Download Murray & Skinner: Historical Analysis of Behavior & Inner States and more Schemes and Mind Maps Philosophy in PDF only on Docsity!
Psychological Reports , 1992, 70, 407-450.
NEEDS (MURRAY, 1938) AND
STATE-VARIABLES (SKINNER, 1938)^1 ,^2
PAUL E. MEEHL
University of Minnesota Summary.—Skinner’s concept of drive as a state-variable and his powerful rationale for introducing it agree closely with Murray’s treatment of need. Operant behaviorists’ usual deprecation of motivation in favor of stimulus control arises partly from features of para- meters, insufficiently explored in some regions, of Skinner box research. For human adults on rich reinforcement schedules, response selection is chiefly controlled by the regnant motive. Skinner’s life-long interest in inner events and translating psychodynamic concepts into behaviorese was obscured by his metalanguage philosophy of science (behaviorism). It was my good fortune to be an undergraduate and graduate student at Minnesota (1938-1945) during B. F. Skinner’s period on the faculty (1936-1945), and later my first job offer came from Skinner the year he began his chairmanship at Indiana. While a student, in addition to hearing his classroom lectures on verbal behavior, I spent many exciting hours of discussion in his office and was privileged to belong to the little group of TAs who met weekly evenings at his home (Skinner, 1979, p. 266). The only clini- cians in that group were Howard Hunt and myself, both markedly (and permanently) influenced by Skinner but not “orthodox believers.” I had come to psychology via the potent bibliotherapy of Karl Menninger’s The Human Mind (1930; Meehl, 1989), and the impact of such a first-class intellect as Skinner on my Freudian commitments led to major cognitive dissonance. I did my best to resolve it, and that resolution, imperfect as it must of course be, is the subject of this paper.^3 Skinner’s magnum opus , The Behavior of Organisms (1938), appeared the same year as Henry A. Murray’s Explorations in Personality (1938), and the psychodynamic clin- ical students at Minnesota freely spoke both languages, the latter under the influence of a young PhD, Robert E. Harris, and two students with MAs from Radcliffe. I suspect the Minnesota group was almost unique in this theoretical bilingualism. Even today, it is hard to conceive a solid-gold third-generation operant behaviorist and a corresponding repre- sentative of the Murray tradition engaging in meaningful communication; but since I do (^1) I am indebted to David Faust, Celia Gershenson, William Grove, David Lubinski, Richard Meisch, Bruce Overmier, Gail Peterson, and Travis Thompson for stimulation, advice, references, and helpful criticism of this paper in draft, although I have not always followed their advice, and they do not all agree with all of my conclusions. (^2) Address reprint requests to Paul E. Meehl, University of Minnesota, Department of Psychology, N Elliott Hall, 75 East River Road, Minneapolis, MN 55455. (^3) My psychotherapeutic orientation is now roughly one-third Freud, one-third Albert Ellis, and one-third Skinner. These fractions refer not (mainly) to subsets of patients, but to aspects of the same patient, and— via approximate translations—to the same state or process in a given patient. The term ‘eclectic’ means, for me, not an indiscriminate heap of thoughts and tactics (Schofield, 1988) but the organized result of conceptual parsing and integration by the therapist.
2 P. E. MEEHL
both, I am here going to explain how and why. Motivation is the variable selected, and Murray will serve as an adequate proxy for Freud, especially since I do not want my needs-defense-behavior analysis to be entangled with Freud’s metapsychology. The latter could hardly be translated into Skinnerese, and I myself do not subscribe to much of it. Some may consider this a feckless endeavor, it being “obvious” that the clash between behaviorism and psychodynamics is basic, pervasive, and irresolvable, involving issues of both substance and method. The glaring difference is in what are acceptable as explanatory entities: Murray invokes inner causes , whereas Skinner usually rejects such in favor of external stimulus control. In almost all his later writings, and increasingly over the years, Skinner castigates non-behaviorists (not only Freudians) for having recourse to postulated internal causes of behavior, labeling such pejoratively as “fictional,” “unobservable,” “lacking in physical dimension,” “useless for control,” “circular,” “redundant,” “tautological,” “imaginary.” He devoted one book, probably the most controversial he ever wrote, almost wholly to the omnipotency of external stimulus control (“schedules”), going so far as to say that if we admire a person’s achievements we should bestow our praise not on the individual but on the shaping and maintaining environment (Skinner, 1971). A whole section of Contingencies of Reinforcement (1969, pp. 64 - 67) is devoted to criticizing the need concept. About Behaviorism (1974, pp. 164 -
- has a section on “The uselessness of inner causes.” Referring to powerful theoretical constructs (which inner states or events must be, for behavior science) in the physical sciences as “third-stage” concepts, Skinner writes: There are few, if any, clear-cut examples of comparable third-stage concepts in psychology, and the crystal ball grows cloudy. But the importance of the stage is indicated by the fact that terms like wants , faculties , attitudes , drives , ideas , interests , and capacities properly belong there. When it is possible to complete a theoretical analysis at this stage, concepts of this sort will be put in good scientific order. This will have the effect of establishing them in their own right. At present they need external support. Some of them, like wants and attitudes, come to us trailing clouds of psychic glory, and a wisp or two of the psychic can usually be detected when they are used (1961, p. 235). In traditional terms an organism drinks because it needs water, goes for a walk because it needs exercise, breathes more rapidly and deeply because it wants air, and eats ravenously because of the promptings of hunger. Needs, wants, and hungers, are good examples of the inner causes. … Sometimes the inner operation is inferred from the operation responsible for the strength of the behavior … it is sometimes inferred from the behavior itself.… So long as the inner event is inferred, it is in no sense an explanation of the behavior and adds nothing to a functional account (1953, pp. 143 - 144). The commonest inner causes have no specific dimensions at all, either neurological or psychic. When we say that a man eats because he is hungry, smokes a great deal because he has the tobacco habit, fights because of the instinct of pugnacity, behaves brilliantly because of his intelligence, or plays the piano well because of his musical ability, we seem to be referring to causes. But on analysis these phrases prove to be merely redundant descriptions (1953, p. 31). The practice of looking inside the organism for an explanation of behavior has tended to obscure the variables which are immediately available for a scientific analysis. These variables lie outside the organism, in its immediate environment and in its environmental history … such terms as “hunger,” “habit,” and “intelligence” convert what are essentially the properties of a process or relation into what appear to be things. Thus we are unprepared for
