Paul E. Meehl

Why Summaries of Research on Psychological Theories are Often Uninterpretable

Null hypothesis testing of correlational predictions from weak substantive theories in soft psychology is subject to the influence of ten obfuscating factors whose effects are usually (1) sizeable, (2) opposed, (3) variable, and (4) unknown. The net epistemic effect of these ten obfuscating influences is that the usual research literature review is well-nigh uninterpretable. Major changes in graduate education, conduct of research, and editorial policy are proposed.

Clarifications about taxometric method

Taxometrics is a statistical procedure for determining whether relationships among observables reflect the existence of a latent taxon (type, species, category, disease entity). A formal-numerical definition is needed because intuitive, commonsense notions of “carving nature at its joints” or “identifying natural kinds” cannot resolve disagreements as to taxonic reality for hard cases. Specific etiology (e.g., major gene, germ, traumatic event) is often unknown and is not appropriate in nonmedical domains. Lacking an infallible criterion, the taxonic inference relies on the internal configural relations among the conjectural fallible indicators. An essential feature is multiple consistency tests that will not be satisfied if the latent structure is not taxonic or the parameters are badly estimated. Common misconceptions are that the taxon must be “sharply” distinguished, quantitative indicators must be bimodal, the causal origin must be biological, emergence of a large dimensional factor refutes taxonicity, and adopting a taxon is a mere matter of convention or preference.

Assessing the legal standard for predictions of dangerousness in sex offender commitment proceedings.

Advocates and courts legitimize sex offender commitment laws by claiming the laws confine only those who are “highly likely” to engage in sexual violence. This article proposes a definition of “probability” of future harm and assesses the legal probability thresholds commitment courts actually use. Using published information about recidivism and actuarial prediction, the authors concludes that moderate, but not extravagant, claims about legal probability thresholds are supportable but only on a rather optimistic set of assumptions. The authors recommend that sex offender commitment courts use the proposed methods to quantify judicial standards and findings about prediction. This will allow the claims for legitimacy to be more readily assessed.

The Selfish Voter Paradox and the Thrown-Away Vote Argument *

The probability that an individuals voting in a presidential election will determine the outcome being negligible, it is argued that participation is irrational if predicated on principles that are either egocentric or act-prospective. Voter participation, if rational, must rely on some over-arching principle that is (a) Sociotropic, (b) Axionomic, (c) Collective-distributive, and (d) Neutrofactual. A distinctively ethical component must be involved, such that all purely “economic,” “cost-benefit” models postulating selfish voter rationality are incoherent. The notion of “helping” to elect ones candidate is criticized and rejected unless formulated in a special way. An important pragmatic consequence of the analysis is that the idea (relied on by the two major parties) of “wasting ones vote” on a third party candidate is shown to be invalid or of more limited application than generally assumed. If a sizeable minority (e.g., college students) were educated to reject that argument, politics might be profoundly affected.

A funny thing happened to us on the way to the latent entities.

Inferred latent entities, whether those of psychoanalysis, factor analysis, or cluster analysis, have declined in value for many clinical psychologists, both as tools of practice and as objects of theoretical interest. Behavior modification, rational-emotive therapy, crisis intervention, psycho-pharmacology, and actuarial prediction all tend to minimize reliance on latent entities in favor of purely dispositional concepts. Behavior genetics is, however, a powerful movement to the contrary. As regards categorical entities (types, taxa, syndromes, diseases), history reveals no impressive examples of their discovery by cluster algorithms; whereas organic medicine and psychopathology have both discovered many taxonic entities without reliance on formal (statistical) cluster methods. I offer eight reasons for this strange condition, with associated suggestions for ameliorating it. Adopting a realist instead of a fictionist approach to taxonomy, I give high priority to theory-based mathematical derivation of quantitative consistency tests for all taxometric results. I urge a large scale cooperative survey of taxometric methods based on Monte Carlo runs, biological pseudoproblems where the true axon is independently known, and live problem in genetics, organic medicine, and psychopathology. An empirical example of taxometric bootstrapping and consistency testing was presented from my own current research on schizotypy.

The path analysis controversy: a new statistical approach to strong appraisal of verisimilitude.

