COMPOSITION DURING SERIAL-LEARNING - A SERIAL POSITION EFFECT

Composition is a computational learning mechanism that merges serially performed elementary processes into hierarchically organized knowledg e structures. The main goals of this research were to explore (a) the role of serial position in composition and (b) the relation between de gree of composition and explicit serial recall in serial learning. In 3 experiments, Ss performed a rule-based serial reaction time task in which they had to categorize a sequence of 12 stimuli shown simultaneo usly on a video monitor. A procedure based on the comparison of reacti on times to random sequences and a repeating sequence identified a ser ial position effect of composition that was, however, moderated by Ss' explicit, postexperimental recall of the repeating sequence. A produc tion-system-based computational model of composition is described that qualitatively reproduces the empirical findings. Implications for the mechanisms governing serial learning are discussed.