Representativeness and Causal Judgments

Representativeness and Causal Judgments

This chapter from Heuristics and Biases examines how resemblance can make a causal claim seem plausible before it has been tested. Its assessments of research and clinical practices reflect the literature cited below.

In many cases, it can be difficult to establish with certainty that a judgmental error has been made. It is partly for this reason, we believe, that there has been less empirical research on representativeness and causal judgments than, say, representativeness and base-rate neglect. Furthermore, the research that has been conducted on representativeness-based causal judgments is more suggestive than definitive. Nonetheless, the suggestive evidence is rather striking, and it points to the possibility that representativeness may exert at least as much influence over causal judgments as it does over other, more exhaustively researched types of judgments. To see how much, we discuss some examples of representativeness thinking in medicine, in pseudoscientific systems, and in psychoanalysis.

Representativeness and Medical Beliefs. One area in which the impact of representativeness on causal judgments is particularly striking is the domain of health and medicine (see Gilovich, 1991). Historically, people have often assumed that the symptoms of a disease should resemble either its cause or its cure (or both). In ancient Chinese medicine, for example, people with vision problems were fed ground bat in the (typically) mistaken belief that bats have particularly keen vision and that some of this ability might be transferred to the recipient (Deutsch, 1977). Evans-Pritchard (1937) noted many examples of the influence of representativeness among the African Azande (although he discussed them in the context of magical thinking, not representativeness). For instance, the Azande used the ground-up skull of the red bush monkey to cure epilepsy. Why? The cure should resemble the disease, so the herky-jerky movements of the monkey make the essence of monkey stand out as a promising candidate to settle the violent movements of an epileptic seizure. As Evans-Pritchard put it:

Generally the logic of therapeutic treatment consists in the selection of the most prominent external symptoms, the naming of the disease after some object in nature it resembles, and the utilization of the object as the principal ingredient in the drug administered to cure the disease. The circle may even be completed by belief that the external symptoms not only yield to treatment by the object which resembles them but are caused by it as well. (quoted in Nisbett & Ross, 1980, p. 116)

Western medical practice has likewise been guided by the representativeness heuristic. For instance, early Western medicine was strongly influenced by what was known as the “doctrine of signatures,” or the belief that “every natural substance which possesses any medicinal virtue indicates by an obvious and well-marked external character the disease for which it is a remedy, or the object for which it should be employed” (quoted in Nisbett & Ross, 1980, p. 116). Thus, physicians prescribed the lungs of the fox (known for its endurance) for asthmatics, and the yellow spice turmeric for jaundice. Again, disease and cure are linked because they resemble one another.

Or consider the popularity of homeopathy, which derives from the 18th century work of German physician Samuel Hahnemann (Barrett, 1987). One of the bedrock principles of homeopathy is Hahnemann’s “law of similars,” according to which the key to discovering what substance cures a particular disorder lies in noting the effect that different substances have on healthy people. If a substance causes a particular reaction in an unafflicted person, that substance is seen as a likely cure for a disease characterized by those same symptoms. As before, the external symptoms of a disease are used to identify a cure for the disease – a cure that manifests the same external characteristics.

Of course, there are instances in which substances that cause particular symptoms are used effectively as part of a therapeutic regimen to cure, alleviate, or prevent those very symptoms. Vaccination exposes the immune system to an antigen, or instructions for producing one, to help it develop a response. Allergy immunotherapy uses controlled exposure to an allergen to develop tolerance. These approaches rest on tested mechanisms and outcomes, not on resemblance alone. The problem with the discredited medical practices described above is the general assumption that the symptoms of a disease should resemble its cause, its cure, or both. Limiting the scope of possible cures to those that are representative of the disease can seriously impede scientific discovery. Such a narrow focus, for example, would have inhibited the discovery of the two most significant developments of modern medicine – sanitation and antibiotics.

