Availability Heuristic: Why the Vivid Risk Feels Like the Common One

Asked how common something is, the mind does not count. It searches, and it reports how the search went. Amos Tversky and Daniel Kahneman named this shortcut availability in 1973 and pointed out why it usually works: frequent events really are easier to recall than infrequent ones, so ease of retrieval is a decent clue to frequency.

The shortcut goes wrong when something other than frequency is driving the ease. A plane crash is covered for a week; the 100 people who died on the roads that same day are not covered at all. A shark attack has pictures. A burglary on your street is retold at every barbecue. Each of those makes its category easier to search, and the search result is read as a frequency. It works in reverse for quiet killers: asthma, diabetes and stroke are hard to picture and rarely reported, so they come to mind slowly and the mind reads the slowness as rarity.

Norbert Schwarz and colleagues sharpened this in 1991 by separating how many examples people produce from how hard it was to produce them. Participants asked to recall 12 instances of their own assertive behavior rated themselves as less assertive than participants asked for only 6, because 12 was a struggle. The content said assertive, the effort said otherwise, and the effort won. A later meta-analysis by Weingarten and Hutchinson found the effect itself holds up while the evidence that subjective ease carries it is weaker than the original account assumed.

The applied claim needs the same care. The 1978 study below is the standard evidence that people overrate dramatic causes of death and underrate quiet ones, and Thorsten Pachur reexamined it in 2024 against several later datasets. The pattern was reliable in the original data, did not reappear in the others, and pooled across all of them there was evidence for no special distortion of dramatic causes. News coverage was clearly skewed toward dramatic deaths, but what judgments tracked best was deaths among the people a respondent knew. Availability as a mechanism is not in doubt. Whether the media are the source that feeds it is.

Tversky and Kahneman published "Availability: a heuristic for judging frequency and probability" in Cognitive Psychology in 1973. Asked whether more English words begin with K or carry K in the third position, 105 of their 152 participants chose the first position for a majority of the five letters tested, although third-position words are the more numerous in ordinary text. In another study they played recorded lists of 39 names, one every two seconds, each list pairing 19 very famous names of one sex with 20 less famous names of the other. Of the 99 people asked which sex appeared more often, 80 picked the group with the famous names, which was the smaller group.

The best-known application came five years later. Sarah Lichtenstein, Paul Slovic, Baruch Fischhoff, Mark Layman and Barbara Combs put 41 causes of death to 111 University of Oregon students and 77 members of the Eugene chapter of the League of Women Voters, in 106 pairs, asking which cause was more likely to kill a randomly chosen American in the next year and by what ratio. The true rates were annual deaths per 100 million residents, taken from National Center for Health Statistics reports for 1968 to 1973.

The errors were large and consistent. Tornadoes were judged more frequent killers than asthma, which was about 20 times more lethal. Death by lightning was judged less likely than death by botulism, which it outnumbered 52 to one. Accidents of all kinds were judged about as deadly as disease, though disease killed 15.4 times as many. Stroke caused 85 percent more deaths than all accidents combined, yet only 20 percent of the students and 23 percent of the league members put stroke ahead. Two patterns ran through the data: small frequencies were pushed up and large ones pulled down, and particular causes were distorted beyond that on top of it.

The authors then counted causes of death in two newspapers over 184 days. Tornadoes were the eighth most reported cause and are in fact rare; homicide, 23 percent less frequent than suicide, was reported 9.6 times as often and given 15 times the space; diabetes, breast cancer and tuberculosis were never mentioned. In a final experiment they told students how these errors arise, including the role of coverage and imaginability, and asked them to correct for it. The judgments did not improve.

The heuristic shows up wherever a risk has to be judged without a table to look it up in. The rows below are illustrations, not findings from the studies cited here.

The 1978 debiasing experiment is the warning label on this one: being told how the error works did not fix it. What helps is replacing the search with a count.

• Look for a published rate for the question you are asking. Reading one takes less time than arguing about an anecdote.
• Ask what made this example easy to recall. If the answer is that it was on the news, was recent, or happened to someone you know, you have learned about coverage and memory rather than about frequency.
• Name the quiet version of the risk before judging it: not the shark, the rip current; not the crash, the blood pressure.
• Watch for the effort signal. If you are straining to produce examples, you will conclude the thing is rare even when your list is long. Decide in advance how many instances count, then judge the list, not the strain.
• In groups, assign someone the base rate. The vivid case will always be in the room; the denominator needs an owner.

None of this makes vivid events feel less vivid. The point is to stop treating the feeling as evidence about how often something happens.

Availability is ease of retrieval standing in for frequency. Its neighbors differ in what does the standing in.

• It is not the frequency illusion, where something you have just learned about seems to appear everywhere. That is a change in what you notice after the fact; availability is a judgment of how common a category is.
• It is not the base rate fallacy, which is the failure to use a frequency you have been given. Availability is what fills the gap when no frequency has been given.
• It is not attentional bias, the tendency for what you are preoccupied with to capture your attention. Attention feeds availability by deciding what gets stored, but it is a property of perception rather than of judgment.
• It is not survivorship bias, where the cases available to be counted have already been filtered by an outcome. There the data are missing; here the data are in memory, unevenly.