Illusory Correlation Chapman and Chapman (1967) first demon- strated the existence of illusory correlation in Ashley L. Watts, Sarah Francis Smith, the domain of psychological assessment. They and Scott O. Lilienfeld asked clinical psychologists to list the charac- Emory University, U.S.A. teristics (“signs”) of various Draw-a-Person (DAP; Machover, 1949) protocols that are Many individuals hold strong, often unyield- associated with specific psychiatric symptoms ing, convictions about the association between and personality traits. The DAP is a widely occurrences that have been shown to be sta- used projective technique that asks partici- tistically unrelated, such as the full moon pants to draw three people—a man, a woman, and strange behavior, or joint pain and rainy and themselves—which purportedly allows weather. Indeed, these and many other ven- clinicians to infer a wide range of normal erable legends are held by the masses in the and abnormal personality traits. Chapman face of clear contrary evidence (Redelmeier & and Chapman found that clinicians exhibited Tversky, 1996). strong agreement on certain pairings of DAP Why do many of us perceive statistical asso- signsandcertainsymptomsorpersonality ciations between certain events when these traits. For example, 91% of clinicians reported associations are objectively absent? The answer that atypical eyes on the DAP tend to be a valid is of substantial importance to clinical psy- indicator of suspiciousness, and 82% of clini- chology and allied fields. Clinical psychologists cians agreed that a large or emphasized head dub this phenomenon illusory correlation is a valid indicator of intelligence. Chapman (Chapman & Chapman, 1967, 1969). More and Chapman observed that these associations specifically, an illusory correlation is reflect potent associative linkages between certain drawing signs and certain indicators the report by observers of a correlation between of psychological disturbance. For example, as twoclassesofevents,which,inreality,(a)arenot clinical psychologist Paul Meehl (1920–2003) correlated, or (b) are correlated to a lesser extent noted, paranoid people are often characterized than reported, or (c) are correlated in the opposite by wide-open eyes. Yet these intuitive relations, direction from that which is reported. (Chapman, which mirror those found in DAP manuals, 1967, p. 151) had been disconfirmed in many previous Psychologicalresearchdemonstratesthat studies. humans tend to be poor at estimating the Inthesecondpartofthestudy,theChapmans statistical relations among events; this faulty presented students with DAP protocols along statistical barometer can predispose us to with pairs of symptom statements, and asked illusory correlations (Dawes, 1989). them to inspect the pictures and statements carefully. For example, a drawing presented Illusory Correlation and Clinical to a participant was attached to two symptom Assessment statements, such as “The man who drew this is either (a) suspicious of other people or (b) Although illusory correlation is a domain-gen- worried about how manly he is.” The pairs of eral phenomenon, Chapman (1967) believed symptom statements were drawn from the six that it bears especially important implications symptoms that the clinicians had listed as DAP for clinical settings, such as biases (systematic correlates;thepairsofstatementswereran- errors)inclinicaljudgment.Inaclassicstudy, domly assigned to each drawing so that each The Encyclopedia of Clinical Psychology, First Edition. Edited by Robin L. Cautin and Scott O. Lilienfeld. © 2015 John Wiley & Sons, Inc. Published 2015 by John Wiley & Sons, Inc. DOI: 10.1002/9781118625392.wbecp573 2 ILLUSORY CORRELATION symptom statement appeared only once with likely that strong semantic associations persist each of the drawings. After viewing all of the between some scale labels and test responses drawings, the researchers asked participants (e.g., Depression and “feels blue or sad”). The which characteristics of the drawings were authors hypothesized that many clinicians associated with which psychological charac- continued to use the original labels, which may teristics. Remarkably, these results mirrored lead to illusory correlation when judges are the relations perceived by clinical psychol- asked to infer relations between MMPI scales ogists (e.g., broad shoulders and manliness, and psychiatric symptoms. atypical eyes and suspiciousness). That is, the Dowling and Graham tested this hypothesis undergraduate participants had “recaptured” in a twofold manner. First, they compiled the identical illusory associates perceived by a pool of items in which an MMPI scale clinicians. name was paired with two of 10 behavioral In a follow-up study, Chapman and Chap- statements (e.g., often complains about his man (1969) replicated their findings using a health, sleeps whenever possible, and keeps his different projective task, the perennially popu- clothes neat). Each statement possessed some lar Rorschach Inkblot Test (Rorschach, 1927). semantic association with at least one of the They sought to identify illusory correlations MMPI scales, but not necessarily the scale with between Rorschach responses and homosex- which it was paired. To ensure that there was uality,whichatthetimewasadiagnosticcat- no actual relation between the MMPI scale egory in DSM-II (American Psychiatric Asso- and the behavioral statements, the researchers ciation, 1968). As predicted, an overwhelming chosestatementsthatwerenotstatistically majority of clinicians reported that a number of associated with any clinical scale more than specific Rorschach responses were associated another. Each subject received a brief intro- with homosexuality (e.g., responses referring to duction to the MMPI, including the names of the anus, feminine clothing, genitals, or sexual the 10 scales; nevertheless, they received no ambiguity). As in the Chapman and Chapman descriptions of these scales. Next, one group of (1967) study, the researchers asked under- graduate students, who had completed a course graduates to render judgments after viewing on the MMPI, and one group of undergrad- Rorschach protocols paired with personality uate students each studied the pool of items. statements; the personality statements were Later, they indicated with which MMPI scale rigged so that there was no statistical relation each of the 10 behavioral statements was most between the card and the statement. Despite associated. the absence of any statistical association Despite the fact that the behavioral state- between the Rorschach cards and the symptom ments were no more likely to co-occur with statements, undergraduates, like clinicians, a specific MMPI scale over any other, under- reported that the aforementioned signs (e.g., graduates reported that certain behavioral anus, feminine clothing, etc.) were indicators of statements co-occurred with certain MMPI homosexuality, despite the fact that the actual scales six out of 10 times. Interestingly, graduate correlations between the signs and homosexu- students reported a significant co-occurrence ality in the stimulus materials were zero. for nine out of 10 statements, suggesting even Dowling and Graham (1976) extended higher levels of illusory correlation. Dowling work on illusory correlation to more objective and Graham’s findings demonstrate that illu- assessment techniques, namely, the widely sory correlation is not limited to projective used Minnesota Multiphasic Personality techniques, and suggest that the tendency Inventory (MMPI; Hathaway & McKinley, to perceive relations between uncorrelated 1943). Despite the fact that the MMPI’s clin- phenomena does not necessarily decrease with ical scale labels were replaced with numbers clinical experience. More broadly, findings (e.g., Psychopathic deviate with 4), it seems concerning illusory correlation have been ILLUSORY CORRELATION 3 replicated using different assessment tasks and (Douglas, Guy, & Hart, 2009), mentally ill different groups of participants (see Starr & individuals account for a negligible portion of Katkin, 1969, for similar findings regarding the violent crimes committed (Quinsey, Harris, Rotter Incomplete Sentences Blank task; and Rice, & Cormier, 2006). Furthermore, the link King & Koehler, 2000, for findings regarding between mental illness and violence is mod- graphology, or handwriting analysis). erated by abuse of drugs and alcohol (Fazel, Gulati, Linsell, Geddes, & Grann, 2009). Fazel Illusory Correlation and Erroneous and colleagues (2009) found that the relation Cultural Beliefs between psychosis and risk was due largely or entirely to co-occurring substance use The implications of illusory correlation extend disorders. The increased risk for violence in well beyond psychological assessment. Indeed, psychotic individuals was not significantly dif- many widespread cultural beliefs reflect illu- ferent from the risk for violence in individuals sory correlations. For example, numerous with substance use disorders alone. Although superstitions, such as linkages between the the reasons for these findings are unknown, number 13 and negative outcomes, are illu- certain illicit substances may trigger violent sory correlations. Another striking illusory behavior by impairing judgment or decreasing correlation is the “lunar lunacy effect” or inhibitions. Additionally, among individuals “Transylvania effect”: the purported relation with severe mental illness, substance abuse may between the full moon and mental illness, exacerbate