Working memory (WM) has been proposed to act as a gateway into long-term memory (LTM), such that only information successfully stored in WM can be transferred into LTM. Accordingly, the capacity limit of WM should constrain the acquisition of LTM traces. Evidence for that prediction is mixed: Some studies showed that increasing WM load translated into weaker LTM of the memorized information. Others found no effect of WM load on LTM learning. Here we present the results of three experiments that comprehensively test the gateway hypothesis. All experiments included a WM test and a delayed LTM test for the same information. WM set size was varied from 2 to 8 or 10 items. Further, we varied the type of materials (verbal or visual), whether the task requires memory for items or bindings, and the type of test (recognition vs. recall). The WM set size influenced performance in the LTM test only when two conditions were met: (1) The WM task involved tests of item memory through old-new recognition with concrete objects or words, and (2) LTM was tested for items not tested in the WM test. Yet even these effects were not found consistently across all experiments. This Registered Report indicates that increasing WM load rarely impairs LTM encoding, implying that WM capacity does not act as a bottleneck to the formation of memory traces in LTM under most conditions.
Bartsch et al. (Tue,) studied this question.