. Visual Cognition, 8: 593-608, 2001. HERRMANN CS, MECKLINGER A and PFEIFER E. Gamma responses and ERPs in a visual classification task. Clinical Neurophysiology, 110: 636-642, 1999. HILLYARD SA and PICTON TW. Electrophysiology of cognition. In VB Mountcastle (Ed), Handbook of Physiology – the Nervous System v. Bethseda: American Physiology Society, 1987. JASPER HH. Report of committee on methods of clinical examination in electroencephalography. EEG and Clinical Neurophsyiology, 10: 370-375, 1958. JOLLIFE T and BARON-COHEN S. Are people with autism and Asperger syndrome faster than normal on the embedded figures test? Journal of Child Psychology and Psychiatry and Allied Disciplines, 38: 527-534, 1997.

375

KEIL A, GRUBER T and MÜLLER MM. Functional correlates of macroscopic high-frequency brain activity in the human visual system. Neuroscience and Biobehavioral Reviews, 25: 527-534, 2001. LOWENHARD P. Study of Witkin’s visual discrimination test and its electroencephalographic concomitants. Journal de Psychologie Normale et Pathologique, 4: 427-438, 1972. LUCK SJ, HEINZE HJ, MANGUN GR and HILLYARD SA. Visual event-related potentials index focused attention within bilateral stimulus arrays .2. Functional dissociation of P1 and N1 components. Electroencephalography and Clinical Neurophysiology, 75: 528-542, 1990. MOTTRON L and BELLEVILLE S. A study of perceptual analysis in a high-level autistic subject with exceptional graphic abilities. Brain and Cognition, 23: 279-309, 1993. MÜLLER MM. Oscillatory cortical activities in the gamma band in the human EEG induced by visual stimuli – representation of the stimulus? Acta Neurobiologiae Experimentalis, 60: 49-65, 2000. MÜLLER MM, BOSCH J, ELBERT T, KREITER A, SOSA MV, SOSA PV and ROCKSTROH B. Visually induced gamma-based responses in human electroencephalographic activity – a link to animal studies. Experimental Brain Research, 112: 96102, 1996. MÜLLER MM and GRUBER T. Induced gamma-band responses in the human EEG are related to attentional information processing. Visual Cognition, 8: 579-592, 2001. MÜLLER M, GRUBER T and KEIL A. Modulation of induced gamma band activity in the human EEG by attention and visual information processing. International Journal of Psychophysiology, 38: 283-299, 2000. MÜLLER MM, JUNGHOFER M, ELBERT T and ROCHSTROH B. Visually induced gamma-band responses to coherent and incoherent motion: A replication study. Neuroreport, 8: 2575-2579, 1997. O’DONNELL BF, SWEARER JM, SMITH LT, HOKAMA H and MCCARLEY RW. A topographic study of ERPs elicited by visual feature discrimination. Brain Topography, 10: 133143, 1997. O’RIORDAN M and PLAISTED K. Enhanced discrimination in autism. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 54A: 961-979, 2001. O’RIORDAN MA, PLAISTED KC, DRIVER J and BARON COHEN S. Superior visual search in autism. Journal of Experimental Psychology: Human Perception and Performance, 27: 719730, 2001. PLAISTED K, O’RIORDAN M and BARON COHEN S. Enhanced visual search for a conjunctive target in autism. A research note. Journal of Child Psychology and Psychiatry, 39: 777783, 1998. PLAISTED K, SWETTENHAM J and REES L. Children with autism show local precedence in a divided attention task and global precedence in selective attention task. Journal of Child Psychology and Psychiatry, 40: 733-742, 1999. PLAISTED KC. Reduced generalization in autism: An alternative to weak central coherence. In JA Burack and T Charman (Eds), The Development of Autism: Perspectives from Theory and Research. Mahwah: Lawrence Erlbaum Associates, 2001, pp. 149-169. PRIESTLEY DL. Analysis of Non-stationary and Non-linear Signals. New York: Academic, 1988. PRING L, HERMELIN B and HEAVEY L. Savants, segments, art and autism. Journal of Child Psychology and Psychiatry and Allied Disciplines, 36: 1065-1076, 1995. PROVERBIO AM and ZANI A. Electrophysiological indexes of illusory contours perception in humans. Neuropsychologia, 40: 479-491, 2002. PULVERMÜLLER F, LUTZENBURGER W and PREISSL H. Nouns and verbs in the intact brain: Evidence from event-related potentials and high frequency cortical responses. Cerbral Cortex, 9: 497-506, 1999. RAVEN JC. Raven's Coloured Progressive Matrices: London, H.K. Lewis, for publisher, 1993. RODRIGUEZ E, GEORGE N, LACHAUX J-P, MARTINERIE BR and VARELA FJ. Perception’s shadow: Long-distance synchronization of human brain activity. Nature, 397: 430433, 1999. ROPAR D and MITCHELL P. Are individuals with autism and Asperger's syndrome susceptible to visual illusions? Journal of Child Psychology and Psychiatry and Allied Disciplines, 40: 1283-1293, 1999. ROPAR D and MITCHELL P. Susceptibility to illusions and performance on visuospatial tasks in individuals with autism.

