Pain and Behaviour (Rachlin, 1985)

THE BEHAVIORAL AND BRAIN SCIENCESPrinted in the United States of America

1985) 8, 43-83

Pain and behavior

Howard RachlinDepartment of Psychology, Sta.o University of New York at Stony Brook,Stony Brook, N.Y. 11794

Abstract: There seem to be two kinds of pain: fundamental "sensory" pain, the intensity of which is a direct function of the intensity ofvarious pain stimuli, and "psychological" pain, the intensity of which is highly modifiable by such factors as hypnotism, placebos, andthe sociocultural setting in which the stimulus occurs.

Physiological, cognitive, and behavioral theories of pain each have their own view of the nature of the two kinds of pain. Accordingto physiological theory and cognitive theory, "psychological" pain and "sensory" pain are both internal processes, with the formerinfluencing the latter as central processes influence peripheral processes. According to behavioral theory, "sensory" pain is a reflex (arespondent) while "psychological" pain is an instrumental act (an operant). Behavioral theory claims that neither kind of pain is aninternal process - that both are overt behaviors.

Although both physiological theory and cognitive theory agree with common sense that pain is internal, they disagree withcommonsense intuitions at other points. They are no better at explaining the subjective experience of pain than is behavioral theory.They have not generated treatments for pain that are superior to those generated by behavioral theory. There is no basis for thefrequent claim by antibehaviorist philosophers and psychologists that behaviorism, because it cannot explain pain, is less capable ofexplaining mental phenomena than physiology or cognition.

Keywords: behavioral psychology; cognitive psychology; mental events; operant psychology; pain; physiological psychology;psychophysics

The purpose of this paper is to describe three theoreticalapproaches to the study of pain - physiological, cognitiveand behavioral - and to evaluate each theoretical ap-proach in terms of its conformity with experimentalobservation, the effectiveness of clinical treatment devel-oped from it, and its position in the current philosophicaldebate about pain.

Modifiability of pain

One aspect of pain that all theories have to account for isits modifiability by "psychological" factors. This wasmade clear by Beecher's (1956) classic observations ofsoldiers' responses to their wounds at Anzio during WorldWar II. The soldiers were not in severe pain, and most ofthem did not request narcotics even though they wereavailable. Beecher contrasted this behavior with that ofcivilian hospital patients who had similar wounds. Thecivilians were in severe pain, and most of them didrequest narcotics.

The painfulness of a stimulus can be increased ordecreased over wide ranges by such factors as placebos(McGlashen, Evans & Orne 1969), hypnosis (Hilgard &Hilgard 1975), acupuncture (Melzack 1973a), and so-ciocultural factors (Tursky & Sternbach 1967). Thesefactors are often so strong that they completely overridethe presence or absence of normally painful stimuli: "Itcan be said with certainty that psychological factors quiteoften cause pain and frequently augment its severity.They may also serve to abate or abolish it even in thepresence of extensive trauma" (Mersky 1968). In thetreatment of severe pain the success of hypnotism (Hi-lgard & Hilgard 1975) has been as good as or better than

the success of surgery (Melzack 1973b) - and this seemseven more true with respect to lasting than to temporarysuccess!

While it is clear that pain is highly modifiable byvarious "psychological" factors, pain is not normally elim-inated by those factors. Beecher's soldiers, for instance,did not report that their pain had disappeared. Mostsoldiers do feel pain, and athletes with injuries oftenclaim that they play despite pain, not without pain. Inthese instances people commonly say that the pain existsbut does not bother them. For instance, Hilgard andHilgard (1975) report a "hidden observer" within subjectshypnotized to be analgesic. One subject said that "whileher hypnotized self had felt no pain, the hidden part hadfelt pain of about the same sensory intensity as thatproduced by the cold water [her arm had been sub-merged in ice water] without hypnosis. However, thecovert pain bothered her much less - at this hidden levelwithin analgesia - than overt pain bothered her in thenormal waking state. " Thus, there seem to be at least twocomponents of pain: "sensory" pain, which is a responseto a given stimulus and is relieved by the removal of thatstimulus, and "psychological"' pain, which can be modi-fied by external contextual events such as war or anintense sport. Certain procedures, like hypnotism, seemable to alter "psychological" pain (the aversiveness) butleave the underlying pain (the sensation) unaffected. It isnot clear at the present time whether narcotic drugs(morphine, for instance) affect one component of pain oranother or both.

Reports of the effect on pain of narcotics, biofeedback,acupuncture, and other treatments often follow this pat-tern - that the pain is there, is intense, but is notaversive. Stories of Lawrence of Arabia, G. Gordon

© 1985 Cambridge University Press 0140-525X185/010043-41 IS06.00 43

Kachlin: Pain and behavior

Liddy, Buddhist monks, and others training themselvesto endure pain all seem to emphasize the separation ofpain as a sensation (which remains) and pain as an aver-sive, terrible thing (which has somehow vanished). Turk,Meichenbaum, and Genest (1983) quote Freud's descrip-tion of the cancer in his jaw as "a little island of painfloating in a sea of indifference."

The word "psychological" is in quotes because "psy-chological" is used often in opposition to "physiological"or "behavioral," and no such opposition is intended here.Each of the three theories of pain to be considered -physiological, cognitive, and behavioral - has, in itsmodern form, an explanation of both components of pain,and each explains both components wholly within its ownterms. A physiological theory of pain contains a physio-logical explanation of both "psychological" pain and "sen-sory" pain; cognitive theory has a cognitive explanation ofboth components of pain, as does behavioral theory.

Another reason for putting "psychological" in quotes isto deny its opposition to "real." "Psychological" (aver-sive) pain is no less real and may be much more importantin pain treatment than "sensory" pain. It is by its aver-siveness, not by its sensory quality, that pain causesdisability. Development of a treatment by which themillions of chronic pain sufferers could view their pain asan island in a sea of indifference would be a vast step inpain treatment, a step that will be more difficult to take aslong as "psychological" pain is seen as false pain.

Psychophysics of pain

The separation of aversive and sensory pain is reflected inthe words that people use to describe it. The mostcommonly used pain assessment technique is the McGillpain questionnaire developed by Melzack and his co-workers (described in Melzack 1983). A similar question-naire that uses somewhat more sophisticated measure-ment techniques (ratio rather than interval scaling) hasbeen developed by Tursky, Jamner, and Friedman(1982). They have shown that after identifying the qualityof pain with words such as "stinging," "grinding, "shoot-ing," and "throbbing" there remain two separate sets ofwords by which people describe the degree of their pain.One set of words has to do with the intensity of the pain asa sensation. In this set there are fourteen words rangingfrom "just noticeable" through "mild," "uncomfortable,""strong," and "severe" to "excruciating" at the upperend. Tursky et al. have found that people are capable ofassigning line lengths and numbers to these words thatare consistent from person to person and time to time.Furthermore, the assignment of numbers, line lengths,and the words themselves to various intensities of electricshock is also consistent from person to person and time totime. Thus a pain-intensity scale that ranges from a scoreof 8 units for just noticeable pain to 227 units for excruciat-ing pain has been formed.

The second scale of pain is called a reaction scale byTursky et al. The words in this scale range from "beara-ble" through "uncomfortable," "distressing," "awful,"and "intolerable" to "agonizing" at the upper end. Num-bers are assigned to the words in the reaction scale by thesame methods used with the intensity scale. The num-bers of this scale range from 23 units for bearable pain to

153 units for agonizing pain. (As with any ratio scale, theunits of the intensity and reaction scales are arbitrary. It isthe ratio of the numbers rather than their absolute valuesthat provides the test for consistency.) While measures onthe intensity scale seem to remain constant across thepopulation of speakers of English, reaction measuresvary. For instance.. Blanchard, Andrasik, Arena, andTeders (n.d.) found that while migraine headache suf-ferers did not differ from nonsufferers in the numbersthey assigned to words of the intensity scale, they diddiffer in the numbers they assigned to words of thereaction scale, assigning proportionately higher numbersthan normal to the higher words. Similarly, Elmore(1979) found that biofeedback treatment for pain alteredpatients' reaction measures (proportionately reducing thenumbers assigned to the higher words) while leavingtheir intensity measures constant. Tursky et al. (1982)found the same result with a combination of behavioraland cognitive treatments of chronic pain sufferers. Theyalso found that morphine altered reaction measures whileleaving intensity measures constant but that aspirin al-tered both intensity and reaction measures. The intensityand reaction scales are separable and seem to correspondto the two components of pain discussed previously. Theintensity scale seems to measure the sensory componentof pain, while the reaction scale seems to measure themore malleable, aversive, "psychological" component ofpain. Whether this convenient division of words will holdup under future psychophysical research is still far fromcertain, however.

Assuming that Tursky et al.'s division of words intointensity and reaction scales is valid, the object of clinicaltreatment would be to vary the reactive component ofpain under conditions where the intensity component isdifficult or impossible to vary. If the intensity componentis seen as a measure of the pain stimulus itself (the "islandof pain"), then the object becomes to vary the reactivecomponent (the "sea of indifference"). Each of the threedifferent theories of pain that we will consider - physio-logical, cognitive, and behavioral - provides a differentperspective on the nature of the reactive component.

Physiological theory and treatment

Pain does not seem to be a simple sensation. Aristotolcclassified pain, not among the five primary senses, but asa passion of the soul that resulted from sensation. Accord-ing to Aristotle, if a pattern of sensation is out of harmonywith a person's psyche it will cause pain, hence the desireto avoid the source of the sensation. Thus, Arisotleemphasized the aversive "psychological" component ofpain. The discovery early in this century of various kindsof sensory receptors in the skin led to considering pain asimple sensation. Pain is identified as the sensation re-sulting from patterns of stimulation of the free nerveendings (Dallenbach 1939).

One problem with this viewpoint is stated by Melzackand Wall (1970): "To call a receptor a pain receptor . . . isa psychological assumption: it implies a direct connectionfrom the receptor to a brain center where pain is felt, sothat stimulation of the receptor must always elicit painand only the sensation of pain. The facts of physiologicalspecialization provide the power of specificity theory

44 THE BEHAVIORAL AND BRAIN SCIENCES (1985) 8:1

Rachlin: Pain and behavior

[that pain is a specific sensation]; its psychological as-sumption is its weakness. '

Melzack and Wall (1965) proposed an alternative thatthey call a "gate" theory, according to which pain is aperception rather than a sensation. That is, there exists anidentifiable sensation of pain, but it is rarely, if ever,purely felt. Rather, it is modified by input from variousother sources. Gate theory and other current physiologi-cal theories of pain consist essentially of a hypotheticalmechanism by which the modification occurs. Gate theo-ry says that large and small cutaneous fibers interact at anearly stage in the nervous system, at "gate cells" in thespinal cord, which in turn regulate the transmission ofpain to both the motivational and the sensory systems.According to the theory, these gate cells themselves areinfluenced by central processes. There is no need todescribe the physiological details of the gate controltheory here because the current consensus seems to bethat, in its physiological details, it is probably incorrect(Kelly 1981). The theory is nevertheless important be-cause it "reversed the historical research emphasis uponpain as solely an afferent sensory experience." Kelly adds,"Pain also disrupts ongoing behavior, demands immedi-ate attention, and serves as a primary negative reinforcerin a variety of situations. It suppresses behavior whenmade contingent upon it and supports a broad repertoireof avoidance and escape responses. To emphasize onlythe sensory features of pain in the study of its neural basesand to ignore its unique affective and motivational prop-erties is to confront only part of the problem."

The Melzack—Wall theory, by its postulation of theinteraction of large and small neurons at a relativelyperipheral level, implies that pain can be inhibited at thatlevel by touch. Stimulation of large-diameter fibers(touch) can, according to the theory, close the gate. Thisfeature of gate control theory has provided some physio-logical support for acupuncture as a clinical technique andhas given rise to some treatments by which cutaneouselectrical stimulation near the site of the pain have suc-cessfully inhibited the pain. But the success of acu-puncture at sites far removed from the pain and thesuccess of electrical stimulation on at least one occasionwhen a patient forgot to hook up his battery (Kelly 1981)indicate that pain inhibition by cutaneous stimulation canalso occur at a more central level. How the central processmight work is still unknown.

More recent physiological theory centers around theaction of opiates. Opiate receptors have been discoveredon neurons distributed throughout the brain, includingthe midbrain. Chemical stimulation of the latter site withmorphine results in analgesia. It has also been discoveredthat certain substances with pharmacological propertiesthat resemble those of morphine are produced in thenervous system and that these substances cause anal-gesia, among other effects, when they are released. Someof these naturally occurring substances, called en-dorphins, are much more potent than morphine. This hasgiven rise to the hope that the administration of en-dorphins to pain sufferers would alleviate pain withoutthe problems associated with morphine. "Unfortunate-ly," according to Kelly, "chronic administration of (3-endorphin produces progressively weaker analgesic ef-fects (tolerance) and also gives rise to withdrawal signscomparable to those of morphine (dependence)."1

The status of surgery in the relief of pain is summarizedby comments of Weisenberg (1975): "No matter whattechnique is used, the percentage of failures is signifi-cant. . . . Surgery must be done over a large area toproduce longer-lasting pain relief. However, the largerthe area cut, the more other functions, such as bladdercontrol and strength in walking, tend to be lost. . . .Frontal lobotomy or leucotomy abolishes the aversivefeelings associated with pain. That is, the patient feels thepain, but it does not bother him. [But!] The majorproblem associated with this type of surgery is the changein personality - the patient can become an emotionalvegetable."

In summary, physiological theory and treatmentsbased on it need to address themselves to the strong"psychological" or, in physiological terms, the strong"central" influence on pain. "Psychological" pain canoccur with no apparent stimulus whatsoever. The prom-ise of physiology is that when the factors (in the environ-ment) that normally control "psychological" pain arediscovered it may be possible to bypass those factors andproduce analgesia directly, selectively, nonaddictively,and without side effects.

Cognitive theory and treatment

Before discussing cognitive theories and treatments ofpain it is necessary, because of the current state of flux incognitive psychology, to define cognitive theory. We willconsider any theory cognitive that is physiologically non-committal and postulates functional elements inside theorganism mediating between environment and behavior.By this definition, Melzack and Wall's (1970) conceptualmodel of pain is a cognitive theory. The fact that someelements of this model are presently reducible to physio-logical mechanisms and that all elements of the model arepotentially reducible to physiological mechanisms is notan unusual feature of cognitive theories.2 Another, muchmore complicated, cognitive theory of pain was proposed(perhaps not seriously) by the philosopher Daniel Den-nett (1978).3 In Dennett's model the Melzaek-Wall gateis just one internal element, interacting with such otherelements as "ratiocination," "belief," "desire, " "reticularformation," "perceptual analysis," and many more. Un-fortunately, except for the physiological research and thetreatment based on the physiological aspects of Melzackand Wall's theory, these cognitive theories of pain havenot been tested by experiments, nor has any treatmentbeen devised on their basis.

Both research and treatment, however, have beenbased on simpler but unformulated cognitive models.One simple model assumes that "psychological" pain maybe mediated by imagery so that certain images willattenuate pain while other images will magnify it (Turk,Meichenbaum & Genest 1983). Subjects are asked topractice creating images, for instance, "imagining theaffected area's being numbed with Novocain or seeingoneself as a television character such as the Six MillionDollar Man or the Bionic Woman, with mechanical limbsinsensitive to pain." According to the theory, images maywork directly on pain, as with the Bionic Woman tech-nique, or indirectly through a mechanism of attention, aswhen the patient is told to imagine a pleasant day at the

THE BEHAVIORAL AND BRAIN SCIENCES (1985) 8:1 45


beach or attending a party. In either case, pain therapybased on images has not been successful as a treatment byitself. Turk et al. (1983), after a careful analysis of extantlaboratory studies, concluded, "The data . . . do notconvincingly establish the efficacy of any cognitive copingstrategy relative to the strategies that subjects bring toexperiments, nor is there sufficient evidence to supportthe use of any one strategy compared to any other."According to Turk et al., results of clinical studies withimagery have been more positive than results of laborato-ry studies, but the problem with clinical studies is that inthe clinic treatments based on physiological and behav-ioral theories are often used in conjunction with cognitivetechniques, so when pain is reduced it is not clear whichtreatment or which combination of treatments waseffective.

