Availability heuristic

The renovation horror stories

Over a few months, three friends tell long, funny stories about contractors who vanished mid-job or flooded a bathroom. Someone planning a kitchen remodel concludes that “most renovations turn into disasters” and puts the project off.

Disasters make good stories and uneventful renovations don’t, so the stories that reach you are filtered for drama. How easily those three come to mind says a lot about what gets retold and very little about the share of renovations that go wrong. That share would need a count, not a memory.

The complaints that stick

A support-team manager is asked for the most common customer complaint. She answers “the export button” at once, remembering several furious calls about it last month. The ticket log shows login problems were logged three times as often, mostly in short, polite emails.

This is the everyday professional form. The angry, recent calls were easy to retrieve, and the routine emails blur together, so recall ranked the complaints by how memorable they were. The records she could have checked answer the question she was actually asked.

Ten reasons is too many

Someone is asked to list ten reasons they like their car. After four, they get stuck and have to reach for “the cup holders are fine”. Afterward, they rate the car lower than they would have before the exercise.

The less obvious form, studied as the ease-of-retrieval effect. The content of the list is entirely positive, yet the difficulty of producing it is read as evidence (“if it’s this hard to think of reasons, I can’t like it that much”). But struggling to reach ten reasons says more about how many reasons are easy to put into words than about how much they like the car. Research on this version reports that asking for many examples can shift judgments this way (see Evidence).

• Retrievability of instances: famous, vivid or recent cases are easier to recall, so their category seems larger. In a classic demonstration, people heard a list of names and were then asked whether it had more men or women; they tended to pick the sex whose names on the list were more famous.
• Ease of search: some things are easier to look for in memory than others. It’s easier to think of words that start with a letter than words with it in third place, which Tversky and Kahneman proposed makes the first kind seem more common (a demonstration later disputed; see Evidence).
• Imaginability: when examples have to be constructed rather than remembered, the ones that are easy to picture seem likelier. Asked how many distinct committees of 2 or of 8 people could be formed from a group of 10, people gave a median of 70 for 2 and 20 for 8; the answer is 45 for both.
• Ease of retrieval proper: the felt difficulty of recalling examples, not only what gets recalled, feeds the judgment (the third example above).

• When memory is a fair sample. If you have had long, direct, unfiltered exposure to the thing you’re judging, how often examples come to mind can track how often it happens. Tversky and Kahneman themselves described availability as a useful clue for exactly this reason.
• When nothing makes some cases more memorable than others. The trouble comes from vividness, recency, publicity and emotional impact; where those don’t differ between the kinds of case, ease of recall is less distorted.
• When a recent case is real evidence of a change. Recency can mislead, but sometimes the recent event is itself information about a changed situation.
• When it’s a first guess you then check. Using recall to form a quick estimate is reasonable if you treat it as provisional and look for a count when one matters.

The test: would the examples that come to mind be the same if the thing were rare but memorable, or common but dull?

The café regular

A barista who has worked the morning shift five days a week for two years is asked which hour is busiest. “Eight to nine, easily,” she says, without looking at the sales data.

She is judging from recall, but her memory is a large, unfiltered sample of exactly the thing in question: she was there for every hour of every shift, and a busy hour isn’t more memorable in some way that has nothing to do with being busy. That’s the memory is a fair sample condition.

Two break-ins on the street

A neighbor’s shed and then another’s are broken into in the same month, and the police confirm a string of similar thefts nearby. A resident who never bothered before buys a lock for his own shed.

The break-ins are recent and vivid, which is where availability usually misleads. But here they aren’t standing in for a long-run rate; confirmed thefts on your own street are evidence of a change in the risk you face now. Updating on them is reasonable even though they’re easy to recall.

Tversky and Kahneman’s account is that people use ease of retrieval as a stand-in for frequency because, over a lifetime, the two are correlated: instances of common classes are usually recalled better and faster than instances of rare ones. The shortcut is cheap and often accurate, and nothing in the feeling of recall signals when it has come apart from the real frequency.

It comes apart whenever something other than frequency makes examples easier to retrieve: personal involvement, vividness, recency, repetition in conversation or media, and the way memory is organized. It fits Kahneman’s “what you see is all there is”: the examples that come to mind feel like the whole picture, and the ones that don’t come to mind don’t register as missing. The same pull toward the memorable case is what gives anecdotes their force, and why a vivid description can crowd out the statistics in base rate neglect. Tversky and Kahneman presented availability alongside two other judgment shortcuts, representativeness and anchoring (see Anchoring effect).

• Ask what made the examples memorable. If the answer is drama, recency, publicity or your own involvement, the examples are probably not a fair sample.
• Look for a count. Records, logs and base rates answer frequency questions that recall can only guess at.
• Think about the cases that don’t make stories. For every renovation disaster retold at a party, ask how many uneventful ones nobody mentioned.
• Be wary of “I can’t think of many”. Difficulty recalling examples can reflect how memory works rather than how rare something is.

