A novel method for assessing rival models of recognition memory (Q2437273)

The author discusses two types of mathematical models of a human memory process -- the strength-based recognition memory models and the multinomial processing tree (MPT) models. A multi-state (MS) version of MPT is proposed which leads to a finite mixture model. It is proposed to use hazard rates and reverse hazard functions for the mixture detection by a general signal detection model. Some theorems on the monotonicity properties of the hazards of two-component normal mixtures are established. Examples of experimental data are considered which represent departure from hazards monotonicity evidence in favor of the mixture model.