Abstract
Attention modulates the activity of sensory neurons in early visual areas such as V1 and MT, and is thought to improve the neural representation of behaviorally relevant stimuli. However, direction-selective neurons in MT depend on input from upstream cells in V1 to gain their complex direction-selective tuning (Born and Bradley, 2005, Ann Rev Neurosci; Rust et al., 2006, Nat Neurosci). Thus, it is important to understand not only how attention modulates neural responses in specific areas, but also to understand whether and how attention impacts information transmission between visual areas. Using fMRI and a forward encoding model (Brouwer and Heeger, 2009, J Neurosci), we observe an increase in the response of direction-selective neural populations whose tuning preference is similar to the attended direction of motion, and a decrease in the response of neural populations with dissimilar tuning, in both V1 and MT. The pattern of modulations is predicted by an established linear/non-linear model that describes how information from V1 is integrated by MT neurons during motion processing (Rust et al., 2006). We used this model as well as the pattern of attention modulation of within-area correlations (Cohen and Maunsell, 2011, Neuron) to predict how cross-area correlations between V1 and MT would change as a function of attention. In consonance with the model’s prediction, we find that that correlations between like-tuned populations of direction-selective neurons decrease, whereas correlations between unlike-tuned populations increase. Finally, we use a multivariate Mutual Information metric to show that the observed changes in gain and cross-area correlations have a constructive functional role; they lead to an increase in the synergistic information coupling between direction-selective populations of V1 and MT.
Meeting abstract presented at VSS 2012