4 P. E. MEEHL
Behavior (1953) we find a whole chapter devoted to it (pp. 141ff), although entitled “deprivation and satiation”—denoting the input operations—rather than “drive,” the state. Furthermore, in his later writings the same basic rationale as was offered in 1938 for introducing drive (also the other state-variable emotion) is presented, substantially unchanged. [A “drive”] is simply a convenient way of referring to the effects of deprivation and satiation and of other operations which alter the probability of behavior in more or less the same way. It is convenient because it enables us to deal with many cases at once. There are many ways of changing the probability that an organism will eat; at the same time, a single kind of deprivation strengthens many kinds of behavior. The concept of hunger as a drive brings these various relations together in a single term.… A drive is a verbal device with which we account for a state of strength … (1953, p. 144). There is nothing wrong with an inner explanation as such, but events which are located inside a system are likely to be difficult to observe.… Eventually a science of the nervous system based upon direct observation rather than inference will describe the neural states and events which immediately precede instances of behavior. We shall know the precise neurological conditions which immediately precede, say, the response, “No, thank you.” These events in turn will be found to be preceded by other neurological events, and these in turn by others. This series will lead us back to events outside the nervous system and, eventually, outside the organism (1953, pp. 27 - 28). Given our Murray/Skinner theme and the fact that Freud’s theory is obviously one of the most “inner”-oriented (and mentalistic) theories, Skinner’s attitude to Freud is of special relevance. Here again we find paradoxes (I do not say contradictions , for reasons given below). On the very first page of Behavior of Organisms appears the interesting classification of Freud’s ego, super-ego and id as concepts “which have remained the subject of scientific investigation” (1938, p. 3). True, this surprising remark occurs in the context of criticizing our failure to develop a science of behavior at its own level. Nevertheless, Freud’s kind of “inner organism,” Skinner says, has “become in turn the subject matter of a science,” unlike the freely willing inner organism, or the vague self , of the man on the street. I can only interpret his position as being that Freud’s theory is science, but rather poor science, like the science which invokes the “conceptual nervous system (C.N.S.)” (1938, pp. 4, 421). In Cumulative Record (1961, p. 243) Skinner goes so far as to say of Freud (paraphrasing the economists on Keynes), “Certain general points have been made—in some sense we are all Freudians—but the facts and principles which have been salvaged can be stated in relatively non-technical language.” The chapter on psychotherapy in Science and Human Behavior offers behavioral translations of several defense mechanisms, plus brief accounts of wit, dreams, and parapraxes (1953, pp. 376 - 378). In that book we find (p. 293) a truly astonishing concession to Freud, about dream interpretation: “Freud could demonstrate [!!] certain plausible relations between dreams and variables in the life of the individual. The present analysis is in essential agreement with his interpretation.” As a 33% Freudian who interprets dreams (see, e.g., Meehl, 1983) but who is also semi-Popperian in metatheory, I would never use such a strong verb as ‘demonstrate’ for what Freud accomplished. Skinner took Freud seriously. I can supplement textual evidence with many clear recollections from personal contact during Skinner’s Minnesota period. He was fascinated by “Freudian slips” (he called them that, unabashedly) and had collected an
MURRAY AND SKINNER: 1938 5
amusing batch of good ones to illustrate thematic strengthening of verbal operants in his verbal behavior class. Queried by me as to their being “unconscious,” he didn’t bat an eye, responding quickly, “Of course, there’s no law that a human can always tact the variables of which his behavior is a function.” Meehl: “Surely not. But Freud’s ‘uncon- scious’ means more than that. Suppose one has learned to tact such a variable, inner or outer, but avoids doing so ‘in the interest of the defense,’ as we say. That’s the kind of unconscious we clinicians care about—which differs from a ball-player’s inability to report on how his brain does a numerical integration of the baseball’s path. He never learned calculus. But the neurotic has learned to tact mother, father, and inner states, but doesn’t now do it, defensively.” Skinner: “Yes, that’s a more interesting case, requiring a more complex analysis. It’s a case of aversive control. What you call ‘defense’—and I have no objection to your term, as long as we’re quite clear what is being described by it—is either a case of the (expected, otherwise available) tacting operant being kept at low strength by punishment, or an incompatible operant—your ‘defensive maneuver’— being at high strength because it has a history of negative reinforcement (escape or avoid- ance).” This kind of dialogue occurred regularly between Skinner and the psychodynamic clinicians. In fact, I often found it easier to discuss Freud’s concepts with Skinner than with my strongly anti-Freudian doctoral advisor Starke Hathaway. The fact is that Skinner was very much interested in Freud, to the point of applying as a candidate for psychoanalytic training at the Boston Psychoanalytic Institute in the late 1940s. I shall charitably refrain from comment on their refusal. If