Clinical versus Statistical Prediction: A Theoretical Analysis and a Review of the Evidence

Meehl argues that for a given predictive task, if the same data are available to both, a formal statistical or actuarial rule will generally predict as well as or better than a clinician's informal judgment. He rejects the assumption that clinical training and experience confer a predictive advantage that cannot be captured by mechanical combination. Crucially, Meehl separates two questions that were often conflated: how data are gathered (where clinicians may have genuine advantages) and how data are combined into a prediction (where statistical rules tend to win). His preface states the thesis carefully: there is no convincing reason to assume formal rules and clinical creativity are equally suited to every task, and comparative effectiveness may vary by task type .

The book's most enduring conceptual contribution is the distinction between data collection and data combination. Meehl argued that the clinical-statistical controversy was often muddled because critics defended clinical methods on the grounds that clinicians gather richer information—interviews, observations, rapport—while the actual comparison concerned how information is combined once available. A clinician may elicit unique data; that does not mean the clinician combines those data optimally in their head. Grove and Meehl's later work formalized this distinction, and Grove's retrospective assessment concluded that Meehl's conceptual analyses of the prediction problem had not been significantly improved since the 1950s . This distinction remains foundational in assessment research and algorithmic decision-making debates.

The conclusion has held up remarkably well. Grove's 2006 review in the Journal of Abnormal Psychology stated that Meehl's conclusion that statistical prediction consistently outperforms clinical judgment 'has stood up extremely well for half a century,' and that his conceptual analyses remained unsurpassed . A later reference work summarized that follow-up studies have generally provided strong support for Meehl's original findings across diverse domains, with statistical prediction equal or greater in accuracy . Meehl himself, looking back in 1986, said he would change at most five percent of the book, acknowledging he had overemphasized the potential advantage of configural judgment . The empirical record has vindicated the core claim while refining its scope.

The broken-leg case is Meehl's own hypothetical illustrating a genuine limit of actuarial prediction: if an actuarial rule predicts a professor goes to the movies on Tuesday nights with .9 probability, but we learn he broke his leg that morning, the clinician rightly overrides the rule. Meehl used this to acknowledge that clinical judgment has a legitimate role in exceptional, theory-driven circumstances. However, he immediately qualified the concession: broken legs are rare, so such overrides do not substantially improve overall accuracy; the example relies on a well-corroborated physical theory, whereas psychological theories are far less reliable; and whether an override is justified in any individual case is itself an empirical question . The broken-leg case is therefore not a refutation but a boundary condition that Meehl himself articulated.

Reception among clinicians was often defensive and sometimes distressed. Robert R. Holt, in a 1958 reformulation, captured the mood: clinical students complained of 'a vague feeling that a fast one has been put over on them'—that Meehl had sold the clinical approach 'up the river' under a show of objectivity . Meehl later recounted hearing that half the Freudian clinical staff at a large Midwestern university 'lapsed into a long reactive depression' after reading it . Yet Meehl was himself a practicing psychodynamic clinician, not a hostile outsider. He described having 'strong identifications on both sides,' which he believed allowed him to be relatively fair-minded about a charged topic . The book's reception reflects a profession confronting uncomfortable evidence about its own judgment.

A frequent misreading treats Meehl as claiming that statistical prediction is always and everywhere superior, or that clinicians should be replaced by formulas. Meehl's actual position was more careful. His preface emphasizes the 'mental conflict' of doing clinical work while trying to be a scientist and insists the matter 'is just not that simple' . He acknowledged genuine clinical powers, including the broken-leg exception, and his target was the unexamined assumption that clinical judgment is adequate, not the existence of clinical skill. A related misreading conflates his argument with a blanket dismissal of clinical data collection. Meehl's point was about combination rules, not about the value of interviews or rapport. Clinicians who dismiss the book as anti-clinical miss its actual scope.