Law and the Fireside Inductions

522 Meehl: Law and the Fireside Inductions THE PSYCHOLOGY OF THE FIRESIDE Lawmen will immediately see the point of my title, but for social science readers I should explain. The phrase “fireside equities” is legalese for what the legal layman feels intuitively or commonsensically would be a fair or just result (Llewellyn, 1960). Sometimes the law accords with the fireside equities, sometimes not; and lawyers use the phrase with derisive connotation. Analogously, by the language “fireside inductions” I mean those commonsense empirical generalizations about human behavior which we accept on the culture’s authority plus introspection plus anecdotal evidence from ordinary life. Roughly, the phrase “fireside inductions” designates here what everybody (except perhaps the skeptical social scientist) believes about human conduct, about how it is to be described, explained, predicted, and controlled. One source of conflict between the social scientist and practitioner of law— especially the legislator—is the former’s distrust of common knowledge concerning human conduct and the latter’s reliance upon this common knowledge. Such reliance is often associated among lawyers with doubts about the value of generalizations, arising from systematic behavioral science research involving quantification and experimental manipulation in artificial situations. Reliance upon “what everyone knows” (simply by virtue of being himself a human being) was hardly critically scrutinized prior to the development of the experimental and statistical methods of contemporary social science. This historical fact provides a built-in preference for the commonsense knowledge of human behavior embodied in positive law. But psychologists mistakenly suppose that the lawyers’ continued reliance upon the psychology of the fireside is wholly attributable to inertia, and these misunderstandings warrant consideration. Without being honorific or pejorative, I shall use “fireside inductions” to refer broadly to those expectations and principles, largely inchoate although partially embodied in proverbs and maxims (e.g., “The burnt child dreads the fire,” “Blood is thicker than water,” “Every man has larceny in his heart,” “Power always corrupts”) arising from some mixture of (1) personal anecdotal observations, (2) armchair speculation, (3) introspection, and (4) education in the received tradition of Western culture prior to the development of technical social science method. It is not clear where nonquantitative, nonexperimental but psychologically sophisticated ideas, such as those of contemporary psychoanalytic theory and therapy, should be classified, but for the moment I will set this aside. With my fellow psychologists I share a considerable skepticism concerning the fireside inductions. Even universally-held generalizations about the origins and control of human conduct should be subjected to (at least) quantitative documentary research and, where feasible, to systematic experimental testing. Obviously the degree of skepticism toward a dictum of commonsense psychology should increase as we move into those areas of social control where our efforts are hardly crowned with spectacular success. For example, there is no known system for the prevention or cure of crime and delinquency that is so strikingly successful that anyone can suggest we are doing so well at this social task that Meehl: Law and the Fireside Inductions 523 it is hardly necessary to call our techniques into question, absent specific research that casts doubt upon them (Meehl, 1970c). That the psychological presuppositions underlying the criminal law should be subjected to merciless armchair scrutiny and quantitative research is not said pro forma but expresses a sincere conviction. UNFAIR CONTROVERSIAL TACTICS SOMETIMES USED BY LAWYERS AND PSYCHOLOGISTS Nor is this merely a platitude—“we need research”—that everyone accepts. One does come across rational, educated persons who disagree, at least when presented with concrete instances. I know, for instance, a very able law professor (formerly a practicing attorney) whose ignorance of the behavioral sciences was systematic and deliberate and who, although he regarded me highly as an individual intellect, made no secret that he thought most scientific research on law, such as quantitative studies on jury behavior, had little point. Over several months, I realized that he had a foolproof heads-I-win-tails-you-lose technique dealing with intellectual threats from the social sciences, to wit: If I introduced a quantitativedocumentary or experimental study of some behavioral generalization having relevance to the law, the findings either accorded with his fireside inductions, or they did not. It they did, he typically responded, “Well, I suppose it’s all right to spend the taxpayer’s money researching that, although anybody could have told you so beforehand.” If the results were not in harmony with the fireside mindmodel, he refused to believe them! When I called this kind of dirty pool to his attention, he cheerfully admitted this truth about his debating tactics. Without defending such illegitimate, systematic resistance to the inroads of behavior science data upon legal thinking, I direct my behavior sciences’ brethren to some considerations that may render my law colleague’s tactics less unreasonable than they seem at first glance. Some behavior scientists, particularly those ideologically tendentious and often completely uniformed with respect to the law, reveal a double standard of methodological morals that is the mirror image of my legal colleague’s. They are extremely critical and skeptical about accepting, and applying in practical circumstances, fireside inductions but are willing to rely somewhat uncritically upon equally shaky generalizations purporting to be the rigorous deliverances of modern behavior science. A shrewd lawyer, even though he might not know enough philosophy, logic of science, experimental method, or technical statistics to recognize just what is wrong with a particular scientific refutation of the fireside inductions, may nevertheless be right in holding to what he learned at his grandmother’s knee or through practical experience, rather than abandoning it because, say “Fisbee’s definitive experiment on social conformity” allegedly shows the contrary. Example: Punishment as a General Deterrent Consider the threat of punishment as a deterrent, one of the most socially important and widely disputed issues relating behavior science to law. While I 524 Meehl: Law and the Fireside Inductions have not kept systematic records of my anecdotal material (fireside induction!), the commonest reaction of psychologists upon hearing of my interest in studying law and teaching in the Law School, is a surprised, “Well, Meehl, I have always thought of you as a hard-headed, dustbowl-empiricist, quantitatively-oriented psychologist—how can you be interested in that medieval subject matter?” When pressed for an explanation of why they consider law medieval, my behavior science colleagues generally mention the outmoded and primitive (sometimes they say “moralistic”) reliance of the criminal law upon punishment, which “is out of harmony with the knowledge of modern social (or medical) science.” This kind of rapid-fire sinking of the lawyer’s ship quite understandably tends to irritate the legal mind. However, the same psychologist who says punishment doesn’t deter relies on its deterrent effect in posting a sign in the departmental library stating that if a student removes a journal without permission, his privilege to use the room will be suspended