Recall as a cost

The procedure

Michael Anderson, Robert Bjork and Elizabeth Bjork published the paradigm in 1994. Participants study category–item pairs — FRUIT–orange, FRUIT–banana, TOOL–hammer, and so on across several categories. They then complete a retrieval-practice phase in which they repeatedly recall half the items from half the categories, cued by the category name and a word stem. After a delay, recall is tested for everything studied.

Three sets of items result. Practised items from practised categories are recalled best, which is unremarkable. Items from categories that received no practice form the baseline. The finding concerns the third set: unpractised items from practised categories, which were recalled below baseline. Having retrieved orange several times appeared to leave banana less accessible than if the FRUIT category had been left alone entirely.

Anderson and his colleagues interpreted this as inhibition. On their account, competitors that intrude during a retrieval attempt are actively suppressed so that the target can be reached, and the suppression persists.

The cue-independence claim

An inhibition account makes a strong prediction, and Anderson and Barbara Spellman tested it in 1995. If the item itself has been suppressed, the impairment should appear even when it is probed through a completely different route. A participant who practised FRUIT items should have trouble recalling banana even when cued with something unrelated to fruit — for instance, a category such as YELLOW.

They reported that impairment. The result mattered because the obvious alternative explanation — that practised items simply block access at the moment of testing, by being stronger and coming to mind first — predicts nothing of the kind. Blocking should be confined to the cue that was strengthened. Cue-independence therefore became the load-bearing evidence for inhibition, and much of the subsequent literature is a dispute about whether it holds.

The dispute

Three lines of objection developed.

Timothy Perfect and colleagues, working through the 2000s, argued that independent probes are rarely as independent as designs assume, and that covert cueing — participants privately reinstating the original category — could reintroduce blocking through the back door. Where they tightened those controls, the effect was inconsistent.

Tanya Jonker, Paul Seli and Colin MacLeod proposed in 2013 a context-based account with no inhibitory mechanism at all. On their reading, retrieval practice shifts the mental context, and at test the practised context is reinstated preferentially, disadvantaging items studied in the original context. The account reproduces several signature findings, including some cue-independent ones, without suppression.

The replication record is the third strand. Kou Murayama, Toshiya Miyatsu, Dorothy Buchli and Benjamin Storm published a meta-analysis in 2014 covering a large body of published studies. Their overall estimate supported a reliable effect, but the analysis also indicated that several theoretically important moderators — competition dependence and interference dependence among them — were less consistently supported than the theory required. Individual replication attempts have since divided, with several well-powered failures for the cue-independent version specifically.

Reading the state of the evidence

The basic phenomenon — practised items up, related unpractised items down relative to baseline — is among the more reproducible results in the memory literature. What remains open is its cause, and the gap between those two statements is wider than most secondary coverage acknowledges. An inhibitory account, a context account and a strengthening-and-blocking account all remain live, and they differ in what they imply about the architecture of retrieval.

The paradigm has been extended well beyond word lists — to eyewitness accounts, to social stereotyping, to conversational remembering between people. Those extensions inherit the unsettled mechanism, and in some cases inherit smaller samples as well. Coverage that presents the effect as a settled property of memory is running ahead of what the published record establishes.

• Anderson, M. C., Bjork, R. A. & Bjork, E. L. (1994). Remembering can cause forgetting. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 1063–1087.
• Anderson, M. C. & Spellman, B. A. (1995). On the status of inhibitory mechanisms in cognition. Psychological Review, 102, 68–100.
• Perfect, T. J. et al. (2004). Transfer appropriate forgetting: the cue-dependent nature of retrieval-induced forgetting. Journal of Memory and Language, 51, 399–417.
• Jonker, T. R., Seli, P. & MacLeod, C. M. (2013). Putting retrieval-induced forgetting in context. Psychological Review, 120, 852–872.
• Murayama, K., Miyatsu, T., Buchli, D. & Storm, B. C. (2014). Forgetting as a consequence of retrieval: a meta-analytic review. Psychological Bulletin, 140, 1383–1409.

Editorial note: this article summarises published findings. It is not advice, and it makes no claim about any individual reader.