Even if treatments based on imagery were shown to beineffective, imagery and pain may still interact. It may bethat certain images reduce pain but that current tech-niques of imagery training do not work. A frequently usedmethod is first to teach subjects relaxation techniques andthen to suggest the images to the relaxed subject, butthere is no evidence that the imagery treatment has anyeffect beyond whatever reduction of pain is due to relaxa-tion itself.

Some evidence that cognitions (images or not) play arole in pain tolerance comes from studies reported byTurk et al. (1983) in which female subjects immersedtheir arms in ice-cold water:

The subjects naturally clustered into two clear groupsaccording to their tolerance time. In each sample . . .the distribution tends to be bimodal, with tolerancesgenerally either less than 100 seconds or approximately300 seconds (the ceiling employed by the experiment-ers). . . . Those in the high-tolerance group seemed tofeel that they could use [imagery] strategies to affectboth the pain and their power to persevere despite thepain, whereas those in the low-tolerance group usedstrategies with less conviction of their usefulness andwith less sense of their own ability to influence theirsituation other than by removing their hands from thewater.Another cognitive treatment for pain in common use is

called cognitive restructuring. Cognitive restructuring isdirected not at a person's images but at a person's beliefs.Beliefs are said to mediate between a painful stimulus anda response. Turk et al. identify a large proportion of non-pain-tolerant subjects as "catastrophizers. ' Citing theresults of a study in which dental patients described theirfeelings, they say:

Catastrophizing took several forms, including negativeself-statements regarding their competence, anxiety-arousing thoughts, and extremely aversive images. Toillustrate, one patient stated, "How I hated it. I hatehaving injections. I think, 'Oh, no, here we go again.' Ihate it with a passion. Just to see that great big needlecoming down at you, the next thing you know you startgoing bananas. I just can't hack it."

On the other hand, one of the pain-tolerant arm-immer-sion subjects described her thoughts this way:

I sat here and thought, "This isn't happening to me. I'msomewhere else. It's not my arm; it's just an arm sittingin the water. I can go on as long as this has to go on.". . . Every time I was ready to give up I thought, "No,

it's not killing me because it's not my arm. It's just anarm in the water."This report, like reports of hypnotized subjects, de-

scribes a dissociation between the person and the pain.Such subjects, one guesses, would describe their painwith words high on the intensity scale but low on thereaction scale, for instance, severe (132) but tolerable(23).

The essence of cognitive theory is that images andbeliefs can be changed directly. Currently there exist noexplicit cognitive methods of instilling images or beliefsexcept telling the subject what to imagine or what tobelieve. But there is no evidence that telling a personwhat to believe will change that person's beliefs. Ittherefore behooves the cognitive theorist to developmethods by which cognitive units may be manipulated.In other words, more attention must be paid to the inputand output of cognitive mechanisms before the cognitiveunits in the middle can be meaningful. There is currentlya wide chasm between cognitive theory of pain andcognitive therapy for pain.

The current state of cognitive theory with regard topain is that images and beliefs are said to mediate be-tween painful stimuli and the aversiveness of pain. Thefact that pain tolerance seems to go along with certainimages and certain beliefs is some evidence for suchmediation. But there is as yet no way to tell whetherimages and beliefs mediate between the environmentand pain tolerance, whether they are the result ratherthan the cause of pain tolerance, or whether they are theresult of some other set of events that also, independent-ly, causes pain tolerance. What those other events mightbe is suggested by the behavioral theory of pain, to whichwe turn next.4

Behavioral theory and treatment

The current behavioral model of pain, like the imagery-based and belief-based cognitive models, has arisen fromclinical practice. Wilbert E. Fordyce (1976) has devel-oped a behavioral treatment for chronic pain that, accord-ing to Turk et al. s (1983) analysis of outcome studies, is atleast as successful as any other treatment and moresuccessful than most others. In the effort to explain andrationalize the treatment he has developed, Fordyce(1978) presents what is essentially the only extant behav-ioral theory of pain. It identifies "sensory" pain withrespondent behavior and "psychological" pain with oper-ant behavior. The theory has wide implications for psy-chology. To understand those implications it is necessary,first, to distinguish behavioral theories in general fromphysiological and cognitive theories; second, to under-stand the difference between operant and respondentbehavior; and third, to understand the relation betweenoperants and respondents. We will take up these threequestions as they apply to behavior in general and as theyapply to pain.

The essential difference between modern behavioraltheories and cognitive or physiological theories is that,according to behavioral theories, whatever actions anorganism takes are actions of the whole organism. Forinstance, a rat's bar press is considered to be an action ofthe whole rat, not its paw or its nervous system or some

4 6 THE BEHAVIORAL AND BRAIN SCIENCES (1985) 8:1


functionally defined internal mechanism. Otherwise be-havioral theory becomes indistinguishable from physio-logical or cognitive theory.

Of course, it is still possible to redefine "the wholeorganism" as something other than the organs enclosedwithin the skin. One might consider an artificial arm aspart of a person, or a diseased tooth or a cancer as astimulus to, rather than part of, the person. Whethersuch behavioral laws as have been discovered applyunder one or another redefinition is an empirical ques-tion; behaviorism is nonetheless distinguishable fromcognitive and physiological theory by its confinement intheory and practice to overt behavior of wholeorganisms.5

Recent criticisms of behaviorism (for instance, Savin1980) have claimed that this confinement to overt, whole-organism responses prevents behavioral theory from ex-plaining the very processes of most interest to psychology- people's mental lives. In philosophical argumentsagainst behaviorism, "pain" has often been used as anexample of a mental term not amenable to behavioralanalysis. Thus, the explanatory power of Fordyce's theoryand its clinical usefulness are important issues for psycho-logy.

In his discussion of stimulus and response as genericconcepts Skinner (1935) defined a reflex in behavioral asopposed to physiological terms as a certain relationshipbetween a set of environmental events and a set ofbehavioral events; for instance, the intensity or rate of thestimulus is related by certain sets of functions to theintensity or rate of the response. Each set of environmen-tal and behavioral events, related to each other in theseways, was considered a single reflex, regardless of ana-tomical connections. A response that was part of a reflexcould be controlled by manipulation of its stimulus.Pavlov's (1927) classical conditioning could, with more orless difficulty, describe the alteration of these reflexes.Unlike Pavlov, however, Skinner (1938) believed that notall behavior was reflexive in this way. He distinguishedbetween this kind of reflex, which he called a respondent,and another kind of reflex, which he called an operant.Just as it might be said that a given respondent is definedin terms of the environmental events that precede it, agiven operant is defined in terms of environmental eventsthat follow it. Like the definition of "respondent," thedefinition of "operant" was generic, not anatomical, and itwas behavioral in the sense that an operant was consid-ered to be behavior of the whole organism.

According to Skinner, the environmental events thatdefine respondents occur just prior to behavior while theenvironmental events that define operants occur justsubsequent to behavior. Thus, for Skinner (1938), tem-poral contiguity was not only important in respondentand operant conditioning but a crucial element of thedefinition of respondents and operants themselves. Cur-rent behavioral theory (Baum 1973; Catania 1971; Gib-bon, Berryman & Thompson 1974; Herrnstein 1970;Maier, Seligman & Solomon 1969; Rachlin 1976; 1978;Rescorla 1967; Staddon 1973; 1980) differs from Skinnerwith respect to the role played by temporal contiguity.While temporal contiguity is unquestionably importantin operant and respondent conditioning, it is not nowgenerally considered a necessary part of the definition of aresponse. Post-Skinnerian behaviorism defines a re-

spondent or operant in terms of temporal correlationrather than temporal contiguity between environmentand behavior. In this sense current behavioral theory ismore molar than Skinnerian theory. (For a molar behav-ioral theory, fundamental processes can occur only oversome significant temporal interval.)

Correlated events may be separated in time. At theextreme, events of a person's early childhood may corre-late with or form a part of the same behavioral pattern asevents of that person's adult life. Although indisputablythere are mechanisms in the nervous system that bridgethe earlier and later events - mechanisms that may bedescribable in physiological or cognitive terms - modern(molar) behaviorism is concerned only with the correla-tions themselves. These correlations are the stuff ofbehavioral laws. It is by widening the scope of the searchfor correlative relationships among environmental eventsand overt behavior further and further back into the past(rather than deeper and deeper into the person) thatmodern behaviorism hopes to explain people's mentallives. For example, the difference between a person withself-control and a person without self-control is seen bybehavioral theory, not in terms of the operation of inter-nal physiological or cognitive mechanisms, but in terms ofthe temporal extent of the environmental variables thatcontrol behavior. The difference between an overweightperson who accepts a sandwich offered between mealsand an overweight person similarly tempted who refusesthat sandwich is seen by behavioral theory as a differencebetween a person whose behavior was controlled, in thatinstance, by immediate reward and a person whose be-havior was controlled, in that instance, by rewards moredistant in time.6

Fordyce (1978) considers operant pain to have evolvedfrom respondent pain. At first pain is correlated withsome antecedent stimulus (a wound, for instance). Aslong as the correlation is in effect, the pain is respondentbehavior. But then, according to Fordyce, "If the chronicpain problem, and the person who has it, exists in anenvironment which indeed delivers effective pain con-tingent reinforcement, a problem of operant pain mayevolve." Here Fordyce is discussing how a particular caseof chronic pain may evolve from a particular case of acutepain. But this theory of pain evolution may be extended topain in general. An infant's crying may at first be entirelydependent on specific stimuli. For an infant, the specifici-ty theory of pain may be essentially correct. But aninfant's crying almost always has immediate conse-quences in terms of parental care and attention, and thereis considerable evidence (Etzel & Gewirtz 1967) thatthese consequences, in turn, affect the frequency andnature of an infant's cries. According to Teitelbaum(1977), the action of the higher levels of the brain involvedin development (and in recovery from brain damage)accompany the development of operant behavior frominitial respondent behavior. Even the most prototypicaloperant response, a pigeon's key peck, has respondentcomponents, in the sense that the nature of the key peckdepends partly on its antecedents and partly on its conse-quences. And this is true with regard to both the topogra-phy of the individual key peck and the pattern of peckingover time (Schwartz & Gamzu 1977).

According to Fordyce, instances of pain in adults nor-mally have both respondent and operant components.



But chronic pain, where physiological causes are appar-ently absent, is operant pain that has, so to speak, lost itsmoorings - has lost its respondent components. Treat-ment of chronic pain in Fordyce's clinic consists, first, ofdiscovering the immediate and distant reinforcers con-tingent on pain, second, either removing those reinfor-cers or breaking up their dependency on pain, and third,reinforcing what Fordyce calls "well-behavior," behaviorincompatible with pain.

It is important to emphasize the behavioral rationaliza-tion of Fordyce's treatment: A person is exhibiting painbehavior. We fail to find any stimulus (external or inter-nal) causing this behavior. (We then call this pain "psy-chological" pain.) Our usual, intuitive supposition in thissituation is that, despite our failure to find a stimulus, aninternal stimulus does exist - the pain itself. Physiologicaland cognitive definitions of "psychological" pain are theo-ries about what sort of internal event the pain itselfcorresponds to. Fordyce's supposition, suggested by be-havioral theory, is that "psychological" pain is operantbehavior and thus there must have existed in the pastevents contingent on this behavior that reinforced it. Inaccordance with modern behavioristic practice, theseenvironmental events need not have been contiguouswith pain behavior but may have occurred at times distantfrom a given instance of the behavior, so long as they werecorrelated with the behavior. When one alters perspec-tive and begins to look outside rather than into theorganism for the causes of "psychological" pain, it turnsout that they are not difficult to discover. The first andmost obvious consequence of pain is access to pain medi-cation, particularly narcotics. Thus, Fordyce removes thedependency of pain medication on pain behavior and putsthe medication on a strict time schedule; then he gradu-ally reduces its strength. Another consequence of pain isusually reduction of physical activity; so an exerciseprogram is instituted. Beyond this, reinforcement isprovided by family and friends in the form of attentionand concern. Often attention and concern habituate, sothat ever-increasing demonstrations of pain are necessaryto keep them going. A vicious circle develops within afamily that Fordyce's treatment endeavors to break. For-dyce also considers whether the dependency of the painsufferer on other family members is reinforcing for them- that is, they may in some sense prefer the pain suffererto be in pain than to be well. Pain is also a way (albeitextreme) to avoid social contact and social obligations.Pain may be a way to avoid temptation of various kinds,such as sexual temptation.

Fordyce's treatment (the details of which are given inFordyce 1976) sometimes gradually eliminates the rein-forcer altogether, as with narcotic medication, but moreoften attempts to help the patient obtain the reinforcer bymethods other than pain behavior while simultanouslymaintaining and increasing reward contingent on wellbehavior. These rewards are likely to differ from patientto patient, so some sort of behavioral analysis is necessaryto fit the treatment to the patient's needs. The behavioralanalysis is, essentially, a search for the consequences ofpain. For all patients taken on by Fordyce, there hasalready been a search for and treatment of the anteced-ents of the pain, but pain is still present. In other words,Fordyce does not begin to treat operant pain until re-spondent pain has been treated.

The treatment also attempts to make the relation be-tween behavior and its consequences more vivid. Be-cause events correlated with pain behavior are oftentemporally distant from the pain behavior itself, they maysupport that behavior without the person's awarenessthat they are doing so. A person is not usually unaware ofan individual instance of pain behavior but may be un-aware of its frequency or intensity as perceived (andreinforced) by others. Thus, Fordyce's treatment empha-sizes record keeping, both of pain behavior itself and ofthe rewards contingent on it.

As Turk et al. (1983) indicate, this procedure seems towork. But, one may ask, what does it accomplish evenwhen successful? A chronic pain patient, initially dis-abled, who now works steadily and maintains a normalsocial life may be said to do those things despite pain. Butone might consider the degree of normal activity itself thebest possible measure of the aversive, if not the sensory,component of pain, and it is the aversive component, the"psychological" component, at which Fordyce's treat-ment is aimed. It would be instructive to apply a psycho-physical scaling procedure that attempts separately tomeasure "sensory" and "psychological" aspects of pain inFordyce's patients before and after treatment. Unfortu-nately, this has not yet been done.

In summary, the clinical observation of two compo-nents of pain, "sensory" and "psychological," seems to becaptured by the intensity and reaction scales of painpsychophysics. Any theory of pain (physiological, cog-nitive, or behavioral) must account in some way for thisdichotomy. Physiological and cognitive theories wouldconsider both components of pain to be the operation ofinternal mechanisms, with sensory pain as a lower-levelor relatively peripheral process and "psychological" paina higher-level or relatively central process. Differentphysiological and cognitive theories would hypothesizedifferent kinds of internal interaction between the twolevels of pain. Behavioral theory considers both sensoryand "psychological" pain to be overt behavior, withsensory pain as respondent behavior and "psychological"pain as operant behavior. For the physiologist or thecognitivist (as well as the mentalist) a baby's crying is amessenger, an ambassador, an indication, or a symptomof pain. For the behaviorist, the crying (plus the kicking,the facial expression, and other overt activities) is itselfthe pain.

What is behind pain behavior?

