Memory Inhibition: How Recalling One Thing Pushes Its Neighbors Down

Remembering is competitive. A cue such as the category name FRUIT is connected to every fruit word you studied, and when you try to retrieve one of them the others come forward too, as competitors. The inhibition account proposed by Michael Anderson and colleagues says that the memory system resolves this competition by suppressing the competitors, and that the suppression outlasts the moment: the next time you need cherry, it is harder to reach than it would have been had you never practiced orange.

The procedure that shows this is called the retrieval-practice paradigm. People study several categories of words, then repeatedly retrieve half of the words from half of the categories using cues such as FRUIT: or____. On a final test, the practiced words are recalled best, as anyone would expect. The unpracticed words from the practiced categories are recalled worse than the words from categories that were never practiced at all. That gap is retrieval-induced forgetting, and the never-practiced categories are the baseline that lets researchers separate it from ordinary forgetting over time.

Two features of the effect, documented in the 1994 paper and in later work, set it apart from simple interference. First, it is competition dependent: in the second and third 1994 experiments, the impairment appeared mainly for strong, high-frequency category members, the ones most likely to compete during practice, while weak members were hardly affected, which is the opposite of what a model in which stronger practiced items simply block weaker ones would predict. Second, later studies found the impairment with cues that had not been used during practice, which a blocking account cannot easily explain and a suppression account can.

The effect is reliable. Murayama, Miyatsu, Buchli and Storm's 2014 meta-analysis of the retrieval-practice literature found a consistent impairment of unpracticed items from practiced categories across many laboratories and materials. What is still argued is the mechanism. Raaijmakers and Jakab argued in 2013 that interference models without any inhibitory process can reproduce most of the findings, and Storm and Levy's 2012 progress report on the inhibitory account acknowledges that some of its predictions, such as strength independence, the claim that the forgetting does not depend on how much the practiced items improved, have mixed support. This page describes the effect as established and the inhibition explanation as the leading account rather than a settled one.

Michael Anderson, Robert Bjork and Elizabeth Bjork published "Remembering can cause forgetting: Retrieval dynamics in long-term memory" in the Journal of Experimental Psychology: Learning, Memory, and Cognition in 1994. In the first experiment, 36 introductory psychology students at the University of California, Los Angeles, studied eight categories of six words each, such as FRUIT: orange. Half of the categories were made of strong, typical members and half of weak, less typical ones.

In the practice phase, participants retrieved three words from each of four categories three times, using the category name and the first two letters of the word as the cue. The other three words in those categories, and all six words in the four remaining categories, were not practiced. Twenty minutes of an unrelated reasoning task followed. Then came a category-cued recall test: participants saw each category name and had 30 seconds to write down every word they could remember from it.

Practice worked as expected. The practiced words were recalled about three quarters of the time, far above the 48.4 percent recall of words from the unpracticed categories. The unpracticed words from the practiced categories came back only 37.5 percent of the time, about eleven points below the baseline. Retrieving some fruit words had made the remaining fruit words harder to recall than the drink words nobody had touched.

The second experiment replaced category-cued free recall with a test that fixed the order in which practiced and unpracticed words were asked for, so that the loss could not be blamed on the practiced words coming out first and getting in the way. The impairment remained. The third experiment separated the strength of the practiced items from the strength of the unpracticed ones and found the forgetting concentrated on strong unpracticed items, the ones most likely to have competed during practice. The authors concluded that the retrieval process itself, not the passage of time or the test order, was causing the loss.

The rows below are illustrations of where selective rehearsal happens before a later recall, not findings from the studies cited on this page.

The forgetting follows practice, so the counter is to decide what gets practiced rather than to let the first few items decide for you.

• When something matters as a whole, retrieve the whole set. Partial practice is what produces the gap; a quiz that covers every item in a category strengthens every item.
• Record first, rehearse second. A written account made before any selective questioning preserves the items that later rehearsal would push down.
• Expect the gap where the practice was uneven. If you have told one version of an event many times, treat the untold parts as likely to be faint rather than absent, and go back to notes or to other people who were there.
• Use the effect on purpose for material you want to leave behind. Retrieving the useful, current version of a fact, a new password or a new address, is one way the old competitor gets quieter; Anderson's 2003 review treats this as a feature of a system that must forget in order to find.
• Do not read a blank as proof. In the laboratory, unpracticed items were impaired, not erased: they were still recalled more than a third of the time, and the loss faded with time in later studies. A moment's failure to recall is weak evidence that a memory is gone.

Memory inhibition has a specific signature, the loss of unpracticed items from a practiced set relative to an untouched baseline. Its neighbors are easy to confuse with it.

• It is not the testing effect, which is the gain on the practiced items themselves. The two come from the same experiments: retrieval strengthens what is retrieved and, on the inhibition account, weakens what competed with it.
• It is not cue-dependent forgetting, where a memory is intact but the cue that would reach it is missing. In the retrieval-practice paradigm the cue is given and the item is still harder to reach.
• It is not the tip-of-the-tongue phenomenon, the felt state of knowing a word without being able to produce it. That state can be caused by a competitor, but it is an experience, not a measured loss relative to a baseline.
• It is not directed suppression, the deliberate effort to keep a memory out of mind that Anderson and Green studied in 2001 with the think/no-think procedure. Retrieval-induced forgetting needs no intention to forget; it happens while you are trying to remember something else.