Representativeness thinking continues to abound in modern “alternative” medicine, a pursuit that appears to be gaining in perceived legitimacy (Cowley, King, Hager, & Rosenberg, 1995). An investigation by the United States Congress into health fraud and quackery noted several examples of what appear to be interventions inspired by the superficial appeal of representativeness (House Subcommittee on Health and Long-Term Care, 1984). In one set of suggested treatments, patients are encouraged to eat raw organ concentrates corresponding to the dysfunctional body part (e.g., brain concentrates for mental disorders, heart concentrates for cardiac conditions, raw stomach lining for ulcers). Similarly, the fingerprints of representativeness are all over the practice of “rebirthing,” a New Age therapeutic technique in which individuals reenact their own births in an effort to correct personality defects caused by having been born in an “unnatural” fashion (Ward, 1994). One person who was born breech (i.e., feet first) underwent the rebirthing procedure to cure his sense that his life was always going in the wrong direction and that he could never seem to get things “the right way round.” Another, born Caesarean, sought the treatment because of a lifelong difficulty of seeing things to completion and always relying on others to finish tasks for her. As one author quipped, “God knows what damage forceps might inflict . . . a lifelong neurosis that you’re being dragged where you don’t want to go?” (p. 90).

A more rigorous examination of the kind of erroneous beliefs about health and the human body that can arise from the appeal of representativeness dealt with the adage, “You are what you eat.” Just how far do people take this idea? In certain respects, the saying is undeniably true: Bodies are composed to a large extent of the molecules that were once ingested as food. Quite literally, we are what we have eaten. Indeed, there are times when we even take on the character of what we ingest: Food can affect body composition, and a person’s skin can acquire an orange tint from carotene. But the notion that we develop the characteristics of the food we eat sometimes goes beyond such examples to almost magical extremes. The Hua of Papua New Guinea, for example, believe that individuals will grow quickly if they eat rapidly growing food (Meigs, 1984, cited by Nemeroff & Rozin, 1989).

But what about a more “scientifically minded” population? Carol Nemeroff and Paul Rozin (1989) asked college students to consider a hypothetical culture known as the “Chandorans,” who hunt wild boar and marine turtles. Some of the students learned that the Chandorans hunt turtles for their shells and wild boar for their meat. The others heard the opposite: The tribe hunts turtles for their meat, and boar for their tusks.

After reading one of the two descriptions of the Chandorans, the students were asked to rate the tribe members on numerous characteristics. Their responses reflected a belief that the characteristics of the food would “rub off” onto the tribe members. Boar-eaters were thought to be more aggressive and irritable than their counterparts – and more likely to have beards! The turtle-eaters were thought to live longer and be better swimmers.

However educated a person may be (the participants in Nemeroff and Rozin’s experiment were University of Pennsylvania undergraduates), it can be difficult to get beyond the assumption that like goes with like. In this case, it leads to the belief that individuals tend to acquire the attributes of the food they ingest. Claims linking milk with phlegm, oily food with oily skin, or spicy food with heartburn need evidence specific to each proposed effect. Resemblance alone cannot establish or refute them.

Representativeness and Pseudoscientific Beliefs. A core tenet of the field of astrology is that an individual’s personality is influenced by the astrological sign under which he or she was born (Huntley, 1990). A glance at the personality types associated with the various astrological signs reveals an uncanny concordance between the supposed personality of someone with a particular sign and the characteristics associated with the sign’s namesake (Huntley, 1990; Howe, 1970; Zusne & Jones, 1982). Those born under the sign of the goat (Capricorn) are said to be tenacious, hard-working, and stubborn, whereas those born under the lion (Leo) are proud, forceful leaders. Likewise, those born under the sign of Cancer (the crab) share with their namesake a tendency to appear hard on the outside, while inside their “shells” they are soft and vulnerable. One account of astrology even goes so far as to suggest that, like the crab, those born under the sign of Cancer tend to be “deeply attached to their homes” (Read et al., 1978).

What is the origin of these associations? They are not empirically derived, as they have been shown time and time again to lack validity (e.g., Carlson, 1985; Dean, 1987; for reviews, see Abell, 1981; Schick & Vaughn, 1995; Zusne & Jones, 1982). They are thus conceptually driven by simple, representativeness-based assessments of the personalities that should be associated with various astrological signs. After all, who is more likely to be retiring and modest than a Virgo (the virgin)? Who better to be well balanced, harmonious, and fair than a Libra (the scales)? By taking advantage of people’s reflexive associations, the system gains some surface plausibility.

It doesn’t stop there. Consider another elaborate “scientific” system designed to assess the “secrets” of an individual’s personality – graphology, or handwriting analysis. Corporations pay graphologists sizable fees to screen job applicants by developing personality profiles of those who apply for jobs. Graphologists are also called on to provide “expert” testimony in trial proceedings and help the Secret Service determine if any real danger is posed by threatening letters to government officials. How much stock can we put in the work of handwriting analysts?

