First evidence of the feasibility of gaze-contingent attention training for school children with autism

Abstract

A number of authors have suggested that attention control may be a suitable target for cognitive training in children with autism spectrum disorder. This study provided the first evidence of the feasibility of such training using a battery of tasks intended to target visual attentional control in children with autism spectrum disorder within school-based settings. Twenty-seven children were recruited and randomly assigned to either training or an active control group. Of these, 19 completed the initial assessment, and 17 (9 trained and 8 control) completed all subsequent training sessions. Training of 120 min was administered per participant, spread over six sessions (on average). Compliance with the training tasks was generally high, and evidence of within-task training improvements was found. A number of untrained tasks to assess transfer of training effects were administered pre- and post-training. Changes in the trained group were assessed relative to an active control group. Following training, significant and selective changes in visual sustained attention were observed. Trend training effects were also noted on disengaging visual attention, but no convincing evidence of transfer was found to non-trained assessments of saccadic reaction time and anticipatory looking. Directions for future development and refinement of these new training techniques are discussed.

Keywords: attention; autism; cognitive training; eye movements.

© The Author(s) 2016.

Figures

Figure 1.

Schematics showing the pre–post tests that were administered: (a) Examples of the ‘boring’ (top) and ‘interesting’ (bottom) stimuli used in the visual sustained attention task. (b) Illustration of the screen layout for the anticipation task (two-location condition). In the four-location condition, objects were presented in the four corners of the screen. (c) Illustration of screen layout for the overlap condition gap-overlap task. In the baseline condition, the central target disappeared as the lateral target was presented. (Colour version available online. DOI: 10.1177/1362361315617880.)

Figure 2.

Schematics of the four training tasks administered. Dashed rectangles and arrows, which were not visible in the original tasks, indicate objects that were moving on-screen: (a) Task 1 (Butterfly). The butterfly (indicated in red) scrolled from left to right as long as the child looked directly at it, with static and moving (indicated in blue) distractors presented in the child’s peripheral visual field. If the child looked to any of the distractors, they disappeared and the scrolling stopped. (b) Task 2 (FlyMe). The purple character (indicated red) scrolled upwards as long as the child looked directly at it. Static and moving (indicated blue) distractors appeared from the top and bottom of the screen. If the child looked to any of the distractors, the purple object sank towards the bottom of the screen and the distractors disappeared. (c) Task 3 (Stars). A target (indicated red) was presented on-screen along with a number of static and moving (indicated blue) distractors. If the child looked to the target within a time window, (s)he received a reward. Both target and distractors changed between trials. (d) Task 4 (Suspects). A target (indicated red) was presented along with a range of distractors. If the child looked to the target within a time window, they received a reward. Once per block of 12 trials, the target changed. Targets from the previous block (indicated yellow) were presented concurrently with the current target, as distractors. (Colour version available online. DOI: 10.1177/1362361315617880.)

Figure 3.

Data quality comparison based on data from the gap-overlap study: (a–c) data from this study, (d–f) data from a comparison study that used identical procedures, in laboratory settings, with typical infants (Wass et al., 2011). (a) and (d) Histograms showing the duration of usable fragment durations that were present in our data (calculated on a block-by-block basis). Markedly, longer usable fragment durations were obtained in the comparison study. (b) and (e) Histograms showing the precision of our data (calculated on block-by-block basis). Markedly, more precise data were obtained in the comparison study. (c) and (f) Gaze maps of usable gaze data obtained during the trial. (g) A schematic of how images were distributed on the screen during the trials. (Colour version available online. DOI: 10.1177/1362361315617880.)

Figure 4.

Scatterplots showing, participant by participant, how performance on the training tasks changed across the course of the training sessions. Best-fit linear regression lines have been superimposed onto the scatterplots.

Figure 5.

Scatterplots showing change in performance in our participants. Individual dots represent individual children. In each case, pre-test performance has been drawn on the x-axis and post-test performance on the y-axis. A 1:1 equivalence line (indicating that performance at pre-test was identical to post-test) has been drawn on each figure. The direction of predicted change following training, based on previous research, has been shown using an arrow in the top-right corner of each graph. (a) Visual sustained attention – first look duration to ‘interesting’ targets. (b) Visual sustained attention – first look duration to ‘boring’ targets. (c) Disengagement latencies. (d) Average saccadic reaction time (gap-overlap task). (e) Anticipations (two-object task). (f) Anticipations (four-object task).