An Oscillating Cortical Network Model of Sensory-Motor Timing and Cordination

We report on the preliminary development of a model of sensory-motor control and rhythmic motor coordination that proposes a solution to the \timing" problem in the theory of motor control and a uniication the \motor program" and the \systems dynamics" points of view. Subsets of oscillating associative memories modeling cortical columns are coupled to form a rhythmic time base of counters that change 40 Hz attractors on the peaks of thalamic cycles to divide down the thalamic clock rate from 10 to .5 Hz. We train a Jordan motor control network to tap the end of a planar three joint arm at target points in time speciied by a \plan" vector representing a target state of this time base. The biomechanics of the arm are modeled by an invertible equilibrium point model, and the arm and controler after learning form an integrated analog feedback system modelling primary cortex and below. The arm dynamics are continuous and behave functionally as a system dynamics type nonlinear oscillator during rhythmic motor activity. A motor program is speciied by a 40 Hz plan vector in premotor cortex unaaected by immediate proprioceptive feedback. These plans may change on a thalamic cycle when activated and entrained by learned cortico-cortico connections to an ongoing 40 Hz \attentional stream" of synchronized sensory-motor loops. Sequences of plan vectors are activated by the output of a nested higher order level Elman network. It's plan vector is given in supplementary motor area by a further nested association cortex Elman net with dorsolateral prefrontal working memory holding the highest master plan. A simple rhythmic tapping sequence is implemented where tap and return programs alternate in premotor cortex according to a plan vector in supplementary motor area held in place by the plan in working memory. We have developed a neural network cortical architecture that implements a theory of attention, learning, and communication between cortical areas by adaptive synchronization of 5-20 Hz and 30-80 Hz oscillationss2, 1]. Using dynamical systems theory, the architecture is constructed from recurrently interconnected oscillatory associative memory modules that model higher order sensory and motor areas of cortex. The modules learn connection weights between themselves which cause the system to evolve under a 5-20 Hz clocked sensory-motor processing cycle by a sequence of transitions of synchronized 30-80 Hz oscillatory attractors within the moduless2]. The architecture employs selective\attentional" control of the synchronization of the 30-80 Hz \gamma band" oscillations between …