JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

IN THE BEGINNING ...

JAMES A. DINSMOOR

INDIANA UNIVERSITY

I stumbled across systematic psychologyalmost by accident. I had entered the graduateprogram at Columbia, intending to specializein social and clinical psychology. But to makeends meet (and master some of the funda-mentals more securely while doing so), I hadaccepted a stipend of $37.50 per month toserve as the "reader" in the introductory coursefor male undergraduates (Columbia College).The first two semesters were fairly uneventfulintellectually, but in the spring of 1945 FredS. Keller returned from his Morse Code re-search with the U.S. Army Signal Corps andresumed his teaching duties, which includedthat course. When he handed out a mim-eographed outline that proposed a "consid-eration [of] principles derived from the analysisof human and animal behavior by the ex-perimental method," I realized that somethingwas up. Soon he was into such esoteric mattersas the static laws of the reflex, Type S andType R conditioning, extinction, generaliza-tion, stimulus discrimination, response dif-ferentiation, chaining, and their applicationto verbal behavior and thinking.By then I was beginning my fourth semester

(plus two summers) of graduate study. Thatwas just enough training for me to have madetwo disturbing observations: First, in R. S.Woodworth's laboratory course I had dis-covered that even in "experimental psy-chology" many of the best known theoriesyielded no predictions sufficiently specific tobe subjected to empirical test. If they madeno predictions, they could not be very useful.Second, the different areas within psychologyseemed to be completely compartmentalized,

The quotation at the end of this review was takenfrom an unpublished manuscript dated October 12, 1987,and is here reproduced by permission. I thank DonaldA. Cook, Kay S. Dinsmoor, Eliot Hearst, and AlexandraW. Logue for their comments on an earlier version ofthe present manuscript. Reprints may be obtained fromJames A. Dinsmoor, Department of Psychology, IndianaUniversity, Bloomington, Indiana 47405.

with no broader principles crossing theirboundaries to link them together. By contrast,the processes Keller was talking about wereobjective descriptions of what happened inthe laboratory and could readily be replicated.When he further suggested that "the modernpsychologist ... seeks to achieve a systematicinterconnection of the basic facts of humanand animal behavior," I was intrigued. Whenhe proceeded to do so, I was hooked. "Re-sponse" as an abstract category seemed tome to be a promising stand-in for whateverspecific activity might be of interest in a givensituation and laws of "response strength" apromising system for psychology as a whole.As most readers will recognize, Keller's

section on verbal behavior came from a manu-script version of the book that did not appearin print until 12 years later (Skinner, 1957).But almost without exception, the funda-mental concepts throughout the rest of theoutline could be found in The Behavior ofOrganisms: An Experimental Analysis (1938).Aside from a series of articles that were rarelyconsulted because the important material wasduplicated in the book, that was all therewas. In its black cover, the "B of 0," aseveryone called it, served in the early daysas the bible of our movement.

In the years that followed, a large numberof writers have commented on Skinner's work,and they have credited him with many andvaried contributions-methodological, empir-ical, and philosophical. The purpose of thepresent review, however, is to single out theone or two contributions that in my opinionhave been most critical to his impact, notonly on my own thinking but on the fieldof psychology as a whole.

A RESEARCH TECHNOLOGYOne factor that could readily be overlooked

nowadays is Skinner's contribution to ex-perimental technology. When an innovation

287

1988, 50,287-296 NUMBER 2 (SEPTEMBER)

JAMES A. DINSMOOR

is widely adopted, it becomes commonplace,and no one any longer gives it much thought.This is what has happened to the arrangementthat serves as the discriminative stimulus forthat ubiquitous tact, "Skinner box"-the pro-vision of a switch that springs back into placeafter it is operated by the subject. As controlof the experimental routine by an electronicsystem requires some sort of switching deviceas an input, it seems likely that when com-puters finally entered the behavioral labora-tory psychologists would inevitably have dis-covered lever pressing, disk pecking, or somesimilar device. But Skinner adopted the leverback in 1931, and it was the switch that ledto the complex networks of electromechanicalprogramming modules-crude equivalents totoday's computers-rather than the other wayaround. In the decades immediately followingpublication of the book, there were many in-vestigators who carefully dissociated them-selves from the rest of Skinner's thinking butwho made use of his lever or his key in theirexperimental work. Today, a very large partof the research on learning and cognition con-ducted with nonhuman subjects depends onthese devices, including, for example, the useof conditioned suppression and autoshaping tostudy Pavlovian processes and the use of thematching-to-sample technique to study me-morial processes and attention. Research tech-nology with human subjects has also caughtup, but in most cases this owes more to theadvent of the computer than to the historicalinfluence of Skinner.

