Paul Meehl

Introduction

Paul Everett Meehl (1920-2003) was an American clinical psychologist, philosopher of science, and one of the most cited psychologists of the twentieth century. He spent nearly his entire career at the University of Minnesota, where he worked on psychological assessment, the causes of schizophrenia, the logic of theory testing, and the philosophy of science. A 2002 survey by Haggbloom and colleagues, published in Review of General Psychology, ranked him the seventy-fourth most cited psychologist of the century, tied with Eleanor J. Gibson.

Meehl is best known for a single provocative claim. In Clinical versus Statistical Prediction (1954) he argued that when a body of cases with known outcomes exists, a simple statistical formula combining the available data will usually predict the outcome at least as well as, and often better than, a trained clinician who has access to the same information. The argument was received with open hostility by many practitioners, and Meehl spent the rest of his life asking why evidence that he considered overwhelming had so little effect on professional practice.

His influence reaches well beyond that debate. With Lee Cronbach he gave psychology the concept of construct validity and the nomological network; his 1978 essay on significance testing anticipated concerns that later surfaced during the replication crisis; and his taxometric methods were, for two decades, the standard tool for asking whether a psychological construct is a category or a continuum.

Early Life and Education

Meehl was born on January 3, 1920, in Minneapolis, Minnesota, to Otto and Blanche Swedal. The surname Meehl came from his stepfather. When he was sixteen his mother died as a result of poor medical care, an event he later cited as formative: it permanently damaged his confidence in the diagnostic accuracy of experts, including physicians and, by extension, clinicians. After her death he lived briefly with his stepfather, then with a neighborhood family for a year so that he could finish high school, and finally with his maternal grandparents, who lived near the University of Minnesota.

As a teenager he intended to become a lawyer, but a copy of Karl Menninger's The Human Mind changed his plans overnight and he decided to become a psychotherapist. He had read widely in Freud, Jones, and Ferenczi while still in high school, which put him at odds with the intellectual environment he entered. Minnesota's psychology department was behaviorist and strongly statistical, its psychiatry department was neurological and descriptively oriented, and neither was sympathetic to psychodynamics.

Meehl began undergraduate study at Minnesota in March 1938 and earned a bachelor's degree summa cum laude in psychology in 1941, advised by Donald G. Paterson, with a minor in biometry. His graduate cohort included Marian Breland Bailey, William K. Estes, Norman Guttman, William Schofield, and Kenneth MacCorquodale. He completed his PhD in psychology in 1945 under Starke R. Hathaway, the co-author of the Minnesota Multiphasic Personality Inventory. The arrival of the Vienna positivist Herbert Feigl in the philosophy department, and a 1943 article by Theodore Sarbin on clinical versus actuarial prediction, pushed him toward the problem that would define his career.

Academic and Professional Career

Meehl joined the Minnesota faculty immediately after taking his doctorate and never left. He became a full professor in 1952 and chaired the psychology department from 1951 to 1957, at the age of thirty-one. He was named Regents' Professor, the university's highest academic rank, in 1968. In addition to psychology he held appointments in law, psychiatry, neurology, and philosophy, and he helped found the Minnesota Center for Philosophy of Science with Feigl and Wilfrid Sellars.

His rise through the profession was rapid. He was president of the Midwestern Psychological Association at thirty-four, received the American Psychological Association's Award for Distinguished Scientific Contributions to Psychology at thirty-eight, and served as president of the American Psychological Association in 1962. Later honors included the Bruno Klopfer Distinguished Contributor Award in personality assessment in 1979, election to the National Academy of Sciences in 1987, the APA Award for Lifetime Contributions to Psychology in 1996, and the James McKeen Cattell Fellow Award in 1998.

Retiring in 1990, he continued teaching as Hathaway-Meehl Professor of Psychology until 1993 and practiced psychotherapy from 1951 until 1994. In 1958 he collaborated with a group of Lutheran theologians and psychologists on What, Then, Is Man?, a project commissioned through Concordia Seminary by the Lutheran Church-Missouri Synod. In 1995 he was one of the signatories of the "Mainstream Science on Intelligence" statement written by Linda Gottfredson. He died at his home in Minneapolis on February 14, 2003, of chronic myelomonocytic leukemia.

Major Contributions to Psychology

Clinical versus statistical prediction

Meehl's 1954 book asked a narrow question: given a set of cases with known outcomes, can a skilled clinician combine the available information into an accurate prediction as well as a formal rule that uses the same information? His answer, based on roughly twenty studies spanning academic performance, parole violation, and response to treatment, was no. Subsequent tallies grew steadily; by 1965 he knew of fifty-one such comparisons, and he and William Grove later assembled 136, of which only eight clearly favored the clinician.

The canonical meta-analytic confirmations came after his death. Grove and colleagues, in a 2000 meta-analysis in Psychological Assessment, found that mechanical techniques were on average about ten percent more accurate than clinical predictions, substantially outperforming them in roughly a third to a half of studies while clinical predictions were substantially better in only six to sixteen percent. Stefanía Ægisdóttir and colleagues, in a 2006 review of sixty-seven studies spanning fifty-six years, reported a thirteen percent accuracy advantage for statistical methods in their most stringent sample.

Meehl himself granted one famous exception. He called it the "broken leg" case: a formula may say that a professor goes to the cinema on Tuesday nights with probability 0.9, but if he has just fractured his femur and cannot fit into a theater seat, no sensible person sticks with the formula. Meehl argued that such cases are real but rare, that clinicians systematically overestimate how often they occur, and that the correction should itself be evaluated actuarially.