but his use fee not returned. This same psychologist suspends his children’s TV privileges when they fight over which channel to watch; tells the truth on his income tax form (despite feeling that the government uses most of the money immorally and illegally) for fear of the legal consequences of lying; and drives his car well within the speed limit on a certain street, having been informed that the police have been conducting speed traps there. It will not do for this psychologist to say that as a citizen, parent, professor, taxpayer, automobile driver, etc., he must make such judgments upon inadequate evidence, but when contemplating the legal order he must rely only on scientific information. [Editor’s note: Dr. Meehl goes on to provide examples of the faulty reasoning or evidence social scientists may offer to defend their dismissal of punishment as a deterrent, and notes that such practices may fuel the legal profession’s skepticism about social scientists’ pronouncements. Dr. Meehl then indicates that varying levels of sophistication in social science methodology may be needed to avoid misinterpreting data that relates punishment to crime rate.] LEVELS OF SOPHISTICATION IN SOCIAL SCIENCE METHOD This levels-of-sophistication question is of great importance in interdisciplinary work and in legal education. Any lawyer knows that having the more meritorious case does not guarantee winning it, a main interfering factor being skill of counsel. Differing levels of sophistication in any technical domain, even possession of a special vocabulary, often lead to misleading impressions as to who has the better of a theoretical or practical dispute. The parish priest can refute the theological objections of an unlettered Hausfrau parishioner. The priest, in turn, will lose a debate with the intellectual village atheist. C. S. Lewis will come out ahead of the village atheist. But when C. S. Lewis tangles with Bertrand Russell, it gets pretty difficult to award the prizes. This dialectical-upmanship phenomenon has been responsible for some of the friction between lawyers and social scientists, especially when the social scientist tendentiously presents what purport to be the findings of modern social science but is expressing the particular psychologist’s, psychiatrist’s, or sociologist’s ethos or theoretical (ideological) Meehl: Law and the Fireside Inductions 525 prejudices. Undergraduate sophistication sufficiently questions the efficacy of criminal law sanctions as a deterrent (although some college-educated legislators appear to be naive about this!), and recognizes the desirability of adequate statistics on comparable jurisdictions or within the same jurisdiction before and after a change in severity (or certainty) of penalty. However, to understand threshold effects, asymptotes, second-order interactions, nonlinear dependencies, rate changes at different points on a growth function—considerations hardly profound or esoteric—already takes us beyond the level of sophistication of many social and medical writers who have addressed themselves to legal problems. [Editor’s note: Dr. Meehl then provides examples of overly quick, or unwarranted generalization on the part of social scientists. He observes that the presumed ineffectiveness of the death penalty as a deterrent does not justify “the rapid-fire dismissal of the general idea that increasing a penalty will be effective.” He notes that such problematic generalizations reflect lack of knowledge or consideration of methodological and technical issues. In order to illustrate some of these methodological complexities, Dr. Meehl notes that increased punishment might decrease prohibited behavior under certain conditions, e.g., when apprehension is a near certainty and penalties are well publicized. Evidence is cited that supports this possibility, such as the decreased rate of military mutiny under the above cited conditions and variations in rate of crime when these conditions are and are not present. Dr. Meehl observes that in contrast, social scientists may attempt to countervail fireside inductions about punishment by invoking studies, such as those involving infrahuman animals, that are of doubtful application to everyday human affairs and require highly tentative extrapolations.] THE CONCEPTS EMPIRICAL, EXPERIMENTAL, QUANTITATIVE Rational discussion of the law’s reliance on the fireside inductions may be rendered needlessly difficult by an unfortunate semantic habit as to the honorific word “empirical.” Since I have myself been fighting a running battle with my psychological associates for some years against this bad semantic habit, I would dislike to see it accepted by legal scholars. The following methodological equation, often implicit and unquestioned, is being taken over by lawyers from behavior science: Empirical = Experimental-and-Quantitative = Scientific This equivalence is objectionable on several grounds. It is epistemologically inaccurate since there is a great deal of the empirical (i.e., arising in or supported by observations or experiences, including introspective experiences) that is neither experimental nor quantitative. Furthermore, the middle term assumes a false linkage because (1) not all experimental research is quantitative and (2) not all quantitative research is experimental. Third, several disciplines (to which hardly anybody would refuse the term scientific) exhibit varying amounts of experimental manipulation conjoined with varying amounts of the qualitative/quantitative 526 Meehl: Law and the Fireside Inductions dimension, e.g., astronomy, ecology, botany, human genetics, paleontology, economics, meteorology, geography, historical geology, epidemiology, clinical medicine. What is an experiment? I am not prepared to give an exactly demarcated definition of the term. Roughly, an experiment is a systematic, preplanned sequence of operations-cum-observations, the system of entities under study being relatively isolated from the influence of certain classes of causal factors; other causal factors being held quasi-constant by the experimenter; and still others manipulated by him, their values either being set for different individuals in the system or changed over time at the experimenter’s will; and output is recorded at the time. Some (under Sir R. A. Fisher’s influence) would add, but I would not, that remaining causally efficacious factors (known, guessed, or completely unidentified but assumed to exist) must be rendered noncorrelates of the manipulated variables by a randomizing procedure permitting their net influence to be estimated (statistical significance test). This definition says nothing about apparatus, instruments, measurement, or even being in the laboratory. I disapprove of stretching the word “experiment” to include clinical or sociological research based upon ex post facto assessment processes, entering files of old data, naturalistic observations in the field or in public places, and so forth. But Campbell’s “quasi-experiment” is useful to denote a subset of these possessing certain methodological features that render them relatively more interpretable (Campbell, 1969). The word “experiment” has become invidious because biological and social scientists tend to denigrate nonexperimental sources of knowledge (such as clinical experience, analysis of documents, file data, or the fireside inductions). Then, by equating “experimental” with “empirical” with “scientific,” they often imply that any knowledge source other than experimental is methodologically worthless (armchair speculation, appeal to authority, metaphysics, folklore, and the like). But the fireside inductions