Recording the opening or closing of a switchas the subject's response has a number ofadvantages. Prosaic but important is the factthat it permits automation of the experimentalprocedure, with its attendant promises ofgreater precision, more complex contingencies,and much more data for the same investmentin human time. And because manipulationof a switch does not ordinarily depend onlocomotion, this technology permits greaterfreedom in programming relationships amongevents. In the runway or the maze, wherethe recorded behavior is a change in the sub-ject's location, the ratio of individual stepsto reinforcer deliveries, time to reinforcer,effort expended, and change in external stim-ulation all are likely to be confounded. Theyare difficult or impossible to disentangle asparameters of the subject's behavior. Fur-

thermore, changing one of these parametersmay require time out for the use of a hammerand saw, disrupting the continuity of ex-perimental events. When antecedent stimuli,responses, and consequences are linked byelectric circuits, all relationships known tonature or that can be concocted by humanimagination should be reproducible and theireffects capable of being examined within thelaboratory setting. Furthermore, in a mannerreminiscent of Watson's remarks about theconditioned response as the unit of "habit"(1924, pp. 207-208), larger patterns of be-havior, such as ratios, chains, simulations offoraging and observing, and those producedby a variety of complex schedules, can beconstructed by linking individual stimuli andresponses together in complex contingencynetworks. A great variety of such patternshave been arranged in which single pressesor single pecks are still recorded as elementaryconstituents. The abstract category "response"serves an integrative function at the theoreticallevel, and in a somewhat different fashionthe concrete instance of a switch closure servesto integrate the data at an empirical level.

EXTENDING THE SYSTEMBut if the closure of a switch provided a

lingua franca at the level of instrumentationand the number of closures per unit of time(rate of responding) provided a lingua francaat the level of measurement, these matterswere of concern primarily to experimentersusing rats, pigeons, and monkeys to studybehavioral processes at the most abstract level.As Skinner noted in the second volume ofhis autobiography (Skinner, 1979, p. 318),until Fred Keller and Nat Schoenfeld beganteaching laboratory courses based on TheBehavior of Organisms, only 550 copies hadbeen sold.The technology did make it relatively easy,

once the necessary equipment was obtained,to conduct research on basic behavioral pro-cesses, and during the period when grantmoney was in reasonable supply the equip-ment could indeed be obtained. Soon theredeveloped a community of "operant condi-tioners" who knew each other both profes-sionally and personally, who spoke a commonlanguage, and who felt common loyalties. Thiswas the original foundation of the movement

288

THE BEHAVIOR OF ORGANISMS

that eventually established two of psychology'smost frequently cited journals, Division 25 ofthe American Psychological Association, andlater, an independent organization known asthe Association for Behavior Analysis (see La-ties, 1987; Peterson, 1978). Many of the pi-oneers in behavior modification were trainedinitially in the conditioning laboratory.A technology for studying the behavioral

processes of rats or pigeons, however, couldnever by itself have generated a system thatwould extend beyond the laboratory to theworld of practical affairs. Although the lan-guage of The Behavior ofOrganisms may soundarchaic to contemporary ears, the seeds werethere. To put it in a way that will be accessibleto a broad spectrum of readers, Skinner hadbegun the task of constructing a set of conceptscapable of dealing in a scientific fashion withthe heartland of psychology. His vocabularywas quite forbidding to those encounteringit for the first time. It required an effort tomaster. But it was on the one hand moretightly tethered to concrete observations thanthe language of any other systematic treatmentand on the other was capable of transcendingthe boundaries between the conventional phys-ical categories used in everyday discourse (fora panorama of application, see Lundin, 1974;Martin & Pear, 1983; Miller, 1975).

CONSTRUCTING A THEORYBecause the presidential address he deliv-

ered to the Midwestern Psychological As-sociation in May of 1949 was entitled "Aretheories of learning necessary?" (publishedin 1950), Skinner has almost invariably beendescribed in subsequent commentaries as"atheoretical," or even "antitheoretical." ButI believe this to be a profound misconception.Finding appropriate terms for the descriptionof behavior is in itself a theoretical enterprise.A revealing glimpse into Skinner's views ontheories and how they should be constructedis provided by the remarks he had deliveredjust 2 years earlier at a symposium conductedon March 4th and 5th, 1947, at the Universityof Pittsburgh (Skinner, 1947). There he de-clared quite explicitly and unabashedly that"A theory is essential to the scientific un-derstanding of behavior as a subject matter"(1947, p. 29). In the preceding paragraph,he had specified what he considered a theory

to be: "Theories ... are statements aboutorganizations of facts.... But they have agenerality which transcends particular factsand gives them a wider usefulness" (1947,p. 28; see also Skinner, 1961, p. 230).

