Results

Behavioural data, detailed in Table 2, show that normal subjects and most subjects with autism had little difficulty performing the task. Only one subject, in the autism group, accomplished fewer than 13 of the 14 nominal target detections in each location. The average accuracy, calculated as the ratio between the number of correctly detected targets and the total number of detection opportunities, was over 90% for both groups. In terms of response latency, both groups had a slight right hemifield advantage which was not significant ( F(1, 20) = 0.09, p = 0.77), and the autism group responded slightly more slowly but not significantly so ( F(1, 20) = 0.22, p = 0.64).

Table 2: Numbers of hits, misses, and false alarms, accuracy ratios (hits / (hits + misses)), and response latencies, for targets in each hemifield, for the autistic (top) and normal (bottom) groups. (As the task was designed so that most of the subjects performed at ceiling, d' scores are not given.)

	LEFT TARGETS						RIGHT TARGETS
subject	hit	miss	FA	accuracy	latency	(SD)	hit	miss	FA	accuracy	latency	(SD)
A1	13	0	2	100%	578ms	(89ms)	14	0	0	100%	683ms	(105ms)
A2	4	4	2	50%	1342ms	(200ms)	3	9	3	25%	1108ms	(365ms)
A3	14	0	0	100%	571ms	(171ms)	14	0	1	100%	506ms	(53ms)
A4	14	0	2	100%	557ms	(96ms)	15	0	1	100%	562ms	(76ms)
A5	14	0	2	100%	639ms	(173ms)	13	0	1	100%	601ms	(103ms)
A6	13	0	0	100%	578ms	(32ms)	13	0	1	100%	729ms	(207ms)
average				92%	711ms	(310ms)				88%	698ms	(216ms)
N1	14	0	0	100%	735ms	(265ms)	14	0	0	100%	676ms	(79ms)
N2	14	0	0	100%	950ms	(279ms)	14	0	0	100%	742ms	(108ms)
N3	14	0	1	100%	579ms	(89ms)	14	0	0	100%	583ms	(92ms)
N4	13	1	1	93%	582ms	(57ms)	15	1	0	94%	653ms	(107ms)
N5	14	0	0	100%	702ms	(89ms)	14	0	0	100%	624ms	(74ms)
N6	14	0	0	100%	554ms	(52ms)	14	0	0	100%	598ms	(85ms)
average				99%	684ms	(150ms)				99%	646ms	(58ms)

Talairach coordinates [78] and statistics of activated foci in the whole-brain comparison of task versus fixation are detailed in Table 3. In normal subjects, the task strongly activated a network of cortical regions comprising left superior parietal lobe, left middle temporal gyrus, left inferior and middle frontal gyri, bilateral premotor cortex, and medial frontal gyrus. In addition, right superior parietal lobe in the normal group was significantly activated in comparison with the autism group. Pooled activations in subjects with autism were weaker and largely confined to bilateral ventral occipital cortex and striate cortex, though there was a trend (p=0.059) towards activation in left orbitofrontal cortex. Surface views of these pooled activations for each subject group are shown in Figure 2.

Table 3: Talairach coordinates, t scores, maximum z scores, extents, and tail probabilities for activated foci in the normal group, in the autism group, and for the comparison between groups.

REGION	COORDINATES			t	zmax	mm3	p
NORMAL
L SPL	-28	-56	60	6.58	3.23	376	0.000
L MTG	-64	-64	6	4.40	2.70	208	0.002
R premotor	46	-2	58	4.40	2.70	240	0.001
L premotor	-36	4	58	5.05	2.88	272	0.000
L IFG	-44	16	-8	3.58	2.41	280	0.000
L MFG	-28	66	22	3.05	2.19	160	0.014
MedFG	4	12	54	4.10	2.60	112	0.014
AUTISTIC
striate	-8	-74	16	6.74	3.27	168	0.044
L vO	-33	-76	-18	3.61	2.42	224	0.022
R vO	28	-76	-19	3.52	2.39	224	0.022
L OFC	-30	34	-12	3.52	2.39	144	0.059
COMPARISON
L SPL	-32	-56	60	4.48	3.24	944	0.000
R SPL	28	-66	52	3.39	2.70	504	0.003
L MTG	-62	-66	2	4.53	3.27	952	0.000
L PoCG	-48	-32	60	3.41	2.72	504	0.003
R premotor	46	-2	60	3.64	2.84	608	0.002
L IFG	-50	10	-8	3.61	2.82	896	0.000
L MFG	-28	56	26	3.08	2.52	448	0.006
L MFG	-46	48	12	2.75	2.32	336	0.010
MedFG	0	14	54	3.40	2.71	552	0.002

Figure 2. Surface views of significant activated clusters of voxels for normal (left) and autistic (right) subjects. While normal subjects activated a network of cortical regions including left superior parietal lobe, left middle temporal gyrus, left inferior and middle frontal gyri, bilateral premotor cortex, and medial frontal gyrus, subjects with autism activated bilateral ventral occipital cortex.

