Availability heuristic vs representativeness heuristic: the difference, and the question that sorts any example

The availability heuristic estimates how frequent or how probable something is from how easily examples come to mind. The representativeness heuristic estimates how probable something is from how much it resembles a typical case. The difference is the source of the judgement: memory for the first, similarity for the second. Amos Tversky and Daniel Kahneman described both heuristics in the early 1970s.

Most summaries give one invented example for each term and stop there. This article gives you one question that sorts any example. It also gives you the original experiments with their numbers, and the criticism that most summaries do not mention.

• Availability: you judge frequency or probability by how easily you can recall or imagine examples.
• Representativeness: you judge probability by resemblance to a stereotype, a typical case or a typical process.
• The sorting question is: did the answer come from recalled examples, or from a match with a type?
• Both heuristics usually work. Tversky and Kahneman called them "quite useful" and said that they sometimes cause "severe and systematic errors".
• The classic results are not all equally solid. A 1998 study of letter positions did not find the pattern that availability predicts.

The difference in one table

The two terms answer the same type of question in two different ways. The table shows where they differ.

The one question that sorts any example

Exam questions and textbook examples describe a person who makes a judgement. To name the heuristic, ask where the answer came from. Use these steps.

1. 1.Find the judgement. It is usually an estimate of frequency or probability.
2. 2.Find the information that the person used to make the estimate.
3. 3.If the person used examples that came to mind easily, the heuristic is availability.
4. 4.If the person compared a description with a typical case, the heuristic is representativeness.

All six situations in the next table come from the original papers.

Availability: the judgement comes from memory

Tversky and Kahneman named the availability heuristic in 1973. In their definition, a person who uses it judges frequency or probability "by the ease with which relevant instances come to mind." The idea is sensible. Frequent things are usually easier to recall than rare things. The authors wrote that availability "is correlated with ecological frequency, but it is also affected by other factors." Those other factors cause the errors.

The letter study

The best-known test used five consonants: K, L, N, R and V. In English text, each of these letters occurs more frequently in the third position of a word than in the first position. A typical text has twice as many words with K in the third position as words that start with K. But it is easier to search memory for words by their first letter.

The 152 participants decided, for each letter, which position was more frequent. For most of the letters, 105 participants chose the first position and 47 chose the third position. The median estimate was that the first position was twice as frequent as the third. The correct answer was the opposite for all five letters.

The fame study

A second study changed how easy the examples were to recall. Participants listened to a recorded list of 39 names. In some lists, 19 names were famous women and 20 names were less famous men. In the other lists, the famous names were men.

One group recalled the names. On average, they recalled 12.3 of the 19 famous names and 8.4 of the 20 less famous names. A second group decided if the list had more men or more women. Of 99 participants, 80 chose the sex that had the more famous names. That class was the smaller one in each list.

Representativeness: the judgement comes from resemblance

Kahneman and Tversky described representativeness in 1972, one year before the availability paper. A person who uses this heuristic judges the probability of an event by two properties. The first is how similar the event is "in essential properties to its parent population". The second is how well it shows the important features of the process that generated it.

Samples that look random

One problem in the 1972 paper is about families with six children. The paper tells you that 72 families in a city had the birth order girl, boy, girl, boy, boy, girl. You estimate how many families had the order boy, girl, boy, boy, boy, boy. The two exact orders are almost equally probable. But 75 of 92 participants judged the second order to be less probable, and the median estimate was 30 families. The second order does not look like a typical family, and it does not look random.

Base rates

The 1974 paper in Science reports a test of base rates. A base rate is how frequent a category is in the population. Participants read short descriptions of people from a group of 100 engineers and lawyers. One set of participants read that the group had 70 engineers and 30 lawyers. The other set read that it had 30 engineers and 70 lawyers.

One description, of a man called Dick, contained no information about his job. The correct estimate is then the base rate: 0.7 in one condition and 0.3 in the other. Participants in the two conditions gave a probability of 0.5. The description did not resemble an engineer or a lawyer, and the participants ignored the numbers that they had.

The Linda problem

The best-known demonstration is from a 1983 paper in Psychological Review. Linda is "31 years old, single, outspoken and very bright", and as a student she was "deeply concerned with issues of discrimination and social justice". Tversky and Kahneman asked 142 undergraduates which statement was more probable. The first was that Linda is a bank teller. The second was that Linda is a bank teller and is active in the feminist movement.

The second statement cannot be more probable, because all feminist bank tellers are bank tellers. But 85% of the participants chose it. The description resembles a feminist more than it resembles a bank teller. The name of this error is the conjunction fallacy.

Where the two heuristics overlap

The two heuristics are not two separate boxes. The same 1983 paper says that the two heuristics can each "make a conjunction appear more probable than one of its constituents". The paper starts with a conjunction error that comes from availability, not from representativeness.

One group of participants estimated how many seven-letter words that end in "ing" are in four pages of a novel. A second group made the same estimate for seven-letter words that have n as the sixth letter. All of the "ing" words are in the second category, so the second category cannot be smaller. The median estimates were 13.4 for the "ing" words and 4.7 for the larger category. It is easier to think of words that end in "ing".

Thus the conjunction fallacy is not a synonym for representativeness. It is an error, and each of the two heuristics can cause it. The same rule applies to all these terms: a heuristic is a method, and a bias is the error that the method causes. Our list of ten cognitive biases explains that distinction in more detail.

What the criticism says

Short summaries frequently present the two heuristics as facts that nobody disputes. The research literature shows more disagreement.

• The concepts are loosely defined. Gerd Gigerenzer argued in 1996 that the heuristics "are too vague to count as explanations". In his words, they "at once explain too little and too much". After an experiment, a researcher can apply one of them to almost any result.
• Kahneman and Tversky disagreed. In the same journal issue, they replied that representativeness, like similarity, "can be assessed experimentally; hence it need not be defined a priori."
• The letter result is disputed. Sedlmeier, Hertwig and Gigerenzer did three studies of letter positions in 1998. Judgements usually followed the actual proportions. Participants overestimated small proportions and underestimated large proportions. The authors concluded that this pattern explains the data better than availability does.
• The format of the question changes the error rate. The 1983 paper shows this with a problem about heart attacks. One version asked about one man, and 58% of participants without statistics training made the conjunction error. A second version asked how many of 100 men were in each category, and 25% made the error.
• The Linda result replicates. A preregistered replication with 1,032 participants found the conjunction effect for Linda again in 2021. It did not find the effect for a second description, James. The authors noted that conjunction effects can be sensitive to the context.

For an exam, the definitions and the classic studies are sufficient. For your own understanding, remember that the two terms are labels for patterns in judgements. Researchers do not agree on how much the labels explain.

How to keep the two terms apart

Two memory aids come from the words themselves.

• Availability: the examples are available in your memory. Ask what came to mind.
• Representativeness: the case represents a type. Ask what the case looks like.

Then test yourself. Do not read the definitions again. Take a new example, cover the answer, and apply the sorting question. Our guide to memorising psychology terminology explains why tests help you remember more than a second reading does.

Psychology World is a dictionary app with more than 18,000 psychology terms, and it supports this type of practice. Search the dictionary for the two heuristics and the related terms, and save the terms that you find to one bookmark list. Then study the list as flashcards, as multiple-choice questions, or type each answer. The multiple-choice mode needs a list with a minimum of four terms. The home page shows the four steps of the learning loop and what each study mode does.