Neural network models of serial order and handwriting movement generation

This dissertation describes two neural network models that address the role of time in behavior. A model of serial order suggests how temporal lists of items, such as phonemes or words, can be learned, remembered, and performed. The handwriting model, called VITEWRITE, explains how serial directional commands can be converted into the smooth, curvilinear trajectories of handwriting. The serial order model stores temporal lists using short term memory (STM) activation patterns that factorize the order and timing information in these lists into independent processing channels, as suggested by Grossberg. This enables the model to store and perform sequences of prescribed timing using STM alone. Serial order effects such as primacy and recency, list length, and position effects in sequence discrimination by human subjects are emergent properties of the network structure. The STM model is embedded in a real-time, self-organizing system that uses Adaptive Resonance Theory circuits of Carpenter and Grossberg to stably learn recognition codes for sequences. Learned expectations account for expectational completion, disambiguation, and serial recall of learned lists. Since the model maps sequences of items into unitized list chunks, it is possible to arrange multiple modules in a hierarchy. The second part of the dissertation describes the VITEWRITE model for generating handwriting movements. The model consists of a sequential controller, or motor program, that interacts with the Vector Integration To Endpoint (VITE) trajectory generator to move a model hand. VITE properties enable a simple control strategy to generate complex handwritten script if the hand model contains redundant degrees of freedom. The controller launches directional commands to independent hand synergies at times when the hand begins to move, or when a velocity peak in a synergy occurs. The VITE model translates these temporally disjoint commands into smooth trajectories among temporally overlapping synergetic movements. Each synergy exhibits a unimodal velocity profile during any stroke, generates letters that are invariant under speed and size rescaling, and enables effortless connection of letter shapes into words. Psychophysical data such as the isochrony principle, asymmetric velocity profiles, and the two-thirds power law relating movement curvature and velocity arise as emergent properties of model interactions.