Abstract
Information about visual categories is widely available across the brain (Haxby et al. 2001) and these representations can be modulated by both explicit learning (Hammer & Sloutsky 2016) and implicit neurofeedback training (Jackson-Hanen et al. SfN 2014). However, the causal link between the neural representation of categories and their perception remains unclear. To address this question, we seek to induce neural plasticity of visual representations via real-time fMRI neurofeedback (deBettencourt et al. 2015) and test whether this drives categorical perception. We hypothesize that increasing neural separation between categories should also differentiate the categories perceptually. To this end, we seek to use neurofeedback to emphasize non-overlapping (unique) features and suppress overlapping (shared) features of novel abstract visual categories. To do so, we constructed a stimulus space of complex artificial shapes that vary along multiple dimensions simultaneously (Op de Beeck et al. 2001). Extensive behavioral norming (n=750) suggests that each stimulus dimension is perceived in an equivalently graded manner, as are manipulations of the space along multiple dimensions simultaneously. Additionally, we developed a novel approach (KL-Evidence) for computing the neurofeedback provided to differentiate categories, based on mutual information between the distributions of neural responses they elicit. Simulations showed that our method accurately pinpoints non-overlapping features to be emphasized during neurofeedback to induce desired plasticity. We'll present preliminary results from an fMRI study in which we use this feedback method to induce plasticity in representations elicited by arbitrary categories from the stimulus space. By collecting perceptual similarity ratings pre- and post-feedback, we examine a potential causal role for these induced neural representations in similarity judgments. More generally, the approaches we develop for inducing neural plasticity may open up a new platform to investigate and understand human learning with fMRI.
Meeting abstract presented at VSS 2018