Purchase this article with an account.
Fraser Smith, Lucy Petro, Lars Muckli, Vicky Adams; Decoding facial expressions across non-overlapping face features in early visual cortex. Journal of Vision 2018;18(10):913. doi: 10.1167/18.10.913.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Neurons in the early sensory areas are subject to multiple sources of influence, spanning information coming from the sensory input, local processing and higher brain regions. Visual occlusion is a particularly challenging problem that is thought to require the involvement of such top down and recurrent connections. In the present study we investigated the role of early visual and higher level brain regions in processing occluded facial expressions, using a novel paradigm that allows the dissociation of bottom-up from recurrent influences (see Greening et al, Cortex, in press). We presented observers with occluded face stimuli revealing different portions of the face in different conditions (eye region, mouth region, or face with eyes hidden, or face with mouth hidden) and investigated the similarity of the brain responses to non-overlapping visual inputs (e.g. eyes to face minus eyes). Participants fixated and performed either an expression or gender categorization task on the same stimulus set. Retinotopy was used to define V1-V3 and Neurosynth to define higher level regions of interest. MVPA decoding analyses revealed that strikingly in primary visual cortex, similar responses could be observed to non-overlapping face regions (e.g. eye region to face minus eye region) while in EVC similar responses were present across the most non-overlapping inputs (eye region to mouth region) but only during emotion recognition. Evidence was also present in high level regions (e.g. STS) of generalization across non-overlapping face inputs (replicating Greening et al., in press) suggesting cortical feedback may generate the spatial generalization effects found in early visual cortex. Importantly, low level control analyses provided no evidence whatsoever of reliable decoding across non-overlapping face inputs. In sum our results suggest that recurrent connections (top down or local) faciliate the reactivation of occluded parts of face stimuli, even in the early visual areas.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only