Achievement

Bringing together mathematical analysis of neural computation and the cognitive psychology and cognitive neuropsychology of the production over time of series of elements, IGERT trainee Manuel Vindiola has developed the first formally principled model of serial production that explains a range of error patterns. While previous models have also deployed neural networks, they have been largely ad hoc constructions with poorly understood properties. Vindiola's theory, in contrast, provides explicit analysis of the behavior of his network, based on highly general principles for encoding and decoding symbol structures (e.g., sequences) in neural networks. In collaboration with fellow IGERT trainee Simon Fischer-Baum, Vindiola has shown that the theory also naturally incorporates Fischer-Baum's principles of order representation, and simulates the error patterns in spelling that Fisher-Baum has used these principles to explain.