After presenting his usual objections to suchexplanatory fictions as a controlling mind ora hypothetical neural structure, he went onto suggest that "The first step in buildinga theory is to identify the basic data" (1947,p. 34; 1961, p. 233). To illustrate this stageof theory construction, he cited the work ofGalileo, who chose "to deal with the positionsof bodies at given times, rather than withtheir hardness or size. This decision ... wasnot so easy as it seems to us today" (1947,p. 38; 1961, p. 235). As the psychologicalcounterpart, he pointed to the problem ofdelineating "the parts of behavior and en-vironment between which orderly relationsmay be demonstrated" (1947, p. 34; 1961,p. 233). In its second stage, he suggested,theory should "express relations among data"(1947, p. 37; 1961, p. 235), preferably interms as general as possible. Then, in thethird stage, new concepts might emerge, asthe concept of acceleration had emerged fromthe second-stage relation between position andtime.

Although he never made it to the thirdstage, I believe this account of the first twostages to be a reasonably accurate descriptionof what Skinner considered his own activitiesto have been when he was writing The Be-havior of Organisms some 10 years earlier.After choosing "stimulus" and "response" torepresent relevant facets of the environmentand the behavior, respectively, he began toexamine some of the kinds of relations thatwere possible between these categories ofevent.To those who have not read Skinner, the

terms stimulus and response have an un-fortunate connotation. They were borrowed,via Pavlov, from the science of physiology,and to the casual observer, the physiologyof the reflex appears extremely atomistic.Although "stimulus" and "response" mayconfer an aura of material respectability upona seemingly more ephemeral subject matter,they also conjure up visions of much too simplydetermined relations between physical ener-gies impinging on the sensory receptors, onthe one hand, and muscular contractions and

289

JAMES A. DINSMOOR

glandular secretions on the other (e.g., Breger& McGaugh, 1965). This stereotyped pictureof Skinner's views may be responsible forthe demand that he and other behavioriststake up the burden of accounting for complexpatterns of activity in physiological terms-a task that is actually quite distinct from thatof the experimental analysis of behavior-andfor some of the charges frequently leveledby critics who do not seem to be familiarwith what has actually been said. If the termsstimulus and response were not available,however, others would have to be inventedin their stead. As Skinner has pointed out(e.g., 1938, p. 9), there is no way to studychanges in the totality of behavior as a functionof changes in the totality of the environment.In order to measure, or even to record, itis necessary to select some part of that be-havior. And when it is shown that a givenpart of the subject's behavior can be changedby changing some part of the environment,that part is known as a "stimulus."

There are other concepts of systematic im-portance to be found in The Behavior of Or-ganisms. Somewhat to my surprise, when Ireturned to the book after a number of years,I found that these concepts were not repre-sented by formal definitions. Most of the con-tent is devoted to empirical determinations ofthe behavioral processes governing the rate oflever pressing by the albino rat. This is thesecond stage of theory construction, as outlinedby Skinner. But each of the "dynamic laws ofthe reflex" has embedded within it a classifi-catory scheme. One may look in vain, for ex-ample, for a direct and explicit definition ofwhat is meant by a "discriminative stimulus."But what might be termed a low-profile andincomplete definition can be found on p. 178:"Only in the presence of SD is RO followed byS5"; and a more complete definition is impliedin "THE LAW OF THE DISCRIMINA-TION OF THE STIMULUS IN TYPE R.The strength acquired by an operant throughreinforcement is not independent of the stimuliaffecting the organism at the moment, and twooperants having the same form of response maybe given widely different strengths throughdifferential reinforcement with respect to suchstimuli" (1938, p. 228). From The Behavior ofOrganisms it is possible to extract such broadconcepts as positive reinforcement, negative re-inforcement (in those days meaning punish-

ment), extinction, stimulus generalization (in-duction) and discrimination, response inductionand differentiation, successive approximation(shaping), chaining, and conditioned rein-forcement. In later years, as the system hasexpanded, other concepts have been added.

CONCEPTS AND DEFINITIONSThe full meaning of the word "concept"

is difficult to grasp unless one is familiar withat least one illustrative study of the concretebehavioral relationships to which the termis applied. In laboratory studies, the exper-imental subject is often human, and somenew and undefined word has typically beenused as the response (e.g., Hanfmann & Ka-sanin, 1937; Hull, 1920; Kuo, 1923); butthe subject may also be nonhuman, in whichcase the closure of a certain switch will dojust as well (e.g., Kelleher, 1958). What isrecorded in either case is the subject's responseto each member of a population of stimulithat differ in a number of respects. In essence,to have a concept is to react in the same wayto a certain group of stimuli that presumablyhave something in common but not to reactin that way to other stimuli that do not sharethe common characteristic. A variety of dif-ferent groupings may be possible but onlyone is currently acceptable to the experi-menter.

