Race and mixture models of individual differences in memory-based learning in older and younger adults: Effects of number of facts to learn and inter-item interference

The research addressed the mechanisms underlying developmental differences in cognitive skill acquisition. The primary aim of the experiment was to examine the effects of number of facts to be learned and level of inter-item interference on age-related differences in cognitive skill acquisition. In addition to standard methods of analysis, individual and group data were fit to non-linear functions so as to describe possible age differences in rates of learning. Mathematical modeling was used as a descriptive tool for examining the transition from computational performance to memory-based retrieval. Predictions were drawn from Logan\u27s race model of instance-based learning and Rickard\u27s mixture model of Component Power Laws and were used to fit models that described both individual differences between age groups and learning conditions and intraindividual changes in performance across practice. The study attempted to provide a complete description of age-related differences in the transition from computational learning to instance-based memory retrieval using an alphabet-arithmetic task. Forty-eight participants (24 young adults and 24 older adults) were given extensive practice on an alphabet-arithmetic task (e.g., A + 2 = C). Eight participants in each age group were given a problem set consisting of six arithmetic facts to learn, and 16 participants in each age group practiced a problem set consisting of 12 arithmetic facts to learn. For half of the participants in each age group receiving the 12-fact problem sets, the items to be learned were constructed to be minimally interfering with each other, whereas for the other half, the items were constructed to be maximally interfering. Traditional ANOVA of mean response times were supplemented by fits of learning curves across blocks of practice. Further, the modeling of group data was rejected in favor of developing mathematical models that described individual data. Two models that accounted for the observed data were pitted against each other in a test of the race model and the mixture model. It was discovered that a race model best described learning of some facts whereas a mixture model better described learning of other facts