Abstract
Pupillometry has been almost completely neglected in perceptual learning research despite the well documented correlations between pupil size and cognitive variables such as mental effort. We explored whether and how pupil size varies with perceptual learning of motion direction discrimination.
Method: The stimuli were moving filtered-noise textures with constant mean luminance presented at eccentricity 7 deg either to the Left or Right of fixation. Twenty one observers discriminated motion directions that differed by either 8(Easy) or 4(Hard) degrees. Each participant (N=21 counterbalanced in 4 groups) trained on a particular combination of difficulty and position on days 2 through 5, and was tested on all 4 combinations on days 1 and 6. The pupil diameter was recorded with an Eyelink 1000 eye tracker in a darkened room with consistent ambient light. It was averaged across all fixations within a trial and then normalized across trials within subjects.
Results: The group-averaged d' increased monotonically (Δd'=1) and the reaction time decreased monotonically (ΔRT=200 ms) during training. However, the pupil diameter changed in a non-monotonic inverted-V pattern across sessions (Takeuchi et al., 2011). Moreover, its time course was modulated by task difficulty – the pupil size peaked on Day 3 in the Hard-training group and on Day 4 in the Easy-training group. Switching to the untrained retinal position increased pupil size in the Hard-training group. There was also a pronounced decrease in pupil diameter within each session, replicating the common time-on-task effect. The individual differences in pupil diameter and overall d' were positively correlated (r=.4 calculated on Day 1 when all groups followed an identical protocol).
Conclusion: The inverted-V pattern in the pupil data is qualitatively different from the monotonic learning curves in the behavioral data. This suggests that pupillometry captures a unique aspect of perceptual learning, modulated by task difficulty and stimulus specificity.
Meeting abstract presented at VSS 2012