Purchase this article with an account.
Michael Landy, Todd Hudson; Error statistics reflect movement coding and prior movement history. Journal of Vision 2011;11(11):970. doi: 10.1167/11.11.970.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Two movement codes appear to underlie reach planning: an endpoint code based on final desired hand position, and a vector code defining the desired movement distance and direction. Previous work on movement coding relies on modified sensory information concerning each code or different movements for studying each code. Instead, we examine predictions for these two coding systems while keeping the task, biomechanics, and sensory inputs constant. We do this by controlling recent movement history, which allows us to examine predictions for the error statistics of the two systems. A vector code describes movements in incommensurate (distance-angle) units. One therefore expects anisotropic reach errors with error covariance aligned with the reach direction, as is often found experimentally. Endpoint-coded reaches are given in spatial position (i.e., x–y) units, which one expects to have roughly equally scaled internal representations and corresponding isotropic xy-errors. Subjects performed 12 repetitions of 36 reaches (six targets, and six start positions equally spaced around each corresponding target), in two ‘groupings’. In ‘target grouping’, all reach repetitions to one target are performed (6 start positions randomly chosen), and then all reaches to another target, until all targets have been completed. In ‘vector grouping’, all repetitions of one vector (i.e., same relative start position) are performed (6 targets randomly chosen), and then another vector, until all six vectors are completed. Although the same 36 reaches are performed in both groupings, the latter provides better practice for a vector coding system, while the former provides practice for an endpoint coding system. Not only do ‘vector-grouped’ and ‘target-grouped’ reaches display the predicted error anisotropy and isotropy (respectively), but target-grouped errors are isotropic even if covariance is computed after sorting by reaches with a common movement vector - suggesting that the more-practiced endpoint code dominates target-grouped reach plans, resulting in isotropic error covariance.
This PDF is available to Subscribers Only