[en] Working memory (WM) and the linguistic system interact in a complex manner. It is firmly established that semantic knowledge supports the short-term maintenance of verbal information. This is shown by the recall advantage observed for semantically related (e.g. leaf – tree – branch) over unrelated (e.g. house – dog – wall) lists of items. However, the mechanisms responsible for these influences of the linguistic system remain unknown.
From a psycholinguistic perspective, interactive activation models of language processing assume that semantically related items reactivate each other, for instance via their shared semantic features (Dell et al., 1997). Hence, semantically related items might be better recalled because they benefit from these mutual reactivations. In this study, we not only demonstrate the plausibility of this mechanism, but we also show that semantic knowledge can be used to save WM resources, and this using a convergent approach involving computational and behavioral methods.
In an experiment requiring participants to perform immediate serial recall of lists composed of six memoranda (e.g. leaf - tree - branch - wall - sky - dog), we observed that the presence of triplets composed of semantically related items (i.e. leaf - tree - branch) enhanced recall performance for items that directly followed the semantic triplet (i.e. wall - sky - dog), and this compared to a condition in which all the items were semantically unrelated (e.g. hammer - jacket - horn - wall - sky - dog). These results show that semantic knowledge can be used to save WM resources. This phenomenon is successfully captured by an adaptation of TBRS*, a decay and refreshing computational architecture (Oberauer & Lewandowsky, 2011). This architecture was adapted by supplementing to it core assumptions derived from interactive activation models (see Figure 1).
