Purchase this article with an account.
Hang Zhang, Nathaniel Daw, Laurence T. Maloney; Testing whether humans have an accurate model of their own motor uncertainty in a speeded reaching task. Journal of Vision 2011;11(11):953. doi: https://doi.org/10.1167/11.11.953.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The distribution of the endpoint of a speeded reaching task is typically bivariate Gaussian in form. This distribution determines the subject's probability of hitting targets of any sizes and shapes. We used staircase methods to determine whether subjects can correctly order the probability of hitting targets differing in size and shape in a speeded reaching task, testing whether they had an accurate model of their motor uncertainty.
Experimental phases: Training. Subjects completed 300 trials of a speeded reaching task, in which they touched a circle on a touch screen within 400 milliseconds. Probability Judgment. On each trial, subjects judged which of two shapes, a rectangle and a circle, was easier to hit. The rectangle was either 4:1 (horizontal) or 1:4 (vertical), and had five possible sizes, tuned to each subject's motor variance. We used a staircase method to determine the radius of the circle the subject judged to be as easy to hit as each of the ten rectangles. Area Judgment. An experimental control identical to the previous phase but subjects judged area. Twelve na&ıuml;ve subjects participated.
Results: (1) While the motor error distributions for all subjects in training were anisotropic bivariate Gaussian, elongated in the vertical direction, none of the subjects correctly compensated for their anisotropy in judging the probability of hitting rectangles differing in orientation. (2) The probability judgments of six subjects were consistent with judgments based on the best-fit isotropic Gaussian approximating their own motor error. (3) The remaining six subjects made judgments consistent with a flatter motor error distribution than their actual. (4) Subjects who accurately estimated probability failed at estimating area and vice versa. Such a relationship between two, in principle, distinct estimation tasks suggests some common process or representation underlying them both.
This PDF is available to Subscribers Only