Fordyce, quite wisely, ignores the question of the sub-strates of pain behavior because his main interest is intreatment, not philosophical dispute; but the question isworth considering here because it is exactly on this issuethat philosophers have claimed behaviorism to be invalid.The point was put quite baldly (and personally) by Searle(1980): "Are there no pains underlying Rachlin's painbehavior? For my own case I must confess that thereunfortunately often are pains underlying my pain behav-ior, and I therefore conclude that Rachlin's form ofbehaviorism is not generally true." Even Wittgenstein(1958, p. 102) seems to reject the equation of pain andbehavior: " 'But you will surely admit that there is adifference between pain-behavior accompanied by painand pain-behavior without any pain?' What greater dif-



fere nee could there be?" Some philosophers, I fear,might be tempted to try to refute a behaviorist the wayJohnson supposedly refuted Berkeley - but by kickingthe behaviorist rather than the stone.

One interpretation of this antibehaviorist argumentwould make it an antiphysiological and anticognitiveargument as well. According to this interpretation, painsare fundamental mental events, purely subjective, per-sonal "raw feels" that cannot be reduced to or explainedby anything else. Let us call this the "pure mentalist"argument. Note that a physiological or cognitive theoryof pain has as much (or as little) trouble with the purementalist argument as a behavioral theory. Searle (1980)believes all mental events to be identifiable with physio-logical events in the brain. For him, pain is just aphysiological event. A pure mentalist might say to Searle,"Are there no pains underlying the neural events inSearle's cortex?" For my own case I must confess thatthere unfortunately often are pains underlying my ownneural events, and I therefore conclude that Searle's formof physiological reductionism is not generally true. Simi-larly (as Dennett himself anticipates), a pure mentalistmight say, "Are there no pains behind the operation ofyour computer mechanism?"

There is, of course, no way to answer an argument thatasserts its truth as a fundamental axiom except to showthat another axiom might lead to more useful beliefs.Thus, the best answer to the pure mentalist is the con-tinued development of physiological, cognitive, and be-havioral theories of pain.

Yet the pure mentalist cannot just be dismissed, be-cause pure mentalism, at least as regards pain, is part ofthe common sense of our time. It would be difficult, itseems, for any psychological theory to try to alter afundamental common belief without showing why thatfundamental belief is so common. Thus, it is up to thepsychological theorist, of whatever orientation, not onlyto provide a more useful alternative to pure mentalismbut also to explain why the pure mentalistic view of painhas become a fundamental belief in the first place. Weleave it for the physiological and cognitive theorist to dothis in terms consistent with physiological and cognitivetheories and only attempt here to explain briefly, inbehavioristic terms, why the pure mentalist theory ofpain is so compelling.

For behavioral theory it is enough to show that a beliefis useful (to explain why it is generally held) and to showwhy another theory, if adopted, would be still moreuseful (to explain why the generally held theory should begiven up). In what way, then, is the pure mentalist view ofpain a useful one?

Pain functions in society much as a fire alarm does. Itcommands instant attention. In a baseball game, a playerwho falls, clutching himself, stops the game. Just asfiremen must respond even though there are always acertain number of false alarms, people continue to re-spond to expressions of pain even though there may be nodiscovered tissue damage. Our expressions themselvestake on standard forms within our society so as to ensurethat response. Pain behavior, whatever else it may be, is aspecial kind of communication - like a fire alarm - thatdemands a social response first and only allows questionsto be asked later. If a person claims to be in pain, thatperson is normally allowed to be right, whether or not

tissue damage is ever found. Because pain as communica-tion works best and quickest when we do not questioneach other's pain, society gives each of us the right toclaim to be in pain regardless of any evidence to thecontrary. Thus pain is generally recognized to be a funda-mental mental event, a purely subjective, personal "rawfeel" that cannot be reduced to or explained by anythingelse.7

From the behaviorist viewpoint, however, we do notinfer the privacy of another person's pain from the factthat our own pain is private; we learn simultaneouslyabout the privacy of our pain and the pain of others. To saya pain is private is, from this point of view, only to say thateach person in our society has the right to ask for attentionand help without being questioned about it. It is useful forsociety to give its members that right, just as it is useful torespond to fire alarms or to move out of the way of anyvehicle with a siren and a flashing red light. Just as most ofus obey laws even when no one can catch us disobeyingthem, so we feel pain even when no one is there. For amolar behaviorist extreme pain is not an immediatesensation (deeply felt) but a long-term pattern of behav-ior, widely performed.

The more convincing we want to be (and the socialobject of pain is to be convincing) the less conscious (lessverbal, less deliberate) we must be about what we aredoing, and the wider we have to spread the pain behaviorin time. To be most convincing we have to infuse thepattern of pain into all our overt behavior, public andprivate.

Thus, the pure mentalist view of pain is useful. Itconfers a sort of privilege on those organisms (human andotherwise) who, by their expression of pain, automaticallysummon our help or at least our sympathy. The problemwith that view, however, is that the privilege may beabused. Children abuse it frequently. The story of theboy who cried wolf in its various forms is an attempt toinstill nonabuse. But the temptation to abuse the priv-ilege is always with all of us. What is wrong with this? Toomany of us are in pain too much of the time. According toKoenig (1973), the average number of aspirins alone (notcounting other pain medication) swallowed each year byeach American (man, woman, and child) is 225. There arecurrently 900 pain clinics in the United States, and thenumber is rapidly growing (Turk et al. 1983). A viewpointon pain other than the pure mentalist view of currentcommon sense might help us to better discriminate"sensory" pain from "psychological" pain and to makethis discrimination of our pain as well as that of others.Whether the best viewpoint for this purpose is physiolog-ical, cognitive, or behavioral remains to be seen.

We turn now to a stronger objection to the behavioralpoint of view - one that arises not from commonsensementalism alone but from commonsense mentalism incombination with the physiological and cognitive views.That objection says that pain, whatever it is, is fundamen-tally internal. The behavioral viewpoint stands aloneagainst the others in where it sees pain. For the othertheories pain occurs inside the organism, and pain behav-ior is only the expression of that internal pain. For astrictly behavioral theory pain occurs as overt behavior, atthe point of interaction between the organism and theenvironment.

It is important to emphasize again that a behavioral



theory cannot internalize its terms and remain a behav-ioral theory. In their desire to expand their theories toexplain mental terms, the three great behaviorists, Wat-son, Hull, and Skinner, have postulated the internaloperation of functional entities originally defined as ac-tions of the whole organism. But internal functions, eveninternal operants, are physiological or cognitive con-cepts, not behavioral ones. The question we ask here iswhether a purely behavioral theory of pain makes sense.

A way in which behaviorism may be extended toaccount for mental terms and still remain behavioristic isto consider instances of overt behavior that have occurredin an organism's past as part of a single pattern of behaviorextending into the present. Even molecular behavioraltheories do this, in miniature, when they consider tem-porally extended events such as response rates or inter-response times as fundamental behavioral variables.8

Modern molar behavioral theory extends this procedureto wider intervals, encompassing events that are far in thepast.

In physiological and cognitive theory, by contrast, theevents defining pain occur inside the organism. Althoughthese events are ordinarily supposed to cause behavior,through "motor mechanisms," it is within the power ofphysiological and cognitive theories to suppose that al-though the internal events do occur, the behavior isinhibited. According to those theories, an individualinstance of pain may never be revealed in behavior. Ifinternal pain does exist in this sense, behaviorism cannotexplain it. It is on the grounds of this restriction thatbehaviorism has been most strongly attacked - first withrespect to its failure to explain internal pain, and, byextension, with respect to its failure to explain othermental processes.

It is worthwhile, therefore, to discuss this kind of attackand to attempt to provide a behaviorist answer to it. If thisattempt succeeds, the way will be clear to further developa wholly behavioral theory of pain.

The essence of the antibehaviorist argument is that it ispossible to conceive of a person in pain who does notexhibit pain behavior. The argument has been made mostvividly by Putnam (1980), who asks us to imagine acommunity of super-Spartans who are trained from birthto inhibit all pain behavior except that which is verbaland, even when they say they are in pain, to say it in acalm voice. Using our previous terminology, super-Spar-tans, by virtue of childhood training, have learned tomold their responses to painful stimuli into a form (thewords "I have a pain in my toe" spoken in a calm voice)that is useful for functions such as telling a doctor where ithurts but not useful for functions such as stopping abaseball game (clutching and falling), warning other peo-ple of danger (yelling loudly), or avoiding tissue damage(pulling ones hand quickly from the fire). In the super-Spartan's world, "psychological" pain would not be aproblem (as it is in our world), but they would pay a pricein burned hands for their lack of it.

Super-Spartans calmly say they are in pain but do notexhibit any other pain behavior. We can certainly con-ceive of such people and imagine how they might havebeen trained. But does that first conception necessarilyentail a second and quite separate conception - thatsuper-Spartans have pain exactly when they say they do?Putnam's argument against behaviorism hinges on the

second conception. But there is a good deal of evidence(reviewed by Nisbett & Wilson 1977) that what people sayabout their internal states does not correspond to whatthe best available cognitive theories (theories that explainbehavior in a wider context) claim that their internalstates actually are. (And behavioral theories insist evenmore forcefully on this distinction. Fordyce (1983b) hasfound it necessary, in pain assessment, to distinguishsharply between what people say about their pain andwhat they do.) How should one interpret what super-Spartans say? Consider the following instance.

Let us suppose a super-Spartan warrior is struck in theshoulder by a spear. At that time he neither makes asound nor performs any action expressing pain. The nextday, visiting the super-Spartan doctor, he says, in a calmvoice, "I have an excruciating pain in my shoulder, and Ihave had it since yesterday when I was struck with aspear." The doctor may then look for tissue damage.Suppose a wound is found. Is the doctor then forced toimagine that the warrior was carrying within him, notonly the wound, but something called "pain" and that thewarrior's current statement is caused, not by the wound,but by the pain? It does not seem logically necessary forthe doctor to imagine this; nor, more importantly, wouldsuch an image be useful. The doctor (and the super-Spartan psychologist) could just imagine that the war-rior's statement is a response to his wound. After all, it isthe wound, not the pain, that needs to be treated. But letus suppose no wound is found. The doctor might thenpunish the soldier for reporting pain not confirmed bytissue damage, suggest changes in the childhood trainingof super-Spartans so as to avoid such instances in thefuture, or just shrug his shoulders and dismiss his patient.As a super-Spartan himself he would not show any sympa-thy to the warrior or give him a day off. What to us seemscruel would, to a super-Spartan, be the normal behavior,cruel or not, of his compatriots. Otherwise we would haveto conceive a kind of childhood training that compelledsuper-Spartans to inhibit pain behavior even though thatbehavior was reinforced in super-Spartan society. Ifsuper-Sparta entails suspension of the normal laws ofhuman learning as well as normal pain behavior, it be-comes harder to conceive of. Within the limitations ofchildhood training, as we know it, there is nothing aboutsuper-Spartan behavior that compels abandonment of thebehaviorist viewpoint.

Sensing this, perhaps, Putnam imagines the evolutionof super-Spartans into swper-super-Spartans. The super-super-Spartans never even say they are in pain, and theyneed no childhood training. They act this way from birth.There are numerous problems in conceiving of super-super-Spartans. First, there is no way (given Darwinianevolution) for super-super-Spartans to evolve. While it isconceivable (however unlikely) that pain behavior shouldbe so maladaptive that people exhibiting the slightesttrace of it (let us call them "crybabies") would die an earlydeath, there would be no conceivable way to distinguishthe super-super-Spartans who inhibit all pain behaviorfrom others who are completely insensitive to pain (let uscall them "numbones"). By hypothesis, neither naturenor super-Spartan kindergarten teachers could selec-tively kill off numb ones and still allow super-super-Spartan children to survive. But let us put aside thispuzzle and try our best to conceive of super-super-


Spartans. (If they could not evolve, perhaps they could bebuilt.) Suppose that a super-super-Spartan genius dis-covers the neurological substrate of the pain that thesuper-super-Spartans always inhibit. (Putnam says thatsuper-super-Spartans still have internal physiologicalpain.) To be specific, say that Dennett's cognitive theoryof pain is essentially correct and reducible to physiologybut that the connections from the pain mechanism to themotor apparatus have atrophied in super-super-Spartans.Suppose that the genius realizes what a great advantage interms of tissue-damage avoidance it would be if super-super-Spartans could communicate to one another whenthe pain mechanism (which from his point of view couldonly be a more or less accurate internal indicator of tissuedamage) was active. The genius then devises a red lightthat can be mounted on people's heads, the intensity ofwhich reveals the output of this internal mechanism. Letus say that all super-super-Spartans were fitted with thisred light from birth. Now, a super-super-Spartan baseballgame could stop if a player's red light were lit; super-super-Spartans would be sympathetic and pay lots ofattention to other super-super-Spartans whose red lightswere frequently shining; super-super-Spartans wouldhave days off from work when their red light was on; insevere cases, they would receive narcotics and otherdrugs that made them feel good; and so on. How longwould it be, do you suppose, until super-super-Spartanswould have to establish clinics to deal with excessive red-light shining? Put the overt function back in the pain andthe super-super-Spartan world would be our world, withthis difference: that the behavior we call pain, they wouldcall shining one's red light. Perhaps, eventually, thesuper-super-Spartans would learn to inhibit the shiningof the red light on certain occasions even though theirinternal mechanism was active. They might then come tospeak of a red light shining within them even though nonewas shining outside. But this would be just an illusion.There might emerge, in super-super-Sparta, two ways ofstudying red-light shining: (a) the external factors - thestimuli, rewards, and punishments that control it - and(b) the internal physiological-cognitive mechanism dis-covered by the super-super-Spartan genius plus the othermechanism (acquired by super-super-Spartans since thedeath of the genius) by which the output of the firstmechanism was inhibited or augmented so as to controlthe red light. (These correspond to our own externalstudy of pain behavior and internal study of pain mecha-nisms). A third discipline, initiated by super-super-Spar-tan mentalists, the study of internal red-light shiningbased on introspective reports, would be pointless. Intro-spective reports of red-light shining could serve as datafor super-super-Spartan behavioral studies (which mightattempt to discover the reinforcers of such reports) orfor super-super-Spartan physiological-cognitive studies(which might attempt to discover the internal mechanismby which such reports were generated), but the content ofthe introspective reports, their testimony as to the exis-tence and the nature of the internal red light, is noevidence whatever either for the existence or for thenature of an internal red light, which, as we have alreadysaid, would be an illusion - not only according to behav-ioral theory but according to physiological and cognitivetheories of red-light shining as well.

Returning to our own world, a person who says, "I can


feel pain without behaving," deserves to be heard with nomore (and no less) credence than the person who says,"The moon is larger when it's on the horizon than whenit's high up in the sky. " Only the mentalist will accept thelatter introspection at face value, and it is only he who canaccept, at face value, the former.

What is pain?

The super-super-Spartans do not prove that behaviorismis wrong or illogical. But the analogy was worth pursuingbecause it highlights what is partly a semantic dispute.The physiologist and the cognitivist want to speak of realpain (identified with internal pain mechanisms) on theone hand and pain behavior on the other. The behavioristwants to speak of internal pain mechanisms on the onehand and real pain (identified with behavior) on the other.If, for instance, an animal were discovered whose painbehavior was exactly analogous to ours but whose internalpain mechanism was entirely different (as a given com-puter output may be instantiated by different programs),the physiological cognitivist would say that the animal'spain was different from ours but its pain behavior was thesame, while the behaviorist would say that its pain waslike ours but its pain mechanism was different.

