Attentional filtering in neocortical areas:: a top-down model

Two-comparator-based rate code models-a control architecture and the corresponding reconstruction network-are merged. The role of reconstruction is information maximization and bottom-up (BU) noise filtering, whereas top-down control 'paves the way' for context compliant BU information that, together, we identify with attentional filtering. Falsifying prediction of the model has gained experimental support. (c) 2004 Elsevier B.V. All rights reserved.