are empirical. No logician would hesitate to say this. Their subject matter is the domain of empirical phenomena, and one who invokes a fireside induction will, when pressed to defend it, appeal to some kind of experience which he expects the critic will share with him, whether personally or vicariously. Even the traditional law review article which traces, say, the development of a juridical concept like “state action” or “substantive due process” through a historical sequence of appellate court opinions is empirical, since its subject matter is the verbal behavior, recorded in documents, of a class of organisms, and the researcher studies the changes in this verbal behavior over time. The presence of analytical discourse in such a traditional law review paper does not render it nonempirical, but to argue this is beyond the scope of the present paper (see Feigl & Meehl, 1974; Meehl, 1970b; Skinner, 1969, Ch. 6). There are important differences between the traditional law review article and the kind of article we expect to find in Law and Society Review. But we have some perfectly good words, more precise and less invidious, to characterize the difference. For a study of files or documents utilizing the statistical techniques of behavior science we can say simply “statistical,” a straightforward word that Meehl: Law and the Fireside Inductions 527 means pretty much the same thing to most people and which is not loaded emotionally. If structural statistics (such as factor analysis or multidimensional scaling, see Meehl, 1954, pp. 11–14) are employed, we have the word “psychometric.” Distinguishing the quantitative or statistical from the experimental dimension is particularly important in discussing methodology of research on law because—as in clinical psychology and personology—one research method in these fields is the application of statistical and psychometric techniques of analysis to documents (e.g., diaries, interview transcripts, jury protocols, Supreme Court opinions). It would be misleading to say that one “performs an experiment” if one plots a curve showing the incidence of concurring opinions over time in the behavior output of an appellate court, but it is equally misleading to say that a traditional law review article which draws no graphs and fits no mathematical functions but traces through a set of opinions over time with reference to the incidence of split votes and dissents, presented in ordinary text, is not empirical. Research does not cease to be empirical, or even behavioral, when it analyzes behavior products instead of the ongoing behavior flux itself. Since the control of variables influencing a dependent variable is a matter of degree, situations arise in which one is in doubt as to whether the word “experiment” is applicable. But this is merely the familiar problem of drawing an arbitrary cut whose location matters little. For research designs methodologically more powerful than studying a slice of cross-sectional file data because we have changes over time in relation to a societal manipulation (e.g., amendment of a penal statute), we have the expression “quasi-experiment” (Campbell, 1969). MENTAL TESTS AND SOCIAL CLASS: THE LEVELS-OF-SOPHISTICATION POINT EXEMPLIFIED The sophistication-level effect is beautifully illustrated by the vexatious problem of interpreting socioeconomic differences on mental tests. I suppose the minimum sophistication level necessary in order even to put the interesting questions is that of understanding why and how intelligence tests were built and validated, including basic concepts of correlation, content-domain sampling, reliability, validity, developmental growth curves, etc., that one learns about in an elementary psychology class. Exposure to basic psychometric theory and multiple strands of validation data (Campbell & Fiske, 1959; Cronbach & Meehl, 1955; Jackson, 1969; Loevinger, 1957) should eliminate some common antitest prejudices and excessive reliance upon anecdotal refutations. (Example: “I knew a kid with an IQ of 196 who became a bum.”) I find it odd, by the way, that some lawyers will pronounce confidently to me about what intelligence tests do and do not measure, when the pronouncer could not so much as define IQ, factor loading, or reliability coefficient. I cannot conceive myself asserting to a lawyer that “the Hearsay Rule is silly” without having at least taken a course or read a treatise on the law of evidence, where the rationale of the rule and its exceptions was discussed. But I am uncomfortably aware that some psychologists permit 528 Meehl: Law and the Fireside Inductions themselves strong—usually negative—views about the law without knowing any thing about it. This bottom level of psychometric sophistication would suffice for an employer to consider using psychological tests in screening job applicants. At a higher level, one thinks about the social class bias in tests. If one stops there, relying upon the class-score correlations as definitive proof of test bias, one will perhaps avoid using tests, either because one may “miss” some good candidates (a poor reason, statistically) or from considerations of fairness, justice, equal opportunity (a good reason, provided the psychometric premise concerning bias is substantially correct). Moving one step higher in sophistication, we realize that the SES-IQ correlation is causally ambiguous, and that the limitations of statistical method for resolving this causal ambiguity are such that no analysis of file data can tell us what causes what. No one knows and, worse, no one knows how to find out to what extent the SES-IQ correlation is attributable to environmental impact and to what extent it is attributable to genetic influence. This causal ambiguity, while rather obvious (and clearly pointed out over 40 years ago in Burks & Kelley, 1928) is, as I read the record, somewhat above the sophistication level of many sociologists and psychologists, who talk, write, and design experiments on the (implicit) assumption that social class is entirely on the causal input side of the equation. The terrible complexity of this problem cannot be discussed here, but I have treated it elsewhere (Meehl 1969, 1970a, 1971a). I can briefly concretize it by reference to the Coleman Report (Coleman, Campbell, & Hobson, 1966) on equality of educational opportunity. In the course of an interdisciplinary discussion at the University of Minnesota branch of the Law and Society Association one law professor argued: Since the Coleman Report showed that the psychological characteristics of a student’s peer group were more closely correlated with his measured ability and achievement than either the school’s physical plant or the characteristics of the teaching staff, these “empirical data of behavior science” would indicate that the way to achieve equal educational opportunity should be mandatory busing to provide disadvantaged pupils with the presumably better stimulation from abler peers. Whatever the legal merits of mandatory busing in relation to de facto segregation, the methodological point is important and requires a level of psychometric sophistication a notch above my law colleague’s. It is possible that the higher statistical correlation between peer-group attributes and student’s academic level is attributable mainly to geographic selective factors mediated by the family’s social class, rather than causal influence of peer-group stimulation. Parental intelligence, personality, and temperament factors are transmitted to the child in part genetically (no informed and unbiased person today could dispute this, but many social scientists are both uninformed