the arch-behaviorist, with his intel- lect, had had some reinforcing couch time (given an analyst not defensive, doctrinaire, or dumb), who knows how both disciplines might have been benefited? I do not argue that Murray and Skinner have no real differences, that it’s “merely semantics,” nor that everything Murray (or Freud) asserts can be translated without loss of meaning into behaviorese. (Translation in the other direction is never discussed, an interesting fact in itself.) What I do argue is (a) that there is more adequate translatability than is generally thought; (b) the surplus meaning—while theoretically interesting—is usually of slight clinical importance; and (c) the constraints Skinner imposes that prevent more complete translation are metatheoretical, reflecting his behaviorist philosophy more than they do scientific substance. I permit myself this distinction, relying on the opening sentence of About Behaviorism (1974, p. 3): “Behaviorism is not the science of human behavior; it is the philosophy of that science [italics added].” In the language of logicians, I might express the point by saying that cognitive collision occurs between Murray’s object language statements and certain metalinguistic proscriptions by Skinner that attempt to impose his philosophy of science on others. The crucial distinction here is between the modest line, “My conspicuous technological success—as shown by these object language generalizations—leads me to advocate such-and-such a research strategy as conducive to success,” and a more dogmatic triumphalist line, “Whatever your theoretical interests may be, you are mistaken , you sin against scientific method , if you allow yourself conjectures outside the class of constructs to which my strategy restricts itself.” It is, of course, this second negative, forbidding, exclusionary kind of statement that arouses the ire of psychologists who want to study cognitions, traits, psychodynam- isms, and social institutions. In reconciling the paradoxical quotations above, I adopt the working assumption that Skinner is, with rare exceptions, consistent. I would say he was one of the most consistent
MURRAY AND SKINNER: 1938 7
“distended seminal vesicles” due to deprivation are no more the basis of sexual motiva- tion than “stomach contractions” are the core of hunger. (I once asked a Benedictine monk at St. John’s University how he dealt with the attractive female secretaries, and his reply was “Simple—one learns not to look.”) For the various complicated relations that arise, which for extralaboratory cases often render the analysis largely programmatic, see Skinner (1953, Chapters 9, 14, et passim ) and Reynolds (1975). Where does this leave us in examining the relation between Murray needs and Skinner state-variables? I shall give primacy to Behavior of Organisms over some (not all!) of the later writings and try to avoid what appear to be inconsistencies by parsing theoretical object-language and metatheoretical dicta. Of course, Skinner may at times be literally inconsistent or change his mind without so announcing. (I suspect all daring, creative Grand Theorists fall into inconsistencies. Einstein down-played the corroborative 1919 eclipse in favor of the beauty and armchair plausibility of his theory; yet he said that if the red-shift did not occur, his theory would be dead. But I am applying Wilson’s Charity Principle, avoiding seeing contradictions whenever possible.) Putting Behavior of Organisms , Science and Human Behavior , and other writings together, we distill the following:
- Motivational variables control families of operants and respondents, which may thereby covary in strength despite negligible overlap in their (stimulus and/or response) properties.
- Motivational variables are controlled by various operations, including deprivation, current stimulation, and various physiological interventions; while the paradigm experi- mental “drives” (thirst and hunger) are customarily manipulated via deprivation, motiva- tional states differ in the extent of their “spontaneous growth” versus exteroceptively controlled excitation.
- The main determiner of motivational operant covariation is historical. Those oper- ants which have been strengthened, shaped, and brought under discriminative control by food reinforcement will tend to covary with hunger, and hence with operations that influence hunger.
- In the real physical causal chain, the motivational variable is a state of the organ- ism—hence, an “inner condition,” the immediate causal ancestor of the changes in operant strength.
- Despite its undoubted existence inside the organism, to control the operant via the motivational variable we must manipulate it by external operations (e.g., feeding and fasting). Insofar as our aim is control, a research strategy emphasizing reinforcement schedules, deprivation, discriminative stimuli, and drive conditioning is preferable to a strategy that focuses attention on the inaccessible state-variable itself. Note that (1) through (4) are empirical generalizations belonging to the science of behavior, whereas (5) is a metatheoretical move, of course influenced by that science’s success, but, in its content, part of what Skinner calls his philosophy of science, i.e., behaviorism. (A philosopher would classify it as belonging to the realm of pragmatics. It may be suggested by the facts, but its content is a kind of metatheoretical advice.) Aversive control is of special importance in psychodynamics and psychotherapy, as Skinner points out (1953, pp. 359 - 383). We could almost define neurosis in terms of appetitive/aversive balance—a behaviorist (partial) translation of Freud’s “economic” generalization that the neurotic’s ego suffers a depletion of psychic energy because so
8 P. E. MEEHL
