Abstract
Our behavior is guided by the statistical regularities in the environment. Prior research on temporal context effects has demonstrated the dynamic processes through which humans adapt to the environment’s temporal regularities. However, learning temporal regularities not only entails dynamic adaptation to traces of previous individual events but also often requires the extraction and retention of summary statistics (e.g., the mean) of temporal distributions. To investigate these summary representations for temporal distributions and to test their sensitivity to distributional changes, we explicitly asked participants to extract the mean of different distributions of time intervals, which shared the same mean but varied in their variability specifically operationalized by the width and presentation frequency of the intervals. Our findings showed that the variability of the estimated mean increased with the distributions’ variability, even though the actual mean remained constant. We further examined how such learning of temporal distributions modulates EEG signals during subsequent temporal judgments. Analysis revealed that the contingent negative variation (CNV), predictive of single-trial RTs, was correlated with how much individuals’ estimates of the mean were affected by the distributions’ variability. Conversely, the post-interval P2 was not modulated by the distributions but predicted participants’ responses, suggesting that P2 reflects the perceived duration of an interval. Taken together, our results demonstrate not only that humans can accurately estimate the mean of a temporal distribution, but also that the representation of the mean becomes more uncertain as the variability of the distribution increases, as reflected neurally in the preparation-related CNV during temporal decisions.
Competing Interest Statement
The authors have declared no competing interest.