OUP user menu

Pursuit eye movement deficits in autism

Yukari Takarae, Nancy J. Minshew, Beatriz Luna, Christine M. Krisky, John A. Sweeney
DOI: http://dx.doi.org/10.1093/brain/awh307 2584-2594 First published online: 27 October 2004

Summary

Oculomotor studies provide a novel strategy for evaluating the functional integrity of multiple brain systems and cognitive processes in autism. The current study compared pursuit eye movements of 60 high-functioning individuals with autism and 94 intelligence quotient, age and gender matched healthy individuals using ramp and oscillating target tasks. Individuals with autism had normal pursuit latency, but reduced closed-loop pursuit gain when tracking both oscillating and ramp targets. This closed-loop deficit was similar for leftward and rightward pursuit, but the difference between individuals with autism and their age-matched peers was more apparent after mid-adolescence, suggesting reduced maturational achievement of the pursuit system in autism. Individuals with autism also had lower open-loop pursuit gain (initial 100 ms of pursuit) and less accurate initial catch-up saccades during a foveofugal step-ramp task, but these deficits were only seen when targets moved into the right visual field. Pursuit performance in both open- and closed-loop phases was correlated with manual praxis in individuals with autism. Bilateral disturbances in the ability to use internally generated extraretinal signals for closed-loop pursuit implicate frontostriatal or cerebellar circuitry. The hemifield specific deficit in open-loop pursuit demonstrates a lateralized disturbance in the left extrastriate areas that extract visual motion information, or in the transfer of visual motion information to the sensorimotor areas that transform visual information into appropriate oculomotor commands.

  • sensorimotor control
  • laterality
  • neurodevelopmental disorders
  • developmental disabilities
  • eye movements
  • IQ = intelligence quotient

Introduction

Autism is a complex neurodevelopmental disorder characterized by odd social interactions and communication, cognitive rigidity, abnormal language development and stereotypical behaviours. Structural imaging and histological studies have identified grey and white matter abnormalities in multiple brain regions (Bailey et al., 1998; Casanova et al., 2002; Cody et al., 2002), suggesting that the disorder is caused by maturational abnormalities in widely distributed brain networks rather than isolated focal pathology. The long-term consequences of abnormalities in brain maturation are not clear. However, one possible consequence of phenomena such as accelerated early brain growth (Courchesne et al., 2001, 2003; Aylward et al., 2002; Sparks et al., 2002) is a disruption of organization in the terminal fields of developing long fibre tracks that are needed to integrate nodes of widely distributed brain systems to support normal sensorimotor and cognitive function.

This disruption in neural development could compromise functional connectivity, the capacity to coordinate activity across many brain regions to produce complex behaviour. Abnormal connectivity of brain regions could have greater impact on the function of the heteromodal association cortex because its contribution to adaptive behaviours typically depends more heavily on the integration and interdependency of function within more complex and widely distributed brain systems. While such a neocortical system integration deficit could account for impairments in higher order cognitive and adaptive behaviours, it would also affect other systems that are dependent upon widely distributed brain systems such as those that perform sensorimotor transformations. Because sensorimotor systems are well understood both neurophysiologically and anatomically, and sensory inputs and motor responses are more easily quantifiable than most higher cognitive processes, an examination of these systems offers advantages as an approach for testing models of functional connectivity deficits in autism, and for determining whether some neural circuits are selectively affected in this illness.

Sensorimotor functions are supported by extensive cortical and subcortical networks despite the relative simplicity of behaviours they subserve. Past studies have documented disturbances in sensorimotor processing in autism involving postural control (Kohen-Raz et al., 1992; Gepner et al., 1995; Molloy et al., 2003). The current study examined pursuit eye movements in individuals with autism. Visual pursuit requires rapid, temporally precise integration of activity within several brain areas in order to visually track moving targets through space. The relevant brain circuitry includes the extrastriate areas of visual cortex devoted to the processing of visual motion information, cortical eye fields and cerebellum that are involved in translating sensory information to motor commands, and the striatum and brainstem which are involved in initiating motor commands. This neurocircuitry is well established based on lesion studies, unit recording studies with behaving non-human primates, and human functional neuroimaging studies (Lisberger et al., 1987; Keller and Heinen, 1991; Ilg, 1997; Berman et al., 1999; Rosano et al., 2002). Due to the well understood circuitry and high demands for functional integration across brain areas, studies of visual pursuit provide a strategy for evaluating functional brain connectivity in autism.