Outside the laboratory, too, nature maybe sliced in a variety of ways. To take aliteral example, picture two butchers slicingup a leg of beef, one in Bloomington, Indiana,and the other in Laconia, New Hampshire.They will cut the beef in different placesand divide it into different parts. If you askfor "round steak" in Bloomington, you willget a certain cut with certain properties, tobe prepared in certain ways, but if you askfor the same cut in Laconia you will get apuzzled look. And it is not merely a matterof the name by which they identify it: Thebutcher in Laconia will have nothing entirelyequivalent, with quite the same properties.The concept does not exist.

Similarly, any population of instances canbe divided in a number of ways-for example,according to number, according to shape,according to size, according to color, or ac-cording to much subtler properties (e.g., Gard-ner & Gardner, 1984; Herrnstein & Loveland,

290

THE BEHAVIOR OF ORGANISMS

1964; Smoke, 1932). Some of these distinctionsmay be useful for a given purpose, othersnot. Technically speaking, what we are deal-ing with here is the set of discriminative stimulicontrolling a given response. In verbal be-havior, this relation between stimuli and aresponse is what Skinner (1957) called a tact.Usually, selective reinforcement determinesthe boundaries. When the rules as to whatis included and what is excluded can them-selves be verbalized-which is not always thecase-we have a definition. Definitions arerules governing the application of the wordsthey define.A science is no better than its definitions.

Rules for the application of words to eventsare critical steps both in testing a theory andin applying it. In some sense-this is notthe place to spell out the details-all scientifictheory must ultimately be reducible to a setof propositions of the form that circumstancesA (for antecedent) lead to outcome C (forconsequent). These are descriptive statementsthat at their most basic level merely summarizesequences of events. When testing or applyingsuch statements, we make the added as-sumption that the A-C relationships observedin the past will continue to hold for the future.The statements relating verbal antecedent

(VA) and verbal consequent (VC) are iso-morphic with the relations between observedantecedent (OA) and observed consequent(OC), and the role of definition is to translatewords into observations and observations intowords, moving from one level to the otherand back again, just as we might translateback and forth between two languages. Whentesting a theory, we begin with the verbalspecification of antecedent circumstances (VA),use a definition to select an experimentalprocedure (OA), carry out that procedure,observe the outcome (OC), and translate thedata by way of definition back into the termsof the theory (VC). If the outcome of theexperiment is consistent with the outcomespecified by the theory (OC = VC), we keepthat theory; if not, we modify what we havesaid.

In a very broad and abstract sense, theuses we make of theories come under theheadings of prediction and control. To makea prediction, we observe the current circum-stances (OA), find a theory that applies tothose circumstances-has a corresponding an-

tecedent (VA)-take the outcome specifiedby that theory (VC), and translate that backinto an observable outcome (OC), which con-stitutes the prediction. When we use a theoryto control events, we begin with the desiredresult (OC), find a theory with a corre-sponding outcome (VC), backtrack to theconditions stated by the theory for such anoutcome (VA), and translate those conditionsback into the necessary practical operations(OA).

Although definitions were not at all con-spicuous in The Behavior of Organisms, Skin-ner's behavior at the first Conference on theExperimental Analysis of Behavior less thana decade later confirmed my budding suspicionthat they are extremely important. Quite abit of his time was allocated to tentative spec-ifications for a variety of basic terms to beused in the science of behavior (see also Figure2 in Dinsmoor, 1987). In the early days ofJEAB, a number of readers complained abouttechnical vocabulary that made material inthe new journal difficult to follow. But withoutthat vocabulary, our authors could not makesensible statements about their procedures,let alone their conclusions. I have discoveredthat undergraduate students and cognitivepsychologists also experience difficulty withbehavioral language and often feel that it ismerely a pedantic version of ordinary English,capable or almost capable of a literal, one-to-one translation into their mother tongue.

INADEQUACY OF POPULARVOCABULARY

This is not the case. The linguistic categoriesthat develop and are passed along from gen-eration to generation within a given cultureare presumably of some effectiveness in deal-ing with the exigencies of daily life, but theyare not the same as those needed for validand general scientific statements. Conse-quently, the words that are reinforced andthe circumstances under which they are rein-forced in everyday conversation are not thesame as those in various scientific commu-nities. To put it succinctly, different conceptsare needed.That Skinner recognized the inadequacies

of the boundaries established by popular usageis evident from the following passage:

291

JAMES A. DINSMOOR

[The] science of behavior ... must not takeover without careful consideration the schemeswhich underlie popular speech. The vernacularis clumsy and obese; its terms overlap eachother, draw unnecessary or unreal distinctions,and are far from being the most convenientin dealing with the data. They have the dis-advantage of being historical products, intro-duced because of everyday convenience ratherthan that special kind of convenience char-acteristic of a simple scientific system. It wouldbe a miracle if such a set of terms were availablefor a science of behavior, and no miracle hasin this case taken place. (1938, p. 7)

It is difficult to conceive of a way in whichany of the physical or biological sciences couldhave achieved even a semblance of their cur-rent effectiveness without the construction oftechnical vocabularies. The observations mustbe classified in appropriate ways to makepossible the generation of valid descriptivestatements of any substantial degree of gen-erality. If the whale were to be classified as

a fish, for example, or the tomato as a veg-etable, otherwise correct statements about fishor about vegetables would for these cases berendered incorrect. However, when whalesare classified as mammals and tomatoes as

fruit, all is well. Consistency is preserved.