and prejudiced against behavior genetics) and partly through social learning. If physical-plant characteristics and teacher characteristics are correlated with the biological and social inputs of the child’s family only via the (indirect) economic-neighborhood-location and political (tax-use) factors, they will have a lower statistical relationship with the child’s cognitive level than is shown by indicators of the cognitive level Meehl: Law and the Fireside Inductions 529 of other children attending the same neighborhood school. Roughly, peer-group attributes happen to be a better (indirect) measure of average family and neighborhood causal factors—genetic and environmental—than teacher or physical plant attributes. The differential correlation would reflect more a psychometric fact (about the factor loadings and reliabilities of certain measures) than a causal fact (about peer-group influence). Whether one analyzes these data by inspecting a correlation table or by more complex statistical devices such as regression equations or analysis of covariance, neither the crude zero-order relationships, beta-weights, nor sums of squares tell us much about the direction of causal influence. We cannot infer whether social stimulation from other students is causally more efficacious than having better qualified teachers or a newer, better lighted, cleaner school building. The correlations with peer-group attributes cannot even tell us whether the impact upon a lower-IQ child of being in a classroom with more bright, dominant, articulate, and intellectually self-confident children does more harm than good. In the opinion of Judge Skelley Wright in Hobson v. Hansen (1967) one cannot find a single sentence indicating recognition of this methodological problem. I do not suggest that awareness of it would have led to a different result. But a single sentence obiter would surely occur somewhere in his careful, scholarly opinion of 109 pages had the judge thought of it or counsel argued it in connection with the Coleman Report’s significance. One may feel, as I do, that the problems of racial discrimination and educational disadvantagement are so grave that the society should lean over backwards—within limits set by principles of distributive justice to individuals—to change things, since we are confronted with a frightful combination of gross inequities and a major social emergency. Under such pressing circumstances the adoption and implementation of policy cannot await definitive solution of difficult scientific questions, especially when the kind of controlled experiment or even semicontrolled quasi-experiment (Campbell, 1969) capable of yielding clear answers cannot be performed and statistical techniques presently available are not adequate for the purpose of unscrambling causal influence. The Coleman Report shows that minority-group children receive substandard educational treatment, and I for one am willing to call that discriminatory, ipso facto. What concerns me here is the legal generalizability of a causal inference methodology. The kind of reliance upon social science data found in Hobson v. Hansen, lacking adequate clarification of the concept “unfair discrimination” in relation to correlational findings, might produce some untoward results in other contexts where the interpretative principles would be difficult to distinguish. And if judges should become cynical about the trustworthiness of what psychologists and sociologists assert, we might be faced with a judicial backlash against the social sciences. One can hardly blame Judge Wright for making a flat statement that intelligence tests do not measure innate intelligence or for repeating the old chestnut that intelligence is “whatever the test measures (Hobson v. Hansen, 1967, p. 478).” The Coleman Report states flatly, with no hint that disagreement exists among psychologists on the highly technical and obscure issues involved, “recent research does not support [the] view” that 530 Meehl: Law and the Fireside Inductions intelligence tests measure more fundamental and stable mental abilities than achievement tests (Coleman et al., 1966, p. 292). “Ability tests are simply [sic] broader and more general measures of education” follows at the same locus, again without the faintest whiff of doubt or qualification to warn the legal reader that this is a complex and controversial issue. We read further, “The findings of this survey provide additional evidence that the ‘ability’ tests are at least as much affected by school differences as are the ‘achievement’ tests,” the causal language “affected by” being again unqualified, with no mention made that some psychologists would interpret the parallelism of ability-score differences and achievement-score differences as suggesting that the achievement differences are not primarily due to school differences but to intelligence differences (an interpretation that would fit well with the report’s other findings). It is well recognized among psychologists concerned with the ex post facto design’s deficiencies that the report is dangerous reading for the nontechnically trained, because of its pervasive use of causalsounding terms: influence, affect, depend upon, account for, independent effect (sic). This causal atmosphere cannot be counteracted by the brief methodological sections, which contain the usual caveats. One wonders whether the report’s authors were really clear about what regression analysis can and cannot do (Guttman, 1941, pp. 286–292). I am not arguing the merits, except to show that an unresolved scientific controversy exists which we psychologists have no right to sweep under the rug when we talk or write for lawyers and judges. If we present a distorted picture even in a good cause, implying that certain technical matters are settled when in fact they are obscure and controversial, the powerful forces of the lawyers’ adversary system will, sooner or later, ferret out the secret. Could we then complain if the findings of social science were treated with less respect than those of chemistry, geology, or medicine by less tractable, more wised-up judges? [Editor’s note: Dr. Meehl then provides an additional example of the complexities involved in analyzing correlational data, and concludes: “The problem of interpreting correlations and the influence of ‘nuisance variables’ is not a hairsplitting academic exercise, it is a major methodological stomach ache, arising in many legal contexts where social science findings are relevant to fair treatment or equal protection.”] AWAY FROM THE FIRESIDE AND BACK AGAIN The levels-of-sophistication problem has a time component reflecting the stage of scientific knowledge. We psychologists should be cautious when an alleged principle of modern behavioral science appears to conflict with the fireside inductions. There are some embarrassing instances of overconfident generalization and unjustified extrapolation which were subsequently corrected by movement back toward the fireside inductions. It would be worth knowing how often such back-to-the-fireside reversals have taken place, and whether there are features of subject matter or methodology that render the counter-fireside pronouncements of social science prone to reversal or modification. Sometimes psychologists seem to prefer negating the fireside inductions, especially those embedded in Meehl: Law and the Fireside Inductions 531 the received scholarly tradition (e.g., Aristotle should be beaten up wherever possible). For example, the experimental psychologists’ revival of the constructs “curiosity” and “exploratory drive” seems strange to a nonpsychologist who has