Previous studies have reported impairments in smooth pursuit in autism (Rosenhall et al., 1988; Scharre and Creedon, 1992). These studies, however, used small samples of individuals with autism, and did not employ currently used criteria and procedures for assigning diagnoses. Further, these studies did not systematically examine performance across a range of pursuit tasks and target speeds to examine different aspects of pursuit responses. Visual pursuit is typically described using a two-stage model. During the initiation or ‘open-loop’ stage (first 80–100 ms after pursuit onset), control of pursuit is almost exclusively dependent on sensory analysis of visual motion that is performed primarily by contralateral extrastriate cortex (V5/MT) (Newsome et al., 1985; Carl and Gellman, 1987; Lisberger and Movshon, 1999). Analysis of visual motion information is passed from the MT/MST complex to the parietal and frontal eye fields, each of which project through the pons and the cerebellum to deep brainstem oculomotor areas. After the initial open-loop period of pursuit, the subsequent ‘closed-loop’ stage of sustained pursuit (after the first 100 ms of visual tracking) relies primarily upon memory for target velocity, predictions about target movement and feedback about performance, rather than an analysis of visual motion per se, to control visual tracking. This is important because once a moving stimulus is being tracked, its ‘retinal slip’ or movement relative to eye velocity becomes near zero, thus providing a very poor signal to drive ongoing pursuit of moving targets. The cortical eye fields and cerebellum are believed to play a crucial role during closed-loop pursuit by way of driving pursuit based on predictive signals and an analysis of recent performance.

The current study used three tasks to examine multiple aspects of visual pursuit including the initiation and maintenance stages of smooth pursuit eye movements using a large sample of high-functioning individuals with autism and matched healthy individuals. Open-loop and closed-loop measurements were obtained, and performance on each task was evaluated across a range of target velocities. Clinical neuropsychological tests of manual speed and dexterity, and of visual attention were also administered, and scores on these tests were examined in relation to visual tracking performance.

Methods

Participants

Participants in the current study were 60 individuals with autism and 94 healthy individuals matched for age and verbal intelligent quotient (IQ). All participants were administered an age-appropriate Wechsler Intelligence Scale to assess their intellectual ability (Table 1), and had verbal, performance and full scale IQ scores above 70. There were no significant differences between the autism and healthy groups in gender ratio, verbal IQ, handedness or years of formal education (Table 1).

View this table:
Table 1

Demographic characteristics of individuals with autism and matched healthy individuals

VariableHealthy (n = 94)Autistm (n = 60)Statistics
Age19.32 (11.27)20.05 (11.24)t(152) = 0.39, n.s.
Verbal IQ108.23 (12.18)104.80 (17.00)t(97) = 1.36, n.s.*
Performance IQ106.23 (11.90)97.58 (14.53)t(152) = 4.03, P < 0.001
Full-scale IQ107.89 (12.55)101.67 (15.78)t(152) = 2.71, P < 0.01
Years of formal education9.13 (4.92)9.04 (4.81)t(147) = 0.11, n.s.
Male participants (%)8488χ2(1) = 0.55, n.s.
Right handed (%)9284χ2(1) = 1.86, n.s.
  • Means (standard deviations) are given, except for percentages.

  • * Degree of freedom was adjusted because the assumption of homogeneous variance was not met. n.s. = not significant.

The diagnosis of autism was established by expert clinical opinion verified by results from the Autism Diagnostic Interview—Revised and the Autism Diagnostic Observation Schedule (Lord et al., 1989, 1994, 2000). Individuals with autism were excluded if they had an associated infectious, genetic or metabolic disorder known to cause autistic features such as fragile × syndrome or tuberous sclerosis.

Healthy participants had no personal history of psychiatric or neurological disorder, no family history of autism, and no first-degree relatives with any neuropsychiatric disorder considered to have a genetic component. They had no history of developmental delay, significant problems in school performance, or any sign of learning disability in extensive psychoeducational testing performed with these individuals as described in our previous reports (Minshew et al., 1997). In both the autism and healthy groups, no participants were taking medications known to affect cognitive or oculomotor abilities. No participant had a history of head injury, birth injury or seizure disorder. Informed consent was obtained from all participants, with children and adolescents providing informed assent along with the consent of their parent or guardian, using procedures approved by the Institutional Review Board of the University of Pittsburgh. Far acuity of all participants was normal or corrected to at least 20/40.

In addition to tests of general intelligence, neuropsychological measures of motor skills and attention were administered:

  1. the Grooved Pegboard and Finger Tapping tests to measure manual praxis and motor speed (Mathews and Klove, 1964; Reitan and Wolfson, 1993);

  2. Form A of the Trail Making Test (Reitan and Wolfson, 1993), a Number Cancellation task (Mesulam, 1985) and a Continuous Performance Test (Conner, 1995) to assess visual attention.

We report only those neuropsychological data when participants completed neuropsychological and oculomotor testing within a period of 6 months. Eighty-four percent of the healthy individuals and 92% of the individuals with autism had both assessments within this time frame.

