Can pupillometry reveal perturbation detection in sensorimotor adaptation during grasping?

Humans adjust their motor actions to correct for errors both with and without being aware of doing so. Little is known, however, about what makes errors detectable for the actor, and how researchers can know when errors are detected. Here, we investigated pupillometry as an unobtrusive, no-report marker of perturbation detection. We also replicate and extend prior work showing that motor adjustments may mask the very errors they correct for. Participants (<i>n</i> = 48) grasped objects while a visuo-haptic size mismatch was applied either sinusoidally (with a continuously changing perturbation) or abruptly (with a constant perturbation that could more easily be adapted to). When mismatches started abruptly and thereafter stayed the same, participants adapted well but also showed decreasing discrimination performance and decreasing confidence in their responses. This was not the case for sinusoidally introduced perturbations. As hypothesized, parameters that characterize phasic and tonic pupil responses were predicted by stimulus parameters and differed depending on participants' grasping and behavioral responses. In particular, tonic responses were smaller for stronger perturbations but larger in trials with correct and confident responses, whereas phasic responses were correlated positively with perturbation magnitude and response correctness, but negatively with response confidence. However, predicting response characteristics from pupil-dilation features using support-vector machine classifiers was not successful. This shows that although pupillometry may yet prove to be a useful no-report marker of perturbation and error detection, there are some challenges for trial-by-trial prediction.<b>NEW & NOTEWORTHY</b> Actors detecting the need to adjust motor actions can make adjustments more efficiently. We show that pupil dilation during perturbed actions reflected perturbation properties and participants' responses, but trial-wise prediction of responses using pupil-dilation parameters was close to chance. Error signals, in addition to perturbation magnitude, play a specific central role in humans' detection of and meta-cognition about motor perturbations. This is a step toward potentially using pupillometry as a no-report marker of perturbation detection.