Organisation temporelle et structuration de la mémoire – OTSM

This project in cognitive psychology aims at describing the relationship between the temporal organization of stimuli and memory processes. Two experimental and theoretical research subjects are developed: 1) the study of the effect of the manipulation of presentation orders of stimuli, and the modeling of categorization processes 2) the modeling of immediate serial-recall tasks and the evaluation of the memory span using sequence alignment algorithms. A single model will be then evaluated in order to account for the discriminability of stimuli in sequential presentations in the experiments of both programs. First, we aim at investigating the mechanisms by which categorical knowledge is abstracted by manipulating the presentation order in which stimuli are presented to subjects. We hypothesize that presentation orders which respect the subject's internal conceptual organization are beneficial to acquiring knowledge. We aim at showing that orders based on rules substantially facilitates learning as compared with previously known beneficial orders such as similarity-based orders. This year, we develop a solution to the issue raised in our last year ANR review. Also, a more careful examination of an exemplar-based model (GCM-gamma) taking into account the temporal organization of stimuli will be provided. A second part of the project aims at reexamining Miller's (1956) estimation of short term memory capacity. The rather high evaluation of Miller (i.e., 7±2) results from the fact that simple items can be recoded by subjects into chunks in memory span tasks Miller examined. Recent research designed to prevent recoding has given evidence for a lower estimation of 3 to 4 chunks. Recently, we showed that both evaluations can simultaneously be found correct in a given task. Nevertheless, this project goes beyond the study of chunking mechanisms and proposes an analysis of scoring systems. Classical scoring produces a debatable estimation of the number of items correctly recalled : it is most often simple and rigorous, but maybe too crude. We aim at applying sequence alignment algorithms applied in genetics in order to develop a scoring system offering a typology of the errors arising in memory and their statistics. In comparison to genetics, these errors will be identified as deletions, mutations, translocations and duplications. This method will be applied to a series of replications of classical memory span tasks, beginning by the pioneer studies related by Miller.