Neural Correlates of Learning in Biological Motion Recognition

Previous studies have proposed that learning contributes to the recognition of biological movements. We investigated the neural correlates of these learning processes in an event-related fMRI adaptation experiment. This paradigm entails repeated presentation of a stimulus resulting in decreased fMRI responses, compared to stronger responses after a change in a stimulus dimension indicating selectivity of an area to this changed dimension. Novel biological movements were generated by linear combination of triples of prototypical trajectories of human movements (Giese Poggio, 2000). Subjects had to discriminate between identical, very similar, moderately similar and completely different point-light stimulus pairs. By choosing appropriate weight vectors the difficulty of the discrimination task could be accurately controlled. The subjects were scanned before and after a training period of three days. Areas relevant to the processing of biological motion (early visual areas, hMT+/V5, KO, FFA, and STSp) were localized as regions of interest using standard mapping techniques. Before training, the observers discriminated successfully between the stimuli in the moderately similar and completely different stimulus pairs but not between the very similar pairs. Consistent with these psychophysical data, we observed significantly stronger fMRI responses for moderately similar and completely different stimulus pairs but not for very similar pairs compared to identical pairs in the FFA and STSp. However, after training the observers’ discrimination performance for very similar stimulus pairs improved and significantly stronger fMRI responses were observed for this condition compared to identical stimulus pairs in all the higher visual areas. These results suggest that learning tunes neural populations in the STS and FFA for the discrimination of novel biological movements