Abstract
The sensorimotor recruitment model proposes that information in visual working memory (VWM) is maintained in posterior visual areas (e.g. D'Esposito & Postle, 2015). This model has found support in recent neuroimaging studies demonstrating decoding of VWM content in the posterior sensory cortices. However, there has been little behavioral evidence to support this model. Here, we hypothesized that if VWM storage takes place in visual cortices, the effectiveness of an interfering stimulus on the memory representation depends on the similarity of the population response across neurons in these areas. For example, decoding of the orientation of gabor patches in VWM has been reported in early vision cortex (Harrison & Tong, 2009), wherein neuronal populations are tuned for specific orientations. The more similar an interfering stimulus to the memory item along the dimension which defines that tuning, the more overlap in neuronal responses and the more interference. To investigate this hypothesis, we had participants memorize either an oriented gabor, color patch, or motion direction and then perform an orthogonal task between study and report. During this orthogonal task, an unrelated distractor was presented which varied systematically in similarity to the remembered stimulus along the relevant dimensions. In every case we found a monotonic relationship between similarity and the effectiveness of the distractor as measured by 1) the degree to which it biased the report towards itself and 2) the precision/certainty/stability of the resulting VWM representation as measured by the variability in the report. Thus, more similar distractors more strongly affected VWM representations as predicted by the tuning curves of present in perceptual cortices. These results strongly support the sensorimotor recruitment model and suggest a novel behavioral method for investigating the content and nature of VWM representations in the perceptual cortices.
Meeting abstract presented at VSS 2018