Putnam's super-Spartan analogy extends the argu-ments of Geach (1957) and Chisholm (1957) against thebehaviorism of Ryle (1949). According to Geach andChisholm, Ryle claimed that a mental state is equivalentto a disposition to behave in certain ways; that is, when weattribute a mental state to a person we are saying some-thing about how that person would behave under certaincircumstances. Geach and Chisholm argued that whenwe attribute mental states (especially perceptions) topeople we are referring to something that is going onright then, not what would go on in the future. Sincehaving a perception corresponds to no immediately ob-served overt behavior, we must be referring, according toGeach and Chisholm, to something going on where wecannot immediately observe it - that is, inside the orga-nism. Hence, they argue, behaviorism is false. The be-haviorism I have been defending here, contrary toGeach, Chisholm, and Putnam's interpretation of Ryle(although not, perhaps, contrary to Ryle), identifies men-tal states with overt behavior in the past and present, notin the future (although behavior in the future may proveus wrong about a particular identification of a mentalstate, as events may prove us wrong about any particularidentification). A rat in a Skinner box may be said to beresponding at a certain rate right now even though the ratis not at this very moment pressing the lever. Judgment ofthe rat's rate is made on the basis of lever presses in therat's past (although future lever presses may prove uswrong). Molar behaviorism extends this mode of analysisto mental events such as pain. (Lacey & Rachlin 1978made a similar point about mental events in general.) Thiskind of behaviorism has never, to my knowledge, beenacknowledged, let alone refuted, by antibehaviorist phi-losophers or psychologists.

Is the sense in which a rat is responding at a certain rateright now (even though not right now pressing the lever)different from the sense in which a person can be in painright now (even though not exhibiting pain behavior rightnow)? Geach, Chisholm, Putnam, and other anti-



behaviorist philosophers would claim that the very mean-ing of pain demands a distinction between the two cases.Whether such a distinction is valid will depend onwhether pain is defined as an internal mechanism or asovert behavior. If pain is defined as an internal mecha-nism then people in pain right now but not exhibitingpain behavior right now will be thought of as inhibitingthe output of that mechanism, whereas a rat not pressinga lever at a given instant is not ordinarily thought to beinhibiting lever presses at that instant. But if pain isdefined as overt behavior a person could be in pain, eventhough not at the moment exhibiting pain behavior,exactly as the rat is responding at a given rate even thoughnot at the moment pressing the lever.

The rat possesses internal mechanisms that mediatebetween pain stimuli and lever presses just as the personpossesses internal mechanisms that mediate betweenstimuli and pain behavior. But lever presses are usuallydefined in behavioral terms (as operants) rather than interms of those (poorly understood) internal mechanisms.The concept of a rat pressing a lever at a given rate at agiven moment (even though not, at that very moment,pressing the lever) has led to a search for correlates oflever presses in the rat's environment rather than insidethe rat. It has proved to be a very convenient concept forthose of us interested in controlling and (it seems to us)understanding rats' lever presses.

Just as the behavioral definition of lever presses (de-spite the existence of internal mediating mechanisms) hasled to control of lever pressing, so a behavioral definitionof pain (despite the existence of internal mediating mech-anisms) may lead to control of pain.

It is on empirical work such as that of Fordyce and noton the conceivability of Putnam's super-super-Spartansthat the behavioral definition of pain must stand or fall. Ifyou have previously defined pain in terms of internalphysiological or cognitive mechanisms then super-super-Spartans will be conceivable to you, and you must believebehaviorism to be wrong (at least as applied to pain). Ifyou have previously defined pain in terms of overt behav-ior then super-super-Spartans will be inconceivable toyou, and you need not believe behaviorism to be wrong.If you have made no previous commitment to a behavioralor physiological-cognitive definition of pain then Put-nam's super-super-Spartans are irrelevant.

What is gained or lost by a commitment in one direc-tion or another? A point apparently in favor of the physio-logical-cognitive definition is its correspondence, as re-gards the internality of pain, with the mentalistic viewresting on introspection and "linguistic intuition." Butthis is a double-edged sword because, when used (as itoften is) against the behaviorist definition of pain, itrequires arguments based on introspection and intuiti-tion - the very arguments rejected by physiologists andcognitivists vis-a-vis the mentalists. It seems inconsistentto argue on the one hand that pain is an internal eventbecause our introspections tell us so and on the otherhand that pain is the output of a computerlike mechanismor a series of neural discharges even though our intro-spections (and "linguistic intuitions" for that matter) tellus that pain is a "raw feel."

The best argument for the physiological-cognitive defi-nition of pain would be a truly effective physiological-cognitive treatment of pain. The best argument for a

behavioral definition of pain would be a truly effectivebehavioral treatment of pain. On this issue, unfortunate-ly, the jury is still out.

ACKNOWLEDGMENTSI would like to thank George Ainslie, William Bauin, DanielDennett, Marvin Goldfried, George Graham, Stevan Hamad,Patrick Heelan, Richard Herrnstein, Dennis Kelley, PeterKilleen, Alexandra Logue, W. Lycan, John Staddon, H. S.Terrace, and Gerald Zurifffor patiently reading and comment-ing on previous drafts of this manuscript. Preparation of themanuscript was supported by a grant from the National ScienceFoundation. Reprints may be obtained from the author, Psy-chology Department, State University of New York, StonyBrook, N.Y. 11794.

NOTES1. Placebos may exert their effect through endorphins releasedby the central nervous system (Levine, Gordon & Fields 1979).Thus, ironically, the placebo effect may be mimicked by drugs.However, there is some evidence (Siegel, Hinson & Krank1981) that, as with drugs themselves, people may developtolerance to and dependence on placebos.

2. But the functional units of a cognitive theory need notnecessarily be reducible to physiology. Fodor et al. (1974) haveargued that, even in principle, cognitive units need not be soreducible.

3. Dennett (1978, Chap. 11) was showing that a determinedmentalist will never accept the idea that a machine can be inpain. He makes the computer more and more complicated so asto handle each of the mentalists objections in turn. Yet thementalist is still not satisfied and, Dennett speculates, neverwill be satisfied. From a behaviorist point of view, Dennett ismodifying his computer in the wrong direction. What is wanted(by the behaviorist) is that the function of pain in humans bepreserved in the machine - that the machine's behavior be asignal of actual damage which the observer needs to do some-thing about. A behaviorist (and perhaps many other people)would sooner view a wagon with a squeaky wheel as an exampleof a machine in pain than Dennett's computer (with all itscomplicated internal machinery). The computer, howevermuch it might simulate pain, will not be judged as in pain(according to the behaviorist) unless it compels (by eventuallyreinforcing them) the kind of social responses that pain does.

4. Ainslie (in press) presents an as yet untested cognitivetheory of pain that is interesting enough to mention. Ainslieclaims that attention is an internal operant response that ispositively reinforced (also internally) by the painful stimulus.According to Ainslie, such stimuli provide a very brief reinforce-ment, strong enough to reinforce attention but not strongenough (and not long-lasting enough) to reinforce a motorresponse. This reinforcement, Ainslie claims, is followed by arelatively long refractory period during which no reinforcementis possible. The brief reinforcer plus the refractory period isworth less on the whole than the absence of both. Thus, a personwith a painful blister, say, is like an addict, constantly temptedto attend to the blister by the brief reinforcement but constantlylacking in other forms of reinforcement normally available.Pain, for Ainslie, is just this lack.

Ainslie's theory is cognitive rather than behavioral in thesense that the processes governing pain are said to be internal.The behavioral implication of Ainslie's theory is that the aver-siveness of painful stimuli is on a continuum with the aver-siveness of tempting stimuli. At one end of the continuum thecycle of pain, the brief reinforcer (too brief for consciousawareness) plus the longer refractory period, last only a fractionof a second. Next come itches with a longer cycle of bothreinforcer and refractory periods. Then come addictions. Final-ly, at the other end of the continuum, are certain moral deci-


sions, what Ainslie calls "sell-outs," with reinforcer and refrac-tory period lasting months to years.

5. In the history of behaviorism, when behavioral explana-tions have run into trouble, they have tended to hide inside thebehaving organism. Thus Watson (1913), who began with stim-uli as they affected the whole organism and responses of theorganism as a whole, ended (1924/1970, following Pavlov) bypostulating reflexes wholly interior to the organism. Hull (1943),who began with a set of axioms regarding stimuli to, andresponses of, the whole organism, ended in reply to Tolman's(1948) persistent and effective criticism by postulating "frac-tional-anticipatory-goal-responses " wholly inside the organism(Hull 1952). Skinner, who began by introducing the operant as aclass of overt activities defined by its function, more recentlyspeaks of covert stimuli, responses, and reinforcers (see Zuriff1979 for a discussion of "inner causes" in Skinner's work).

6. It is not inconsistent that molar behaviorism identifiesmental states with events in the past and present rather than thefuture but operants arc defined in terms of events that followthem. Molar behaviorism is a theory of an observer of behavior,while operants are activities of the observed organism. Theobserver may observe a series of acts, perhaps distant in thepast, and consequences of those acts less distant in the past.Current operant behavior is thought to be a function of pastbehavior and its more recent consequences.

7. A good example of the social use of the essential privacy ofpain is the following passage from a recent novel by HilmaWolitzer (1983):

"Once . . . Kenny broke his leg in two places, skiing. A few weekslater, he and Joy had three other couples for dinner. One of the mensaid, admiringly, that Kenny's leg must have hurt like a sucker, and Joysmirked. The other women nodded at her, knowingly, without a wordpassing among them. Kenny was amazed and outraged. He said that thepain had been excruciating, and one of the women actually laughed outloud. She folded her arms and said, It wasn't anything like labor,huster, you can bet on that.'

"Kenny said he wouldn't know, but that the bone had literally piercedthe skin. It was whitish yellow, like a huge, emerging tooth. A seasonedski pro had fainted when he looked at it.

"Ha! the woman said. Ha, ha, the other women added, even Joy,who had wept in the ambulance and said, Darling darling darling,' untilthey'd put him under for surgery.

"The mildest of the men said, 'We get more heart attacks, more lungcancer, more sports injuries. I mean, that's statistics.

'My water broke with Steven hours before I began to dilate, Joysaid. It was a dry birth. She ripped ofl a ragged piece of bread andchewed it.

" Have you ever been shot?' Kenny asked no one in particular. Guysin Vietnam got it in the guts, in the head.' It was a weak, even a speciousargument, since nobody at the table, including Kenny, had been inVietnam. Joy lifted the heavy meat platter with one hand and left theroom.

There was a long troubled silence during which the wine was finishedand crusts of bread were shredded into little hills of crumbs. Then one ofthe men said, 'Hemorrhoids! '

8. Integration such as this is to some extent a characteristic ofpsychology in general - as when a tone of a certain frequency orlight of a certain wavelength is considered to occur at an instanteven though frequency and wavelength require finite intervalsfor their definition.

Open Peer CommentaryCommentaries submitted by the qualified professional readership ofthis journal will be considered for publication in a later issue asContinuing Commentary on this article. Intenrative overviews andsyntheses are especially encouraged.

Commentary/Rdch\in: Pain and behavior

Behavior is what can be reinforced

George AinslieJefferson Medical College at the Coatesville V.A. Medical Center,Coatesville, Pa. 19320

The difference between prescriptive (normative) laws and de-scriptive ones is clear and fundamental, yet behaviorism has hadtrouble with it from the start. Rachlin's article is a case in point.

He says that behaviorism is distinguished from other schoolsby the doctrine that "whatever actions an organism takes areactions of the whole organism." This is not the kind of doctrinethat can be proven or disproven, and Rachlin acknowledges thathe means it to be accepted or not on the basis of whether thebehaving scientist, foraging, perhaps, for facts, finds it "moreuseful" than other doctrines. It loses nothing if put in normativeform: "Do not postulate part-organisms as intervening vari-ables, and avoid introspections as data (since they imply one partof an organism observing another part).

Constraints on scientists' spontaneous foraging to preventtheir seduction by false doctrines were the original concern ofbehaviorism. Watson feared the "medieval" orthodoxy of intro-spectionism, which was "dominated by a kind of subtle religiousphilosophy" (1924, p. 3). He did not deny that thoughts could beobservable processes, but he complained that introspectionistsusually fooled themselves about what they observed: "Whenthey come to analyze consciousness, naturally they find in it justwhat they put into it" (pp. 5-6). Skinner likewise worried that"the whole weight of habit and tradition" kept "raising theghosts of dead systems" (1938, p. 5). He did not specifically saywhy behavior had (o be limited to "what [an organism] isobserved by another organism to be doing" (p. 6), but thecontext made it clear that this was a self-control measure neededto keep these ghosts at bay. Later, perhaps believing that he hadlaid the ghosts to rest, he conceived a number of controllingvariables that were beyond direct observation by another orga-nism (Zuriff 1979). Rachlin in turn chides Skinner for succumb-ing to the temptation to "hide inside the behaving organism. 'Rachlin's even stricter behaviorism represents a prescriptioi forresisting such temptations.

Thus when Searle (1980) asks, "Are there no pains underlyingRachlin's pain behavior?" one would expect Rachlin's answer tobe, "Perhaps, but I have no trustworthy way of studying them."Instead, he sets out to show that introspections actually containno information beyond what can be observed by another orga-nism. He wants the behaviorist doctrine to be taken not just asprescriptive but as descriptive as well.

I believe he is making a mistake. It has been thoroughlyargued elsewhere that a rule against introspection can onlypartially be obeyed, that even behaviorists must depend on thisprocess in some cases (Alston 1974). A normative law can livewith violations, but a descriptive one cannot. Furthermore,Rachlin's position requires me to believe that my perception ofmy own pain is different in some essential property from myperception of the injured baseball player who clutches himselfor of the rat who works to avoid shock. It is hard to think whatthat property might be. Different sensory modalities? Theexistence of distinct pain tracts is doubtful (Weisenberg 1977). Iprobably see the baseball player with the same neurons thatsubtend painful brightness. Different replicability? Given theproper instructions, other people might be able to repeat myintrospection, an accepted method for consensual validation inGestalt psychology, linguistics, and bargaining research. Ofcourse, my perception might be distorted by incentives bearingon the observation process itself- perhaps my wish for sympa-thy, perhaps my love of an outworn theoretical system - but thishas been known to happen to observations made upon otherorganisms as well, and sometimes even to the replication ofthose observations. Perhaps the subjects of my introspectionsare not behaviors? That would be to say that my "pain behavior"is not behavior until someone else has seen it. But that is just to


Commentary/Rachlin: Pain and behavior

restate Rachlin's position, not to supply the property that setsmy introspection apart. There seems to be no distinct qualitythat separates internal modalities of observation from externalones. Introspective information may indeed be suspect forcertain scientific purposes, but it cannot be sharply differenti-ated from other kinds of information, much less be provedredundant.

I might seem to be arguing that Rachlin's point and perhapsthe point of behaviorism in general is a narrow procedural one. Iam not. In the course of obeying their prescriptive law aboutmaking observations and explanations on the level of behavior,Watson's followers cultivated the empirical hunch that mostpsychological processes were maintained by differential rein-forcement. Stimulus-driven processes like conditioned reflexesbecame decreasingly important as the description of verbaloperants, autonomic operants, pain operants, and so forth con-firmed this hunch. These were revolutionary findings.

These findings have contributed to the controversy surround-ing behaviorism, for in many bargaining situations people find itadvantageous to seem unfree (Schelling 1960); the revelationthat some kinds of process are operants disturbs social andpsychological balances. The extent of the population of operantsis a matter of descriptive law, but it is not generally discussedseparately from the prescriptive law about avoiding introspec-tion. The confounding of these quite separate laws probablysprings from the use of "behaviorism" to name the prescriptiveone. After all, a behavior in ordinary speech is not somethingpublic as opposed to private, but something active as opposed topassive, something purposive, an operant. Behaviorist doc-trines like the statement '"Mental states' are dispositions ofbodies to behave" (Graham 1982) can be read as expressingeither the prescriptive law or the trend of the descriptivefindings, or both indiscriminately. I would argue that the heartof behaviorism is its hypothesis that "mental states" are oper-ants, not its prescriptive law against introspection.