Procedures

Participants were tested in a darkened flat black room and task instructions were provided via intercom. Visual targets were presented in the horizontal plane at eye level on an acrylic hemi-arc with 1-metre radius. Participants were seated at the centre of the arc, and a chin rest with forehead and occipital restraints was used to minimize head movement. A laser diode module was mounted immediately over participants' heads to produce a 3 mm point source of light. The light source was reflected onto the arc by the use of a mirror attached to a rotary stage assembly (catalogue number 1121-142 from New England Affiliated Technologies, Lawrence, MA), which manipulated velocity profiles and locations of targets.

Eye movements were monitored using infrared reflection sensors mounted on spectacle frames (Model 210 from Applied Science Laboratories, Inc, Bedford, MA). Prior to performing experimental tasks and later during testing, fixation data were collected to calibrate eye movement recordings by presenting targets for 5 s at 0, ±3, 6, 9, 12, 15° of visual angle. An experimenter monitored eye movement activity throughout testing to insure that participants were alert and performing tasks according to instructions. Each trial began with presentation of a 0.5 s auditory tone at centre fixation concurrently with the presentation of the visual target to alert participants to prepare to track the target.

Foveofugal step-ramp task

Targets were presented at centre for 2–4 s, and then stepped 3° to the left or right and immediately continued moving in the same direction away from centre at a constant speed of 4, 8, 16, or 24°/s (Fig. 1). The target disappeared after reaching ±15°. One second later, the target reappeared at centre to begin the next trial. The duration of central fixation and the speed and direction of targets were pseudo-randomly assigned for each trial.

Fig. 1

Schematic presentation of a foveofugal step-ramp task. The diagram shows a time (horizontal axis) by target position (vertical axis) plot for a single trial of this task. The dashed line represents target movement with an initial 3° step followed by smooth movement to the left (negative) 15° position. The line above illustrates a typical eye movement response in which an initial catch-up saccade (dotted line) to the target is followed by smooth pursuit.

In the foveofugal step-ramp paradigm, a saccade typically occurs ∼200 ms after the onset of target motion so that the eyes can catch up to the moving target, followed immediately by smooth pursuit. The initial phase of the pursuit response (defined here as the first 100 ms of pursuit after the end of the first catch-up saccade) is referred to as the open-loop stage because it is driven by sensory input rather than by the use of internally generated feedback about performance accuracy and predictions about target motion. The closed-loop stage encompasses all pursuit after the open-loop period. Primary measures of performance taken from the open-loop stage were latency and gain of the first catch-up saccade and pursuit gain during the first 100 ms after the initial saccade. Pursuit gain was defined as the ratio of the average velocity of pursuit eye movement to target velocity. The gain of the initial saccade was calculated as the ratio of eye movement amplitude to target distance. The primary measure for the closed-loop stage was pursuit gain. This task consisted of a total of 32 trials (4 trials × 4 velocities × 2 directions).

Pure ramp task

Targets were presented at the central location for 2–4 s, then swept to the left or right from centre at one of several speeds (4, 8, 16, 24, or 32°/s) and then terminated after reaching ±15° (Fig. 2). The two primary measures for this task were pursuit gain and pursuit latency, with the latency of pursuit initiation defined as the time for the eyes to develop and maintain a velocity of at least 2°/s for 20 ms. Forty trials were performed (4 trials × 5 velocities × 2 directions).

Fig. 2

Schematic presentation of a pure ramp task. The diagram shows a time (horizontal axis) by target position (vertical axis) plot for a trial with a target moving from centre to a left (negative) 15° position. The dashed line shows target movement. The line above shows eye movement in a typical trial where smooth pursuit is sometimes interrupted by catch-up saccades that correct for tracking error (dotted line).

This task differed from the foveofugal step-ramp paradigm in that the target began moving smoothly from the centre without first stepping abruptly from the centre. This task allows examination of pursuit initiation before the occurrence of catch-up saccades. However, because participants sometimes make saccades to the target before initiating smooth pursuit in this task, which delays pursuit onset in proportion to the duration of the first catch-up saccade, the average latency to initiate pursuit was computed based only on trials in which smooth pursuit initiation preceded the first catch-up saccade.

Oscillating target task

In this task, the target oscillated back and forth across the arc for ∼22 s at each of four speeds (8, 16, 24 or 32°/s). When the target reached ±12° from the centre, it gradually decelerated to reverse directions at ±17°, at which point it immediately accelerated until reaching ±12° at which time the original constant speed was resumed as the target swept back across the arc. The 8°/s stimulus was repeated to obtain a sufficient number of sweeps back and forth across the arc for analysis. Pursuit gain was the primary measure taken from this task. Since the target was accelerating or decelerating between ±12 and ±17° of visual angle from the centre, pursuit gain was calculated only from pursuit performance between ±10° when target speed was constant.