A SCIENCE OF BEHAVIORBecause they do not use a mentalistic vo-

cabulary, Skinner and other behaviorists fre-quently have been accused by cognitive psy-chologists of ignoring or denying all that ismost interesting and most central to psy-chology. But what Skinner was avoiding wasnot the processes themselves, as should beobvious from his later writings, but only wordsthat were inadequately related to their con-trolling stimuli:

The traditional description and organizationof behavior represented by the concepts of"will," "cognition," "intellect," and so on,cannot be accepted so long as it pretends tobe dealing with a mental world, but the be-havior to which these terms apply is naturallypart of the subject matter of a science of be-havior. What is wanted in such a science isan alternative set of terms derived from an

analysis of behavior and capable of doing thesame work..... Traditional concepts are based

upon data at another level of analysis and

cannot be expected to prove useful. They haveno place in a system derived step by step fromthe behavior itself. (1938, p. 441)

In common with those used by other learn-ing theorists, Skinner's concepts occupy aposition that is central to psychology as awhole. It was no accident that the first attemptsto communicate his views to a broader au-dience (Keller & Schoenfeld, 1950; Skinner,1953) took the form of textbooks to be usedfor introductory courses rather than for somespecialty area within the discipline. Almostall behavior of interest to psychologists insuch areas as personality and social, clinical,industrial, educational, and developmentalpsychology is behavior that is greatly influ-enced if not wholly shaped by a process oflearning; although some interventions by ap-plied psychologists take the form of changesin the situation, more often they take the formof changing the individual's reactions to thatsituation, a process that in most cases canbe classified as learning.

Although they differ greatly in the databases they employ and in their scientific rigor,there is a curious similarity between behavioranalysis and psychoanalysis. Perhaps it is morethan a coincidence that they both apply theword "analysis" quite broadly, one to behaviorand the other to the psyche. Both have rejectedthe conscious mind of common-sense psy-chology as the arbiter of behavior and havepresented themselves as alternative systemscapable of integrating a large part of theirsubject matter. The frequent use of the word"dynamic" in The Behavior of Organisms isreminiscent of the use of the word "psy-chodynamic" in clinical writings. Perhapssome reference was intended to the underlyingemotions or motivations-whatever thesewords may mean-that affect the behaviorin a way that is orthogonal to its immediatecontext (see Skinner, 1938, pp. 21-26). Buta more rigorous and more inclusive char-acterization is that both systems attempt todescribe the laws of motion governing thecharacteristic reactions of the individual tocertain situations that recur from time to time.That is, there is reference to the prior historyof the organism. Writers like Dollard andMiller (1950) and Mowrer (1950), closelyassociated with Hull, devoted considerableattention to the behavioral interpretation of

292

THE BEHAVIOR OF ORGANISMS

psychoanalytic mechanisms, and so, in laterwritings, did Skinner (e.g., 1953, 1961).

INTERVENING VARIABLESSome of the terminology used in The Be-

havior of Organisms has virtually disappearedfrom current usage (e.g., Type S and TypeR conditioning, now called by other names,or the reflex reserve, a construct that hassince been discarded); but the fruitfulness ofother concepts is indicated not only by theircontinued use in research on conditioning butfrequently by their extension to applied set-tings (e.g., see Bijou & Baer, 1961, 1965;Lundin, 1974; Martin & Pear, 1983; Miller,1975). The same cannot be said for mostof the terms suggested by rival theorists fromthe same period (e.g., Guthrie, 1935; Hull,1943; Tolman, 1932), although they, too, wereattempting to deal with the central problemsof psychology at the most general level.One of the reasons for the popularity of

Skinner's concepts among those interested inthe application of psychology may lie in hissparing use of mediating constructs. True,his attitude toward such constructs has neverbeen quite as proscriptive as that ascribedto him by many current writers, both criticaland friendly. So far as I can determine, hehas never denied their scientific legitimacy;he has only questioned their strategic valuefor the conduct of research in the science ofbehavior. In fact, in The Behavior ofOrganisms,he himself made use of such "hypotheticalmiddle term[s]" as reflex strength, the reflexreserve, and particularly the "intermediatestate[s]" (p. 24) of emotion and drive-con-cepts that he explicitly identified as corre-sponding in their status within the systemto the mediating constructs that had recentlybeen proposed by Tolman: "What has herebeen treated as a 'state,' as distinguished fromthe operation responsible for the state, is calledby Tolman an 'intervening variable"' (1938,p. 437; note also on p. 25 his recognitionof the distinction between learning and per-formance). What is more, in a footnote heestablished his precedence by citing an earlierpublication (Skinner, 1931).

