A nonlinear dynamical systems perspective on response time distributions

Theories of word recognition have rarely explored the modeling potential of nonlinear dynamical systems. This dissertation explores whether a nonlinear dynamical model can simulate response time data from two lexical decision experiments. In these experiments, adult readers quickly decide whether a letter string spells a word. Correct response times are assumed to estimate the time required for a dynamical system to reach a stable solution, that is relaxation time. The ranges of shapes of response time distributions are assumed to index the stability of solutions or attractors. The first experiment illustrates the ranges of shapes of distributions associated with the word-frequency effect. Distributions for frequently occurring words are less skewed than distributions for infrequently occurring words. The second experiment uses the same high- and low-frequency words to illustrate the ranges of distributions associated with the semantic priming effect. Distributions for words following a semantically related word (e.g., doctor: NURSE) are less skewed than distributions for words following an unrelated stimulus (e.g., xxxx: NURSE). The shapes of distributions from the experiments were used to determine the ranges of parameter values of the model. The results of the simulation indicate that the response time data from the experiments are mimicked by the relaxation time data generated by the model. The researcher concludes that nonlinear dynamical systems provide a promising approach to investigating and understanding response time behavior.