Modeling Articulatory Rehearsal in an Attention-Based Model of Working Memory

Maintenance of verbal information within human working memory occurs through two main complementary mechanisms: articulatory rehearsal and refreshing. Both are well-described in the literature but very few computational models have attempted to describe the ways they interact to maintain information. Because these interactions are difficult to apprehend without computer simulations, this paper presents a possible implementation of rehearsal within TBRS* (Time-Based Resource-Sharing Theory), a computational model operating only with refreshing and based on the TBRS verbal theory. Computer code is available at https://​osf.​io/​taqmv/​. The implementation was tested on different benchmark findings and could replicate all main effects attributed to rehearsal, while still being accountable for the same effect as TBRS*. Four aspects of our rehearsal implementation are discussed with respect to the human behavior that is intended to be described: first, the primacy and recency effects in relation to the short-term property of rehearsal and the long-term purpose of refreshing; second, the moment at which long-term memory influences working memory performance during working memory tasks in reference to the redintegration hypothesis; third, the interplay between the two maintenance mechanisms; finally, positional coding as the appropriate representation for rehearsal.