Eye movement measurement

Eye movement recordings were digitized at 500 Hz with a 14 bit A/D converter (DI-210 from Dataq Instruments, Akron, OH). The position trace was differentiated to calculate eye velocity and acceleration, and each of these signals was smoothed using a custom finite impulse response filter to increase the signal to noise ratio. The filter had a non-linear transition band (from pass to no pass) between 20 Hz and 65 Hz for velocity and position data, and 30 Hz and 65 Hz for acceleration data. Data from each trial were reviewed to detect and eliminate artefacts (e.g. blinks, periods of inattention/discontinued task performance). Saccades were identified from the point when eye acceleration exceeded 1000°/s2 until 25% of peak deceleration. The resolution of measurement permitted detection of saccades on the order of 0.20–0.25°. Saccades were excluded from epochs used to measure pursuit gain. Data obtained within the 10 ms after the end of saccades were also excluded from data used to measure pursuit responses to minimize the effects of post-saccadic drift and signal ringing from digitization. Blinks were identified using electrodes placed immediately above and below the left eye.

Data analysis

Measures obtained from performance of the eye movement tasks were averaged across trials and analysed using 3-way mixed ANOVA (analyses of variance) [Group (autism versus healthy) × Target Speed × Direction (rightward versus leftward)]. To correct for the possible sphericity violation, degrees of freedom were corrected using the Huynh-Feldt correction. All post hoc comparisons were performed using Bonferroni probability correction with a family-wise Type 1 error rate of 0.05. Degrees of freedom for t tests for the post hoc analyses were corrected when the homogeneity of variance assumption was violated. The focus of the subsequent analyses was on effects pertinent to subject group differences. Therefore, task effects such as the robust effect of target velocity on tracking performance are not presented in detail unless their interactions with the group effect were significant.

Results

Foveofugal step ramp task

Significant Group × Direction interactions were identified in the analysis of performance during the open-loop stage [F(1,138) = 5.07, P < 0.05 for open-loop pursuit gain and F(1,146) = 9.33, P < 0.01 for the gain of primary catch-up saccades]. For both measures, individuals with autism showed poorer performance only when tracking targets into the right visual field (Figs 3 and 4). The Group × Direction interaction was not significant for the latency of the primary catch-up saccade; nor were there significant overall group differences in saccade latencies.

Fig. 3

Open-loop pursuit gain while tracking foveofugal step ramp targets moving at several target speeds in individuals with autism and matched healthy individuals.

Fig. 4

Gain of initial catch-up saccades when tracking foveofugal step-ramp targets moving at several target speeds in individuals with autism and matched healthy individuals.

Individuals with autism had lower closed-loop pursuit gain on this task than healthy individuals [F(1,146) = 5.04, P < 0.05]. The Group × Target Direction interaction was not significant, indicating that the reduced closed-loop pursuit gain was similar for leftward and rightward pursuit. The Group × Target Speed interaction was also not significant, indicating that the relative reduction in closed-loop pursuit gain in individuals with autism was consistent across target speeds, rather than progressively increasing with target speed.

Pure ramp task

The proportion of trials where smooth pursuit initiation preceded the first catch up saccade was modestly higher in individuals with autism (55%) than healthy individuals (46%) [F(1,143) = 5.42, P < 0.05]. Because only approximately half the trials could be used to compute pursuit latency and because target velocity effects on pursuit latency were modest, pursuit latencies were aggregated across target velocities for each participant. A Group × Target Direction ANOVA was then performed. Neither Group nor Group × Direction effects were significant (F < 1), failing to provide evidence for reduced pursuit latencies in autism. No group difference in the latency to initiate the first catch up saccade was observed (F < 1).

Individuals with autism had lower pursuit gain than healthy individuals [F(1,147) = 18.64, P < 0.001] (Fig. 5). As with closed-loop pursuit data in the foveofugal ramp task, the Group × Direction and Group × Target Speed interactions were not significant (F < 1).

Fig. 5

Smooth pursuit gain during tracking of pure ramp targets moving at several target speeds in individuals with autism and matched healthy individuals.

Oscillating target task

Consistent with the analysis of sustained pursuit from the ramp tasks, individuals with autism had lower pursuit gain than healthy individuals when tracking predictable oscillating targets [F(1,150) = 11.21, P < 0.01] (Fig. 6). This effect did not differ between leftward and rightward pursuit, nor did the group difference vary with target speed (F < 1).

Fig. 6

Smooth pursuit gain during tracking of oscillating targets moving at several target speeds in individuals with autism and matched healthy individuals.