Intervening variables (by which I meanto include hypothetical constructs, which aresometimes considered a separate category) playa central and a critical role in such matured

sciences as physics, chemistry, and genetics.But examination of the difficulties encounteredby Tolman and by Hull with their respectivesystems suggests that Skinner took the wisercourse in staying as close as possible to hisdata, holding such constructions to a min-imum. The problems with Tolman's andHull's systems were opposite in nature, andthe failures at both extremes suggest that sucha fate may have been inevitable at any pointalong the continuum.

In Tolman's case, the proposed variableswere borrowed from traditional common-sensepsychology and hence were intuitively ap-pealing, but they were largely programmatic.He offered only a tentative list of what theappropriate theoretical terms might be, andhe was understandably reluctant to labor overthe details of their definition. That could comelater. Tolman's strategy allowed considerableflexibility and made it easy for him to "ex-plain" results that posed difficulties for Hulland perhaps even for Skinner. But as a con-sequence of that strategy, Tolman's systemnever achieved the specificity of predictionneeded to compete as a viable alternative tothose offered by Skinner and Hull.

Hull went to the opposite extreme. Heconstructed a highly formalized network ofintervening variables in which most of therelationships, including those with observableevents, were specified with a degree of rigorthat was, within psychology, quite rare. Look-ing back, the image that often occurs to meis that of Icarus, who strapped on wings offeathers and wax to escape from an islandprison. He flew too high, the sun melted thewax, and he fell into the sea. Hull's ambitionwas too lofty, and his goals could not beachieved. Perhaps he was naive to think thatsuch an enterprise could succeed, but one canstill admire his daring and the height to whichhe flew.

Attempts were made by others to extendHull's system to the realm of practical affairs.Dollard and Miller (1950) constructed agreatly simplified version of Hull's system-in itself a significant move-and used it to"explain" a variety of psychodynamic pro-cesses noted in the clinic. (Unfortunately theseprocesses were themselves open to questionon empirical grounds.) Although in the UnitedStates the first behavior modifiers were in-spired by Skinner (see Ullmann & Krasner,

293

JAMES A. DINSMOOR

1965; Ulrich, Stachnik, & Mabry, 1966), inBritain they usually cited Hull (e.g., Eysenck,1960). But Skinner's system is still with us,whereas the Hullian contribution seems tohave petered out. The essential difficulty, inmy view, was that Hull's structure was toocomplex to be practical in application. It wasdifficult and time-consuming to learn andcumbersome to use, and, errors aside, thebenefits simply were not of sufficient mag-nitude to compensate for the costs.By avoiding such lengthy verbal linkages

between theory and data, Skinner was ableto keep his concepts precise, simple, and rel-atively few in number. They map fairly di-rectly onto the observations themselves, ina manner that Skinner suggested was "purelydescriptive" (1938, p. 426). But Skinner'sdescriptions were certainly couched in termsdifferent from those used by the layperson,based on different criteria.

FUNCTIONAL DEFINITIONSAs stated earlier, the stimuli impinging

upon an organism can be categorized in avariety of ways. The classificatory schemeused in The Behavior ofOrganisms is especiallyeffective for a systematic science of behaviorbecause it cuts across a number of importantboundaries. What Skinner's definitions ignoreis almost as important as the criteria to whichthey attend. They ignore, for example, thespecies of the subject, the distinction betweenthe laboratory and the natural world, differentsettings within the natural world, and differenttopographies of behavior in each of thosesettings. From a conceptual point of view,at least, there are no barriers to extrapolationacross the face of psychology.Compare the concepts in Skinner's system

with the categories normally used by thenonpsychologist. When asked to describe ascene, the average person might simply listthe objects that are present, giving each thename supplied by the national language andperhaps its spatial relation to other objects:"two chairs and a table, a window throughwhich one can see some trees outside...,")and so on. (Significantly enough, when askedto describe someone's actions, the same personmight give a highly inferential account, couchedin terms belonging to what might best bedescribed as a contemporary folk psychology.

