Recognizing biological movements is a fundamental skill for survival and social interactions. Using combined psychophysics and fMRI, we provide novel evidence that learning shapes biological motion processing across stages of visual analysis in the human brain. We report improved performance after training in discriminating biological movements whose similarity varied parametrically along a spatio-temporal morphing continuum, coupled with increased fMRI selective adaptation to the movement differences. Learning novel human-like movements shaped higher-level processing of known action categories: global movement analysis in hMT+/V5, V3B/KO and generalization of existing representations for prototypical actions to novel exemplars in STSp, FFA. However, learning artificial movements bolstered the formation of novel category representations: integration of local configurations in retinotopic areas, global movement analysis in hMT+/V5, V3B/KO, and processing of biological properties in STSp, FFA. These findings propose distributed experience-based plasticity mechanisms that mediate recognition of complex movements and action understanding in the human visual cortex.

In collaboration with Martin Giese, University of Tuebingen.