These are sensible practices rather than remedies tested as cures for this bias.

Status: replicates robustly, on the strength of a meta-analysis of the ease-of-retrieval version whose effect survives correction for publication bias. That is the version tested most often, not the famous frequency demonstrations, which have a thinner and more mixed replication record. Neither version appears to have had a multi-lab or registered replication, and the mechanism behind specific findings is still debated. A stricter reading could call the label generous; the caveats below narrow what the heuristic explains rather than removing the basic effect.

The original work. Tversky and Kahneman introduced the heuristic in a 1973 paper and summarized it in their 1974 review. The best-known demonstrations, as later researchers describe them:

• Famous names. People heard lists of names in which one sex’s names were more famous. In the most cited version, a list of 19 famous men and 20 less famous women was judged to contain more men, and the reverse list more women, and people recalled about 50% more of the famous names.
• Letters. Asked whether consonants such as R and K are more common as the first or the third letter of English words, most people said first, although the five consonants used (K, L, N, R and V) are all more common in third position. Of 152 participants, 105 judged the first position more likely for most of the letters.
• Committees. Estimates of how many distinct committees could be formed from 10 people fell steadily with committee size, although the true numbers rise and fall symmetrically.

Ease itself, not just content. In 1991, Schwarz and colleagues separated what is recalled from how hard it is to recall. In their first experiment, 40 female students at a German university described 6 examples (easy) or 12 (hard) of their own assertive or unassertive behavior. Those who described 12 assertive examples rated themselves less assertive than those who described 6, and the pattern reversed for unassertive examples. That crossover was only marginally significant (p < .07). In a later experiment, everyone listened to music while recalling, and some were told the music made the kind of memory they were recalling easier to reach. When the music could explain how easy or hard recall felt, the crossover disappeared, which fits the idea that people were using the feeling of ease as information.

The deciding evidence. Weingarten and Hutchinson (2018) meta-analyzed 263 studies (582 effect sizes) that used this few-versus-many manipulation. In studies designed to reproduce the standard effect, and where the judgment was directly about what people recalled, the average effect was medium-sized (r = .25). Two corrections for publication bias shrank it, to r = .16 and r = .19, but both left it clearly above zero. Subjective ease accounted for about a third to half of the effect, so it is part of the explanation but not all of it. For the minority of studies where the judgment was about something one step removed from what was recalled, one correction left an effect indistinguishable from zero. At the time of their search, the authors found no replications with public results on the main replication registries.

Where the record is weaker.

• The original crossover. Yeager and colleagues (2019) tested seven classic effects in large, randomly sampled groups of American adults. Six appeared, with smaller effects than in student studies. The ease-of-retrieval crossover did not (d = 0.09, not significant), even though recalling 12 examples was again reported as harder than recalling 6. They noted that few earlier studies had tested that interaction, and the ones they found that did had not replicated it in students either. The meta-analysis above mostly measures a simpler comparison, and its authors report that leaving out these large national samples barely changed their estimate.
• Limits in new settings. In two small preregistered experiments, Tamayo and colleagues (2026) found that asking for six rather than two reasons made the task feel harder but did not change how credible people found news headlines. (This entry’s summary is based on the abstract.)
• The letter study. Sedlmeier, Hertwig and Gigerenzer (1998) found only one brief published replication of it. They noted an earlier critique that the five consonants were an unrepresentative sample, since most consonants really are more common in first position. In three studies with German speakers and a larger set of letters, people’s judgments generally followed the actual proportions, compressed toward the middle. That fit a “regressed frequencies” account, in which people track real frequencies with some blurring, better than the versions of availability they tested.
• Famous names. In two experiments with 195 Canadian students, McKelvie and Drumheller (2001) did not find famous names judged more numerous in a direct comparison; measured against the true difference, there was an effect, but a smaller one (d = 0.34) than earlier studies. (Based on the abstract.)
• Risk judgments. Hertwig, Pachur and Kurzenhäuser (2005) found that “availability by recall” and regressed frequencies fit people’s judgments of which risks kill more people about equally well. Pachur, Hertwig and Steinmann (2012) found that recall of cases in people’s own social networks fit student risk judgments best, better than feelings about the risks, and that this kind of direct experience was fairly accurate. Encounters with the risks in the media played a negligible role, which cuts against the popular version in which news coverage drives risk estimates. In an experiment with 143 participants, Efendić (2021) found that asking people to think of more deaths from a cause among people they knew raised their mortality estimates for it. (Based on the abstract.)

What remains uncertain. Whether a given judgment runs on the felt ease of recall, on what is recalled, or on stored frequency knowledge that simply regresses toward the middle; how much of the ease effect is explained by ease; and how large the effect is outside student samples. The general claim, that how easily things come to mind shapes judgments of how common they are, is well supported; several of its textbook demonstrations are not.