Unlike astrology, graphology is not worthless. It has been, and continues to be, the subject of careful empirical investigation (Nevo, 1986), and it has been shown that people’s handwriting can reveal certain things about them. Particularly shaky writing can be a clue that an individual suffers from some neurologic disorder that causes hand tremors; whether a person is male or female is often apparent from his or her writing. In general, however, what handwriting analysis can determine most reliably tends to be things that can be even more reliably ascertained through other means. As for the “secrets” of an individual’s personality, graphology has yet to show that it is any better than astrology.

This has not done much to diminish the popularity of handwriting analysis, however. One reason for this is that graphologists, like astrologers, gain some surface plausibility or “face validity” for their claims by exploiting the tendency for people to use the representativeness heuristic. Many of their claims have a superficial sensible quality, rarely violating the principle that like goes with like. Consider, for instance, the “zonal theory” of graphology, which divides a person’s handwriting into the upper, middle, and lower regions. A person’s “intellectual,” “practical,” and “instinctual” qualities supposedly correspond to these different regions (Basil, 1989). Can you guess which is which? Could our “lower” instincts be reflected anywhere other than the lower region, or our “higher” intellect anywhere other than the top?

The list of such representativeness-based connections goes on and on. Handwriting slants to the left? The person must be holding something back, repressing his or her true emotions. Slants to the right? The person gets carried away by his or her feelings. A signature placed far below a paragraph suggests that the individual wishes to distance him- or herself from what was written (Scanlon & Mauro, 1992). Handwriting that stays close to the left margin belongs to individuals attached to the past, whereas writing that hugs the right margin comes from those oriented toward the future.

What is ironic is that the very mechanism that many graphologists rely on to argue for the persuasive value of their endeavor – that the character of the handwriting resembles the character of the person – is what ultimately betrays them: They call it “common sense”; we call it judgment by representativeness.

Representativeness and Psychoanalysis. Nisbett and Ross (1980) contend that “the enormous popularity of Freudian theory probably lies in the fact that, unlike all its competitors among contemporary views, it encourages the layperson to do what comes naturally in causal explanation, that is, to use the representativeness heuristic” (p. 244). Although their claim may be difficult to put to empirical test, there can be little doubt that much of the interpretation of symbols that lies at the core of psychoanalytic theory is driven by representativeness. Consider the interpretation of dreams, in which the images a client reports from his or her dreams are considered indicative of underlying motives. There are, of course, an almost infinite set of potential relationships between dream content and underlying psychodynamics, and it is interesting that virtually all of the “meaningful” ones identified by psychodynamically oriented clinicians are ones in which there is an obvious fit or resemblance between the reported image and inner dynamics. A man who dreams of a snake or a cigar is thought to be troubled by his penis or his sexuality. People who dream of policemen are thought to be concerned about their fathers or authority figures. Knowledge of the representativeness heuristic compels one to wonder whether such connections reflect something important about the psyche of the client, or whether they exist primarily in the mind of the therapist.

One area of psychodynamic theorizing in which the validity of such superficially plausible relationships has been tested and found wanting is the use of projective tests. The most widely known projective test is the Rorschach, in which clients report what they “see” in ambiguous patches of ink appearing on various cards. As in all projective tests, the idea is that in responding to such an unstructured stimulus, a person must “project,” and thus reveal, some of his or her inner dynamics. Countless studies, however, have failed to produce evidence that the test is valid – that is, that the assessments made about people on the basis of the test correspond to the psychopathological conditions from which they suffer (Burros, 1978).

The research findings notwithstanding, clinicians frequently report the Rorschach to be extremely helpful in clinical practice. Might representativeness contribute to this paradox of strongly held beliefs coexisting with the absence of any real relationship? You be the judge. A person who interprets the whole Rorschach card, and not its specific details, is considered by clinicians to suffer from a need to form a “big picture,” and a tendency toward grandiosity, even paranoia. In contrast, a person who refers only to small details of the ink blots is considered to have an obsessive personality – someone who attends to detail at the expense of the more important holistic aspects (Dawes, 1994). Once again, systematic research has failed to find evidence for these relationships, but the sense of representativeness gives them some superficial plausibility.