Ease-of-Retrieval Effect

What it is

The availability heuristic, as Amos Tversky and Daniel Kahneman described it in 1973, is the habit of judging how common something is by how easily examples come to mind. The natural reading is about quantity: the more examples you can recall, the more common the thing seems. Schwarz and colleagues, in a 1991 paper in the Journal of Personality and Social Psychology titled "Ease of retrieval as information: Another look at the availability heuristic", tested a different reading, in which the signal is the experience of retrieving and not the content retrieved.

In the central experiment, participants recalled either 6 or 12 examples of assertive or unassertive behaviour and then rated themselves. Recalling 12 is difficult. Participants who recalled 12 assertive behaviours rated themselves as less assertive than those who recalled 6. Participants who recalled 12 unassertive behaviours rated themselves as more assertive than those who recalled 12 assertive ones. The self-assessments tracked the content recalled only when recall felt easy. In a further manipulation, a misattribution of the difficulty to another source discredited the informational value of the ease, and the effect disappeared. The authors discussed the informative functions of subjective experiences.

Kahneman takes the result seriously enough to propose that the process might be called an unexplained unavailability heuristic, because the inference depends on retrieval feeling harder than it should.

In effect

Thinking, Fast and Slow reports three variations. In the first, participants told that background music would hinder recall rated themselves as equally assertive after twelve examples as after six, because the difficulty now had a cause that was not about them. Kahneman reports that curved versus straight text boxes and screen background colour worked in the same way. In the second, students were asked to recall either three or eight behaviours bearing on the risk of cardiac disease. Those with no family history went by fluency. Those with a family history showed the opposite pattern and went by content. The book gives these studies without sample sizes or effect sizes, so they illustrate the effect and do not measure it.

The meta-analysis is the quantitative record. Evan Weingarten and J. Wesley Hutchinson pooled 142 papers, 263 studies and 582 effect sizes. They found that the standard few-versus-many manipulation produces a medium-sized effect on average. In the standard paradigm, subjective ease mediated about a third to a half of the total effect, which supports the standard explanation and indicates that other mediators are also at work. They found evidence of publication bias that reduces the standard effect by up to a third. Moderators outside the standard design produced smaller and often reversed effects, while several theory-based moderators produced strong effects that matched the theory.

The practical lesson for anyone using the effect in communication is conditional. Asking people to list many reasons may backfire if listing feels hard, and asking for a few may persuade if listing feels easy. The cardiac result suggests the pattern flips when the topic is personally relevant. Nothing in the sources held here tests this in a purchase setting, so the commercial use is an extrapolation.

What it does not say

It does not say that availability is only fluency. Ease mediated roughly a third to a half of the effect in the meta-analysis, so other processes contribute.

It does not say the effect is large or settled. The pooled estimate is medium-sized before correction and smaller after it, because of publication bias. A critical reanalysis posted by Ulrich Schimmack on the Replicability-Index blog in 2021 argues that the bias-corrected estimate is close to zero and cites two large failed replications. That post is not peer reviewed, and this page has not traced the replication papers it cites, so it is recorded as a dissent and not as a finding.

It does not say that the original 1991 sample sizes were large. The same post gives a figure of 146 for the second study, derived from the reported statistics, which this page has not confirmed against the paper.

It does not say that more examples never help. The cardiac result shows the content of examples can dominate when the topic is personal.

It does not claim an effect on buying. The ease-of-retrieval work is about self-ratings and judgements, and extending it to purchases is a further step.

---