much must be expended in warding off anxiety by means of the rigid and hyperalert defensive system. Aversive control is based on the contingencies of negative reinforce- ment. A negative reinforcer is defined by the fact that its removal (or, later, avoidance) exerts the usual family of reinforcing effects, i.e., it strengthens, shapes, and chains operants, and establishes their discriminative controls. In addition to this “operational” defining property, negative reinforcers typically possess several other efficacies as well. Their contingent presentation depresses appetitively controlled operants, the effect com- monly called punishment. They may elicit certain “emotional” respondents, chiefly (auto- nomically mediated) smooth muscular and glandular respondents. They can confer posi- tive and negative reinforcing power on neutral stimuli. Of particular relevance for our discussion is that non contingent presentation of a negative reinforcer also exerts a depres- sing effect on positively maintained operants. This interesting property at least suggests (as it did to Estes and Skinner, 1941) as a sixth property the elicitation of an aversive state-variable, commonly labeled ‘anxiety.’ To forestall misunderstanding, let me emphasize that I am stating Skinner’s 1938 position, which he continued to hold in its essentials, for the purpose of relating it to Murray’s views of the same year. I am not here attempting to summarize, let alone explain, the vast body of experimental research bearing on—and importantly qualifying —these basic postulates. For example, a paradoxical strengthening effect of a normally “punishing” stimulus (foot shock) was discovered over half a century ago (Muenzinger,
- in the old discrimination apparatus, and later independently re-discovered in the Skinner box (Kelleher & Morse, 1964). The parametric complexities arising here are so great as to defy any simple categorization of stimulus events or even of highly general second-order functional relationships determining their reinforcing properties (cf. Barrett & Katz, 1981). But I conjecture that parametric modifications of the simple positive/ negative reinforcer dichotomy may make the above list of principles, as amended, more compatible with certain concepts in psychodynamics (see, e.g., Freud 1920/1955; Meehl, 1990a; Menninger, 1938; Rado, 1956; Reik, 1941; Sincoff, 1990). Similarly, on the positive reinforcement side, I do not discuss Premack’s differential probability principle (1959; and see, e.g., Staddon & Ettinger, 1989), Herrnstein’s match- ing principle (1961; and see, e.g., Staddon & Ettinger, 1989), Nevin’s behavioral momen- tum (Nevin, Mandell, & Atak, 1983), the conservation or molar equilibrium principle, or the various “economic” formulations of behavior choice (Allison, 1979, 1983; Allison, Miller, & Wozny, 1979; Collier, Johnson, Hill, & Kaufman, 1986; Mazur, 1979; Rachlin, 1989; Staddon, 1979a, 1979b, 1979c; Staddon & Ettinger, 1989; Timberlake, 1980). Whether these developments affect a Murray-Skinner rapprochement is unclear, although my hunch is they would facilitate more than impede (partly because of Murray’s teleo- logical “adaptive” emphasis, anathema to Skinner). As Staddon points out (1979a), Skinner was not enthusiastic about formal approaches to operant reinforcement principles (Skinner, 1961, pp. 250 - 252; 1966, pp. 27 - 29).^4 4 Despite my disclaimers, a reviewer complains that I do not deal with the developments in operant behavior analysis in recent years. Although I have not run a rat since the 1960s, and do not monitor that literature, I am of course aware that much has happened by way of expansion, amendment, detail, and application. But this paper (as its title clearly indicates) is about the conflicting, polarizing views of two eminent thinkers as set forth in their classic works over a half-century ago. I believe psychologists’ per- spectives can be improved by reflecting upon the history of ideas (e.g., it may save us from reinventing the
10 P. E. MEEHL
their concurrent and historical relations specifies a motive. This is why Tolman (1932) said that purpose is immanent in behavior. SOME HELP FROM THE PHILOSOPHERS To his credit, Skinner never employed “official” philosophy of science as a bludgeon against substantive opponents, a polemical vice to which some psychologists—disciples of Hull, Spence, Tolman, and even Freud—were prone. In fact, he took a rather dim view of the philosophers’ contribution to psychology, holding that both it and the Fisher design-of-experiments emphasis were unhelpful to the behavioral scientist, if not harmful. In reviewing Smith’s Behaviorism and Logical Positivism (1986), Skinner (1987a) agreed with the main thesis that the influence of logical positivism on the development of American behaviorism was slight, contrary to the received opinion. The logical positivism of the early 1940s had liberalized itself to such an extent that in the Minnesota discussions Feigl and I would often be defending the hypothetical constructs of Hull, Freud, and the factor analysts against Skinner’s disapproval of them in favor of an astringent “pure intervening variable” approach (cf. MacCorquodale & Meehl, 1948). The only quasi-philosophical concept Skinner relied on was operationism, which he carefully defined for behavioral science thus: “Operationism may be defined as the practice of talking about (1) one’s observations, (2) the manipulative and calculational procedures involved in making them, (3) the logical and mathematical steps which intervene between earlier and later statements, and (4) nothing else ” (Skinner, 1945). Whether that methodological prescription comports with his insistence—against Boring’s contribution to the symposium—on internal states and events being included in behavior science, I do not discuss (but see my query about it [Meehl, 1984] and his reply, “I pass” [Skinner, 1984]). It is fortunate for Skinner’s position that he carefully refrained from reliance on “general philosophy of science,” because by 1950 (at the latest) one could hardly find a competent logician, philosopher, or historian of science who subscribed to operationism in anything like Bridgeman’s 1927 formulation, which Bridgeman himself had radically revised years later (1936). Skinner was fully aware of the extent to which other sciences (astronomy, physics, chemistry, genetics) employ hypothetical constructs to advantage. But he also knew that at certain stages of their development they did quite well at a descriptive level. Chemistry flourished, using Mendeleev’s table, before valence—a dispositional concept—was explained in terms of completing electron shells; and the electron itself, while a hypothetical construct in the philosopher’s eyes, was almost “operationally defined” by the experiments of another science (physics) quite apart from chemical combining properties. Phenomenological thermodynamics was a powerful quantitative enterprise before its explanatory “reduction” to statistical mechanics. Population genetics was a rich quantitative science before Watson and Crick. Although the gene—earlier, merely a “factor”—is, strictly speaking, a theoretical construct, it was very closely tied to direct cytological observation via the statistics of trait linkage (cf. geneticist Snyder’s 1951 arguments, listed by Meehl, 1990b, pp.25-27). Genetics provides a nice example for Skinner’s position because it illustrates (pre-Watson and Crick) how an obvious mapping of concepts at two levels in Comte’s Pyramid of
