Knowlton, B.
UCLA
While much of the focus of sensorimotor sequence learning has been on sequence-specific learning, a primary goal of most real-world sequence learning is the transfer of knowledge to new sequences. For example, when a child learns to play the piano, she may practice a set of pieces, but this practice will also result in better performance on new pieces as well. In this talk, I will describe work investigating sequence learning transfer. In all experiments, we used the serial reaction time task, in which subjects press keys corresponding to cued locations that appear according to fixed sequences. We found that practice interleaving sequences resulted in better performance on new sequences than when practice on training sequences was blocked. Interleaved practice may result in a more generalized motor program that can better support the learning of new sequences. In a second study, subjects practiced interleaved sequences in an fMRI scanner followed by a transfer test on new sequences. We found that the magnitude of transfer was correlated with activation in a large region of the cerebellar midline; Thus, those subjects that engaged the cerebellum more during practice showed better subsequent transfer. In contrast, there was no correlation between the degree of cerebellar activation and sequence-specific learning, which appeared to be more related to activation of motor cortical regions during practice. These results will be discussed in terms of the role of the cerebellum in motor learning.