376

Caroline Brown and Others

Journal of Child Psychology and Psychiatry and Allied Disciplines, 42: 539-549, 2001. SCHOPLER E, REICHLER RJ and ROCHEN RENNER B. CARS (the Childhood Autism Rating Scale). Los Angeles: Western Psychological Services, 1988. SHADLEN MN and MOVSHON JA. Synchrony unbound: A critical evaluation of the temporal binding hypothesis. Neuron, 24: 67-77, 1999. SHAH A and FRITH U. An islet of ability in autistic children: A research note. Journal of Child Psychology and Psychiatry, 24: 613-620, 1983. SHAH A and FRITH U. Why do autistic individuals show superior performance on the block design task? Journal of Child Psychology and Psychiatry and Allied Disciplines, 34: 13511364, 1993. SHIBATA T, SHIMOYAMA I, ITO T, ABLA D, IWASA H, KOSEKI K, YAMANOUCHI N, SATO T and NAKAJIMA Y. Attention changes the peak latency of the visual gamma-band oscillation of the EEG. Neuroreport, 10: 1167-1170, 1999. SINGER W and GRAY C. Visual feature integration and the temporal correlation hypothesis. Annual Review of Neuroscience, 18: 555-586, 1995. SINKKONEN J, TIITINEN H and NÄÄTÄNEN R. Gabor filters: An informative way for analyzing event-related brain activity. Journal of Neuroscience Methods, 56: 99-104, 1995. TALLON C, BERTRAND O, BOUCHET P and PERNIER J. Gammarange activity evoked by coherent visual-stimuli in humans. European Journal of Neuroscience, 7: 1285-1291, 1995. TALLON-BAUDRY C and BERTRAND O. Oscillatory gamma activity in humans and its role in object representation. Trends in Cognitive Sciences, 3: 151-162, 1999. TALLON-BAUDRY C, BERTRAND O, DELPUECH C and PERNIER J. Stimulus specificity of phase-locked and non-phase-locked 40Hz visual responses in human. Journal of Neuroscience, 16: 4240-4249, 1996. TALLON-BAUDRY C, BERTRAND O and PERNIER J. Evoked and induced 40 Hz visual responses in man: From feature binding to selective attention. Journal of Psychophysiology, 12: 193, 1998.

TALLON-BAUDRY C, BERTRAND O, WIENBRUCH C, ROSS B and PANTEV C. Combined EEG and MEG recordings of visual 40 Hz responses to illusory triangles in human. Neuroreport, 8: 1103-1107, 1997. TOMBERG C. Cognitive N140 electrogenesis and concomitant 40 Hz synchronization in mid-dorsolateral prefrontal cortex (area 46) identified in non-averaged human brain potentials. Neuroscience Letters, 266: 141-144, 1999. TREISMAN A. Solutions to the binding problem: Progress through controversy and convergence. Neuron, 24: 105-110, 1999. TYMCHUK AJ, SIMMONS JQ and NEAFSEY S. Intellectual characteristics of adolescent childhood psychotics with high verbal ability. Journal of Mental Deficiency Research, 21: 133-138, 1977. WECHSLER D. Selected Papers of David Wechsler. Oxford, England: Academic Press, 1974. WING L. The Autism spectrum: A Guide for Parents and Professionals. London: Constable, 1996. WITKIN HA, OLTMAN PK, RASKIN E and KARP S.A. The Embedded Figures Tests. Palo Alto, CA: Consulting Psychologists Press, 1971. WORLD HEALTH ORGANIZATION. International classification of Diseases and Disorders. Geneva, Switzerland: Author, 1992. YORDANOVA J, KOLEV V, HEINRICH H, BANASCHEWSKI T, WOERNER W and ROTHENBERGER A. Gamma band response in children is related to task-stimulus processing. Neuroreport, 11: 2325-2330, 2000. YORDANOVA J, KOLEV V, HEINRICH H, WOERNER W, BANASCHEWSKI T and ROTHENBERGER A. Developmental event-related gamma oscillations: Effects of auditory attention. European Journal of Neuroscience, 16: 2214-2224, 2002. YORDANOVA J, KOLEV V and DEMIRALP T. The phase-locking of auditory gamma band responses in humans is sensitive to task processing. Neuroreport, 8: 3999-4004, 1997.

Caroline Brown, School of Psychology, University of West of England, 8 Woodland Road, Bristol BS8 1TN. e-mail: [email protected]