The eight autistic subjects were members of a long-standing subject pool recruited from a regional centre for developmental disabilities and from clinical referrals. All the autistic subjects were male, with a mean age of 26.7 years and a standard deviation of 5.4 years (range 19.4 to 32.3). Of these, seven were right-handed and one was left-handed. None were medicated, and none had any history of comorbid psychiatric disease. The diagnosis of autism was made by experienced clinicians, including a clinical psychologist and a pædiatric neurologist, according to the structured observations of the Autism Diagnostic Observation Schedule (ADOS-G) [Lord et al. 1989]. In addition, all subjects were administered the Wechsler Adult Intelligence Scale--Revised and the Childhood Autism Rating Scale (CARS) [Schopler et al. 1988], and each subject's parent was interviewed with the Autism Diagnostic Interview--Revised (ADI-R) [Lord et al. 1994]. Table 1 summarises these behavioural measures. In all eight cases the DSM-IV [American Psychiatric Association 1994] criteria were met according to both the ADI-R and the ADOS. All subjects with autism were negative for fragile-X, as assayed by cytogenetic analysis. The controls were twelve normal, right-handed subjects with no history of neurological or psychiatric disease, five female and seven male, mean age 22.9 years, standard deviation 4.1 years, range 17.4 to 30.3, recruited from among local college students and hospital employees. (Constraints on resources prevented matching the two samples by sex; however, effects in the normal females were highly similar to those in the normal males both for the SSVEP (F(1,10) = 0.02, p = 0.89) and for background activity (F(1,10) = 0.13 p = 0.73).) Informed consent was obtained from each subject. All subjects were paid for their time.
| Subject | Age | PIQ | CARS | ADI-R Social | ADI-R Communication (verbal) | ADI-R Communication (nonverbal) | ADI-R Restricted & Repetitive Behaviours |
|---|---|---|---|---|---|---|---|
| 1 | 32 | 93 | 42.5 | 29 | 22 | 12 | 11 |
| 2 | 31 | 112 | 23.5 | 22 | 16 | 8 | 7 |
| 3 | 19 | 81 | 35.5 | 45 | 20 | 14 | 10 |
| 4 | 31 | 80 | 32.5 | 25 | 21 | 14 | 7 |
| 5 | 29 | 92 | 36.5 | 26 | 20 | 14 | 6 |
| 6 | 31 | 106 | 36 | 21 | 22 | 12 | 10 |
| 7 | 20 | 108 | 45 | 30 | 20 | 13 | 6 |
| 8 | 20 | 51 | 36.5 | 23 | 27 | 13 | 12 |
| Mean (S.D.) | 26.6 (5.8) | 90.4 (19.9) | 36.0 (6.5) | 27.6 (7.7) | 21.0 (3.1) | 12.5 (2.0) | 8.6 (2.4) |
The paradigm was identical to that used in our previous work on attentional shifts in normal subjects [Belmonte 1998] and consisted essentially of two `oddball' detection tasks running side by side. Stimuli were coloured squares, 1.8° on each side, centred 3.0° superior and 5.1° lateral to a fixation cross. Each square was flashed for 56ms and was followed by another 56ms of blank display, producing an SSVEP with a 112ms cycle (8.9s-1). On detecting a target in the currently attended location, subjects had as rapidly as possible to move a joystick to the opposite side and to shift their attention to that opposite side. Stimuli were presented in 144 trials, and ran continuously within each trial. Each trial contained 32 targets and was about 50s in duration. Fixation was monitored by electrooculography and by closed-circuit television. Both detection accuracy and response latency were recorded as behavioural measures. Although the latter measure is subject to influence by delays in motor implementation, the former depends solely on detection and not on the speed of motor implementation. Before EEG recording began, subjects were observed during practice trials until it was clear that they were performing the task as instructed. In addition, the behavioural data served as an ongoing check for appropriate performance of the task..
Due to the small size of the scalp region of interest, a uniformly sized electrode array was applied to all subjects regardless of variation in head shape. Within such small regions, it has been our experience that errors introduced during the process of electrode placement are comparable to the small errors introduced by variation in head circumference. Ag-AgCl electrodes, 1cm in diameter, were placed on each hemisphere along three parallel lines 1.5cm, 3.5cm, and 5.5cm superior to the line joining the inion and the preauricular point. On the upper and lower lines, electrodes were placed 4cm and 8cm lateral to the midline. On the middle line, electrodes were placed 6cm lateral to the midline. This scheme produced, over each hemisphere, one central electrode located over occipitoparietal scalp and surrounded by four equally spaced neighbours at a radius of 2.8cm.
EEG was recorded at a sampling rate of 286s-1 (i.e., 32 samples during each 112ms cycle of the flashing stimulus) using Scan386 digitisation software (Neuroscan, El Paso, Texas) and a Scientific Solutions analogue-to-digital converter with a Grass Model 12 Neurodata Acquisition System. Half-amplitude cutoffs were 0.1s-1 and 100s-1. Bipolar derivations were transformed off-line into a measure consisting of four times the voltage at the central electrode minus the voltages at each of the four neighbours. This procedure emphasises sources that underlie the centrally placed electrode and diminishes effects of volume conduction [Hjorth 1975]. EEG intervals in which the range of the median-filtered horizontal electrooculogram exceeded 25µV within 75ms were rejected. A similar procedure was applied in the case of the vertical electrooculogram with a threshold of 100µV in 300ms. These rejection parameters were selected based on their reliable identification of horizontal saccades and eyeblinks, respectively, in pilot data.
Using the Gnuroscan system [Belmonte 1997], a set of extensions to the Neuroscan software, artefact-free intervals consisting of the 900ms epoch following correctly-detected targets were averaged into four separate bins depending on the amount of time since the previous correctly detected target: 56ms-728ms, 840ms-1512ms, 1624ms-2296ms, and 2408ms or longer. Since 2296ms was the longest possible interval between target presentations, the 2408ms bin contained only responses that followed at least one missed target. For each of the eight successive 112ms periods within this epoch, SSVEP and background amplitudes were derived from the Fourier transform [Mast & Victor 1991], and the difference between amplitude in response to left targets and amplitude in response to right targets was computed.
Each of the two sets of behavioural measures (latency and accuracy) was subjected to a 2×4×2 analysis of variance (BMDP program 2V) with factors diagnostic group, SOA, and target location. Each of the two sets of electrophysiological data (SSVEP and background) was subjected to a 2×4×2×8 analysis of variance with factors diagnostic group, SOA, target location, and latency.