Emmanuel Procyk has invited Ignasi Cos from ISIR (Institut des systèmes intelligents et de robotique) in Paris to come and give a talk:
When to dwell and when to move: finding comfort in variability.
Understanding how the CNS predicts the duration of time intervals and defines their motor content is a non-trivial task. For example, professional baseball batters adjust the duration and speed of their strike movement according to their prediction of ball movement time, as expert waltz dancers do to synchronize the end of each loop on beat with the music, as do orchestra players guided by the tempo dictated by their director. Irrespective of the differences in behaviour, these tasks highlight the common problem of predicting a temporal interval and of producing a set of controlled movements that fits therein. Although experimental evidence shows that the CNS can estimate time intervals, these predictions become increasingly inaccurate as the duration of the interval expands (Ivry, 1996; Hazeltine et al., 1997). Despite this limitation, skilled musicians or dancers seldom fail in their predictions and on defining the content of their intervals. How does the CNS do to make this possible? On the way to answer this question, I will show the results of a recent set of experiments in which subjects were instructed to tap repetitively at different frequencies and amplitudes, synchronized with a metronome or in an autonomous fashion. Our main result shows that subjects predicted the duration of intervals within cycle by trading-off the variability of intervals of movement and dwell during each cycle to minimize the error of the overall estimate. I will first describe the experimental results, focusing on the subjects’ behavioural distribution of variability. Second, I will introduce a theoretical model that reproduces these results by seeking the optimization of variability across each tapping cycle. Although this demand further analysis, the average subjects’ behaviour was compliant with the principle of overall variability minimization implemented by the model. This suggests that subjects are subjectively knowledgeable about the variability of their temporal estimates, both for movement and rest intervals, and that they exploit the statistical specificities of each of them to predictively improve the overall performance of their temporal estimates.