August 2012
Volume 12, Issue 9
Vision Sciences Society Annual Meeting Abstract  |   August 2012
"What" and "when" in action prediction
Author Affiliations
  • Junzhu Su
    Department of Psychology, UCLA, Los Angeles, USA
  • Jeroen J.A. van Boxtel
    Department of Psychology, UCLA, Los Angeles, USA
  • Hongjing Lu
    Department of Psychology, UCLA, Los Angeles, USA\nDepartment of Statistics, UCLA, Los Angeles, USA
Journal of Vision August 2012, Vol.12, 462. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Junzhu Su, Jeroen J.A. van Boxtel, Hongjing Lu; "What" and "when" in action prediction. Journal of Vision 2012;12(9):462.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Research shows that humans can predict future actions with high temporal precision (Graf et al., 2007). We hypothesized that in many real-life situations (particularly when observing an interaction between others), it may be more important to predict what will happen in the future rather than precisely when it will happen.

In Experiment 1, we showed subjects brief movie sequences of two interacting actors, or just one of those actors, followed by a 400ms occluder. Two static figures then followed: a target depicting the posture as the action would have appeared had it continued for 400ms, and a distractor posture that occurred earlier or later than the elapsed time. Observers were asked to select the figure representing a correct action continuation. When the distractor had occurred in the preceding movie sequence, observers were more accurate in the interaction condition than in the single-actor condition. This result indicates that in the interaction condition, observers could better distinguish between backward and forward action trajectories. However, when target and distractor were both sampled from future postures, subjects were more accurate in the single-actor condition, suggesting a greater temporal precision in the single-actor condition. Experiment 2, was similar in design, but observers were asked to indicate whether after the occlusion the actor, sampled from the future action trajectory, was rotated in depth. Accuracy was higher in the interaction condition than in the single-actor condition, indicating more accurate "what" estimates in the former condition. However, postures that were a natural temporal progression and delayed/advanced postures resulted in similar performance, indicating loss of temporal precision.

Overall, these findings suggest that humans excel in predicting when an action posture will occur in the future when viewing a single actor, but are better able to predict what posture will occur based upon a meaningful interaction between two actors.

Meeting abstract presented at VSS 2012


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.