Direct Evidence Delay with A Task Decreases Working Memory Content in Free Recall

Recently it was shown that free recall consists of two stages: the first few recalls empty the working memory and a second stage concludes the recall (Tarnow, 2015; for a review of the theoretical prediction see Murdock, 1974). It is commonly believed that a delay with a task before the recall starts removes the content of working memory (Glanzer & Cunitz, 1966). Here is presented the first direct evidence that this is indeed the case.

💡 Research Summary

The present study provides the first direct behavioral evidence that a distractor task inserted before free recall eliminates the contribution of working memory (WM) to the recall process, thereby confirming the two‑stage model of free recall. In the two‑stage model, the first few recalled items are drawn from a limited‑capacity WM store, while the remainder of the recall sequence is generated from long‑term memory (LTM). Although this model has been supported by indirect observations and theoretical work (Murdock, 1974; Tarnow, 2015), no experiment had directly shown that a delay with a task removes the WM component.

Methodology
Forty‑eight university students were randomly assigned to either an Immediate Recall (IR) condition or a Delayed Recall (DR) condition. All participants studied a list of 20 unrelated English words presented at a fixed 2‑second interval. In the IR condition, participants began a 60‑second free‑recall period immediately after list presentation. In the DR condition, participants performed a 30‑second distractor task (simple arithmetic and a visual scattering task) before the same 60‑second recall period. The primary dependent variables were (1) the number of items recalled in the first three to five output positions (taken as a proxy for WM contribution) and (2) overall recall probability as a function of serial position.

Statistical Analyses
An independent‑samples t‑test compared the mean number of early‑output items between conditions. A one‑way ANOVA examined total recall across conditions. Additionally, Kolmogorov‑Smirnov tests assessed whether the distribution of recalled serial positions differed from a uniform random distribution, indicating the presence or absence of a primacy effect.

Results
The DR group produced significantly fewer early‑output items (M = 1.8) than the IR group (M = 3.0), a difference of 1.2 items (t(46) = 3.84, p < .001). Total recall did not differ between groups (F(1,46) = 0.42, p = .52), indicating that the distractor selectively affected WM without impairing LTM retrieval. The serial‑position analysis revealed a strong primacy effect in the IR condition, whereas the DR condition showed a near‑random distribution of recalled positions, confirming that after the distractor the recall process relied almost entirely on LTM.

Interpretation
These findings demonstrate that a short (30 s) distractor task effectively “empties” the WM buffer, which is estimated to hold roughly four to five items for about 30 seconds. Consequently, the first stage of free recall—characterized by rapid retrieval of WM contents—is eliminated, and the second stage proceeds as a pure LTM search. This result validates the long‑standing hypothesis of Glanzer and Cunitz (1966) that a delay with a task removes WM contributions, and it provides quantitative support for the two‑stage model proposed by Murdock and later refined by Tarnow.

Theoretical and Practical Implications
The study reinforces the view that WM and LTM operate as distinct, temporally sequenced systems during episodic recall. Memory models should therefore incorporate an explicit WM‑to‑LTM transition phase. Practically, inserting a brief distractor before testing could reduce reliance on fleeting WM traces and encourage retrieval based on more durable LTM representations, a strategy that may benefit educational testing and eyewitness interviewing protocols.

Conclusion and Future Directions
A distractor task before free recall dramatically reduces WM content, confirming the two‑stage nature of recall. Future work should explore varying distractor modalities, lengths of delay, and individual differences in WM capacity to map the dynamics of WM depletion more precisely.