Purchase this article with an account.
Gabriel Diaz, Brett Fajen, Dennis Ehlinger; Learning to anticipate the actions of others: The goal-keeper problem. Journal of Vision 2009;9(8):608. doi: 10.1167/9.8.608.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
When humans observe the actions of others, they can often accurately anticipate the outcome of those actions. This ability is perhaps best exemplified on the sports field, where athletes must anticipate the actions of their opponents based in part on the complex movement of the opponent's body. In this study, we investigated the process by which actors learn to better anticipate the actions of others. Specifically, we tested the hypothesis that learning in this domain can be understood in terms of the ability to perceptually tune to more reliable sources of information. We investigated this matter within the context of blocking a penalty kick in soccer. Because of the extreme time constraints in a soccer penalty kick situation, the goal-keeper must anticipate the direction in which the ball is kicked before the ball is contacted, forcing him or her to rely on the movement of the kicker. We used a motion capture system to record the joint locations of experienced soccer players taking penalty kicks. The data were then used to create videos that depict the keeper's view of a point-light kicker approaching and kicking a ball. Each trial in the experiment displays a video that terminates upon foot-to-ball contact. The subject's task is to judge whether the ball was kicked to the left or right. The stimulus set included kicks from one individual or many individuals, and feedback was provided in some conditions but not others. By correlating judgments with a previously identified set of candidate cues, we can investigate changes in the sources of information upon which observers rely with practice, as well as differences between experienced and inexperienced players. We will also discuss ongoing efforts to use machine-learning techniques as an assumption free method to extract non-local cues that reliably specify the intended action.
This PDF is available to Subscribers Only