A neural mechanism of feature binding based on the dynamical map theory indistributed coding scheme

In order to understand a neural basis for the binding problem, we propose a neural network model that is constructed based on dynamical map theory. In the architecture, object stimulation activates sets of neurons of the diff erent sensory networks, where all the features of an object are stored as p oint attractors in the networks. Relations among these features are encoded as a point attractor in the integration network. Model simulations show th at dynamic linkages among point attractors across the different networks pl ay a key role in solving the binding problem, in which the integration netw ork serves to mediate the binding. (C) 1999 Elsevier Science B.V. All right s reserved.