MURRAY AND SKINNER: 1938 11
Sciences^6 can occur. It requires no forcing, and little conjecture, when the observational evidence is in. Psychologists have often misunderstood Skinner’s view about the nervous system, which is hard to excuse because he has stated it clearly from his doctoral dissertation (Skinner, 1931) on. I know from many hours of discussion exactly how he saw it. The replicable generalizations of behavioral science stand on their own feet. In fact, they constrain the neurophysiologist’s theorizing, since a brain-theory incapable of explaining behavioral laws must have something wrong with it. (That this argument applies in the reverse direction, although in a more complicated way, never apparently occurred to him; but I pass that here.) At an advanced stage of development of both behavior science and neurophysiology, each having been pursued “at its own level,” the presumed isomorphism between the two sets of concepts will become obvious. That scheme is very like what happened in the identification of Mendelian factor = gene at a chromosomal locus = cistron as an ordered sequence of codons. When Skinner defines behaviorism as the philosophy of science for a science of beha- vior, he is not deriving it from a general philosophy of science, a procedure he knows cannot be carried through, for the simple reason that general philosophy of science does not prescribe strict operationism and does not proscribe hypothetical constructs. On the current philosophical scene, it is recognized that we need to distinguish general phil- osophy of science from special philosophy of science, the latter suited for analyzing and reconstructing particular disciplines (Stegmüller, 1979). My suggestion is to consider Skinner’s behaviorism as a research strategy for those interested in controlling behavior , a kind of “methodological advice” given such-and-such specified aims. The proscriptive component of his meta-discourse, which admittedly sounds quite forbidding at times, I charitably construe as, roughly, “We are in the stage comparable to phenomenological thermodynamics, population genetics, Mendeleev’s Table; for the foreseeable future, the best strategy is to analyze our subject matter at its own level of description, with minimal use of hypothetical constructs, whether mentalistic or neurophysiological.” That this is a fair reading is suggested by Skinner’s classification of Freud’s “psychic institutions” (ego, super-ego, id) as scientific explanations of behavior, which astonishes those young operant behaviorists who have never read Behavior of Organisms (Skinner, 1938, p. 1). How can we reconcile this freely volunteered concession with his strong distaste for such constructs? On my reading, easily. Freud’s psychic institutions are not excluded by any general “rules of scientific method.” They are just not good science; they are like phlogiston rather than valence. My friendly disagreements with Skinner over half a century include the value of philosophy of science, which I prefer to call metatheory, given the current view that it is the empirical theory of scientific theorizing, a branch of social science (see logician Sneed, 1976, 1979). Its being an empirical theory, whose data consist of accepted and abandoned scientific theories and episodes in their history, does not, of course, preclude inclusion of logic, mathematics, statistics, and even “armchair epistemology” in meta- (^6) Auguste Comte (1788-1857), inventor of the terms ‘positivism’ and ‘sociology,’ envisaged a pyramid of the sciences with physical sciences as the base, sociology as the apex, and biology in the middle layer. While not a thoroughgoing reductionist, he emphasized dependence of the concepts and laws of one “level” upon the sciences below it (e.g., one cannot fully understand physiology without chemistry). See Comte (1830-42/1974, 1830-54/1983), Oldroyd (1986, Chapter 5), and, for a strong examplar in genetics, Meehl (1990b).
MURRAY AND SKINNER: 1938 13
But there is a price paid for this that can produce oddities. For example, if any concept is “core” (Lakatos, 1970, 1974, 1978; Meehl, 1990a, 1990b) to the system, it is surely reinforcement. The purported operational definition (repeated early, almost ritual- istically, in almost all articles or book chapters by operant behaviorists) is the well known increase in operant strength. That is offered as an explicit definition of the concept, and the fact that reinforcers have other interesting and important powers is regularly treated as a (nondefinitional) empirical finding. Over a quarter-century ago, my colleague Roy Pickens complained that a referee insisted the term ‘reinforcement’ be stricken from a paper on effects of cocaine self-administration in the rat, on the ground that the short term effect was a decrease in rate below the unconditioned operant level (see Pickens & Thompson, 1968). I could not resist needling him a bit, “Roy, those hyper-operational pigeons have come home to roost,” which he took in good spirit. This amusing episode illustrates an important methodological point, relevant to our analysis. Reinforcement as an operation exerts a related family of effects , short-term and long-term, that include strengthening the operant, shaping it, bringing it under discriminative stimulus control, conferring the reinforcing property on discriminative stimuli, chaining, eliciting biologic- ally relevant respondents (Pavlov’s salivating dog!), countervailing aversive stimuli (e.g., Brady & Conrad, 1960; Olds, 1959), and so on. The increase in strength is the obvious “big” effect, and no doubt deserves to be considered a privileged indicator; but the Pickens paradox and semantic hassle can be liquidated by avoiding a strict single- property operational definition and substituting an open concept specification of meaning (Pap, 1953, 1958, Chapter 11; Meehl, 1972, p. 21 [1973, p. 195] and references cited therein, 1978, pp. 815 - 816). With that approach, if cocaine does some of the other things in the “reinforcement family,” we do not scruple to label it a reinforcer, despite the anomaly of decreased rate. In the philosophers’ metatheory of open concepts, a member of the indicator set may be privileged (i.e., receive a heavier weight) but can sometimes be outweighed by some subsets of the others. While the metatheory of open concepts can be treated purely as a statistical problem (cf. Meehl & Golden, 1982), it usually has a substantive