As a behavioral theory in the descriptive sense, the Fordyce-Rachlin argument does not go far enough. Rachlin marshalsgood evidence that human pain behavior often fails to follow apain stimulus, but then he agrees with Fordyce that it is arespondent as long as a pain stimulus is present. Why did thesoldiers at Anzio withhold pain behavior when the stimulus waspresent? Why do most of us fail to withhold it? How canseemingly benign stimuli like a dripping faucet or a very lean VI(variable-interval) schedule of reinforcement come to be aver-sive even though they do not condition respondents? Why doespain seem to reinforce attention but punish other kinds ofbehavior? It is possible to deal with such questions by treatingall mental processes as operants (Rachlin reviews my theoryfairly in his note 4), but Rachlin cannot do so, because he wouldhave to use intervening variables inside the organism. Theprescriptive law may have cleared the way for the descriptiveone, but it also confines it in ways that are not useful.

Behavioral definition of pain:Necessary but not sufficient

Joseph H. Atkinson, Jr. and Edwin F. KremerDepartment of Psychiatry, School of Medicine, University of California, SanDiego, La Jolla, Calif. 92093

Pain is commonly discussed as if it were a unitary disorder, withthe corollary that a discrete theoretical conceptualization (physi-ological, cognitive, or behavioral) sufficiently defines the phe-nomenon. Nevertheless, the term pain, especially chronicpain, may well subsume a family of disorders rather than a singleentity with a uniform etiology or pathophysiology (Merskey1983). The failure of diverse treatment approaches (includingbehavioral interventions) to consistently alleviate chronic painwould argue that critical variables important to the predisposi-

tion, inception, and maintenance of chronic pain have not beenidentified. An important task now is to develop more compre-hensive categorization of pain disorders by defining, wherepossible, homogeneous subgroups of patients by behavioral andbiological markers. Presumably these relatively homogeneoussubgroups could be investigated for their natural history anddifferential responses to treatment. The question, then, iswhether a strictly behavioral approach contributes to this effort.We believe that it is necessary but insufficient to the task, andthat cognitive and physiological data are required to fully in-terpret pain behavior.

Among the major contributions of an operant behavioraldefinition of pain as outlined by Rachlin are that it (a) identifiesthe aversive quality of pain as pathology of goal-seeking andreward behaviors, and (b) views pain as a response of theorganism, rather than a bothersome symptom of medical orpsychiatric illness. The first allows for quantifiable data otherthan self-report of pain intensity for treatment and outcomecriteria, since verbal pain estimates are distorted by manyvariables (Kremer, Block & Atkinson 1983). Even so, the task ofpain assessment remains complex. It cannot be assumed thatbehavioral indices of pain are uniformly reliable and objectivesimply because they are quantifiable and easily scaled. Forexample, there is evidence from studies using automated ac-tivity monitors that some pain patients underreport activitylevels while accurately reporting social behaviors (Sanders1980). It is also known that pain chronicity, affective state, andmedications commonly prescribed for pain patients (e.g., ben-zodiazepines) confound self-assessment and behavioral self-report (Kleinknecht & Donaldson 1975). As more reliable meth-ods of behavioral assessment are developed, patients may bebetter categorized according to levels of physical, social, andother behaviors.

The second contribution, that of viewing pain as a response ofthe whole organism in interaction with its environment, isessential to a comprehensive model of chronic pain. Disease hasbeen described as a failure of the organism to adapt to situationsand experiences that are adverse or are perceived as adverse(Weiner 1977). In this view, no one factor (i.e., social, political,nutritional, or neurological) is an exclusive cause of diseaseitself, and factors may interact to make the adaptive task moredifficult. Organisms may be predisposed to disease, and theprecipitation and perpetuation of disease is a complex processrelated to many factors. The more conventional approach todisease disagrees with the adaptive model; it conceptualizesdisease as the direct result of these factors themselves (Weiner1977).

Implicit in the adaptive approach is the belief that the brainregulates adaptive activity and modulates physiological pro-cesses via the neuromuscular system, the autonomic nervoussystem, and the limbic-hypothalamic-pituitary axis. It is con-ceivable that pain behaviors, as representative of adaptivefailure, might be reflected internally by neuroendocrine dys-regulation. The experimental animal paradigm of avoidanceconditioning is analogous to the more complex human task ofcoping or adapting to specific situations, and complex patterns ofneuroendocrine response to avoidance-conditioning experi-ences are noted in primates (Mason, Brady & Tolliver 1968).Animal studies more specific to pain indicate important rela-tionships between the hypothalamic-pituitary-adrenal (HPA)axis and the endogenous opioid system (Millan, Przewlocki &Herz 1981). For example, stressful situations, which commonlyactivate the pituitary, alter the sensory threshold to noxiousstimuli (Bodnar, Kelly, Spiaggia, Ehrenberg & Glusman 1978;Chesher & Chan 1977); and hypophysectomy blocks stress-induced analgesia to acute and chronic pain (Amir & Amit 1978).Furthermore, there is evidence that opioid peptides influencethe release of growth hormone, prolactin, ACTH, and otherhypothalamic hormones (Meites, Bruni, Van Vugt & Smith1979).


Although there is considerable variability of hormonal re-sponse associated with stress (Mason 1975) and psychiatricdisorders (Winokur, Amsterdam, Caroff, Snyder & Brunswick1982), it is conceivable that neuroendocrine patterns occur inpain disorders. If so, it may be possible to distinguish subtypesof pain patients by neuroendocrine as well as behavioral indices.Nonsuppression of plasma cortisol after dexamethasone (dex-amethasone suppression test, DST) has been reported in pa-tients with so-called psychogenic pain (Blumer, Zorick,Heilbronn & Roth 1982) as well as in pain patients with majordepression and other psychiatric disorders (Atkinson, Kremer,Risch, Morgan, Azad, Ehlers & Bloom 1983). In addition wehave noted that in some pain patients prolactin secretion isrelatively unaltered by dexamethasone (Atkinson, Kremer,Risch & Janowsky 1984) regardless of psychiatric status, where-as healthy volunteers show dexamethasone suppression of pro-lactin and cortisol, and a percentage of patients with affectivedisorder show both cortisol and prolactin nonsuppression afterdexamethasone (Meltzer, Fang, Tricou, Robertson & Piyaka1982). Finally, differing concentrations of cerebrospinal opioidpeptides are reported in patients with psychogenic and organicpain (Almay, Johansson, von Knorring, Terenius & Wahlstrom1978), and reliably greater plasma opioid peptide concentra-tions have been reported to occur in chronic pain patients thanin psychiatric subjects and volunteers (Atkinson et al. 1983),indicating that other biological markers may be available. Giventhat pain and disorders of mood often appear concurrently andpresumably involve common neurotransmitter systems(GABAergic, noradrenergic, cholinergic, serotonergic, andpeptidergic), discriminating biological markers of stress, pain,and affective disorder will be exceedingly complex. Longitudi-nal studies are needed to assess the behaviors of markers invarious phases of pain syndromes and associated affective states.

If chronic pain is to be understood as a failure of adaptivemechanisms, then basic research using animal models is re-quired to identify factors which predispose organisms to chronicpain, as well as to study those which initiate and maintain theprocess. Our primary clinical obligation, however, is to try toalleviate suffering. Affective disorder is one component of suf-fering in pain syndromes, and here the presently availableneuroendocrine markers might have therapeutic application.For example, nonsuppression of cortisol after dexamethasonemay help identify pain patients with major depression in whomtricyclic antidepressants are indicated. In the future it may bedetermined that behavior therapies are appropriate to certaingroups of patients identified by specific clinical and laboratoryfindings, but less productive in other groups. Indeed, thenecessity for such discrimination was acknowledged in theinitial descriptions of behavioral treatments of chronic pain(Fordyce, Fowler, Lehman and DeLateur 1968). Rachlin's sug-gestion that the best argument for a behavioral definition of painwould be a truly effective behavioral treatment of pain over-simplifies the diversity of pain disability and underestimates thetherapeutic challenge.

Internal events as behavior, not causes

Daniel J. BernsteinDepartment of Psychology, University of Nebraska-Lincoln, Lincoln, Neb.68588

In general I find Rachlin's defense of a behavioral analysis of painquite convincing. The parallels between sensory/psychologicalpain and respondent/operant behavior are well drawn, and hisaccount of the likely development of the commonsense (men-talist) approach to pain is a useful component of his position.Rachlin's restrictive criteria for behavioral analysis are lessconvincing, and my commentary will focus on the limitations hiscriteria impose on a behavioral account of pain.

Commentary/ Rachlin: Pain and behavior

Rachlin repeatedly limits a "purely behavioral" account toactions of the whole organism, specifically excluding any inter-nal events. He justifies his rejection of internal events in anyform by asserting that their inclusion makes a behavioral analy-sis indistinguishable from a physiological or cognitive account.This assertion would be true only if the internal event wereidentified as a causal agent (real pain) and the overt behavioridentified as a product (symptom) of the internal state. It wouldbe reasonable to reject physiological reductionism in the pursuitof causes of behavior, because identification of physiologicalmechanisms is an incomplete analysis. The inevitable next stepis identification of the environmental antecedents of the physio-logical events, an analysis that can be made without reference tophysiology by finding the environmental antecedents ofbehavior.

It is less reasonable to reject analysis of internal events thatare under the control of identifiable respondent processes andoperate independently of overt operant behavior. The work ofBykov (1959) and other investigators in the Soviet Union (cf.Razran 1961) suggests that many body processes are sensitive torespondent conditioning procedures. With proper instrumenta-tion for measurement and for delivery of stimuli, control of thebehavior of internal organs is possible, and the causal analysis isconsistent with a behavioral account. It seems unnecessarilyrestrictive to eliminate these processes from an account of painsolely because the behavior under stimulus control can only bemeasured inside the skin. The analysis is readily distinguishablefrom a physiological account because the controlling variablescan be identified in the environment.

Including the behavior of some internal body parts in abehavioral analysis can bolster Rachlin's discussion of the pain"underlying behavior" without embracing mentalism. In re-sponding to Searle's (1980) criticism of a behavioral position,Rachlin conjures up a pure mentalist who asks Searle to identifythe pain underlying his neural activity. Rachlin considers thatrequest a mentalist error that all materialists would reject, andhe implies that Searle is making the same error by asking for theidentification of a physiological substrate. This is a weak re-joinder, for all three perspectives (as represented in the targetarticle) are explicitly materialist, and Searle proposed the iden-tification of entities that are in principle measurable within theguidelines of natural science. Rachlin might have responded bynoting that respondent behavior inside the organism is the cuefor Searle's verbal report, but his limited range of behavioralanalysis precludes such a response.

A slightly broader behavioral account might argue that thebody is full of organs and muscles that are responsive to interac-tions with the environment, and that people have learned torespond verbally and nonverbally to the behavior of these bodyparts (cf. Skinner 1945). Instead of merely identifying them asthe source of real pain, a behavioral analysis investigates theconditions under which these internal organs behave in waysthat are labeled with pain words. The behavior of these bodyparts is not a cause of the other operant pain behavior; instead,this behavior is the pain which Searle labels.

Consider the following analysis of the pain associated withlabor and delivery of a child. For several hours before birth awoman typically reports pain and makes her body tense (operantbehavior) while her uterus repeatedly contracts and remainstaut for short periods (respondent behavior). Most preparedchildbirth techniques involve muscle relaxation and breathingexercises as ways of minimizing the aversive nature of the laborperiod. The aversiveness decreases greatly if the woman's skel-etal muscles remain relaxed and if she responds to trained cueswith patterned breathing instead of responding vocally to thecontractions. This description is consistent with a behavioralaccount, because interventions that reduce operant pain behav-ior produce corresponding reductions in reported aversiveness.A well-prepared woman might remain calm during contractionsbut still report pain when asked. If Searle asked where her pain


Commentary/'Rachhn: Pain and behavior

was, Rachlin could identify the behavior of the uterus as the painwithout being mentalistic.

Limiting behavioral analysis to the actions of the wholeorganism is one way to avoid confusing identification of physio-logical mechanism with explanation of the causes of behavior. Itis not the internality of events that makes them mentalistic,however; it is their identification as causal agents of overtbehavior that is challenged by behaviorism. The historicalchoice of stopping analysis at the skin may reflect the limitedresearch technologies of the time and not a necessary dichoto-my. If the procedures of behavior analysis can be brought tobear on internal parts of the organism, then their behavior canbe studied without introducing inappropriate inferences ofcausal order. It is possible to avoid mentalism without limitingour analysis to the whole organism.

Pain is three-dimensional, inner, andoccurrent

Keith CampbellDepartment of Traditional and Modern Philosophy, University of Sydney,Sydney, N.S.W., Australia 2006

1. The dimensions of pain. Rachlin has done philosophy aservice by further publicizing the view that pain is not simple,despite any apparent intuitions to the contrary. There is com-plexity not only in the organic processes associated with pain butalso in the responses that constitute the pain experience.Rachlin treats pain as two-dimensional, identifying a sensoryand an aversive component, which are to be measured onintensity and reaction scales.

The two scales, for intensity and for reaction, are independentat least to the extent that for different people, different increasesin the one go with a given increase in the other. The indepen-dence of the two scales, and hence of pain's two components,would be further confirmed if for some people in some circum-stances the relationship between the two scales were inverted.The experiments should be repeated using people with a historyof masochism.

To recognize two dimensions to pain is a step in the rightdirection, but it does not go far enough. Tantalizingly, Rachlinrecognizes this point but ignores it. For pains have not onlyintensity and aversiveness but also quality. They can be sting-ing, grinding, shooting, or throbbing pains. The existence ofpains of differing qualities is of high importance to behaviorismsince it provides the basis for those arguments from the pos-sibility of a shift in pain quality without any shift in pain's causesor behavioral manifestations, to the conclusion that cause andmanifestation do not exhaust the reality of pain. The argumentshave very close formal parallels in the spectrum-shift argumentsagainst behavioral (and functionalist) accounts of colour vision.

A spectrum-shift argument supposes that two people can havecolour vision in which the experience of each involves a shiftalong the spectrum relative to the other. What one person seesas a bluish-green may look to the other as (what the first woulddescribe as) more nearly yellow. Especially if the shift is slight,both may use the same colour descriptions, and both may beable to make exactly the same colour discriminations (the be-havioural manifestations of colour vision may match). In bothcases the stimuli (colour vision's causes) will be the same. Theinternal, mental effects, insofar as they are functional/be-havioural, may likewise match. Since the colour experiences donot match, there is more to colour than functional and be-havioural accounts allow.

The force of the qualia argument in the pain case is no lessthan that for the colours.

Rachlin's neglect of the qualitative aspect of pain derives, Isuspect, from a general orientation concerned above all with

therapeutic strategy (see the final paragraph of the target arti-cle). Such a semipractical approach, which sets aside any the-oretical issue not connected to questions of treatment, has itsown justification, of course, but it forfeits any claim to becomprehensive.

Compare the case of pain with that of a group of organicdiseases, bacterial in origin, with fever, coughing, and a varietyof skin rashes and lesions as symptoms. For medical entitiesfrom this group, suppose no methods aimed at suppressing theskin rash symptoms have any effect on the course of the disease.Then there will plainly be a difference between an account of thedisease that deals with the aspects relevant to its treatment and adescription of it in its entirety. The former, but not the latter,will leave out the matter of skin condition.