observed children or pets—or who remembers the opening sentence of Aristotle’s Metaphysics, “All men by nature desire to know.” Consider three examples relevant to law as a means of social control. Traditional reliance upon punishment (aversive control) in socialization, both in suppressing antisocial conduct and in education, is horrifying to the contemporary mind. One reads about the execution of a 14-year-old for larceny in the 1700s, or Luther’s description of his schooling, in which corporal punishment was not even confined to infractions of discipline, but was the standard procedure of instruction. If a child didn’t give the right answer, he would be rapped on the knuckles with a rod. Research on white rats and, to a lesser extent, on human subjects led to the generalization that, by and large, punishment is a mode of behavior control inferior to reward (positive reinforcement). Thirty years ago, I was taught that the useful role of punishment was to suppress undesirable responses sufficiently (in the short run) so that alternative competing behaviors could occur, and the latter could then be positively reinforced. This is still a fair statement of the practical situation. Following the publication of Skinner’s epoch-making The Behavior of Organisms (1938), his student Estes’s doctoral dissertation (1944), and especially Skinner’s Science and Human Behavior (1953) and his Utopian novel Walden Two (1948), aversive control fell into extreme disfavor. These writings combined with the gruesome stories told to us clinical psychologists by adult neurotics about their aversively controlled childhoods to produce a rejection of both the general deterrent and rehabilitative functions of the criminal law. But this could be an illicit extrapolation, conflating the rehabilitative and general deterrent functions of the criminal law under the generic rubric “punishment.” Supporters of general deterrence need not assume the same psychological process operates on deterrable persons as that involved when punishment is unsuccessful in reforming convicts. Only punishment as a reformer even approximates the laboratory model of an aversive consequence following emission of the undesired response. Furthermore, the criminal sanction is rather more like withholding positive reinforcement (given elimination of flogging and similar practices from the penal system), both fines and imprisonment being deprivations. The distinction is becoming fuzzed up by experimental work with animals because manipulations such as “time out” (during which the instrumental act cannot be performed because the manipulandum is unavailable, or a stimulus signals that reinforcement will now be withheld) has aversive (punishing) properties. Having neither expertise nor space for the details, I refer the reader to Honig (1966), especially the chapter by Azrin and Holtz, which should be read asking, “To what extent do the current experimental findings refute, confirm, or modify the fireside inductions concerning punishment?” How would the psychologist classify a statutory provision that threatens to deprive the citizen of, say, money that the citizen had never learned to expect, e.g., the Agricultural Adjustment Act (see U.S. v. Butler, 1936, p. 81). where Mr. Justice 532 Meehl: Law and the Fireside Inductions Stone’s dissent hinges partly on the semantics of coercion, which he argues must involve “threat of loss, not hope of gain.” Can the experimental psychologist speak to this issue? I doubt it. A second example concerns imitation. The folklore is that both humans and infrahumans learn by imitation. (The criminal law, as lawyers and psychologists agree, is invoked to handle trouble cases, where the normal processes of socialization have not been applied or have failed to work. I trust my selecting the notion of general deterrence as a fireside induction will not be misconstrued as a belief in its major socializing role, which I daresay no psychologist would care to defend.) Our policy concerning TV and movie presentation of social models of aggression or forbidden sexual behavior is influenced by our beliefs about imitation. Despite the related Freudian emphasis upon identification as a mechanism of character formation, when I was a student the tendency in academic psychology was to minimize the concept of imitation to the point of skepticism as to whether there was any such process at all. I was taught that the classic experiments of E. L. Thorndike on the cat (circa 1900) had demonstrated that, for infrahuman organisms at least, there was no learning by imitation. This alleged laboratory refutation was presented as an example of how scientific research had overthrown part of the folklore. The failure of Thorndike’s cats to learn one particular problem-box task, under his special conditions of drive and so forth, was overgeneralized to the broad statement, “Infrahuman animals cannot learn by imitation.” The received doctrine of scientific psychology became so well entrenched that a well-designed experiment by Herbert and Harsh (1944) was largely ignored by the profession (see Barber, 1961). But beginning slightly earlier with Miller and Dollard’s Social Learning and Imitation (1941), a book that cautiously reintroduced the concept and made important conceptual distinctions as to kinds of imitation, the subject came to be restudied, especially by developmental psychologists in relation to aggressive behavior (see, e.g., Bandura & Walters, 1963; Megargee & Hokanson, 1970). A recent article by John, Chesler, Bartlett, and Victor (1969) makes it probable that Thorndike’s negative dictum at the turn of the century was just plain wrong, even for Felis catus. Point: A lawyer in 1930 might have lost a cocktail party debate with an animal psychologist, but the lawyer would have been closer to the truth. A third area important in such legal contexts as presentence investigation is that of forecasting behavior probabilistically. The fireside inductions say that you should rely heavily upon the record of an individual’s past conduct. As I have argued elsewhere (Meehl, 1970c), it may be that a naive judge will (over the long run) make better decisions than one who knows just enough psychology or psychiatry to rely on medical or social science experts making an intensive study of the offender. The efficiency of actuarial prediction is almost always at least equal to, and usually better than, prediction based upon (purported) clinical understanding of the individual subject’s personality (see references in Footnote 4 in Livermore, Malmquist, & Meehl, 1968; and Footnote 8 in Meehl, 1970c). Second, behavior science research itself shows that, by and large, the best way to predict anybody’s behavior is his behavior in the past (known among my Meehl: Law and the Fireside Inductions 533 colleagues as Meehl’s Malignant Maxim). Hence the naive judge’s reliance on the fireside inductions may yield better results than the intermediate-level sophistication, which knows enough to ask a psychologist’s or psychiatrist’s opinion, but does not know enough to take what he says cum grano salis, especially when clinical opinion conflicts with extrapolation from the offender’s record. The subtle interaction between levels of sophistication and the developing state of scientific knowledge is nicely illustrated by the Supreme Court’s attitude toward statutes postulating inherited tendencies to mental deficiency and criminalism. In Buck v. Bell (1927) the court upheld the constitutionality of an involuntary sterilization statute for mental defectives, in an opinion famous for Mr. Justice Holmes’s “Three generations of imbeciles is enough.” The