Varying timescales of stimulus integration unite neural adaptation and prototype formation

Summary

Date: 
February 17, 2016 - 10:00am
Location: 
Northwest 243
About the Speaker
Name: 
Marcelo Gomes Mattar (UPenn)

Neural responses to visual stimulation are modulated by sensory history on different timescales. On short timescales, a response may be modulated by stimuli that recently preceded it (i.e. neural adaptation). On a long timescale, increased responses occur for stimuli distinct from a long-term average (i.e. norm-based responses). In each case, neural responses reflect the deviation of a current stimulus from a "prior" integrating stimulus history. We hypothesized that both neural adaptation and the average representation in norm-based responses are better explained by a single modulatory mechanism that integrates information over intermediate timescales. We formulated a simple model of exponential integration of recent stimulus history to create a continuously updated neural prior. Using face stimuli and BOLD fMRI data collected from 15 subjects, we validated this hypothesis within a fusiform ROI. We then measured the time constant of temporal integration across ventral visual cortex and found a smooth gradient consistent with variable temporal receptive windows. These results were then replicated with 19 subjects and a different set of face stimuli.