component also, because in the life sciences most causal influences are stochastic rather than nomological (Meehl, 1978, pp. 812 - 814) due to the ubiquity of uncontrolled and unmeasured variables that may interrupt the causal chain of interest. That being so, it is scientifically legitimate (and often profitable) to depart from the pure intervening variable approach and state explicitly that the term being introduced designates a hypothetical construct, which is causally linked, but only stochastically, to each of the (fallible) indicators of its presence and magnitude. This analysis brings up the vexed problem of surplus meaning in theoretical terms (Beck, 1950; Feigl, 1950; MacCorquodale & Meehl, 1948; Reichenbach, 1938). Given a set of observational particulars (considered unproblematic for present purposes) and theoretical statements said to be based on them, logicians recognize different kinds and degrees of surplus meaning. First, it is a truism in philosophy that any lawlike statement has surplus meaning over any finite collection of particulars, even if every term is in observational language. That’s the old induction problem and has little scientific interest. Second, when a lawlike statement is extrapolated to a new domain (e.g., to humans from rats), we begin to have a substantive scientific interest in its ampliative character. Third, when nonobservational terms are found in the theoretical network, we have hypothetical constructs. Their surplus meaning arises from two sources. The first is that almost all
14 P. E. MEEHL
empirical theories have one-way derivability, theory-to-general-fact but not the reverse. (This is a logical and mathematical fact, not a matter of taste, as some psychologists seem to think.) The mystery of how a network of theoretical statements can simultaneously assert and define is solved by a technical device known as the Ramsey Sentence, to discuss which is beyond our scope here (interested readers may see Carnap, 1966, Chapter 26 and pp. 269 - 272; Glymour, 1980, pp. 20 - 29; Lewis, 1970; Maxwell, 1962, pp. 15ff, 1970, pp. 187 - 192; Stegmüller, 1979). But in addition to the surplus meaning entailed by logical structure, a further source is interpretive text associated with the formalism, in which the theoretical entities “partially defined implicitly” by their role in the network are characterized substantively and the admissible interpretations further narrowed. This means that the empirical content of the theoretical terms is not provided solely by “upward seepage” from the observation statements, as was mentioned above (see Meehl, 1990a, 1990b). LATER SKINNER VERSUS 1938 The later Skinner never refuted the powerful and incisive analysis in Behavior of Organisms of the rationale for introducing state-variables. Further, while his meta- discourse came increasingly to deprecate them, along with other “hypothetical inner causes,” many passages in the later works admit—nay, insist, against critics—that behav- ioral science can and does treat such inner states and events. We might fault him for an inconsistency between his object-language (dealing with the facts as best he can) and his metatheoretical stance (pushing his “philosophy of science of a behavior science”). More charitably, we could interpret the difference as between his over-all research strategy and preference and his honesty as a scientist when faced with certain facts. Because he shares with other scientists a belief in spatiotemporally continuous causal chains, he, of course, knows that the only way being deprived of food during the preceding 23 hours can influence lever pressing rate is via a present organic state (“hunger”). It is simply that he does not believe in ESP, spooks, or action at a distance. Speculating about the change, my hunch is a convergence of influences on him over the years. I believe he became tired (bored? irked?) at having to answer the same arguments repeatedly. (Having given up explaining statistical prediction to clinicians after so many years of trying [Dawes, Faust, & Meehl, 1989; Meehl, 1954, 1986a] I know the feeling.) I also suspect that Skinner became somewhat intellectually isolated with age, conversing and corresponding less, and mostly with persons who agreed with him (Daniel Wiener, personal communication, April, 1991). In the 1940s’ Minnesota discussions, I had the impression that he was a bit embarrassed by the demise of his reflex reserve concept, which he had introduced (somewhat apologetically) in Behavior of Organisms , as not quite in the strict operational spirit of his system: I shall speak of the total available activity as the reflex reserve , a concept that will take an important place in the following chapters. In one sense the reserve is a hypothetical entity. It is a convenient way of representing the particular relation that obtains between the activity of a reflex and its subsequent strength. But I shall later show in detail that a reserve is clearly exhibited in all its relevant properties during the process that exhausts it and that the momen- tary strength is proportional to the reserve and therefore an available direct measure. The
16 P. E. MEEHL
persons in the clinic, some big parametric problems arise.^7 My point is a quantitative one, urging caution in extrapolating a quantitative claim (“schedules are more potent than drives”) when the extrapolated context exhibits several nonnegligible quantitative differ- ences from the original. Assume that the activities of a healthy nonincarcerated human instantiate Skinner’s behavior laws, that is, his behavior system is qualitatively correct, including verbal behavior. Here are some structural and parametric differences from the rat-in-box case:
- The range of schedules’ reinforcement probabilities is usually small, compared to the rat (or pigeon!) in the box.
- For the class of operants associated with a particular motive-reinforcer combina- tion, schedules remain relatively constant, barring major societal catastrophe.
- Hundreds of different operants are involved in a typical 24 hour period.
- Hundreds of discriminative stimuli are controlling.
- There are scores, if not hundreds, of reinforcing stimuli, mostly secondary.
- Most of these secondary reinforcers are themselves discriminative stimuli control- ling succeeding operants in long chains.
- Many chains share partially overlapping sequences; and the final reinforcers may be associated with disparate motives.
- Human adults locomote, moving amongst various “social Skinner boxes” present- ing different operanda, SDs, and reinforcements.