opinion naturally does not display sophistication about the varieties of mental deficiency, such as the distinction between high-grade familial deficiency (usually nonpathological, being merely the low end of the normal polygenic distribution) and the Mendelizing or developmental anomaly varieties, characteristically yielding a lower IQ, relatively independent of social class, and presenting differing eugenic aspects since some of them have no discernible hereditary loading and others a clearcut one. Without entering into such technical issues in genetics, the court came to what I would regard as the right result (see Reed & Reed, 1965). In Skinner v. Oklahoma (1942) involuntary sterilization of a habitual criminal was disallowed, again a right result in my view. The fireside inductions that underlay Oklahoma’s statute are perhaps as strong and widespread for criminal or “immoral” tendencies as for mental deficiency. But the scientific data on inherited dispositions are much stronger in the one case than in the other, and that much social science knowledge the Court did possess. Suppose that a more refined taxonomy of delinquents and criminals should enable us to discover that some persons disposed to antisocial behavior get that way in part on a genetic basis (see Footnote 10 in Meehl, 1970c), although in most delinquents the etiology is social. A modified form of the fireside inductions underlying Oklahoma’s unconstitutional statute would then be defensible, and a properly redrafted statute combining habitual criminality as a legal category with psychogenetic categories or dimensions could be upheld on the same grounds as Virginia’s sterilization statute. But the court’s new task would demand far more technical sophistication, especially given the ideological components that would saturate social scientists’ opinions in the briefs, than was required for handling Buck v. Bell and Skinner v. Oklahoma. DIRECT APPLICATION OF EXPERIMENTAL RESULTS TO LEGAL PROBLEMS: SOME RATIONAL GROUNDS FOR CAUTION As a clinical practitioner who was trained at a hard-nosed, quantitativelyoriented, behavioristic psychology department (Minnesota has been called the “hotbed of dustbowl empiricism” by some of its critics), I sense a deep analogy between the problem faced by judge or legislator in balancing the fireside inductions against purportedly scientific psychological or sociological findings, and the perennial problem of how far we clinicians are entitled to rely upon our clinical 534 Meehl: Law and the Fireside Inductions experience, lacking (or apparently contradicting) experimental or quantitative research. For myself as clinician, I have not been able to resolve this dilemma in an intellectually responsible way, although I have been steadily conscious of it and engaged in theoretical and empirical research on it for over a quarter century. So I am hardly prepared to clean up the analogous dilemma for lawyers. However, dwelling on this analogy may enable me to offer some tentative suggestions. There is a similarity in the pragmatic contexts of law and clinical practice, in that something will be decided, with or without adequate evidence, good or bad, scientific or anecdotal. A judge cannot leave a case undecided—although a logician could point out that law, being an incomplete postulate set, renders some wellformed formulas undecidable. Let us strip the concept “scientific experiment” to its essentials, as I have tried to do in a rough meaning stipulation supra. Forget the usual images of glass tubing and electronic equipment operated by bearded gents wearing white coats in a laboratory. What, for instance, is the purpose of gadgetry? Scientific apparatus performs one of two functions. Either it plays a role on the input side, contributing to the physical isolation of the system under study and to the control or manipulation of the variables, or it facilitates the recording of observations (output side). We conceive a situation space whose dimensions are all physical and social dimensions having behavioral relevance. In research on human subjects, this set of dimensions will include such minor variables as the material of the experimenter’s desk, since in our society the social stimulus value of an oak desk differs from that of pine. If we are studying the impact of a psychoanalytic interpretation and design an experiment to smuggle this real-life phenomenon into the laboratory, what happens? We move in the situation space front the ordinary-life context of psychotherapy to the experimental context. This movement is in the interest of locating the system studied more precisely, because in the ordinary-life, nonlaboratory situation, the values of certain variables known (or feared) to have an influence are neither assigned by the investigator nor measured by him (with the idea of their influence being removed statistically). There is no mystery about this, no conflict between a scientific and nonscientific view of the subject matter. The problem presented is quite simple—it is the solution that presents complexities. In order either to eliminate certain causal variables, hold them constant, or manipulate them, or to measure them, we move to a different region of the situation space. By (1) eliminating, (2) fixing, (3) manipulating, or (4) measuring (and “correcting for”) the input variables, we intend to test generalizations as to what influences what, generalizations we could not reach in the natural, field, nonexperimental situation. The price we pay is that these generalizations are only known to hold for the new region of the situation space; their application in the ordinary-life context is an extrapolation. This untoward consequence of the experimental method does not flow from tendentious, polemic formulations polarizing a scientific against a non-scientific frame of reference (empirical versus armchair psychology). Such locutions are Meehl: Law and the Fireside Inductions 535 misleading, as they locate our methodological stomach ache in the wrong place. The problem can be stated in general terms within the scientific frame of reference. To concretize it: Suppose I am interested in the behavior of tigers. If I rely uncritically on anecdotes told by missionaries, hunters, native guides, etc., my evidence will suffer both on the input and output sides, i.e., from indeterminacy of the input and inaccuracy of the output. If I have accurate output data (e.g., carefully screened independent, convergent testimony by skeptical, reliable, scientifically trained observers, use of telescopic camera recording, high-fidelity tapes of the tiger’s vocalizations), I may be able by such means to take care of the recordingaccuracy problem (although hardly the recording-completeness problem). But I will still be troubled by indeterminacy on the input side. I don’ know all of the inputs to this tiger when I am photographing him at a distance. I do not know his inputs an hour earlier when he was invisible to me, and I have reason to believe that those previous inputs alter his momentary state and change his behavioral dispositions. Suppose, to get rid of this uncertainty on the stimulus side, I capture my tiger and put him in a zoo or bring him to my animal laboratory. I eliminate the influence of variations in the chirping of a certain bird as part of the tiger’s surround. There is a sense in which I now don’t have to be concerned about birds chirping, the only birds that chirp in a proper psychological laboratory being those that the experimenter himself introduces. But there is another sense in which I should be worried about the influence of bird chirping. An average level