Purchase this article with an account.
Sol Sun, Susanne Ferber, Jonathan Cant; Decoding and reconstructing summary statistical information from human visual cortex. Journal of Vision 2018;18(10):1325. doi: https://doi.org/10.1167/18.10.1325.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Extraction of summary statistics can aid the visual system in overcoming the capacity limits of visual working memory (VWM). By the same token, representations of individual items held in VWM are biased towards the average of the set. While the processing of single items versus ensembles of items has been studied extensively using behavioral paradigms, less is known about the neural mechanisms mediating these processes. Thus, we investigated the processing of single items versus ensembles using fMRI. Participants completed a challenging VWM task where, on any given trial, eight oriented isosceles triangles were presented for study. Participants were instructed to remember the orientation of both the individual items, and the average orientation of the entire set. After a 3 s delay period, participants used a continuous free-recall paradigm to report either the orientation of a randomly chosen single item, or the average orientation of the set. By applying a forward encoding model to the BOLD signal during the study and delay periods from regions of early visual cortex (i.e., V1 – V4), we were able to successfully reconstruct tuning functions that corresponded to the average orientation of the entire set. Importantly, this average orientation was never directly presented to participants, as the individual items were separated from the average by at least 30 degrees. Furthermore, we found that the amplitudes of the reconstructed tuning functions were correlated with behavioral recall precision for the average orientation. The same tuning function amplitudes were not correlated with recall precision for the individual items, suggesting that the reconstructions represent the ensemble summary statistic itself, as opposed to noisy representations of the individual items. These results reveal that the dominant representation in early visual cortex can be based on information that was never physically presented on the retina, at least when VWM capacity is exceeded.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only