Abstract
We contrast two accounts of how novel sequences are learned. The first is that learning changes the signal-to-noise ratio (SNR) of existing cortical representations by reducing noise or increasing signal gain. Alternatively, learning might cause the initial representations to be recoded into more efficient representations such as chunks. Both mechanisms reduce the amount of information required to store sequences, but make contrasting predictions about cortical activity patterns. We applied representational similarity analysis to patterns of fMRI activity as participants encoded, maintained, and recalled novel and learned sequences of oriented Gabor patches. We fit four models of sequence representation to the activity patterns of novel sequences and tested how the representation changed as a function of learning. We found no evidence for the SNR-change hypothesis. Instead, we observed that in three loci in the occipital and parietal cortex the same sets of voxels encoded both novel and learned sequences but using different encoding schemes. Our results suggest that sequence learning induces recoding rather than simply strengthening the initial representations.