Talairach coordinates for each region of interest in the attention comparison are given in Table 4. fMRI time series for each region, averaged within each subject group, are shown in Figure 3, and z-scores are detailed in Table 5. The analysis of variance showed a main effect of group, reflecting a tendency of many areas of the autistic brain to activate more when attending left than when attending right (F(1, 59) = 4.76, p = 0.033). In addition, a group x hemisphere x region-of-interest effect reflected group differences in patterns of regional activation (F(2, 59) = 4.17, p = 0.020). Post hoc t-tests revealed group differences in left ventral occipital cortex, which was relatively activated during left attention in the autistic brain but during right attention in the normal brain (t(10) = 2.55, p = 0.029), and in left intraparietal sulcus, which was activated during left attention in both groups but more highly so in the case of the autistic group (t(10) = 3.29, p = 0.008).

Table 4: Talairach coordinates (mean±SD) for regions of interest derived from the task-versus-fixation comparison, for autistic and normal subjects separately.

	normal	autistic
L vO	(-28±3, -69±6, -15±8)	(-26±14, -76±4, -18±6)
R vO	(27±7, -69±6, -13±10)	(28±10, -76±4, -19±8)
L IPS	(-33±5, -69±6, 29±7)	(-30±7, -77±3, 25±9)
R IPS	(25±4, -69±6, 26±9)	(25±7, -77±3, 26±8)
L SPL	(-28±11, -69±6, 50±2)	(-21±9, -74±6, 50±6)
R SPL	(25±14, -69±6, 49±3)	(21±7, -74±6, 49±5)

Figure 3. Grand averages of fMRI time series from left (left two columns) and right (right two columns) hemispheres, for shifts from left visual field to right visual field (solid line) and from right visual field to left visual field (broken line). The horizontal axis is seconds from the time of an attention shift; the vertical axis is percent above the baseline signal recorded during the fixation periods. Time series were averaged from six seconds following the preceding shift or the beginning of the task period, up to the time of the next shift or the end of the task period. Note in the left hemisphere the autistic lack of attentional modulation in ventral occipital cortex, (top left), and heightened modulation in intraparietal cortex (middle left). A large average difference in superior parietal cortex (bottom right) arose from only three of the six subjects in the autism group and was therefore not statistically significant.

Table 5: z-scores from the comparison of left attention (positive z) to right attention (negative z) in ventral occipital cortex, intraparietal sulcus, and superior parietal lobe for the autistic (top) and normal (bottom) groups.

subject	L vO	R vO	L IPS	R IPS	L SPL	R SPL
A1	+1.70	+0.50	+0.92	+0.14	+1.51	+2.34
A2	+2.19	+1.34	+0.82	-1.58	+1.09	-0.89
A3	+1.43	+1.20	+0.59	-1.08	-0.28	+1.27
A4	+0.45	-0.59	+0.94	+0.00	+0.13	-0.21
A5	-1.41	-1.34	+0.50	-0.11	-0.73	-0.87
A6	+0.40	+0.93	+1.07	-0.70	-0.70	+1.93
average	+0.79	+0.34	+0.81	-0.55	+0.17	+0.59
SD	(1.29)	(1.08)	(0.22)	(0.68)	(0.94)	(1.43)
N1	-1.52	+0.54	-0.16	-0.71	-0.79	-0.41
N2	-0.80	+1.17	-0.19	+0.30	+0.04	+0.21
N3	-1.60	+0.71	+0.31	-0.27	-0.33	-0.13
N4	-1.43	+0.38	+0.28	-0.51	-0.08	-1.04
N5	-1.55	+1.00	+0.50	-0.67	+0.04	+0.17
N6	+1.17	+1.17	+0.71	-0.63	+0.83	+0.07
average	-0.96	+0.83	+0.24	-0.41	-0.05	-0.19
SD	(1.08)	(0.34)	(0.36)	(0.39)	(0.53)	(0.48)

Notably, the three largest z-scores in left ventral occipital cortex came from the three subjects (A1, A2, A3) with clinical diagnoses of autism. The two males with clinical diagnoses of autism (A2, A3) also generated the two greatest right intraparietal activations during right attention as compared to left attention. Right ventral occipital cortex was significantly (folded F(5, 5) = 10.33, p = 0.023) more variable in the autism group than in normals: in the two subjects with clinical diagnoses of Asperger syndrome (A4, A5), it activated more for right attention than for left attention. Although no significant group effects were seen in the superior parietal region, the right superior parietal region was significantly more variable in the autistic group (folded F(5, 5) = 9.06, p = 0.030), with three subjects having large positive (i.e., left-attention) z-scores.