Memory Suppression? How Inhibition May Allow for Flexible Control

Everyone does something they later regret. Can you ever intentionally forget that you did it? The idea of memory repression has rarely been considered within scientific psychology, but the processes involved in intentional forgetting (also covered last week) are the focus of a recent article by Michael Anderson.

In his article, Anderson argues that the "cognitive control" required to suppress an unwanted memory is fundamentally similar to the processes involved in overcoming a habitual or "prepotent" response in favor of a weaker response. A simple example might be naming the ink color of color words (GREEN) in the Stroop task, where the concept of "green" may be actively inhibited so that "red" can be activated more strongly and thus pronounced more quickly. Anderson argues that inhibition may be important for flexible behavior more generally, including scenarios where subjects may not wish to retrieve certain memories.

In the laboratory, this possibility can be tested with the "retrieval practice" paradigm, in which subjects learn lists of category-exemplar pairs (e.g., fruit - apple, drink - coffee, fruit - strawberry, etc) and afterwards are asked to recall half of the words from half of the categories using stem cued recall ("fruit - ap___"). At the very end of the task, subjects are asked to recall all of the words they had studied. Results from the final test typically show that practiced words are remembered best, followed by words from categories that were not practiced. However, recall is worst for those words that belong to practiced categories but were not themselves retrieved - as though these words had been actively inhibited to retrieve other words from the same category, possibly by the same mechanism of inhibition in the Stroop task described above.

Astute readers will note that "active inhibition" is not necessary to explain these results; instead, the impaired recall of unpracticed exemplars from practiced categories could result from interference, competition for representation, or many other mechanisms. To the contrary, Anderson argues that inhibition is the best explanation, given the following characteristics of the retrieval practice paradigm:

1) Memory is not impaired for unpracticed exemplars of practiced categories if the practiced exemplars are simply repeatedly studied (rather than repeatedly recalled), contrary to predictions based on the interference account. Despite identical exposures to all items, the cost to unpracticed items is much higher when competing items were recalled than when they are merely seen.

2) Memory impairment on unpracticed exemplars of practiced categories is "strength-independent" - meaning that it does not appear to be related to how much competing items have been strengthened through recall or exposure. To be fair, there is some room for doubt of this claim, given that similar strengthening has been assumed based on two facts: identical numbers of exposures, and similar increases in recall on a final test [although I cannot find the paper in which this latter result was actually published - if anyone has a citation, please leave a comment].).

3) These effects are interference-dependent, meaning that memory for unpracticed exemplars from practiced categories is not impaired if they are low-frequency exemplars (e.g., penguin is a low-frequency exemplar of the category bird). Anderson argues that low frequency items are not inhibited relative to items from unpracticed categories because they are not likely to interfere, and thus do not require inhibition.

4) These effects are cue-independent, meaning that they are not specific to the precise exemplar-category pairs used in the study. For example, if the pair fruit-banana is unpracticed while the pair fruit-apple is practiced, recall of "banana" is impaired even on a subsequent test like "monkey-b_____." The import of this finding is that inhibition appears to act on the items themselves rather than the associations between them, contrary to the predictions of some non-inhibitory accounts.

Evidence from a second paradigm also suggests that inhibition may be at work in memory. In Anderson's "think/no-think" paradigm (discussed previously), subjects study several weakly-related word pairs (e.g., ordeal-roach) and are subsequently asked to recall one word when provided with the other from each pair. Then, subjects are presented with one word from each pair, and are told to either think about or avoid thinking about the other paired word (this is the "no-think" phase of the experiment). At the very end of the task, subjects are asked to recall all of the paired words.

The results typically show that recall is impaired on words which subjects had avoided thinking about, and that this impairment is worse the more frequently they encountered that word in the "no-think" stage of the experiment. Furthermore, this effect is cue-independent (as in the retrieval practice paradigm). Anderson concludes that the inhibition that is apparently involved in this task may be the same as that involved in the retrieval practice paradigm, as well as in directed forgetting and in task-switching paradigms more generally (where subjects must "deselect unwanted aspects
of the representation of a stimulus to attend to task relevant attributes," according to Anderson).

Anderson then reviews evidence from neuroimaging of the think/no-think task, in which anterior cingulate, dorsolateral and ventrolateral prefrontal cortices (ACC, dlPFC and vlPFC respectively) were more active for words that were to be suppressed than those that were to be recallled. Anderson notes that these regions overlap with those involved in response inhibition tasks (such as Stop Signal and Stroop). Alternatively, it's also possible that these regions are engaged so that subjects can think of something else to "drown out" the to-be-suppressed word, rather than to actively inhibit it.

Hippocampus activation was also lower during the presentation of words whose paired item should be suppressed rather than retrieved. The degree to which these regions showed these trends (higher activity in ACC/PFC and lower activity in hippocampus) for particular words predicted the degree to which they were impaired on a later test of recall, as though these activations were directly reflecting the same mechanism involved in the putative inhibitory processes.

Finally, Anderson considers the directed forgetting paradigm, in which subjects are actually told to intentionally forget certain items, prior to a test of recall on those items at the conclusion of the task.

Directed forgetting differs in at least one important way from the tasks discussed previously - inhibition appears to be a controlled or voluntary process in the case of think/no-think and directed forgetting, but appears automatic in the case of retrieval practice paradigms. Nontheless, frontal patients show both decreased retrieval practice costs and reduced directed forgetting effects, suggesting that the inhibition shared by both tasks is subserved by the frontal lobes, and thus may be an executive function. Anderson argues that the fact that frontal patients still show some inhibitory effects may actually reflect interference, rather than inhibition per se.

To counter the argument that apparently "automatic" inhibition as observed in the retrieval practice paradigm cannot reflect true cognitive control (which is often thought to require effort, attention, or intention), Anderson proposes the "Flexible Control Hypothesis" of inhibition, which states that a common inhibitory process can act on a variety of representations in order to accomplish a variety of goals. Anderson argues that the "lack of intention" to inhibit in the retrieval practice paradigm "does not imply an absence or even a reduction of controlled inhibition" - instead, inhibition is employed in the service of a different goal (to remember some words, rather than to forget others as in the other paradigms discussed above).

Just as inhibition can be engaged by different goals, Anderson argues it can also be targeted at different kinds of representations. For example, in the directed forgetting paradigm, inhibition may be targeted at the entire list (or its context) rather than at individual words. In contrast, Anderson suggests that inhibition may be targeted at individual words in the retrieval practice and think/no-think paradigms.

In conclusion, Anderson has proposed a powerful theoretical taxonomy of inhibition, placing it at the center of a wide variety of tasks, and as a enabling mechanism in cognitive flexibility. On the other hand, there is substantial reason for doubt that Anderson's argument about laboratory memory tasks will truly apply to the "real world" issues of recovered memories, childhood trauma, and post-traumatic stress disorder (PTSD), as reviewed at Mixing Memory. For example, PTSD patients would seem to suffer from an inability to inhibit traumatic memories, and yet do not show response inhibition deficits in stop signal or go/nogo tasks.

In contrast, lateral inhibition resulting from the active maintenance of task-relevant information (or task-irrelevant, in the case of directed-forgetting) may be a more parsimonious explanation of these results.