To suppose that for pain we can conflate the aspects relevantto treatment and the complete entity is to beg all the philosophi-cally important questions.

2. The location of pain. Rachlin's second service to philosophyis his forthright concession-indeed insistence-that a behav-iorist account of pain, or any other mental item, is rigorouslyconfined to outer events. "Behavioral theory cannot internalizeits terms and remain a behavioral theory." Nothing but good cancome from keeping competing theories clearly distinct; it is amatter, one is tempted to say, of mental hygiene.

For this very insistence on the outer status of pain is preciselywhat makes behaviorism so implausible. The two sides of thisimplausibility are, of course, the possibility of pain withoutbehavioral manifestation and its converse, pain behavior with-out suffering.

For pain without pain behavior, the most striking cases arenot Putnam's super-Spartans, which Rachlin discusses, but thecases of pseudo-anesthesia, especially under curare, which arementioned in the medical texts (Buxton Hopkin 1980; Hutchin-son 1960). There seems to be no doubt that all symptoms of paincan be suppressed for at least a moderate period. And, accord-ingly, all symptoms can in principle be suppressed, for thatindividual at least, over an indefinite period. If we then takeRachlin's position that "for a molar behaviorist extreme pain isnot an immediate sensation (deeply felt) but a long-term patternof behavior, widely performed, it would seem to follow thataccording to molar behaviorists, a person maintained in a state ofrelaxation over an extended period, especially if the periodbegins in infancy, would not be in pain, certainly not in extremepain. It is to be hoped that no molar behaviorists becomesurgeons, and vice versa.

As for pain behavior without suffering, Rachlin should give usmore on distinguishing real psychological pain from faking. Thedistinction is straightforward, metaphysically if not epis-temically, for cognitive or physiological theories of pain: In realpain there is a suitable inner cause for pain behavior, while infaking there is no such thing, and in consequence, with fakingthere is no suffering, no literal pain.

But on the behaviorist theory, the difference must be foundon the outside. Now suppose, not a race of super-super-Spar-tans, but one of super-super-crybabies. Super-crybabies put ona pain display in advance of any actual tissue damage or malfunc-tion, thereby averting it. Super-super-crybabies do this as iffrom deeply ingrained habit, spontaneously, without delibera-tion i.. conscious intent. But the "habit" was not formed in theusual way; it is innate.

It would seem that a super-super-crybaby could pass its wholelife in complete comfort, yet be counted, on the molar behav-iorist test of a widely performed behavior pattern, as in per-petual and extreme pain.

Fordyce's (1983a) therapeutic success is compatible withphysiological or cognitivist philosophies of pain; his patientshave formed a habit, but they are not faking. The habit is one offeeling pain although organic causes have lapsed. And habits offeeling can respond to treatments directed at their behavioralmanifestations.


Commentary/Rachhn: Pain and behavior

3. Timing the occurrence of pain. In "What is pain?" para-graphs 2-5 Rachlin discusses the situation of someone exhibit-ing extensive pain behavior but not right now showing any signsof pain. Is this analogous to the case of a rat currently displayinga certain rate of bar pressing even though quiescent right now?He first suggests that a choice of definitions will settle thequestion: pain defined as internal mechanism will distinguishthe two cases; pain defined as overt behavior will assimilatethem. But how pain is to be "defined" is not in any interestingsense a matter of choice or decision. The question is whether thephenomenon indicated by the normal use of the English wordpain involves more real elements than overt behavior. That is aquestion of how the world is, not a matter of semantics.

By paying attention to the phenomenon of pain, we can, Ithink, discover that the behaviorist position is mistaken. Thus:

Setting aside anxiety, embarrassment, and other nonpain sources ofdiscomfort; If I am neither displaying nor inhibiting pain behavior, Iam quite comfortable. If I am quite comfortable 1 am not right now inpain. But if I am neither bar pressing nor inhibiting bar pressing, Imay still be displaying right now a certain bar-pressing rate.

This line of thought depends on common experience, or intui-tion, about the incompatibility of comfort and pain. Whichbrings me to my last point: We must distinguish relying onintuition for knowledge that some mental event occurs fromrelying on intuition or introspection to inform us of that event'snature. All developed theories of pain, especially nonmen-talistic ones, must abandon the second kind of reliance. But onlybehaviorism, whether molar or molecular, backward- or for-ward-oriented, must repudiate the first type. Therein lies itsweakness.

Heuristically, "pain" is mainly in the brain

W. Crawford ClarkNew York State Psychiatric Institute; Department of Psychiatry, ColumbiaUniversity, New York, N.Y. 10032

How the purer spirit is united to this clod, is a knot too hard for fallenhumanity to untie. . . . How should a thought lie united to a marblestatue, or a sunbeam to a lump of clay! . . . to hang weights on thewings of the winde seems far more intelligible.

JOSEPH GLANVIU., (1636-1680)

What and where Is pain? I cannot agree with Rachlin and theneobehaviorists that pain resides only in behavior. Neitherphilosophers nor scientists know any more about how (orwhether) physiological activity is associated with mental images(if they exist), or why they appear to correspond with thephysical world most of the time, than did Joseph Glanvill. Iagree with Rachlin that imaginary super-Spartans do not dis-prove behaviorism, that philosophical discussions of what pain islead us nowhere, and that we must therefore view the empiricalevidence. However, if pressed for a philosophical view anempiricist would find Fodor's functionalism (1981a; 1981b), inwhich a very complex computer could have thoughts and feelpain, to be more congenial than a view that places pain in thebehavior of an organism, or in the behavior of a computer forthat matter. The mentalistie physiological-cognitive model asdescribed and dismissed by Rachlin has proven itself useful; itpromotes fruitful research and advances pain treatment. Thesensory experience of pain cannot be proven to exist, but neuraland cognitive processes that appear to be related to pain providea valuable heuristic. When research is undertaken or treatmentis required, it is necessary to investigate all components ofbehavior, including the physiological. The empirical evidenceclearly demonstrates that most pains arc not the behavior of thewhole organism but appear in identifiable physiologicalsubsystems.

Whether a patient's perceptual world is in his mind, or in hisbehavior, or in his nervous system, or does not exist at all, is lessimportant to an empiricist than the relationships between in-puts and outputs. Behaviorism does not have a monopoly onobjectivity. Even a mentalistie empiricist is quite willing to say,"I do not know whether morphine really reduced the painproduced by these calibrated noxious stimuli, but I do know thatdiscrimination, determined by error rate, between higher andlower intensities was decreased." Or, "I do not really knowwhether the patient felt less pain, but he reduced his intake ofanalgesics." For example, calibrated noxious stimuli or somato-sensory, pain-intensity, and emotional descriptor words (input)can be presented to an individual, and various types of judg-ments (output) can be obtained. These are then analyzed bymathematical models such as signal detection theory (Clark &Yang 1983) and multidimensional scaling (Clark 1984) to dis-cover the subject's attitudes and the dimensions underlying thepain response. Or specific cognitive coping strategies may betaught and responses such as heart rate, gross behavior, and self-reports measured (Mischel, Fuhr & McDonald 1984). Thesemodels lead to a far richer understanding of the complexities ofpain than does the behavioral approach.

Does pain behavior equal pain? If pain is in the behavior, whathappens when operant pain behavior and respondent painbehavior are discordant? Hilgard (1969) reports that althoughhypnosis succeeds in eliminating the report of pain (operantbehavior), the autonomic responses to noxious stimuli (re-spondent behavior) continue. Which behavior counts? Is thesubject in pain or not? The report of pain possesses both operantand respondent characteristics, but behavioral analysis is notthe only means of distinguishing between them. A mentalistiepsychophysical approach may yield more information. For ex-ample, a number of studies using signal detection theory havefound that placebos, acupuncture, and suggestion causechanges in the location of the criterion for reporting pain.However, they generally fail to find any effect on sensorydiscriminability, an index of accuracy related to neurosensoryactivity; this makes it unlikely that internal mechanisms mediat-ing pain have been influenced. Thus, these treatments changethe pain report criterion (operant behavior), while neural ac-tivity in some "pain center" (respondent behavior) remainsunchanged. Similarly, while behavioral therapy is an effectivetreatment for certain chronic pain behaviors, the neural activityin some "pain" center could nevertheless remain unchanged.

Individuals with congenital insensitivity to pain appear topresent a problem for the behaviorist. From birth, these indi-viduals fail to avoid noxious stimulation and do not experiencepain. However, with experience (including severe burns, abra-sions, and fractures) they eventually learn to distinguish mildstimulation from tissue-damaging stimulation and to avoid thelatter. Now that they exhibit pain avoidance behavior, shouldthey not also experience pain? Yet the evidence is that they havenot learned a new dimension of sensory experience but haveonly improved their ability to make distinctions along an inten-sity dimension (Clark 1984).

What about individuals who exhibit pain behavior in theabsence of any noxious stimulus? Are they in pain? Patients,particularly depressed patients, use the word pain to describediscomfort, misery, or anguish. In the laboratory, we find someanxious subjects who report pain to a zero-intensity stimulus. Ifsuch behavior is equivalent to pain, the behaviorist must con-clude that experimentally induced pain may occur in the ab-sence of a noxious stimulus. Surely it is more realistic to admitthat pain behavior may occur in the absence of a painful sensoryexperience.

Finally, a problem is encountered in moving from the animalmodel (where molar behaviorism appears reasonable) to humanbehavior. Molar behaviorism defines a respondent or operant interms of correlated events rather than temporal contiguitybetween environment and behavior. These two behaviors are


Commentary/Rachhn: Pain and behavior

differentiated on the basis of their history. This argument maybe acceptable in the instance of a laboratory animal whose everybar press has been objectively recorded almost from birth;however, it encounters difficulty when it is transposed to theanalysis of human pain behavior. Granted, "a rat in a Skinnerbox may be said to be responding at a certain rate right now eventhough the rat is not pressing the lever," but this is true only if itsentire behavioral history has been objectively recorded. Unfor-tunately the necessary objective antecedent behavioral mea-surements are not available in man. Without this detailedhistory, recorded at the time it occurred, some form of internalconstruct must be relied upon. For the analysis of human painbehavior the molar behaviorist approach appears to be no better(or worse) than a cognitive approach.

Can treatment outcome save the day for anyone? Treatmentmethods based on various theoretical positions, from incanta-tions to neurosurgery, have each had their share of success.Thus, treatment outcome can neither prove nor disprove atheory. Nevertheless, since the issue has been raised, it isworthwhile to scrutinize the real world of pain treatment. Herepain is very seldom "behavior of the whole organism." Therecan be no doubt that behavioral treatments for intractablechronic pain, such as those developed by Fordyce (1976), havebeen quite successful in decreasing pain behavior and benefit-ing many patients who have failed with conventional treatment.However, these cases represent a minuscule proportion of thepatients who are successfully treated for both acute pain andchronic pain. Direct physiological intervention with respect to aspecific subsystem, rather than the whole organism, is ex-tremely effective. The physician and the dentist treat the stim-ulus source or various parts of the neurosensory system, and thepain disappears. A tooth is repaired or extracted, pressure on anerve is released, an unguent is rubbed on a burn, surgerycorrects a malformation, codeine relieves an earache, methy-sergide relieves migraine, aspirin relieves arthritic pain, ananticonvulsant eases the paroxysmal pain of trigeminal neu-ralgia, and so on. Treatment is tailored to specific subsystems,not to the behavior of the entire organism. Few practionerswould offer, or patients accept, behavior therapy for these pains;indeed, to take a behavioral approach would be unethical inmost instances. When direct intervention fails, cognitive tech-niques may have more to offer than granted by Rachlin. Asummary by McCaul and Malott (1984) of recent evidence forthe effectiveness of cognitive strategies for coping with painconcludes that pain behavior can be modified by purely cog-nitive interventions.

Neither philosophical arguments nor treatment outcome cansettle conceptual disagreements. The neobehaviorist positionresults in two types of labeling errors: (1) labeling as not in painpeople who fail to display pain behavior but who have sustainedtissue damage, and (2) labeling as in pain people who displaypain behavior without tissue damage. Physiology and cognition- as well as behavior - need to be included when studying andtreating the complex phenomenon of pain.

ACKNOWLEDGMENT1 wish to thank Harriet Nerlove Mischel, Department of Psychology,Columbia University, for her assistance in formulating and sharpeningthis commentary.

On Rachlin's "Pain and behavior": Alightening of the burden

Wilbert E. FordyceRehabilitation Medicine and Pain Service, University Hospital, Seattle,Wash. 98195

Rachlin and the neurophysiological and cognitive investigatorshe cites seem committed, each in different ways, to presenting a

unified concept of what they seem to assume is a unifiedphenomenon, namely, pain. In my view, pain is not such aunified entity. What is subsumed under the rubric "pain" is a setof events. Further, the borders of the "set" arc perhaps notstable. The unification is in the eyes of the beholder. It is notclear to me that there is any inherent reason to insist that "pain"is some one thing.

I shall not attempt to analyze physiological or cognitiveperspectives - or pure mentalistic ones - on the subject of pain.As a somewhat pragmatic behaviorist who is not above trying touse cognitive methods clinically, I shall attempt to set forth myown behavioral perspective on the matter.

Pa' behaviors are overt events that are likely to causeobservers to identify the person as having a "pain problem." Noassumption need be made about there having been, either at themoment or in the recent history of the person, noxious stimuliwithin or on the surface of the organism but peripheral to thecentral nervous system that served as antecedent stimuli to thepain behaviors. Pain behaviors are capable of occurring becauseof anticipated consequences, immediate or remote. Those con-sequences are assumed to have been anticipated on the basis ofprior experience. Thus, conceivably, in a given person at a giventime, pain behaviors could spring forth de novo, in the sensethat no antecedent stimulus in the form of noxious stimulation isessential to the emergence of pain behaviors.

That pain behaviors may occur de novo does not mean thatthey often do, only that they may. There is present in most painpatients compelling historical information indicating that therehas been at some time in the past an injury leading to thermal ormechanical energy impinging on free nerve endings that in turnactivated A-delta and C fibers. That peripheral stimulation willhave, in the normal course of events, elicited a sensation thatpeople, when asked, are likely to identify as "pain." The pre-sumed transmission of the peripheral stimulus to the CNSprovides a basis for concluding that there was - and may still be- "pain." That "pain," whether measured in intensity or inreactive terms, constitutes another set of events different from,though potentially linkable to, pain behaviors. That "pain" islikely to be followed by pain behaviors. When the pain behav-iors do occur and are responded to by the environment, one canbegin to say that the person has a pain problem.

The basis for saying there is a pain problem is the occurrenceof pain behaviors. Pain behaviors occurring de novo, in thesense described above, still constitute a pain problem. They donot, however, constitute a pain problem that is linked to noxiousstimulation or, as is commonly stated, to "the organic factor."

In the practical case, a more typical pattern is that an injuryresulting in pain and restricted activity has, in the case ofchronic pain, led to disuse of the body parts involved. Muscleand tissue fibers shorten and lose bulk, strength, or elasticity.Attempts to use or move the body part then result in theexperience of "pain." The problem here is that both the suffer-ing person and, all too often, the physician identify this "pain" asindicating that healing has not occurred and that further immo-bilization is indicated. Further rest only worsens the problem.This is basically a problem of failing to realize that hurt and harmare not the same.

Whether modification of pain behaviors also modifies "pain"is a moot question. It depends on what "pain" we are talkingabout. Moreover, the modification of pain behaviors has meritin its own right, irrespective of how much or how little painbehaviors may correlate with some concept of "pain." This pointdoes not concern the debate about whether there is something"inside" the person known as "pain." It merely indicates thatpersons who undergo a treatment program resulting in reduc-tion of pain behavior also show some other worthwhile changes.They do more than they did before. They also have lower healthcare utilization rates. There is ample basis for inferring that thereduction of pain behaviors by activation and reduction ofanalgesic consumption results in neurophysiological changes


Commentary/RacMin: Pain and behavior

that may themselves alter "pain." But it is not necessary topostulate such an effect. In my view, pain behaviors need not beseen simply as extensions of "pain"; nor is it necessary to insistthat ncurophysiological or cognitive events assigned the label"pain" must exist in order for pain behaviors to occur.