One of the things that may be learned fromlaboratory work with rats and pigeons is touse physical rather than inferential categoriesin describing behavior.) For inanimate objects,the conventional categories maintained by ourculture draw sharp distinctions in terms ofimmediate physical characteristics, such asshape, size, and color, because these attributesare quickly and reliably discriminated andthe distinctions among different objects areimportant for everyday uses, such as decidingwhere to sit, what to eat, when to flee, andso on. But if used for a psychological system,this kind of physical specification would im-pose restrictions that would prevent us fromtransferring the concept from one experimentto another, let alone to the settings for suchtarget behaviors as studying, personal groom-ing, inviting on a date, attaching seat belts,or sequences of verbal behavior.

In terms of progress toward a general the-ory, the situation is not much more en-couraging when we turn to the sensory psy-chologist. There the dimensions used to describethe stimulus are less immediate and thereforeless convenient for everyday purposes-theyrequire instrumentation-but serve more ac-curately to interrelate physiology and functionwithin the sensory field. Although dimensionslike wave length, visual angle occluded, lu-minous flux, angular velocity, auditory fre-quency, or sound pressure level may be ap-propriate for the reporting of experimentsin vision or audition, in research on con-ditioning they occupy center stage only inthe methods section; in theoretical discussions,they are relegated to the wings. For the mostpart, we are not much interested in this typeof specification, despite its technical sophis-tication, but in whether stimuli are discrim-inative, reinforcing, eliciting, inhibitory, aver-sive, and so on. These classifications requirefunctional definitions.The type of definition required for a general

science of behavior sounds extremely laboredto the untrained ear, and the need for thistype of definition is by no means apparentto casual observation. Those who teach thisform of analysis encounter substantial re-sistance from many of their students. Thefailure to appreciate the inadequacy of thetraditional vocabulary and the gains to beachieved through a functional classificationmay constitute the single most important bar-

294

THE BEHAVIOR OF ORGANISMS 295

rier to more widespread acceptance of a trulyscientific perspective among professional psy-chologists, let alone the lay public. Functionaldefinitions are not based on immediate senseimpressions but on the location of the stimulusor the response in question within a networkof events stretching across some period of time.To put it another way, a history is required.Such definitions are not the most convenient,to be sure, but they are necessary if we areto break away from the popular mythologyof our culture and establish a scientific treat-ment of our subject matter.

In Skinner's system, stimuli are classifiedin terms of the operations into which theyenter and the effects of those operations onthe subject's behavior. As originally described,at least, a (positive) discriminative stimulusis present when the specified response is el-igible for reinforcement. The response occurssooner or at a higher rate in the presenceof that stimulus than in its absence. In anotherexample, the stimulus may be delivered anumber of times following instances of thedesignated response; if the response occursat a higher rate under that circumstance, thenthe stimulus is classified as reinforcing.

In spite of the fact that these functionaldefinitions transcend the limitations of de-scription in terms of immediately detectablephysical attributes, as indicated above, ul-timately they do refer back to concrete physicaldescription, and in a fairly unambiguous man-ner. Whether a stimulus is present or absentwhen a reinforcing event arrives is a matterthat in most cases can easily, reliably, andobjectively be determined by direct observationor by suitable recording apparatus. It canalso be verified by an independent observer,and agreement among observers should beclose to perfect. Whether the event that takesplace in the presence of the stimulus can beclassified as reinforcing is determined by mea-suring some physically specifiable item ofbehavior (originally, counting the closures ofa switch) across different periods of time.

Although these are not the same obser-vations used by the layperson or the sensorypsychologist, they are equally physical in theirstatus, and it is the return to concrete physicalspecification that provides the objectivity sonecessary to scientific discourse. Skinner him-self has suggested that effectiveness in con-trolling behavior is a better criterion for the

selection of a scientific vocabulary than isagreement among observers. Perhaps the con-trol has to come first, to make the productionof a new term reinforcing to its author. Butto establish the collective body of knowledgeknown as a science, we need the ability tocommunicate with other scientists. Eventually,we must transmit our findings to our studentsand to the general public, who we hope willmake use of our verbal behavior and willprovide economic support for our efforts. Thetransmission of scientific information dependson agreement between the speaker or writerand the listener or reader in the way in whichwords are used.

In this review, I have argued that eventhough Skinner largely avoided the use ofintervening variables in his work, he none-theless constructed a theoretical system; andI have argued that the most important sourceof his influence on the behavior of otherpsychologists is to be found in the natureof the concepts he proposed for classifyingbehavioral events. These concepts are thecontent of the system. Direct testimony con-cerning the importance of Skinner's defini-tional structure has been provided by ArnoldBuss, who took a course with him in the1940s. Preparing a reminiscence for the cen-tennial of Indiana University's psycholog-ical laboratories, Buss recorded the followingobservation: "I took a seminar with him, andhe asked the class not to necessarily agree withhis position but to use his language so that wecould understand one another without confu-sion. By the end of the semester, I discoveredthat his language equalled his position, andonce you became accustomed to talking hisway, you were a Skinnerian."

