Abstract
Perception requires the brain to make decisions. Models of the link between stimulus, decision and behavior have been motivated by this fundamental insight. In the present study, we were motivated by the question, how does the brain transform an internal representation of a stimulus into a decision?
We report two experiments that show reaction times (RT) for an object categorization task can be predicted from perceptual representations of stimuli. We hypothesized that the boundaries between categories of objects in perceptual representations of stimuli (e.g. animate and inanimate objects) also reflect boundaries for decision-making. This hypothesis predicts that individual object representations closer to the decision-boundary in the perceptual representation will be harder to discriminate, requiring more decision-time. To model RTs for the categorization task, distances of individual exemplar representations from the representational boundary were transformed into "decision-times" using a biologically motivated perceptual decision-model (Gold and Shadlen, 2007), in which decision-times are modeled as the number of iterations for the model to converge on a decision.
As a first test of our hypothesis, we reanalyzed the fMRI data of Kriegeskorte et al. (2008). Linear discriminate analysis (LDA) was used to find the boundary between animate and inanimate object exemplar representations in human inferior temporal cortex. In accordance with our hypothesis, we found that model decision-times correlated significantly with behavioral RTs (Spearman's ρ = .333; p <.01). We next extended these findings to MEG time series data. Here, we found RTs correlated with model decision-times based on neural representations of the stimuli 200ms after stimulus onset (Spearman's ρ = .696; p <.01). Our results suggest that the process of settling on an internal representation of a stimulus is itself partially constitutive of, rather than prior to, decision-making in perception.
Meeting abstract presented at VSS 2013