- Many discriminated operants SD.R have been shaped and maintained by distinct motive-reinforcement operations, so the same or similar behaviors come under different motivational control. (This is one of the main reasons why clinical interpretation is so difficult at times.) For example, a driver stops the car upon seeing a neon sign “Fisbee’s Bar and Grill” and locomotes into the place. Why? The reason could be (a) hunger, (b) wanting a beer, (c) urinary pressure, (d) check World Series on TV, (e) make a phone call, (f) seeking social company, or (g) seeking a sex partner. All of these motive states except possibly (g) are on nearly a CRF schedule in the presence of visual stimulus “bar and grill,” and the “Fisbee’s” stimulus component may signal that even (g) is rich.
- A large portion (I conjecture almost all) of instrumental acts performed by adult humans involve temporally remote goals, and the reinforcement is self-administered contingent on achieving “subgoals,” as when completion of an onerous task required by one’s job is reinforced by an internal tact such as, “Well, that is done!”
- Although partial reinforcement schedules are probably the rule rather than the exception, the normal ecology of healthy, freely moving human adults rarely imposes such extremely lean reinforcement probabilities as experimenters tend to employ in the box. A large part of our daily behavior resembles Skinner’s paradigm CRF example SD:visual pencil→R:reaching→S*:tactile pencil. Lacking a large scale (unobtrusive) Barker (1968) type of monitoring of adults’ activities over 24 hours, I can only rely on introspection and opinion for a rough estimate. Inquiry among fourteen psychologists as to what percent of their operants—all kinds, all reinforcers—are reinforced yielded a 7 I am not here making the common clinician’s complaint that the experimental findings don’t apply to the “real world.” As my late colleague, MacCorquodale, used to remind clinical students, everything that occurs in spacetime is real. An event or generalization does not somehow lose reality by taking place in a building called a laboratory! See Skinner, 1961 (pp. 242- 257 ).
MURRAY AND SKINNER: 1938 17
range from 40% to 99%, median 91%. (The only estimate <50% was by a resigned department chair serving his last two onerous months in office, and I’m not sure that he used ‘reinforce’ as Skinner does. I could have deleted his estimate with a clear conscience.) A reinforcement probability of P = .90, say on the commonest kind of schedule in daily life, is VR1.1, far richer than those we are accustomed to see in Journal of the Experimental Analysis of Behavior. When operant behaviorists assert that “Of course we run hungry animals, but the big point is that reinforcement schedules are far more important than motivational variables,” this is a quantitative claim whose ecological validity is parameter dependent. To test that claim “in the box,” we should perform experiments in which schedules running CRF, VR1.1, 1.5, 2.0, (densifying in the small integer region) are set to countervail drive levels. I am told by several knowledgeable operant behaviorists that no such “competitive” experiments have been reported. Until they are, the down-playing of drive by extrapolation from Skinner-box research is, it seems to me, unwarranted. Being a neo-Popperian, I will venture a risky conjecture: For schedules in the range CRF–VR5, drive will be controlling. I predict a hungry, nonthirsty rat will press the food-bar rather than the water-bar, almost exclusively, despite their schedules being VR2 and VR1.1, respectively. And my philosophy of science self cannot resist commenting that the absence of parametric studies in that region—considering the thousands of experiments ringing the changes on complicated schedules of a sort no animal outside the lab will ever come across—is a nice example of how theoretical orientation can determine what facts will be collected. I don’t claim the list is complete, but these eleven differences, especially the eleventh, alert us to the danger in immediate extrapolation of quantitative generalizations about motives versus schedules. But the biggest difference is a twelfth, relating to Murray, Tol- man, and McDougall’s way of identifying purpose by cessation of goal-seeking activity.
- Given stable human schedules (mostly low VR! CRF, very rich), when a chain runs off, the equivalent of a huge food-pellet is usually delivered, reducing the motive to near-zero intensity. In Murray language, a certain need is regnant for a limited time interval and terminated by achieving an adequate subgoal. Example: A clerk is thirsty, suspends clerking behavior to approach a water cooler, drinks, abates thirst, goes back to clerking. The n Water was regnant for a couple of minutes, and during that short interval no “competition” with other needs or operants was involved. It is not as if we were always playing off schedules against schedules, or influence of drive strength against potency of schedules, as some writers imply. The clerking operants usually “work,” as does the water seeking operant. The clerking behavior is in some respects like repetitive lever-pressing under a steady state motive; the drinking chain is not. Whenever a controlling motive is essentially terminated by adequate “satiating” reinforcement, the situation resembles Thorndike’s (1898) cat escaping from a cage or Tolman’s rat reaching the goal box more than it does Skinner’s pigeon repeating the same response rapidly hundreds of times to get a small pellet that leaves hunger unabated (cf. Skinner, 1930). Skinner’s substitution of a continuously available manipulandum for the traditional “trial” (maze, shuttle box, discrimination chamber, obstruction box), with the consequent cumulative record instead of a “learning curve,” had great fertility for several reasons, especially by providing a highly sensitive measure of instantaneous operant strength. While not deprecating that major advance, I observe that—like most good instrumentation ideas—it carried with it a disadvantage; it is an experimental paradigm
MURRAY AND SKINNER: 1938 19