Construct validity and the nomological network

With Cronbach, Meehl argued in 1955 that psychological tests cannot be validated merely by correlating them with a criterion. A test measures a theoretical construct, and evidence for it accumulates through a nomological network: a web of lawful relations linking the construct to other constructs and to observable behavior. This reframing licensed scientific talk about unobservable psychological processes at a time when operationism and behaviorism discouraged it, and it remains the conceptual foundation of modern test validation.

Schizotaxia, schizotypy, and schizophrenia

In a 1962 paper Meehl proposed that liability to schizophrenia is carried by a single dominant gene, producing an integrative neural defect he called schizotaxia. Interaction with environmental and polygenic influences yields a measurable personality organization, schizotypy, and a minority of schizotypes decompensate into clinical schizophrenia. At a time when schizophrenia was widely attributed to parenting, Meehl insisted the disorder was biological in origin. He later estimated the population prevalence of schizotypy at about ten percent.

Taxometrics

To test whether a latent variable is a discrete class or a continuum, Meehl and his colleagues developed a family of procedures including MAXCOV and MAMBAC, collectively known as taxometrics. The approach does not impose a structure the way factor analysis or cluster analysis does; instead it looks for consistency across several mathematically independent procedures. Meehl applied it first to schizotypy, believing it vindicated his taxon, and it was subsequently applied to mood, anxiety, eating, and personality disorders.

Metascience and the critique of significance testing

In "Theoretical Risks and Tabular Asterisks" (1978) Meehl argued that in the soft areas of psychology the null hypothesis is quasi-always false, so that tables of significant differences are largely artifacts of statistical power. Everything correlates with everything to some degree, a phenomenon he called the crud factor. Because most psychological theories predict only a direction rather than a numerical value, more data makes them easier rather than harder to "confirm," the opposite of what happens in mature sciences. His remedy was risky numerical point predictions and consistency tests, a position he refined in later papers on what he called the problem of epistemology rather than statistics.

Major Publications

  • Clinical versus Statistical Prediction: A Theoretical Analysis and a Review of the Evidence (1954), University of Minnesota Press, reissued with a new preface in 1996.
  • An Atlas for the Clinical Use of the MMPI (1951), with Starke R. Hathaway.
  • What, Then, Is Man? (1958), written with a group of Lutheran theologians and psychologists.
  • Psychodiagnosis: Selected Papers (1973), a collection including the construct validity paper, "Schizotaxia, Schizotypy, Schizophrenia," and "Why I Do Not Attend Case Conferences."
  • Selected Philosophical Papers (1991), edited by C. A. Anderson and K. Gunderson.
  • Multivariate Taxometric Procedures: Distinguishing Types from Continua, with Niels Waller.

Impact and Legacy

Meehl's central finding has been repeatedly confirmed and, by his own account, largely ignored. In a 1989 survey he conducted with Robyn Dawes and David Faust, more than fifteen percent of responding clinical psychologists had never heard of the clinical-statistical controversy, and only forty-two percent had covered it in detail during training. He and Grove catalogued the reasons for the resistance: fear of technological unemployment, threats to professional self-concept, attachment to theory, the belief that formulas are dehumanizing, resentment of computers, and poor methodological education.

The literature has also attracted substantive criticism. Drew Westen and Joel Weinberger argued that Meehl's paradigm asks clinicians to do something they are not trained to do, namely predict broad, distal behavioral outcomes, and that clinical expertise is more visible in moderate-inference judgments, case formulation, and treatment planning, where actuarial alternatives are rarely available. Others have noted that the average mechanical advantage is modest rather than overwhelming, that many primary studies are old, and that a formula cannot rescue poor assessment data. Defenders reply that clinicians who are handed an algorithmic prediction and allowed to override it reliably make it worse, and that Grove's analysis of the broken-leg objection shows why: genuine exceptions are rare, behavioral science theories are seldom as well corroborated as the mechanics of a hip cast, and unusual events in human affairs usually shift probabilities only modestly.

Meehl's own taxometrics has fared worse under scrutiny. A quantitative review by Nick Haslam and colleagues in 2012 covered 177 articles, 311 findings, and more than half a million participants. Although thirty-nine percent of findings were taxonic, such findings were far less common in methodologically stronger studies, and after controlling for confounds the estimated true prevalence of taxa was about fourteen percent. Simulation work by John Ruscio and colleagues showed that skewed or dichotomous indicators can produce plots that look taxonic without any taxon existing. The authors concluded that many influential early taxonic findings, including several in the schizotypy literature Meehl founded, were probably spurious.

Meehl was notably willing to criticize himself. In "Causes and Effects of My Disturbing Little Book" (1986) he judged that no more than five percent of the 1954 book needed retraction three decades later, and he added that his errors had run in one direction: he had overestimated how much clinicians might gain from configural, nonlinear pattern recognition, so that correcting the book would make it more actuarial, not less. He also acknowledged that Quinn McNemar had caught a statistical error in one of the studies he reviewed that he had missed. That mix of strong claims and public self-correction is why Meehl remains a reference point in arguments about algorithms, expertise, and the limits of human judgment.

Learn More

References

  • Meehl, P. E. (1954). Clinical versus Statistical Prediction. University of Minnesota Press.

Last updated: 2026-10-20

Menu
Loading