Decay versus interference: a new look at an old interaction.

A question that has been fought over for at least 80 years is whether memory traces decay with time or interfere with one another (McGeoch, 1932). This question is central to interpretations of forgetting and memory capacity and to understanding the design of the human cognitive architecture. There are two camps, which take turns declaring victory (e.g., there is " no temporal decay in verbal short-term memory, " proclaim Lewandowsky, Oberauer, & Brown, 2009; " time causes forgetting from working memory, " respond Barrouillet, De Paepe, & Langerock, 2012). Often lost in the fog of battle is evidence that decay and interference are both at work, with decay more subtle than interference but no less common. Berman, Jonides, and Lewis (2009) made this point when they went " in search of decay in verbal working memory " and ultimately found it when they pooled data across experiments (Berman et. al., Fig. 7). Small but robust decay effects can be measured by controlling interference across conditions (Altmann & Gray, 2002) rather than trying to eliminate it (Reitman, 1974). There is also the theoretical argument that decay plays a functional role in cleaning up episodic detritus (Altmann & Gray, 2008), which suggests that decay effects should be pervasive. In the study reported here, we asked whether this perspective—that large interference effects and small decay effects pervasively coexist—could bring new clarity to the results of a classic study that helped shape and still influences the debate. Waugh and Norman (1965) used a probe-digit procedure , in which a list of digits was presented for study, and memory was tested with a probe that indicated which digit to recall. The last digit of the study list was the second occurrence of the probe, and the target for recall was the digit that followed the first occurrence of the probe. Interference was indexed by the number of items between the target and the end of the list. Decay was indexed by presentation rate. Waugh and Norman's (1965) data appear in Figure 1. There was a large effect of number of interfering items but no effect of presentation rate (1 item/s vs. 4 items/s). The authors did not test the interaction of these factors—and although the interaction invited comment initially, it then slipped beneath the waves. Meanwhile, the large interference effect evolved into textbook evidence for an interference-only perspective (e. And yet, the interaction (see Fig. …