Developmental trends

To test for potential differential maturation of pursuit abilities in individuals with autism, participants were divided into two age groups (8–15 years old and 16 years or older) to examine age-related differences in performance. The age cut-off used resulted in dividing each subject group roughly in half (51 younger and 43 older participants in the healthy group, and 28 and 32 in the autism group, respectively). In addition, this age cut-off approximates the age when pursuit performance typically reaches adult levels (Ross et al., 1993). For open-loop pursuit gain in the foveofugal ramp task, the Age Group × Subject Group × Target Direction interaction was not significant. However, the same three-way interaction for saccade gain in the foveofugal ramp task was significant [F(1,147) = 3.96, P < 0.05]. The poorer performance by individuals with autism making saccades toward targets in the right hemifield was only statistically significant among the younger participants [F(1,74) = 12.35, P < 0.01].

Significant Age Group × Subject Group interactions for closed-loop pursuit gain were observed for all tasks [F(1,147) = 3.90, P < 0.05 for the foveofugal ramp task, F(1,148) = 6.74, P < 0.05 for the pure ramp task, and F(1,149) = 4.00, P < 0.05 for the oscillating target task]. In all tasks, the reduction in closed-loop gain relative to age-matched healthy individuals was only significant in the older group [t(51) = 2.72, P < 0.01 for the foveofugal step ramp task, t(49) = 4.33, P < 0.001 for the pure ramp task, and t(43) = 3.52, P < 0.01 for the oscillating target task] (Fig. 7), and resulted from age-related maturation in healthy individuals that was less robust in individuals with autism.

Fig. 7

Smooth pursuit gain during tracking of oscillating targets in younger (<16 years old) and older individuals (16–53 years old) with autism and matched healthy individuals.

As an alternative strategy for examining age effects, regression models were applied to characterize developmental changes in performance. Inverse regression models were chosen based on our previous research, which demonstrated that the model was best suited to delineate non-linear trajectories in cognitive development (Luna et al., in press). The inverse regression models were significant for closed-loop pursuit from all tasks in healthy participants [F(1,91) = 5.30, P < 0.05 for the foveofugal step ramp task, F(1,91) = 5.33, P < 0.05 for the pure ramp task, F(1,91) = 12.98, P < 0.001 for the oscillating target task]. The models were, however, not significant in participants with autism (F < 1 for all tasks), consistent with the reduced normal developmental trajectories in autism suggested in the other analyses of age effects.

Relationships between neuropsychological test scores and eye movement measures

As we have reported previously (Minshew et al., 1997), individuals with autism performed more poorly than healthy individuals on measures of motor and attention skills (Table 2). One approach for examining the broader significance of oculomotor deficits discussed above is to consider their relationship to performance on neuropsychological measures of manual motor skills and visual attention. Because of the relatively large number of eye movement measures, we focused in the following correlational analyses on selected key pursuit variables based on the results of the primary group comparisons. This reduced the risk of Type I error for the correlational analyses. We also used the Bonferroni probability correction to correct for multiple comparisons as in the previous analyses.

View this table:
Table 2

Neuropsychological test scores of individuals with autism and matched healthy individuals

Neuropsychological parameterHealthy (n = 94)Autistm (n = 60)
Commission errors from CPT16.69 (9.38)18.69 (8.04)t(66) = 0.92, n.s.
Omission errors from CPT8.23 (7.32)11.10 (11.11)t(66) = 1.28, n.s.
Hit reaction time from CPT (ms)394.27 (89.91)381.35 (71.76)t(66) = 0.64, n.s.
Total time from trails A (s)19.58 (8.49) (Median: 19.00)32.88 (19.83) (Median: 29.00)Z = 4.37, P < 0.001*
Total time from number cancellation (s)468.72 (155.73)562.11 (159.99)t(116) = 3.12, P < 0.01
Digit span (forward + backward; age corrected scaled score)11.14 (2.63)10.64 (3.46)t(95) = 0.91, n.s.
Finger Tapping dominant hand47.85 (8.70)43.91 (8.75)Hand: F(1,131) = 45.59, P < 0.001
Finger Tapping non-dominant hand43.68 (7.53)41.63 (8.61)Group: F(1,131) = 4.63, P < 0.05
Hand × Group: F(1,131) = 3.89, n.s.
Grooved Pegboard dominant hand73.62 (14.76)89.13 (24.86)Hand: F(1,130) = 26.19, P < 0.001
Grooved Pegboard non-dominant hand79.74 (15.67)97.35 (32.75)Group: F(1,130) = 19.74, P < 0.001
Hand × Group: F < 1
  • Means (standard deviations) are given.

  • * Mann–Whitney test was used because the distributions were skewed. n.s. = not significant. CPT = continuous performance test.

Rightward open-loop pursuit and primary saccade gain from the foveofugal ramp task and pursuit gain from the pure ramp and oscillating target tasks were examined in the correlational analyses. Closed-loop pursuit gain from the pure ramp and oscillating tasks were averaged across directions and speeds because Group × Target Direction, Group × Target Speed, and Group × Target Direction × Target Speed interactions were not significant in the analyses reported above. Measures from the Finger Tapping and Grooved Pegboard tests were also averaged across left and right hands because group differences on these tasks were similar for right and left hand performance (Table 2).

