Evaluation of self-generated behavior: untangling metacognitive read-out and error-detection.

Abstract

When producing a duration, for instance by pressing a key for one second, the brain relies on self-generated neuronal dynamics to monitor the \"flow of time\". Converging evidence has suggested that the brain can also monitor itself monitoring time. Here, we investigated which brain mechanisms support metacognitive inferences when self-generating timing behavior. Although studies have shown that participants can reliably detect temporal errors when generating a duration (Akdogan & Balci, 2017; Kononowicz et al., 2017), the neural bases underlying the evaluation and the monitoring of this self-generated temporal behavior are unknown. Theories of psychological time have also remained silent about such self-evaluation abilities. How are temporal errors inferred on the basis of purely internally driven brain dynamics without external reference for time? We contrasted the error-detection hypothesis, in which error-detection would result from the comparison of competing motor plans with the read-out hypothesis, in which errors would result from inferring the state of an internal code for motor timing. Human participants generated a time interval, and evaluated the magnitude of their timing (first and second order behavioral judgments, respectively) while being recorded with time-resolved neuroimaging. Focusing on the neural signatures following the termination of self-generated duration, we found several regions involved in performance monitoring, which displayed a linear association between the power of (8-14 Hz) oscillations, and the duration of the produced interval. Altogether, our results support the read-out hypothesis and indicate that first-order signals may be integrated for the evaluation of self-generated behavior.\n\nSIGNIFICANCE STATEMENTWhen typing on a keyboard, the brain estimates where the finger should land, but also when. The endogenous generation of the when in time is naturally accompanied by timing errors which, quite remarkably, participants can accurately rate as being too short or too long, and also by how much. Here, we explored the brain mechanisms supporting such temporal metacognitive inferences. For this, we contrasted two working hypotheses (error-detection vs. read-out), and showed that the pattern of evoked and oscillatory brain activity parsimoniously accounted best for a read-out mechanism. Our results suggest the existence of meta-representations of time estimates.