A progression of approximations to internal models of complex visuo-motor transformations.

Previous research suggests that the acquisition of an internal model of a complex visuo-motor transformation might proceed as a progression of approximations. Here, we test this assumption by comparing performance of three groups of participants during practice and subsequent open-loop tests with three different types of visuo-motor transformation: the kinematic transformation of a sliding first-order lever, a point-symmetric transformation, and a line-symmetric transformation. With the point-symmetric transformation, participants showed the most accurate performance from early practice on and essentially no practice benefit. Movements were not systematically biased toward one of the other transformations. With the line-symmetric transformation, the initially poorer accuracy reached the level of the participants practising the point-symmetric transformation rather rapidly. Early in practice movements according to a point-symmetric transformation were observed. In a similar way the accuracy of participants practising the sliding-lever transformation caught up, but remained worse at the end of practice, nevertheless. With this transformation movements according to a point-symmetric transformation were again observed early in practice, and the movements remained biased toward the line-symmetric transformation. These findings support the notion that the acquisition of an internal model of a complex visuo-motor transformation proceeds in a progression of approximations to the accurate internal representation.