For individuals with autism, pursuit gain from all tasks (open- and closed-loop) was significantly correlated with measures of motor praxis from the Grooved Pegboard (time to complete pegboard task; value for correlation coefficient, r, ranged from −0.42 to −0.49; Table 3), whereas the correlations with the measure of simple motor speed provided by the Finger Tapping test (taps per 10 s periods) were not significant. Primary saccade gain was negatively correlated with the speed of finger tapping (r = −0.48, P < 0.001). In the healthy group, correlations between pursuit gain and the praxis test (Grooved Pegboard) were not significant, but modest correlations between simple motor speed (Finger Tapping) and pursuit measurements were observed (r = 0.34–0.39). No correlation between data from visual attention and eye movement tasks was significant for either group.

View this table:
Table 3

Correlation coefficients between measures of manual motor skills and visual tracking performance in individuals with autism and matched healthy individuals

Foveofugal step-ramp taskPure ramp taskOverall pursuit gainOscillating target taskClosed-loop pursuit gain
Primary saccade gain rightwardOpen-loop pursuit gain rightward
Healthy groupFinger Tapping average0.20 (P = 0.086)0.39 (P < 0.001)0.37 (P = 0.001)0.34 (P = 0.002)
Total time Grooved Pegboard average0.20 (P = 0.075)−0.26 (P = 0.020)−0.39 (P < 0.001)−0.24 (P = 0.033)
Autistm groupFinger Tapping average−0.48 (P < 0.001)0.19 (P = 0.160)0.12 (P = 0.393)0.23 (P = 0.091)
Total time Grooved Pegboard average0.06 (P = 0.678)−0.42 (P = 0.001)−0.49 (P < 0.001)−0.43 (P = 0.001)
  • Bold indicates significant correlation (after Bonferroni probability correction).

Thus, these results suggest that pursuit deficits in autism are related to problems of praxis rather than deficits in visual attention. Because of the correlations with pursuit variables, developmental effects on motor skills were also examined using the same age cohorts used for the pursuit variables. Healthy individuals showed more robust age related improvement in praxis than individuals with autism, although the difference was not statistically significant [F(1,141) = 3.515, P < 0.07]. There was an age related improvement in finger tapping in both groups, but the rate of development was similar (F < 1).

Correlations between IQ scores (verbal, performance and full scale) and eye movement measures were computed using eye movement measurements on which significant group differences were found in previous ANOVAs (pursuit gain from the pure and oscillating tasks, and initial gain of saccades and pursuit in rightward trials from the foveofugal task). None of the correlation coefficients were significant (r = −0.13 to 0.17 for the healthy group, and from −0.05 to 0.21 for the autism group), indicating that oculomotor impairments in individuals with autism did not result from the relatively small between-group IQ differences, and that the eye movement deficits in this sample of high functioning individuals were not related to any general intellectual impairment.

Discussion

The current study investigated the integrity of the pursuit eye movement system in autism by comparing the performance on three pursuit eye movement tasks by large samples of high-functioning individuals with autism and healthy individuals matched on age, verbal IQ, sex, and handedness. The findings document deficits in open-loop pursuit when tracking targets into the right visual field, bilateral deficits in closed-loop pursuit that were not related to target direction, and an association of pursuit deficits with poorer manual praxis in individuals with autism. Closed-loop pursuit deficits were more pronounced in older participants (≥16 years) in all three pursuit tasks, suggesting a maturational disturbance of relevant circuitry in individuals with autism. Lateralized deficits in open-loop pursuit were consistent across age groups, while the lateralized effects in primary saccade gain were more pronounced in younger individuals with autism. Differences in the developmental trajectories and laterality in the open- and closed-loop pursuit deficits suggest that these abnormalities have a fundamentally different pathophysiology.

Open-loop pursuit deficits

Individuals with autism showed deficits in the initial open-loop stage of pursuit only when tracking targets moving into the right hemifield. This hemifield-specific abnormality probably reflects problems in sensory processing or sensorimotor transformation rather than pure motor deficits, because the lateralized pattern of dysfunction did not generalize to closed-loop pursuit. Neocortical control of saccades is primarily organized in each hemisphere to control movement in the contralateral direction (e.g. the left hemisphere controls rightward saccades), while the neocortical pursuit system primarily controls ipsilateral movement (Dürsteler et al., 1987; Kurylo and Skavenski, 1991; MacAvoy et al., 1991; Thier and Andersen, 1998; Schiller and Chou, 2000; Ilg and Thier, 2003). Thus, unilateral lesions in cortical eye fields result in lateralized deficits in opposite directions for pursuit and saccades. Our observation that pursuit and saccade deficits were in the same rightward direction rather than in opposite directions, therefore suggests that the deficit observed during open-loop pursuit results from a dysfunction prior to response preparation performed in the cortical eye fields.

