Text

theses

Smooth pursuit eye movements are used to maintain gaze on moving targets. In order to overcome processing delays, smooth pursuit is capable of predicting future target motion. While work over the years has characterized the nature of some of the cues (verbal, visual, probabilistic) to future motion, the integration of prediction with immediate sensory inputs is not well understood.

The current study tested the extent to which prediction during pursuit can be modeled as a form of optimal Bayesian cue combination, with prior and likelihood cues trading off in proportion to their respective reliabilities. Stimuli were random dot kinematograms (RDKs) consisting of 200 dots moving in a chosen mean direction with varying levels of directional variability (Gaussian, with SDs of 0, 30, 45, or 60 deg). The variability of the prior distribution determining the mean directions of the dots on each trial was also varied (Gaussian, with SDs of 10 or 45 deg, or uniform distribution from 0-360 deg). Subjects were instructed to pursue (pay attention to) the RDK and report the perceived mean direction of the RDK at the end of each trial.

The main experimental findings were: (1) The influence of the prior on pursuit persisted for ~250 to 600 ms after the onset of target motion. (2) The influence of the prior decreased over time, with pursuit depending totally or near totally on the immediate stimulus motion (referred to as the “likelihood”) by about 200-500 ms after the onset of target motion. (3) The influence of the immediate stimulus motion (likelihood) overcame that of the prior later in the trial when RDK directions were more variable. (4) Increasing the variability of the prior led to a greater influence of the likelihood on pursuit direction earlier in the trial. (5) When the variability of the prior direction increased, there were indications that the variability of the likelihood had a greater effect on the variability of pursuit directions. (6) Perceptual tests using short duration intervals of motion (150 ms) also found a greater influence of the prior and more variable reports of target direction when the variability of the RDK directions increased.

These findings suggest that basic principles of Bayesian cue combination can apply to smooth pursuit eye movements. These results can be useful for defining the quantitative characteristics of the cortical areas involved in both the representation of current sensory motion and the representation of the prior, as well as identifying the neural processes that are involved in the formation of the pursuit command via the combination of these two cues.

ETD_10510

Ph.D.

Includes bibliographical references

doi:10.7282/t3-n17b-0s85

ETD doctoral

The author owns the copyright to this work.