Abstract
Anticipatory pursuit allows accurate eye movement initiation with little delay when the target has a predictable behavior, either in its direction, timing or speed. Anticipatory velocity is less extreme when many target speeds are randomly interleaved within a block, compared to anticipatory velocity in a block with a unique target speed (Heinen, S. J., Badler, J. B., & Ting, W. (2005). Timing and velocity randomization similarly affect anticipatory pursuit. Journal of Vision, 5(6), 493–503). We asked how a parametric manipulation of the probability of a given target speed within a block of trials modifies anticipatory eye velocity. We recorded pursuit eye movements of human subjects for targets moving in a fixed direction but with two different speeds in a gap paradigm. Probability of occurrence of a 15 deg/sec speed over a 5 deg/sec one was 0, 10, 25, 50, 75, 90 and 100% in different blocks. Our results showed that mean anticipatory pursuit velocity (evaluated 80 ms after target motion onset) is proportional to the probability of a given target speed within a block, ranging from about 3 to 6 deg/sec from the lowest to highest probability. This result suggests that anticipatory pursuit behavior is built upon a probability-weighted average of different expectations. We further investigated how different expectations are combined to form a single anticipatory behavior from the point of view of statistical inference modelling. In particular, we compared different models of probability encoding (finite state Markov model and Bayesian integration of evidence) that could underlie the observed anticipatory velocity.
AM was funded by an EU Marie Curie individual fellowship. DS was supported by the Swiss National Science Foundation 10011-107768/1.