Similarity-Based Compression in Working Memory: Implications for Decay and Refreshing Models - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Computational Brain & Behavior Année : 2023

Similarity-Based Compression in Working Memory: Implications for Decay and Refreshing Models

Résumé

The ability to compress information is a fundamental cognitive function. It allows working memory (WM) to overcome its severely limited capacity. Recent evidence suggests that the similarity between items can be used to compress information, leading to a rich pattern of behavioral results. This work presents a series of simulations showing that this rich pattern of WM performance is captured using the principles of TBRS*, a decay and refreshing architecture. By assuming that similar items are compressed, the architecture can explain the beneficial effect of similarity on the items themselves. The architecture also explains the fact that when similar items are mixed with dissimilar items, this provides a proactive—but no retroactive—benefit on WM performance. In addition, the model captures fine-grained patterns of transposition errors recently reported. Several analyses are reported showing the robustness of the model’s predictions. We reached the conclusion that decay and refreshing theories provide a plausible explanation for compression effects in WM. These conclusions are discussed in light of recent experimental results. The importance of computational modeling for testing theories is emphasized.

Dates et versions

hal-04189189 , version 1 (28-08-2023)

Identifiants

Citer

Benjamin Kowialiewski, Benoît Lemaire, Sophie Portrat. Similarity-Based Compression in Working Memory: Implications for Decay and Refreshing Models. Computational Brain & Behavior, 2023, ⟨10.1007/s42113-023-00179-0⟩. ⟨hal-04189189⟩
11 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More