Radical behaviorism is a dead end

Jeff FossPhilosophy Department, University of Victoria, Victoria, British Columbia,Canada V8W 2Y2

What is Rachlin up to? He evenhandedly dishes out criticism toall three major theories of pain, but he refuses to explicitly pick afavorite among them, concluding merely that, as regards thequestion "Which theory is best?" the answer is, "The jury is stillout." But if we assume that Rachlin intends more than aderogatory review of pain theories, and if his behavior is ruledby his principles, then he is in favor of a radical behavioristietheory of pain. I will argue briefly that the theory Rachlinpromotes is grossly unrealistic and that the behaviorist funda-mentalism it expresses is uncalled for.

Rachlin says, "The question we ask here is whether a purelybehavioral theory of pain makes sense." If by "makes sense" hemeans merely "has the logical possibility of being true," thenthe answer to his question is yes, though the question is rathersilly. Manifestly false theories, like Ptolemy's or the alchemists',make sense in this trivial way: since they are not self-contradic-tory, they at least have a meaning - all consistent theories aretrue in some possible world, if not in this one. On the otherhand, if by "makes sense" Rachlin means "has a chance of beingtrue in the actual world," then the purely behavioral theory hechampions makes no sense at all.

The theory, adumbrated, is that pain is a sort of behavior.What sort? Pain behavior - we uneasily (if not painfully) awaitRachlin's account of how such behavior is to be noncircularlydefined. A little more fully, "sensory pain is respondent behav-ior, and "psychological" pain is operant behavior. What aboutthe person who feels pain but is not now evincing pain behavior?Well, answers Rachlin, that person, despite appearances, ispain-behaving in the operant sense, the sense in which, "a rat ina Skinner box may be said to be responding at a certain rate rightnow even though the rat is not at this very moment pressing thelever." Those pains which never evince behavior, Rachlin isprepared to dismiss as illusions. By these artifices the funda-mentalist behaviorist does not, at least, contradict himself- histheory makes sense in the weaker of the two ways outlinedabove. But it is well known among historians and logicians ofscience that an empirical theory can be recast in indefinitelymany forms, some of which secure both internal consistency andpeaceful coexistence with the evidence to date. Joseph Priest-ley's book The Doctrine of Phlogiston Established, published in1800, well after rapid oxidation had been accepted by chemistsas the true nature of fire, demonstrates that even very sharpscientific minds can pronounce a cul-de-sac the broad highwayto truth. The final downfall of a theory is sterility: Lavoisier snew chemistry went on to new triumphs while phlogistic theorystagnated. Given that it is obvious that neurophysiologicalmechanisms mediate and produce behavior, to abjure andignore such mechanisms in pursuit of a "purely behavioral"theory of consciousness is to purchase ideological purity at theexpense of truth.

A curarized subject undergoing dental work without anesthe-sia experiences painful sensations. Rachlin could try this forhimself: Since subjects could report the pain after the curarewore off, they would have experienced "psychological" painaccording to Rachlin s theory. These results are indefinitelyreplicablc. Now note that a subject could be kept curarized for

life after the dental work, thus never evincing pain behavior.Only someone with a dull axe to grind would insist, contrary toall of the evidence available from the replicable eases where thesubjects are finally de-curarized, that the permanently cura-rized subject feels no pain - yet Rachlin does so insist. He takesthe trouble to accommodate the folk-theoretic belief that onemay be in pain now while not now evincing (respondent) painbehavior, by using the device of operant behavior: Even whilecurarized, the subjects are pain-behaving at a certain rate, giventhat they will evince pain behavior when de-curarized. But whybother to make this accommodation? If he is willing to rideroughshod over the belief that the permanently eurarized sub-ject (or the one who dies before the curare wears off, etc.) feelspain, why not simply derogate all pain reports whatever? Ifbehavior is all there is, why even bother to try to specify somesort of behavior that corresponds to the folk-psychological no-tion of pain, especially given that according to such notions painis quite distinct from behavior of any sort? Even the consistencyRachlin purchases is merely logical, as is evidenced by itsuncomfortable fit both with extant theory and with the facts.

What could be Rachlin's motive in this exercise? Hull, Wat-son, and Skinner promoted behaviorism as a methodologicalprophylactic for psychology against the unscientific proceduresused in psychoanalysis and other introspective studies. They didnot, as Rachlin himself notes, deny the reality of the internalmechanisms and processes that make behavior possible. Rachlinis concerned to eschew such mechanisms and processes be-cause, as he says, such tilings "are physiological or cognitiveconcepts, not behavioral ones. ' So what? The existence of suchinternal mechanisms and processes is indubitable. And truth ispreferable to ideological purity.

On kicking the behaviorist; or, Pain isdistressing

Myles GenestDepartment of Psychology, University of Saskatchewan, Saskatoon,Saskatchewan, Canada S7N 0W0

I fear that Rachlin would be well advised to guard his shins. I, ofcourse, am not a violent person. Nevertheless, the temptationto kick the behaviorist to which Rachlin refers is born of frustra-tion, and parts of his paper exemplify the stubbornly held viewsthat non-radical-behaviorists find so trying. These views domi-nate the second half of the paper; but before pursuing them, Iwill briefly address two other points.

First, I concur with Turk & Salovey (commentary in thisissue) that Rachlin has completely misrepresented the cog-nitive-behavioral approach presented by Turk, Meichenbaum,and Genest (1983).

Second, Rachlin's division of pain into two types is an over-simplification. Rachlin believes that there are two distinct kindsof pain. The first he calls "sensory pain, " and he asserts that thiskind of pain by itself includes no distress, no anguish, noemotion: "In these instances people commonly say that the painexists but does not bother them"; and again, "the pain is there, isintense, but is not aversive." (Notice that he is not referring tosimple sensations in these instances - heat, cold, etc. - but to"pain.') The "bother" or "aversiveness" Rachlin includes as aseparate kind of pain, which he calls "psychological." Thearguments advanced for this dichotomy are specious and super-ficial. The two-component model is introduced as followingprimarily from anecdotal reports that people can separate painfrom its bothersomeness. Such reports suggest the existence ofan interesting phenomenon, but the readiness of a behaviorist totake them at face value is puzzling, especially in light of evi-dence, which Rachlin himself cites, that people are frequentlyinaccurate in such introspective accounts (e.g., Nisbett &Wilson 1977).


Coinmentan//\{i\ch\m: Pain and behavior

As support for the sensory/psychological dichotomy, Rachlindescribes Tursky, Jamner, and Friedman's (1982) intensity andreaction scales. He notes that the "intensity and reaction scalesare separable and seem to correspond to the two components ofpain discussed previously. " The mere measurability of twoseparable components does not indicate that they mean whatRachlin suggests they do. If subjects are asked to report sepa-rately the "intensity" of pain and their "reaction to it usingscales with intensity-related and aversiveness-related adjec-tives, respectively, what might their reports on these scalesindicate? Faced with having to report separate components, itseems plausible that subjects might look to the intensity of thesensation (e.g., heat, cold, pressure, or whatever is concurrentwith the pain) as a means of providing intensity reports. Theymight, in other words, use the "intensity" scale to report theintensity not of the pain but of the physical stimulus, as reflectedin simple sensation (which pain is not). The "reaction' reports,on the other hand, might actually be reports of pain intensity(which is, of course, not the same thing as sensation intensity).Or the reaction reports could be primarily reports of affect, or acombination of affect and motivation. There are several pos-sibilities besides Rachlin s sensory and psychological pain di-chotomy. The requisite validity information to choose amongthem is not available.

The most compelling argument against the use of the senso-ry/psychological dichotomy is the straightforward issue of defi-nition: "Pain" means an aversive or distressing perception. Tospeak of pain without anguish - affeetless pain - is to stopspeaking about pain. Affect, motivation, belief, and other psy-chological factors play a role in pain. Their role is, however,integral to all pain, not to one type.

But to get back to kicking the behaviorist, it is the final portionof Rachlin s paper that is particularly troublesome. Rachlin'scentral position here is that pain equals certain overt behaviorsof the whole organism. This position necessitates a rejection ofthe integrity of the statement "I can feel pain without behav-ing. "Behaving" refers to overt behavior here, since Rachlinrepudiates previous behaviorists' inclusion of internal or covertevents in behavioral theories. The point is, according toRachlin, that if someone is to claim, credibly, that he is in pain,he must be evidencing overt, observable pain behaviors.Rachlin says "I can feel pain without behaving' is analogous to"the moon is larger when it's on the horizon than when it's highup in the sky." It appears he is suggesting that the person"feeling" but not "showing" pain is simply in error. This state-ment about pain, Rachlin implies, like the moon watcher'sstatement about the size of the moon, is mistaken, based onfaulty evidence. Presumably, the rig/it evidence would be onlyovert behavior. So that if one wishes to determine accuratelywhether he is in pain, he should do something like look in amirror to determine whether he is showing behavioral signs ofpain: grimaces, a limp, clutching, and so forth. The absurdity ofthis seems patent to me, but since it must not be to Rachlin, letme go a step further.

Apparently, Raehlin's position denies one's ability to label anexperience such as pain correctly from internal cues. Yet there isevidence that people can and do use cues unavailable to observ-ers in judging their states. For example, it has been found thatsubjects can make extremely accurate judgments about theirlevel of alcohol-induced impairment when they attend to inter-nal sensations (Mann, Vogel-Sprott & Genest 1983; McCollam,Burish, Maisto & Sobell 1980). (Subjects in these studies wereunable to distinguish a glass of real alcohol and mix from adisguised placebo, and therefore could not have been usingamount of alcohol consumed as an external cue.) In a similarvein, one can use entirely internal cues to make claims aboutones myopia. In these instances, the judgments of individualscan be verified by objective data concerning alcohol digestion,or measurements of the eye, respectively. Similarly, pain pa-tients use internal information to rate their pain, and their

ratings are frequently highly discrepant from those of observers(Lenburg, Glass & Davitz 1970; Teske, Daut & Gleeland 1983).Reliable observer judgments account for only 10-16% of thevariance in chronic and acute pain patients' self-reports (Teskeet al. 1983). Is one then to conclude that patients are largelymistaken about their own pain levels, that trained observersknow better? If one can judge alcohol impairment accuratelyfrom internal information, why should it not be possible to judgepain similarly? Or is one not really impaired, or not reallymyopic, or not really in pain, unless others verify it fromobservable, overt behavior?

ACKNOWLEDGMENTThe author thanks John Mills, lor his enthusiastic discussions of theseissues and suggestions concerning a draft of this paper, and GlennPancyr, for his helpful comments.

Pain's composite wheel of woe

George GrahamDepartment of Philosophy, University of Alabama, Birmingham, Ala. 35294

In his target article Howard Rachlin defines a thesis he calls"behavioral theory for pain and argues that it is compatiblewith alleged counterexamples to the identification of pain withbehavior. He contends that, with some refinement, the follow-ing is true:

(R) S is in pain = S pain-behaves.Rachlin denies that being in pain means that one is pained (or

that there is a pain) in one. In addition to (R), he holds that(Rl) Pain behavior is overt.(Rl) can be separated from (R). A behavioral theory for pain

does not have to restrict itself to overt behavior (see Graham1982). And it shouldn't restrict itself to overt behavior. This isbecause pain includes a sensory component that takes placebeneath the skin. It feels like something to be in pain, and suchfeelings are neither identical with nor always displayed in overtbehavior.

Rachlin recognizes that pain includes a sensory component,holds that the component is (respondent) behavior, but, oddly,does not admit that the sensory component occurs beneath theskin. The sensory element of pain is not incompatible, he says,with (Rl).

In certain respects, restricting pain to overt behavior is soundadaptationist thinking (see Dennett 1983). To an extent, pain issimply a mechanism for keeping creatures alive, and this re-quires appropriate injury-avoiding movements of the wholeorganism. It requires overt behavior. However, there remainsroom for multiple adaptive strategies, and some of these allowfor pain or part of pain to take place beneath the skin. This sort ofpoint can be made quickly by comparing pain with digestion.Digestion is adaptive and leads to movements of the wholeorganism, but it is not overt. Digestion is not some fifth wheel.But this does not mean that digestion is some sort of spatialreorientation of the whole creature.

In general, pain should make a difference in movements ofthe whole organism. But this goal can be satisfied by pain'ssensory component occurring beneath the skin.

The idea that the sensory component of pain is overt behavioris unpersuasive, though motivated by the intelligent desire toavoid thinking of pain as some sort of fifth wheel. Persuasive orat least extremely suggestive, however, is Raehlin's additionalthesis that pain contains both respondent and operant compo-nents. Philosophers sometimes distinguish between sensoryand aversive components of pain or between pain qualia andattitudes directed toward pain qualia (Graham & Stephens 1984;Thomas 1978). Raehlin's hypothesis is that this distinctionshould be understood in terms of the distinction betweenrespondent and operant behavior. This suggests the following


Coimnentary/Rachlin: Pain and behavior

analysis of pain. Pain includes a sensory component that hasquality and magnitude. We are all aware of variations in themagnitude of our pain. We feel certain pains strongly or mildly,deeply or slightly. This quality of pain is respondent behavior.The magnitude or intensity with which a pain will be felt is notaffected by the consequences of the feeling. It also cannot hebrought under voluntary control. According to this view, I willtake my pain to have a certain magnitude even if, for example, Ihave been threatened with death for doing so; or even if I havebeen punished for feelings of that magnitude.

The aversiveness of pain constitutes a second component ofpain; or in related terms, the desire to avoid or withdraw from asource of painful stimulation is a second component of pain. Thisaversiveness, desire, or tendency to withdraw is operant behav-ior. The tendency can be manipulated by manipulating itsconsequences. We wish to withdraw if we have been reinforcedfor doing so. We will not wish to withdraw if we have beenpunished for doing so.

There is a widely prevalent view of pain that conflicts with theview suggested by Rachlin. This is the idea that sensory andaversive components of pain cannot be distinguished. We can-not measure the magnitude of pain independently of the state'sbeing more or less aversive. Aversiveness and intensity com-pose a single dimensional quality. Rachlin plausibly interpretsthe studies of Tursky, Jamner, and Friedman (1982) to discon-firm the prevalent view. It is not that, like pleasures, pains areunaversive; the aversiveness of pain is essential to pain. It israther that aversiveness is not all that there is to pain. Inaddition to pain's aversiveness, pain feels to us to he more or lessintense, and this feeling and pain's aversiveness do not neces-sarily covary.

It is worth noting that Rachlin's contrast between respondentand operant components in pain can be interpreted indepen-dently of the idea that they are behaviors. For those unattractcdto calling the sensory-feeling component "behavior, " Rachlin'scontrast can he read as a contrast between elements of pain withdistinct behavioral properties. Pain's qualia are unmoved byconsequences. Pain's aversiveness is affected by consequences.