REFERENCESBijou, S. W., & Baer, D. M. (1961). Child development:

Vol. 1. A systematic and empirical theory. New York:Appleton-Century-Crofts.

Bijou, S. W., & Baer, D. M. (1965). Child development:Vol. 2. Universal stage of infancy. New York: Appleton-Century-Crofts.

Breger, L., & McGaugh, J. L. (1965). Critique andreformulation of "learning-theory" approaches to psy-chotherapy and neurosis. Psychological Bulletin, 63,338-358.

Dinsmoor, J. A. (1987). A visit to Bloomington: Thefirst Conference on the Experimental Analysis ofBehavior. Journal of the Experimental Analysis of Be-havior, 48, 441-445.

296 JAMES A. DINSMOOR

Dollard, J., & Miller, N. E. (1950). Personality andpsychotherapy: An analysis in terms oflearning, thinking,and culture. New York: McGraw-Hill.

Eysenck, H. J. (Ed.). (1960). Behaviour therapy andthe neuroses. Oxford: Pergamon Press.

Gardner, R. A., & Gardner, B. T. (1984). A vocabularytest for chimpanzees (Pan troglodytes). Journal of Com-parative Psychology, 98, 381-404.

Guthrie, E. R. (1935). The psychology of learning. NewYork: Harper.

Hanfmann, E., & Kasanin, J. (1937). A method forthe study of concept formation. Journal of Psychology,3, 521-540.

Herrnstein, R. J., & Loveland, D. H. (1964). Complexvisual concept in the pigeon. Science, 146, 549-551.

Hull, C. L. (1920). Quantitative aspects of the evolutionof concepts. Psychological Monographs, 28 (1, WholeNo. 123).

Hull, C. L. (1943). Principles of behavior: An introductionto behavior theory. New York: Appleton-Century-Crofts.

Kelleher, R. T. (1958). Concept formation in chim-panzees. Science, 128, 777-778.

Keller, F. S., & Schoenfeld, W. N. (1950). Principlesofpsychology: A systematic text in the science of behavior.New York: Appleton-Century-Crofts.

Kuo, Z. Y. (1923). A behavioristic experiment oninductive inference. Journal ofExperimental Psychology,6, 247-293.

Laties, V. G. (1987). Society for the ExperimentalAnalysis of Behavior: The first thirty years (1957-1987). Journal of the Experimental Analysis of Behavior,48, 495-512.

Lundin, R. W. (1974). Personality: A behavioral analysis(2nd ed.). New York: Macmillan.

Martin, G., & Pear, J. (1983). Behavior modification:What it is and how to do it (2nd ed.). Englewood Cliffs,NJ: Prentice-Hall.

Miller, L. K. (1975). Principles of everyday behavioranalysis. Monterey, CA: Brooks/Cole.

Mowrer, 0. H. (1950). Learning theory and personalitydynamics. New York: Ronald Press.

Peterson, M. E. (1978). The Midwestern Associationof Behavior Analysis: Past, present, future. BehaviorAnalyst, 1, 3-15.

Skinner, B. F. (1931). The concept of the reflex inthe description of behavior. Journal of General Psy-chology, 5, 427-458.

Skinner, B. F. (1938). The behavior of organisms: Anexperimental analysis. New York: Appleton-Century.

Skinner, B. F. (1947). Experimental psychology. InW. Dennis, B. F. Skinner, R. R. Sears, E. L. Kelly,C. Rogers, J. C. Flanagan, C. T. Morgan, & R.Likert, Current trends in psychology (pp. 16-49). Pitts-burgh, PA: University of Pittsburgh Press.

Skinner, B. F. (1950). Are theories of learning nec-essary? Psychological Review, 57, 193-216.

Skinner, B. F. (1953). Science and human behavior. NewYork: Macmillan.

Skinner, B. F. (1957). Verbal behavior. New York:Appleton-Century-Crofts.

Skinner, B. F. (1961). Cumulative record (2nd ed.).New York: Appleton-Century-Crofts.

Skinner, B. F. (1979). The shaping of a behaviorist: Parttwo of an autobiography. New York: Knopf.

Smoke, K. L. (1932). An objective study of conceptformation. Psychological Monographs, 42 (4, WholeNo. 191).

Tolman, E. C. (1932). Purposive behavior in animalsand men. New York: Appleton-Century-Crofts.

Ullmann, L. P., & Krasner, L. (Eds.). (1965). Casestudies in behavior modification. New York: Holt, Rine-hart & Winston.

Ulrich, R., Stachnik, T., & Mabry, J. (Eds.). (1966).Control ofhuman behavior. Glenview, IL: Scott, Fores-man.

Watson, J. B. (1924). Behaviorism. New York: People'sInstitute.