applied across the board, he would be estopped from invoking a history of reinforcement, since reinforcement is defined by its behavior effect. This criticism he anticipated (and answers correctly) in Behavior of Organisms (1938, p. 62; cf. Meehl, 1950). If it were illicit in science to infer the value of an unobserved (causal) particular x from an observed (effect) particular y , having repeatedly corroborated a relation ( x R y ), we could not have ascertained the chemical composition of the stars. Relying on Hooke’s Law (“stress is proportional to strain”), if I see the visible portion of a stretched spring double its length, I infer that the unobserved weight has been doubled. Operant behaviorist colleagues reading this paper in draft took mild umbrage about “control,” as if I were imputing a purely technological interest to Skinner and his disciples. This is not my intention, and I am well aware that Skinner was at some pains to characterize himself as a theorist , albeit not of the hypothetico-deductive kind (Skinner, 1969, pp. vii-xii ). But I do not think my remarks here are misleading, given their context. Given his repeated “theoretical” admissions that a state of thirst really exists, inside the organism, as part of the causal chain from input to output, to then disparage invoking it on the ground that we can only influence the state by deprivation or other input manipu- lations, elevates effective control to special status. This Murray (or Freud) would never have done, and I insist that it reflects a preference , not a defensible epistemological criterion of conceptual legitimacy. THE MURRAY NEEDS Our principal focus being Skinner, I shall treat the Murray needs summarily and only as their conceptualization relates to Skinner’s 1938 notion of state-variables, urging the reader to have a look at the extended discussion in Explorations in Personality. I must enter a caveat to operant behaviorists about to read Murray: His thought modes and, even more, his choice of words will probably turn you off. It is a truism among historians and philosophers of science that if one wishes to grasp the substance of a scientist’s theory so as to analyze and utilize it, and especially to relate it to another theory, one must attend to what the theorist says about what he did and what he observed , rather than what he says about what he says (i.e., meta-talk). In Skinner’s case, we had to parse his arch- behaviorist “philosophical” meta-talk from his scientific discourse, lest his “minimizing” tendency prevent our seeing how much his translations endeavor to include. Skeptical, aseptic, and devoted to a “pure intervening variable” theory, his methodological advice is precisely opposed to a philosopher of science like Popper, who advocates “risky, low probability conjectures” (and even holds—wrongly, as I think—that there is no such process as induction!). One of the few polemical remarks in Behavior of Organisms pokes fun at hypothesizing scientists “whose curiosity about nature is less than their curiosity about the accuracy of their guesses” (Skinner, 1938, p. 44). Murray, a very different sort of mind, obviously luxuriates in speculations, analogies, metaphors, cross- connections. By my lights, much of the meta-talk in the Explorations in Personality theory section is unhelpful and, to one in the “tough-minded” tradition, even counter- productive or obscurantist. Should an operant behaviorist complain that Murray’s talk about a “holistic,” “purposive,” “organismic,” “Gestalt,” “dynamic” viewpoint cuts no ice, I will cheerfully agree. One of the book’s dedicatees is Alfred North Whitehead, thanked for his “philosophy of organism.” I doubt that a single experimental procedure or a single personality questionnaire item found in the book derived from Whitehead’s
20 P. E. MEEHL
philosophy. So I advise the tough-minded reader to ignore most of Murray’s meta-talk, remembering charitably that the date was 1938, when most psychology departments— Harvard’s included—considered “personality” hardly a fit topic of scientific study. Interestingly, Skinner seems to feel obliged to defend his introduction of state-variables against a skeptical, minimizing methodology that he largely shares ; while Murray’s meta-talk has the same defensive intent, against the same methodology, but which he does not share. Bypassing the defensive meta-talk, we note immediately three encouraging points in Murray’s introduction of the need concept. First, while he offers seven main reasons (and 16 adjuvant arguments[!], some rather light makeweight) for introducing such a concept, he insists that the main basis for inferring, defining, and identifying a need is behavioral. Second, the core structure of this behavioral basis is that in a situation the organism emits some behavior B which brings about an effect E, i.e., some stimulus change is contingent on something the organism does. Prima facie, that reads pretty much like a good Skinnerian. Third, when it comes to characterizing the behavior thus identified, he emphasizes that we (normally) do not specify the effector patterns (“actones,” divided into “motones” and “verbones”) but instead the molar effect. It is particularly significant that in defending that level of behavior analysis, Murray cites Skinner’s classic paper (1936) on the generic nature of the concepts of stimulus and response—one of the few mentions of an experimental psychologist in the entire 105 pages of the theoretical chapter in Explorations in Personality! It is an interesting historical fact that 99% of all psychological research since Ebbinghaus, on humans or infrahuman animals, has taken as the dependent variable a molar achievement class (turn in maze, depression of lever, utterance of a “word”) rather than a response class specified by effector pattern (MacCorquodale & Meehl, 1954, “Excursus: The response concept,” pp. 218 - 231). Of course in the Skinner box this “works” because the (mixed) class of effector activities that suffices to get the lever down is, via the apparatus structure, the class that is reinforced. It may be objected that while his tripartite sequence S→R→S* looks familiar, Murray over-emphasizes the termination of activity as a consequence of goal attainment, not a usual operant behaviorist focus. True enough, and I have no wish to obscure that fact. It arises, however, not from any basic conceptual difference, but rather from the difference in the situational contexts typical for Murray and for Skinner. As explained above, in the box we deliver small amounts of reinforcement lest we satiate drive so quickly as to obscure the quantitative features of the cumulative record (and their determining contingencies) that we wish to study. This being merely a routine “technical” procedure in animal experimentation, we are tempted to underestimate its conceptual significance when we are relating two theories developed on largely disjunct domains (species, oper- ants, reinforcers, parameters ). Skinner’s first publication showed the decelerated curve of ingestion due to satiation (Skinner, 1930, and 1938, Figs. 120 and 121, pp. 344 - 345). Murray, having in mind cases such as a free moving adult human going to lunch, proper- ly thinks of reinforcements that satiate the regnant need or at least reduce its intensity so that other competing needs become regnant. The main point is that the content of a need is provided by the empirical fact that, given a certain situation(-class), one or another behavior sequence occurs, having a tendency to bring about certain changes.