Difficulty initiating accurate pursuit during the open-loop stage might involve the cerebellum (Takagi et al., 2000). The cerebellar vermis is involved in controlling saccades and smooth pursuit in the ipsilateral direction (Krauzlis and Miles, 1998), and thus could produce saccade and pursuit deficits in the same direction during the open-loop stage. However, because of its proximity to the final motor pathways, laterality of open- and closed-loop deficits is likely to be consistent if a cerebellar dysfunction were the primary site of pursuit disturbances (Straube et al., 1997). Such a pattern of deficits was not observed in the present study. Thus, the open-loop pursuit disturbance probably results from a problem at another site involved in extracting and passing forward visual motion information for sensorimotor transformation. This implicates extrastriate areas such as MT/V5 dedicated to the processing of visual motion information, or the pathways through which this information is passed onto sensorimotor areas.

Monkeys with experimental unilateral MT/V5 lesions demonstrate motion perception deficits that cause hemifield specific reduction in saccade accuracy and open-loop pursuit responses during foveofugal step-ramp tasks similar to those observed in our participants with autism (Figs 3 and 4) (Newsome et al., 1985; Dürsteler et al., 1987; Dürsteler and Wurtz, 1988). Past studies have reported that some aspects of motion perception are compromised in autism, though the laterality of this deficit has not yet been systematically examined (e.g. Carlin et al., 1999; Milne et al., 2002; Blake et al., 2003). Some of these studies used small samples and included low-functioning individuals with mental retardation. Because individuals with multiple forms of mental retardation are known to have difficulty in visual motion perception (Fox and Oross, 1990; Sparrow et al., 1999), some prior studies are difficult to interpret with regard to specific deficits in autism and their potential relationship to our findings. A recent study with high-functioning individuals with autism by Bertone et al. (2003) did not observe impairment in motion perception for simple, first-order visual motion information such as were used for the current study, and reported deficits that were specific to higher levels of motion processing. This pattern of findings suggests that, while basic motion information can be extracted from visual displays such as those used in the current study, it can not be passed forward or used by brain systems further along in the information processing stream.

In addition to reduced pursuit and saccade gain during open-loop pursuit, monkeys with MT/V5 lesions show longer latency to initiate pursuit that is specific to the receptive field represented by damaged neurons (Lisberger and Pavelko, 1989). Prolonged pursuit latency has been also reported in patients with lesions in MT/V5 (Heide et al., 1996) and in non-human primates when the quality of visual motion signal was degraded (Churchland and Lisberger, 2000). Our failure to detect increased pursuit or saccade latencies in either hemifield provides further support for the view that the open-loop pursuit impairments do not result from a disturbance in the sensory processing of visual motion information. Intact response latencies also indicate that lower gain during pursuit initiation is not related to a disturbance in the speed of information transfer along visual or sensorimotor pathways as might be associated with gross white matter pathology. Lower open-loop gain during the initial stage of pursuit in individuals with autism seems most likely to result from a disturbance in the transfer of visual motion information from sensory to sensorimotor systems, which affects the fidelity or resolution of visual motion information. This is consistent with the model that disturbances of brain maturation in autism compromise the function of long track pathways connecting widely distributed but functionally linked brain regions, and provides evidence that this disturbance may be selective for some pathways.

The associations between the pursuit deficits and impairments in praxis suggest that the observed pursuit deficits result from a more general problem in visual sensorimotor transformation rather than a highly selective disturbance in the sensory analysis of visual motion information or pursuit control. Further, these findings suggest that impairment in sensorimotor integration is not specific to types of visual information (visual motion versus visual spatial information) or types of motor output (eye versus hand movement). The observation that the accuracy of primary catch-up saccades was negatively associated with simple motor speed in autism was unexpected. As motor speed is less dependent on premotor systems than praxis skills, one possible interpretation of this pattern of observations is that it represents a biased neurodevelopmental emphasis on motor relative to premotor systems in autism.

Our finding of abnormal open-loop pursuit suggests that the sensorimotor transformation dependent upon pathways projecting from area V5/MT in the left hemisphere is abnormal, while those dependent upon projections from homologous right extrastriate cortex are not. Consistent with this observation, Rinehart et al. (2002) reported prolonged manual response latencies to stationary visual cues presented in the right hemifield in individuals with autism. Thus, findings from both the current study and those of Rinehart et al. (2002) indicate that individuals with autism may have a problem utilizing visual information from the right visual hemifield to produce accurate motor responses.

