Quantum probability theory has successfully provided accurate descriptions of behavior in decision making, and here we apply the same principles to two category learning tasks, one using overlapping, information-integration (II) categories and the other using overlapping, rule-based (RB) categories. Since II categories lack verbalizable descriptions, unlike RB categories, we assert that an II categorization decision is characterized by quantum probability theory, whereas an RB categorization decision is governed by classical probability theory. In our experiment, participants learn to categorize stimuli as members of either category S or K during an acquisition phase, and then rate the likelihood on a scale of 0 to 5 that a stimulus belongs to one category and subsequently perform the same rating for the other category during a transfer phase. With II categories but not RB ones, the quantum model notably outperforms an analogous Markov model and the order effects on likelihood ratings are significant.

Current studies of the serial reproduction paradigm focused on stimuli that were statistically independent of each other. We explored serial reproductions of random walk series and examined whether Bayesian models previously built for independent stimulus could be adapted to autocorrelated stimulus. We found that Bayesian models captured most of the empirical results qualitatively, but could be further improved by incorporating recency effects. Besides, given that the optimal strategy of iterative prediction of random walk was to reproduce the current stimuli, we also compared serial prediction of random walk to serial reproduction. We found that serially reproduced and predicted series both decorrelate as a function of chain position and that the means of the series increase in both tasks, which matched qualitative predictions of the Bayesian models.

A general state-space model of prototype learning was formulated Interms of a set of Internal states and nonlinear Input-output mappings. The general model Includes several previous models as special cases such as Hintzman's (1986) multiple trace model, Metcalf's (1982) holographic model,and two parallel distributive memory models (Knapp & Anderson, 1984;McClelland & Rumelhart, 1985). Two basic properties common to the threemodels were defined in terms of this general model--addltivlty and time Invarlance. An experiment was conducted to test the basic properties using random spectral patterns as stimuli allowing possible nonlinear input and output distortions. Especially, ordinal tests of addltivlty were performed with few assumptions about Internal features that subjects may use to encode the stimulus Information. The results support addltivlty but tlme-Invarlance was clearly violated. Implications of these findings for models of the human memory system are discussed.