A new approach for using path analysis to appraise the verisimilitude of theories is described. Rather than trying to test a models truth (correctness), this method corroborates a class of path diagrams by determining how well they predict intradata relations in comparison with other diagrams. The observed correlation matrix is partitioned into disjoint sets. One set is used to estimate the model parameters, and a nonoverlapping set is used to assess the models verisimilitude. Computer code was written to generate competing models and to test the conjectured models superiority (relative to the generated set) using diagram combinatorics and is available on the Web (http://www.vanderbilt.edu/quantmetheval/downloads.htm).

PHILOSOPHY OF SCIENCE: HELP OR HINDRANCE?

Scientists routinely employ metatheoretic principles, explicit discussion of which typically occurs in times of intense controversy, theoretical crisis, scientific revolution, and entry into a new domain. The writings of philosophers, including their disagreements, are often helpful in such circumstances. Whether knowledge of formal metatheory helps us in doing “normal science” is researchable. Much scientific thinking is of poor quality, and it could be improved by explicit metatheoretical education. Clinical practice and training programs should emphasize rational skepticism, respect for evidence, objectivity, and quantitative thinking. Because the relation between principles and success is probabilistic, metatheoretical research should implement the case study method by formal actuarial procedures.

Cliometric metatheory: II. Criteria scientists use in theory appraisal and why it is rational to do so.

Definitive tests of theories are often impossible in the life sciences because auxiliary assumptions are problematic. In the appraisal of competing theories, history of science shows that scientists use various theory characteristics such as aspects of parsimony, the number, qualitative diversity, novelty, and numerical precision of facts derived, number of misderived facts, and reducibility relations to other accepted theories. Statistical arguments are offered to show why, given minimal assumptions about the world and the mind, many of these attributes are expectable correlates of verisimilitude. A statistical composite of these attributes could provide an actuarial basis for theory appraisal (cliometric metatheory).

A most peculiar paradox

THE “empirical identity” view of the denotata of neurophysiological and phenomenal terms has been challenged as follows: Assume complete determinism in the physical (brain-state) series, and a parallelism between it and the phenomenal (mind-state) series. Suppose the parallelism is interrupted so that the subject experiences a phenomenal state different from that which has been invariably correlated with the present brainstate. No “interaction” occurs, so that all the physical laws hold as usual; yet the subject “would surely know” that he was having the one experience rather than the other. Thus, if an external observer informed him as to the current state of his brain, he would be “aware” that the usual correspondence had broken down. Such a hypothetical failure of isomorphism, even if in fact it never occurs, does not involve a contradiction. This possibility renders any “identity” of the neural and phenomenal unacceptable. I wish to show that this argument involves certain rather paradoxical consequences. For simplicity, suppose there are only two brain-states, G and R. The (hitherto exceptionless) phenomenal accompaniments of these states are mind-states g (“experiencing green”) and r (“experiencing red”). Causally dependent upon the brain-states G and R are subsequent brainstates G′ (“naming green”) and R′ (“naming red”), and dependent upon these are peripheral-motor events G′′ (“uttering word ‘green’”) and R′′ (“uttering word ‘red’ “). The identity view asserts that if an exceptionless regularity holds coordinating g and r to G and R, the role of the former in the whole law system is indistinguishable from that of the latter, so that empirical identity can be asserted, on the usual grounds. The critic claims that even if the physical sequence continues to be wholly lawful a subject would “know” that he was seeing r rather than the usual g. But just what, and how, would such a subject “know”? The physical sequence runs off as usual, Green light→Retinal state for green→G→G′ →G′′. But corresponding to G occurs phenomenal r instead of the usual g. If no physical laws are violated, what are the consequences? (1) The subject will utter “green” although he “knows” he is seeing red. (2) If asked, “You said ‘green,’ did you mean to say ‘red’?” he will answer “No, I mean ‘green,’” since replying is a physical event and the physical series continues as usual. (3) He will hear himself say “green” and will not contradict himself by a subsequent remark, since to do this would mean that the usual physical consequences of a state, say, HG (brain-state produced by stimulus of own utterance), have been affected by the

Using Meta‐Scientific Studies to Clarify or Resolve Questions in the Philosophy and History of Science

More powerful methods for studying and integrating the historical track record of scientific episodes and scientific judgment, or what Faust and Meehl describe as a program of meta‐science and meta‐scientific studies, can supplement and extend more commonly used case study methods. We describe the basic premises of meta‐science, overview methodological considerations, and provide examples of meta‐scientific studies. Meta‐science can help to clarify or resolve long‐standing questions in the history and philosophy of science and provide practical help to the working scientist.