Abstract
Accurate and precise eye movements are important for visual perception. Yet, both saccades and smooth-pursuit are affected by sensory and motor noise, and are constrained by limited processing time. Can observers report the degree of accuracy in their eye movements, despite typically being unaware of the eye’s constant movement? To investigate this question, we measured sensorimotor confidence for tracking of unpredictably moving targets that elicited both saccades and smooth-pursuit behaviour. We define sensorimotor confidence as the subjective evaluation of task performance that should take into account sensory and motor uncertainty as well as the sensorimotor goal (e.g., maximise tracking accuracy). We asked human observers to track the centre of a noisy dot cloud with their eyes as it moved horizontally. The trajectory followed a sum of sinusoids (average speed: 9 deg/s). There were 20 unique trajectories, repeated 12 times each, half with direction reversed, presented in a randomised order. After 6 seconds of tracking, observers reported if they thought their tracking performance was better or worse than their estimated average performance in the task. From these sensorimotor confidence judgments, we computed their metacognitive sensitivity, the ability to classify objectively better from objectively worse performance, using an ROC-style analysis (Locke et al., 2020, Cognition). Metacognitive sensitivity was significantly above chance-level (area-under-the-curve of 0.50) for the group, however only marginally so (mean and SEM: 0.56 ± 0.02, n = 7). Investigating the individual repeated trajectories by performing a median-split of the objectively better versus worse repeats, all observers were on average more likely to report “better” for the better half of repeats. Together these results suggest humans have marginal metacognitive sensitivity for tracking eye movements. This limited access to sensorimotor uncertainty may be the cost for humans to make fast and accurate eye movements.