Abstract of this talk: A fundamental unresolved problem in consciousness research is how to account in neuronal terms for the qualitative aspects of experience. In the specific case of conscious face perception one could state this question as identifying the neuronal mechanisms that endow a specific face image with its unique appearance. Here I will present evidence supporting the hypothesis that face appearance in consciousness is determined by its unique location in neuronal face-space, i.e. to the similarity distances of its neuronal activation pattern to all other perceptually-driven patterns. I will present evidence obtained from intra-cranial IEEG recordings from the visual system, obtained in patients during clinical diagnosis procedures. These recordings reveal a surprising "convergent evolution" between the face-space geometry of human high order, face-selective, sites and artificial Deep Convolutional Networks trained for face recognition. Furthermore, perceptual similarity measures are found to be correlated to the neuronal face-space geometry revealed through these recordings. These results support the hypothesis that conscious face appearance is indeed defined by its unique position in the individual's face space.
However, it is important to note that the "space" revealed through recording electrodes is in fact an arbitrary structure, constructed by the experimentalist analyzing the recordings. The critical question is how such a space may be constructed intrinsically, through inner brain network dynamics. Here I will propose that the binding of individual neurons into a perceptually-relevant space is done through recurrent signal exchange among all the neurons participating in the relevant space. I will show data obtained from intra-cranial recordings as well as fMRI indicating that such recurrent "ignition-like" dynamics is directly linked to the transition from non-conscious to conscious perception.
Finally I will present data supporting the notion that face perception is linked to recurrent activity confined to high order face-selective regions. Using intracranial recordings during spontaneous blinks that serve as "ecological illusions" I will demonstrate that early visual area activity is not an integral part of conscious visual perception. I will further show that ignitions in fronto-parietal areas are specialized for various post- perceptual processes, such as motor-planning, decisions and arousal but not for visual perception proper.
To summarize: our work points to local recurrent activity, binding patterns of content-selective neurons, as the central dynamics underlying and defining the appearance of perceptual content in the human brain.
Keywords: Mathematical Consciousness Science Online Seminar Series