and fluctuation of bird chirps is part of the normal ecology of tigers in the wild. If I want to extrapolate my laboratory findings to the behavior of wild tigers, this extrapolation is problematic. The background of bird chirps may have a quantitative impact upon the tiger’s behavioral dispositions, and perhaps upon his second-order dispositions to acquire first-order dispositions (Broad, 1933; Meehl, 1972). In research on human subjects, it is frequently found that the influence of variable x upon variable y is dependent upon values of variable z, called by statisticians an “interaction effect.” Interaction effects regularly occur whenever the sensitivity of the experimental design suffices to detect them. It is not absurd to suppose that, in human social behavior, almost all interactions of all orders (for instance, the influence of variable v on the interaction effect of the variable z on the first-order influence of variable x on variable y) would be detected if our experiments were sufficiently sensitive. When we liquidate the influence of a variable, either by eliminating it through physical isolation or holding it fixed, we are in danger of wrongly generalizing from our experimental results to the natural, real-life setting. Law-trained readers unfamiliar with social science statistical methods may have found the preceding rather abstract. Suppose I am a developmental psychologist interested in the deterrent effects of punishment and I argue that the sanctions of the criminal law are inefficacious, relying on “Fisbee’s classic experiments on punishment in nursery-school children.” In Table I, I list differences that might be relevant in extrapolating from the laboratory study to the legal context. 536 Meehl: Law and the Fireside Inductions TABLE I Extrapolating from Fisbee to Real-Life Experiment Criminal Sanction Against Larceny Four-year-olds Adults Mostly upper middle class Mostly lower and lower middle class Mostly biologically normal Numerous genetic deviates in group Time-span minutes or hours Time-span months or years Social context: Subject alone Social context: Criminal peer-group inputs Reward: Candy Rewards: Money, prestige, women, autonomy, leisure, excitement Punishment: Mild electric shock Punishment: Deprivations of above rewards (Punishment more like nonreward or time-out than strictly aversive stimulus onset) Punished response emitted; experimenter aims at reform Punished response not emitted by most; law aims at general deterrence Subject’s perception of situation: Who knows? Subject’s perception of situation: Who knows? One need hardly be obscurantist or antiscientific in one’s sympathies in order to be nervous about extrapolation in the joint organism-situational space from the region of the left-hand column to that of the right as ground for repealing a statute penalizing larceny. In the foreseeable future, lawmakers will unavoidably rely upon a judicious mixture of experimental research, quasi-experiments, informal field observation, statistical analysis of file data, and the fireside inductions. The legislator, prosecutor, judge, and public administrator—like the clinical psychologist—cannot adopt a scientistic purist posture, “I will not decide or act until fully adequate standards of scientific proof are met by the evidence before me,” The pragmatic context forces action. In these matters, not doing anything or not changing anything we now do is itself a powerful form of action. When the fireside inductions are almost all we have to go on, or when the fireside inductions appear to conflict with the practical consequences of extrapolated experimental research or psychological theory, it would be nice to have some sort of touchstone as to pragmatic validity, some quick and easy objective basis for deciding where to place our bets. Unfortunately there is none. Are Some Classes of Fireside Inductions More Trustworthy than Others? This problem is so important in a society that has become sophisticated and self-conscious as to its own modes of social control that one might reasonably argue in favor of support for second-level empirical research aimed at developing a taxonomy of fireside inductions, enabling us to sort them into categories having different average levels of accuracy. We do not even possess a corpus of the explicit fireside inductions upon which our law relies. To ferret these out from statutes, appellate court opinions, the Restatements, and so forth would be a monstrous and thankless task, although I suggest that a random sampling of the documents might be worth doing. One might inquire, apart from whether the fireside inductions are corroborated by social science research, how many legal Meehl: Law and the Fireside Inductions 537 rules, principles, and practices accord even with the fireside inductions of contemporary men. Most law professors will readily agree, I find, that much of our law is predicated on notions about human conduct that hardly anybody would care to defend. Example: A lawyer will advise a testator to leave a dollar to one of his children whom he wishes to disinherit, because courts have held that since a parent naturally tends to bequeath property to his child, failure to do so creates a legal presumption that the testator omitted him by inadvertence. I doubt that this presumption accords with the fireside inductions. Laymen agree with me that if a father is sufficiently compos mentis to write a valid will at all, his failure to mention a son is not attributable to forgetfulness. Example: One class of exceptions to the Hearsay Rule is “declarations against interest,” when the declarant is unavailable as a witness. The unavailability creates a special need for hearsay, and the fireside induction is that the declaration’s having been made against interest renders it more trustworthy. But the interest is required to be pecuniary, not penal (Livermore, 1968, pp. 76–78: McCormick, 1954, pp. 546–551). Surely our fireside inductions do not suppose that a man is more likely to be careless or mendacious in admitting rape than that he borrowed money! Similar oddities have been noted in the Hearsay exception for “Admissions of a party-opponent,” where a litigant’s predecessor or joint tenant falls under the exception, but not a tenant in common or a co-legatee (McCormick, 1954 , pp. 523–525). I cannot imagine that the fireside inductions concerning motivations for accuracy would support these distinctions, which arose not from empirical considerations in the psychology of testimony but through formalistic intrusion of property-law metaphysics into the law of evidence. A taxonomy of fireside inductions based upon their substantive, methodological, and psychological properties might permit a rough ordering of inductive types as to accuracy, comparing the fireside inductions in each category with social science generalizations available in the literature. If strong taxonomic trends existed among researched cases, we would have some basis for judging the probable trustworthiness of those unresearched. We could inquire whether the following methodological features of a fireside induction are associated with a higher probability of its being scientifically corroborated: Hardly anyone entertains serious doubts about the induction. Persons of different theoretical persuasions agree about the fireside induction almost as well as persons holding the same theoretical position. There is a consensus of the fireside that cuts across demographic variables such as education, occupation, social class, religious belief, ethnic background, and the like. Within the legal profession prosecutors, defense lawyers, law professors, and