This observation requires replication, but given the prominent language and communication disturbances associated with autism-spectrum disorders, it is interesting to speculate that both of these deficits may point to greater deficits of functional connectivity in the dominant hemisphere. These findings suggest a potential dysregulation in the local circuitry of the output layers of visual cortex, in the long fibre tracts carrying visual information forward to sensorimotor areas, or intrinsic to the projection zones of these fibre tracts in sensorimotor areas that may be greater in the left hemisphere in autism.

Relevant to this model, several studies have reported evidence for atypical dominance for language processing in autism (Müller et al., 1999; Herbert et al., 2002) and reduced functional connectivity in left hemisphere language areas (Just et al., 2004). While signs of lateralized brain abnormalities have not been consistently reported in autism, the neurophysiological paradigms and measurements in the present study require a high degree of temporal coordination of activity and information processing across brain regions, and thus may be more sensitive to hemisphere-specific abnormalities than neuropsychological testing or volumetric brain measurements.

Of note, however, a recent study reported that white matter volume reduction was greater in the left hemisphere in Asperger's patients (McAlonan et al., 2002). The affected area included posterior cortex near the left MT complex and long fibre tracts connecting posterior sensory areas with parietal and frontal cortex, which include the output pathway from the MT/V5 to the cortical eye fields. While this finding was with Asperger's disorder rather than the typically more severely impaired high-functioning individuals with autism, this type of regional abnormality in white matter is consistent with the interpretation that the unilateral deficit found in the open-loop stage may stem from disturbances in transferring visual information within the left hemisphere for oculomotor control. In this regard, it is interesting to note that there are differences in the timing of early maturational processes across the neocortical hemispheres (Paus et al., 1999). This raises the possibility that a disturbance in genetically determined neocortical maturation could occur at a point in time when it could have greater impact on the maturation of pathways in the left hemisphere.

Closed-loop or sustained pursuit tracking deficits

In addition to deficits during pursuit initiation, individuals with autism had poorer closed-loop pursuit gain in all three tasks. This effect was not hemifield specific. Closed-loop pursuit requires ongoing and sustained modulation of visual tracking to optimize pursuit accuracy, and relies primarily upon perceptual feedback about tracking error and dynamic predictions about target motion, rather than sensory information about target motion. Thus, it relies upon effective communication between the multiple cortical and subcortical areas providing information about error monitoring and anticipation of target movement.

Frontal eye fields situated at the border of premotor and prefrontal cortex are known to play an important role in controlling pursuit eye movements (Lynch, 1987; Shi et al., 1998; Tanaka and Lisberger, 2001; Rosano et al., 2002), especially in the use of predictive information for controlling pursuit (Keating, 1993). Maturational disturbances in the frontal lobe have been reported in autism, as have executive function deficits on neuropsychological and oculomotor tests (Zilbovicius et al., 1995; Ozonoff, 1997; Minshew et al., 1999; Goldberg et al., 2002), consistent with the reduced maturational progress in closed-loop pursuit observed in the present study.

However, disturbances in the basal ganglia and cerebellum might also contribute to closed loop deficits. The cerebellum plays a significant role in closed loop pursuit (Lekwuwa et al., 1995; Straube et al., 1997), and cerebellar abnormalities have been among the most consistent histopathological findings in autism (Ritvo et al., 1986; Bailey et al., 1998; Kemper and Bauman, 1998). Furthermore, polymorphisms in ENGRAILED 2, a homeobox gene mapped to 7q36 that is essential in cerebellar development, has recently been linked to autism (Gharani et al., 2004). The basal ganglia play an important role in predictive and motor skill learning and, therefore, in sustained pursuit, and thus might contribute to pursuit impairments.

Implication for autism

Recent studies on brain growth in autism have reported abnormal developmental trajectories characterized by early accelerated increase in grey and white matter volume (Carper et al., 2002; Herbert et al., 2004). While the long-term consequences of this dysmaturation are not known, it seems likely to interfere with the organization and pruning of the terminal fields of long associational tracts, which would disrupt temporally precise coordination of activity across elements of brain systems that produce adaptive behaviours. The current study provides examples of functional connectivity deficits in the sensorimotor domain by documenting a selective disruption of open-loop pursuit in the right hemifield and a developmental failure in the sensorimotor systems mediating closed-loop pursuit. Even though brain areas that have been implicated in autism overlap with the neural substrate of visual pursuit, no intrinsic abnormalities in any single area could produce the oculomotor deficits we observed with their different lateralization and developmental profiles. The findings are more consistent with reduced functional connectivity within the visual pursuit system caused by brain maturational disturbances, and support the model that dysmaturation of distributed networks is a fundamental characteristic of autism.

Acknowledgments

This research was funded by an NICHD Collaborative Program of Excellence in Autism (HD35469) and grants MH01433 and NS33355 from the National Institutes of Health, and the National Alliance for Autism Research.

References

View Abstract