Representational complexity and dynamics of recurrent processing during visual object recognition

Recognising objects is a crucial feature of human perception, achieved by the brain rapidly and accurately despite the complexity of visual environments. This ability, termed cored object recognition, relies on a set of hierarchical operations by the ventral stream, whereby stimuli are progressively integrated in a feedforward fashion. Evidence shows, however, that modelling object recognition with feedforward activity provides a limited account. Research provides a solution with recurrent processing. Recent evidence from functional imaging indeed supports its role in object recognition. Despite the consensus in the field that core object recognition relies on both feedforward and recurrent processing, there is divergence in how to measure their relative input. Dissective studies estimate the recurrence associated with separate visual phenomena, whereas integrative studies compare non-recurrent with recurrent models. Here, we take a significant step towards reconciling these two approaches, as we design a new paradigm that simultaneously compares non-recurrent and recurrent processes of visual recognition, and separates their respective computational role. We combine functional imaging methods (fMRI & EEG) for an acute spatial and temporal measure of the impact of recurrence upon representational transformations in the ventral stream. This series of studies will significantly enhance our understanding of the role of recurrent processing in object recognition.