judges are in substantial agreement. Personality traits (e.g., dominance, social introversion, hostility, rigidity) are not appreciably correlated with adherence to the induction. The particular fireside induction involves an observation of actions, persons, or effects that occur with sufficient frequency so that most qualified and competent observers will have had an extensive experience as a generalization basis. The fireside induction deals 538 Meehl: Law and the Fireside Inductions with relatively objective physical or behavioral facts rather than with complicated causal inferences. The policy implications of the induction are such that nobody’s political, ethnic, religious, moral, or economic ideology or class interest would be appreciably threatened or mobilized by its general acceptance in the society or by lawmakers’ or administrators’ reliance upon it in decision-making. Sophisticated armchair considerations do not reveal a built-in observational or sampling bias that would operate in the collection of anecdotal support or refutation of the induction. The induction is qualitative rather than one that claims to make quantitative comparisons despite the lack of a reliable measuring device. I do not profess to know the relative importance of items in this list, and I can think of exceptions to any of them as a touchstone. Thus, for example, the height of a person’s forehead is a relatively simple physical fact. But judgments of forehead heights of prison inmates by their guards were shown to be correlated with guards’ estimates of their intelligence, whereas objectively measured forehead heights were not. (Explanation: The erroneous fireside induction that a low forehead—low-brow—indicates stupidity led guards’ perceptions and/or memories of this simple physical feature to be infected with their behavior-based estimates of a prisoner’s intellect.) And Are Some Experimental Findings Safer to Extrapolate? One asks here about the comparability of two groups of organisms as to species, developmental period, and status. What motives, what rewards and punishments, what time relationships are shared between the experimental context and the natural setting? An indirect lead as to extrapolability can sometimes be picked up from the experimental literature itself, asking, “To what extent do the experimental findings replicate over a variety of species, drives, instrumental responses, rewards, and punishments?” If we cannot generalize within the laboratory, moving to the field is presumably risky. How sizable are the relationships? A social scientist who countervails a lawyer’s fireside induction by extrapolating from psychological research yielding two correlation coefficients r = .25 and r = .40 is just plain silly; but an unsuspecting lawman, overly impressed with social science statistical methods, might be taken in. Point: Random sampling errors aside, this correlational difference represents an increment of less than 10% in variance accounted for, and could easily be liquidated by moving to a not very distant region of the situation space. To have the best of both worlds, one would want accurate recording on the output side and proper statistical treatment, but with the situation being very similar to the one to which we wish to extrapolate. Accurate observations, accurate records (instead of memory and impressions), appropriate statistical 1 I cannot trace the reference, which is from the late Professor Donald G. Paterson’s lectures. Meehl: Law and the Fireside Inductions 539 analysis are all attainable in the field or natural life setting, lacking experimenter manipulation of the input. I am therefore inclined to view Campbell’s “Reforms as Experiments” (1969) not as a second-best substitute for laboratory investigation, but as often intrinsically preferable, because the situational-extrapolation problem is so grave that the scientific precision of laboratory experiments with college students or school children is largely illusory. THE LAWMAKER’S DILEMMA The legislator’s, judge’s, or administrator’s situation is most comfortable when there is a sizable and consistent body of research, experimental and nonexperimental (file data and field observation data), yielding approximately the same results as the fireside inductions. While one may be scientifically skeptical even in this delightfully harmonious situation, in the pragmatic context of decision making, rule writing, policy adopting, etc., such rigorous skepticism can hardly lead to pragmatic vacillation. Some sort of action is required, and all we have goes in the same direction. The methodologically unsatisfactory situations can be divided into three groups, differing in degree rather than kind: (1) No quantitative or experimental evidence is available or readily collectable before action must be taken. Here we rely upon the fireside inductions, these being all we have. A healthy skepticism concerning the fireside inductions, engendered by the study of social science, makes us nonetheless uncomfortable. (2) We have a large-scale adequately conducted study in the field situation supplemented by file data from different jurisdictions varying in relevant parameters (e.g., offense rate, community socioeconomic indices); these field-observation and file-data results accord with theoretical concepts developed experimentally on humans and infra-humans; but the conclusion conflicts with the fireside. Such a massive and coherent body of information should countervail the fireside inductions, even those with the admirable properties listed above. It seems difficult to dispute this, since by including file data from the nonlaboratory setting to which we wish to extrapolate, we are in effect comparing two sets of anecdotal data, one of which has the methodological advantage of being based upon records instead of relying upon our fallible and possibly biased memories of observations gathered nonsystematically as regards representativeness of persons and situations. Example: If statistics show that accuseds released without bail pending trial have such a low incidence of pretrial criminal offenses or failure to appear for trial that the bail system has negligible social utility (combined with its obvious inequity to the poor), our fireside inductions to the contrary should not countervail. But a nagging doubt persists, since other relevant statistics (e.g., ratio of reported felonies to arrests) tend to support the fireside induction that some fraction of these defendants have committed further crimes during their pretrial freedom, and we cannot accurately estimate this fraction (see the excellent article by Tribe, 1970). A lively sense of the lawmaker’s dilemma can be had by reading the Senate debate on the District of Columbia crime bill (Congressional Record, 1970). (3) The most difficult situation is that in which there is a collision between 540 Meehl: Law and the Fireside Inductions a fireside induction having several of the good properties listed above, and a smattering of social science research that is strong enough to give us pause about the fireside inductions, but not strong enough to convince us. Thus, the research may not be entirely consistent from one investigator to another; or it comes only from the experimental laboratory and consequently involves considerable extrapolation in the situation-space; or, if a large-scale quantitative nonexperimental survey, it has the causal-ambiguity and variable-unscrambling difficulties intrinsic to such studies (Meehl, 1969, 1971a). One hardly knows what to suggest in such collision situations except the social scientist’s usual “more research is needed.”