Rachlin's contrast is also more general than he makes itappear. The method of distinguishing between respondent andoperant features of pain promises to capture precisely how muchmixture there is in other bodily sensations as well as certainemotions. Consider being afraid. Normally we distinguish be-tween the feeling of being afraid (fear) and other responsesinvolved (such as the tendency to escape). Suppose we rein-terpret this distinction as a distinction between respondent andoperant elements of being afraid. The reinterpretation postu-lates that we should find some elements (e.g., the tendency toescape) to be susceptible to operant conditioning, and otherelements (e.g., the feeling) to be immune to such conditioning.Furthermore, we should expect people to report that changes inthe magnitude of the feeling are not (always) matched bychanges in the tendency to escape. (Fans of horror moviessometimes report having intense feelings of fear hut no desire toleave the theater. Quite the contrary: keeping one's eyes on thescreen is reinforcing.) Emotions such as fear as well as bodilysensations (such as those involved in perceptions of heat andwarmth) might have ingredients of this sort. Philosophers havebeen attracted to viewing emotions and various sensations ascomposites. Rachlin's reference to behavioral properties offerspossible empirical confirmation of these interpretations.

This is not the place to refine Rachlin s contrast among thevarious kinds of states to which it may apply. But I shouldmention that I may have distorted Rachlin s view of the com-posite nature of pain to the point at which he would no longerregard the view as his own. He states his view as a view ofbehavior. I state his view as a thesis about pain qualia andattitudes directed toward those qualia. He also states his view ashaving various inspirations in the behavioral literature on pain. Istate his view as having a contribution to make to the philosoph-

ical literature. My view of his view may not be his view of hisview. Yet I have the pervading sense that if his view is to gainappropriate acceptance it should be the view I attribute to him.Respondent feelings; operant behavior. It's a powerful idea.

Is pain overt behavior?

Gilbert HarmanDepartment of Philosophy, Princeton University, Princeton, N.J. 08544

Behavioral theory considers both sensory and "psychological" pain tohe overt behavior. . . .

RACIII.IN, p. 48

Once, when I was a boy, I pretended to have caught my hand inthe crack of a door in order to deflect my parents' attention froman embarrassing line of questioning. I yelled and jumpedaround a bit. In fact, I had not caught my hand in the door andwas in no pain at all. Did I engage in "pain behavior"? If not,why not? If so, does "behavioral theory" imply, falsely, that Iwas in pain?

When I was somewhat older, I occasionally got very severeheadaches which seemed unaffected by aspirin, so I learned justto wait them out. I do not believe I ever mentioned them toanyone or behaved differently in any way from how I would havebehaved without a headache. More recently, I have learned thatthe headaches are connected with sinus congestion and can berelieved by taking a decongestent or by inhaling water vapor.Before I learned this, was there overt "pain behavior" of mineassociated with these headaches? If so, what behavior? If not,does behavioral theory falsely imply I was not in pain when I hadthe headaches?

Sensory pain and conscious pain

Julian JaynesDepartment of Psychology, Princeton University, Princeton, N.J. 08544

It is widely agreed among clinicians treating pain problems thatthere are at least two kinds of pain, although what pain phe-nomena fit into which category is not entirely settled. Thecategories are usually referred to as acute and chronic ornocioceptive and psychological or, as in the present paper,sensory and operant.

Professor Rachlin has written an exciting and stimulatingpaper. It is particularly commendable because he has been ableto spread his argument so as to bring together experimentalpsychology and the best of modern philosophical thinking. Butthe two main questions, I think, need further exploration.

Is chronic pain an operant? In street language this means thatchronic pain is always fulfilling some purpose of the patient,either getting sympathy or a pension, avoiding work or war,reenactinga hurt-child-caring-parent relationship in surrogate,getting noticed by nurses or family, feeling important withimportant-sounding medicines, or, as emphasized in the paper,obtaining medication, particularly narcotics. It should be point-ed out here, of course, that all such explanations in humanpatients are interpretations after the fact. There is no experi-mental evidence that this is the case, although I would agreewith Rachlin that it often seems so.

A recent study should make us worry about oversimplifiedinterpretations, however. In the burn unit of the HarborviewMedical Center in Seattle, Dr. Alan Dimick has tried a self-administered morphine procedure. Instead of receiving fromthe nurse the standard dose of 2-4 milligrams every 2-4 hours,


CommentarylRaclnUn: Pain and behavior

the patient pushes a button to receive the drug. As Dr. Dimiekstates the results,

We are finding that if the patient is allowed to push the button and gethis own dose of morphine, the patient usually pushes the button lor alesser amount of drugs than if the physician orders it or if the nursegives it. And this method provides a much better pain control.(Dimiek as quoted in Driver, 1984)

I think this shows that what the patient is feeling is a combina-tion of sensory pain and anxiety, and what is reduced by self-administration is the latter. It is well known that feelings ofcontrol over pain diminish it (Bowers 1968). At any rate, it is notexactly what one would expect on the operant model.

There is also a logical fallacy here. The fact that operanttraining, as in the Fordyce therapy,1 reduces pain is an impor-tant and welcome result, but it does not therefore follow that thepain was an operant in the first place. Because an alternativebehavior can be learned does not mean that the original behav-ior was learned as well. For example, adaptation to a sensorystimulus that produces an innate response is indeed a form oflearning, but that doesn't mean that the original response to thestimulus was learned. What is needed to make the point is acontrolled laboratory experiment in which an introspectivelyreal pain in a normal subject is learned in order to obtain somefuture reinforcement.

Is pain merely pain behavior? The central difficulty for most ofus with Rachlin's courageous position is his insistence that painis pain behavior and nothing else. As Rachlin understands it,most of us wish to say that there is a sensory pain and then ourconscious reaction to it, and that psychological pain is at least anevent that goes on in consciousness (where the word "in" is to betaken metaphorically). Indeed, if Rachlin wished to incorporateconsciousness (or a theory of private events) in his perception ofthe pain problem and agree with me that consciousness waslearned on the basis of language at a particular point in history toobtain very specific consequences (see Jaynes 1976), then hecould call consciousness an operant (and correctly so), fittingconsciousness into neobehaviorism in a consistent way. Inanimals pain is pain behavior, but in humans every sensory painhas its analogue in consciousness as what can be called consciouspain. And then he could merge the concepts of operant pain andconscious pain.

But for this to work out to explain some of the data, thereferents of that cardinal term reinforcement, wide enough asthey already are, would have to include reinforcement by otherconscious analogues, particularly ideas of the conscious self.Then we could speak of intrinsic reinforcement by emitting anoperant that is consistent with some preconceived notion of whoone is in various senses. So successfully functioning healthyindividuals maintain their good behaviors as such behaviors areconstantly being reinforced by consistencies with good andreasonable self-images - as consciousness of their own behaviorstell them, something that can only happen with consciousness.On the other hand, patients with what most of us might think ofas detrimental behaviors such as pain or neurotic symptoms maylearn or maintain their distressing symptoms when those symp-toms are reinforced by some long-learned though perhapserroneous self-conception.

Phantom pain. Such a possibility would greatly expand thepain data that Rachlin could explain. The problem of chronicpain in amputated limbs or occasionally in breasts after mastec-tomy and sometimes even in teeth after multiple extractions isan example. Almost all amputees experience some phantomlimb sensations. Even Lord Nelson when he lost an arm in anaval engagement at Tenerife wrote back to a friend that hecould still sense his missing arm, and he took this as evidence forthe existence of his eternal soul - a considerable consciousreinforcement.2

In recent times, approximately half of amputees have phan-tom limb pain of some kind and of varying durations. It is moresevere in youth (though absent in infancy) and roughly propor-

tional to the duration and intensity of suffering before amputa-tion. Others usually have phantom itching or cramps. To me itseems clear that this has something to do with consciousness of abody image, perhaps a denial that anything is missing in thegreat wish to be whole again. I recently interviewed such a case,which I think is of sufficient interest to mention here.3

Seven years ago, B.W., then 30 years ol age, ol high schooleducation, was pushing a stalled car on the highway when a followingcar crashed into him. He was next conscious of his right foot and anklebeing close to his right eye, and wondered what they were doingthere. His pain then began. Some hours laterat the hospital, surgeonsfirst tried to save his leg but then had to amputate it just above theknee. Recovering from the anesthetic and told that his leg had had tobe amputated, he denied that this was so, claiming he could stillwiggle his toes and feel himself doing so in the absent leg. Two monthslater, because of complications, he had to have a second amputationjust below the hip. He then made a complete recovery, both phys-ically and psychologically. Presently he is well employed and is anavid amputee athlete taking part in their Olympics. He also counselsother amputees during their recovery.

But ever since his own amputation, beginning at the hospital, hehas had phantom pain. This consists of a series of 10 or 15 sharpjabbing pains in the ankle or sometimes the arch of the amputated leg.At first such series occurred quite frequently, but now after sevenyears he may go for a month without pain, followed by a month whenit happens almost every day. The pain is as severe as originally,perhaps waking him up at night, or while he's sitting down.

His wife, who is in therapy for anxiety attacks that on the surface donot seem to be related to her husband's problems, was also inter-viewed. She stated that during the attacks, "he almost passes out, is ina cold sweat, sometimes almost jumping off the couch with it. Thewife seemed to exaggerate the pain of her husband while the husbandspoke about it in a matter-of-fact way, not being particularly in-terested in any therapy for his phantom pain.

A possible operant scenario could be constructed, perhaps, that thehusband's problems with his amputation were keeping the coupletogether, that his pain was being reinforced by his wife's concern,perhaps contributing to her own problems. But on further question-ing, he revealed something that neither husband nor wife hadrealized before, that his attacks of phantom pain occurred only whenhis prosthesis was unattached.While an operant explanation is possible for this particular

case, it could not be complete without consciousness and theimportance of a whole body image. This then would allowphantom pain to fit into operant terminology, particularly whenone remembers the denial and the wiggling of toes with whichhis recovery began. It is as if having taken off his prosthesis,B. W. is reminded that he does not have a whole body. And thenoccasionally later, as if he were asking himself like an unbeliev-ing child, "Have I really lost my leg?" the phantom pain, as anemitted response, is proof for a few minutes that the leg is stillthere, the momentary consciousness of which is the reinforce-ment. Thus, through seeing consciousness an operant, thesephenomena could be included in an operant explanation.-4

But when I consider some of the extreme debilitating forms ofchronic pain, such as phantom pain in some paraplegics or afteravulsion of the brachial plexus, I have to part company fromreinforcement ideas. Such pain seems far too severe for anoperant theory to deal with.

An alternative view. I would like to suggest a slightly differentcategorization that sorts things out in a way different from acuteversus chronic or sensory versus operant. I would propose thedistinction I have mentioned above between sensory pain andconscious pain. Animals and early humans had sensory pain. Butwith the advent of consciousness around the beginning of thefirst millennium B.C. began sensory pain plus its analogue inconsciousness, which complicates all pain with anxieties, de-pressions, hopelessness, anger, memory flashbacks, and feel-ings of attack upon the conscious self. And this is complicated bythe evidence that sensory pain can be delayed or even abolished


Commentary/\{a.ch\in: Pain and behavior

by consciousness - evidence that the gate-control theory wouldexplain in a different way. This new type of pain and pain controldevelops in history at the same time as there are large changes inthe nature of emotions (Jaynes 1982). Pain in the conscioushuman is thus very different from that in any other species.Sensory pain never exists alone except in infancy or perhapsunder the influence of morphine when a patient says he has painbut docs not mind it. Later, in those periods after healing inwhich the phenomena usually called chronic pain occur, wehave perhaps a predominance of conscious pain.

Recruitment. But consciousness is not something floatingabove and apart from bodily physiology; it is part of it andconstantly interacting with it, although we are not even close tounderstanding its neural substrate. One of the phenomena ofconsciousness is called recruitment, as when, for example,someone fantasizing about sexual behavior will thereby occasionphysiological responses in erectile tissues, which then continueto recruit other physiological responses together with hormonalchanges, the feedback from which keeps cuing the fantasy untilit is difficult to turn conscious attention to other things. Theprocess is similar in pain. Consciousness of pain, cued by a hostof emotional and situational and hormonal variables, recruitssensory pain until it is difficult to turn attention away.

A common example is a patient saying that on awaking in themorning he is free of pain. But as soon as he thinks about gettingout of bed, the pain begins. Then he remembers that he is ill,that he doesn't feel well the way he used to. And once out of bedhe may experience low-intensity pain until the phone rings, atwhich time there is a sudden dramatic increase.

Or a woman who finds a lump in her breast and worries if it iscancerous. She may. suddenly feel pain in the breast that maythen increase in severity and even spread to the shoulder andarm over time. Later, if her physician assures her that the lumpis of no consequence, there is usually a sudden total relief fromthe pain (Melzack & Wall 1983, p. 247). In this view, thedifference between acute and chronic pain resides in the differ-ent ratios of sensory and conscious pain as well as their instiga-tion. Thus the dichotomy could be sensory-instigated for acutepain and consciousness-instigated for chronic pain. So by chang-ing consciousness through imagery or other means, one shouldbe able to decrease the latter.

Predictions. If this is the case, then certain dramatic predic-tions follow. If consciousness was learned sometime after 1000B.C. (as I argue, Jaynes 1976), there should be no evidence ofconscious pain or chronic pain in texts around or before thatdate. Indeed, in ancient Proto-Indo-European (before 2000B.C.) there is no word for pain or hurt at all, although there arewords for wounds and cuts (American Heritage Dictionary,Appendix). And in the old part of the Iliad, written down about850 B.C. hut relayed by oral tradition from earlier times, thereare extremely gory descriptions of bloody woundings and terri-ble disembowelments, but hardly any notice of discomfort socaused. Rachlin is right for ancient Troy: Pain is merely painbehavior. In contrast, one should look at Plato's Philebus ofabout 350 B.C. for a quite modern-sounding discussion of painand suffering (as well as a remarkable description of itchingbeginning at 46D).

Also, neither chronic pain nor phantom limbs should appearin infants before the age of about three, when consciousnessbegins to be learned. There is some evidence that this is true(Simmel 1962). Also, in subhuman animals there should be nochronic pain and no phantom limbs: Before the present pro-liferation of veterinarians, one used to see three-legged dogsoften, but I know of no observations of such a dog turning to lickor nibble the absent leg, as would a normal dog with a hurt leg.

How does conscious pain generation work? Probably in manyways. Association and recruitment is the most obvious. Perhapslanguage is also a medium of such unfortunate learning. Thepatient asks himself, do I still hurt? and back comes the painwith such immediacy. In fact, I suggest that that is precisely

what it is, Pavlovian conditioning rather than opcrant condition-ing. In the period of sensory-instigated pain, sensory painproduces conscious pain. But sensory pain is being paired withverbalizations, such as the statement "I hurt." So that later sucha verbalization, even in monitoring form, produces the con-scious pain response, which then could be followed by recruit-ment as well as perhaps being maintained by the opcrantreinforcements mentioned by Rachlin.

If this is so, we can see why Fordyce, at least in his originalstudy (Fordyce et al. 1968), did not ask his improving patientswhether they still felt pain, since that is precisely the kind ofquestion and solicitousness that his opcrant training is trying toabolish. It is a Heisenbergian situation. And yet that stillremains the question to which most of us wish to know theanswer.

NOTES1. For a critical review of Fordyce and other operant techniques in

the treatment of pain, (as well as a masterly review of the entire subject)see Melzack and Wall (1983) 333-37.

2. Nelson also lost an eye but never reported on whether he sensed aphantom eye. I once knew a gaunt elderly blind man who roamedaround Wiltshire with the help of a young boy. Both eyes had beenenucleated. One morning he leaned over to me, stretching open one ofhis empty sockets witli his fingers, and said, "You see - they're growingagain! Just about the size of pearls! I saw them in the mirror thismorning. I suggest this is an instance of phantom anatomy and itssubstantiation by "sensation.

3. I met with these patients in my capacity as a consultant with theRichmond Center of Charlottetown, Canada, directed by WilliamLawlor. I am grateful to Dr. Wayne Matheson for discussion on thisproblem, as well as to Dr. Frank Wheelock of Boston.

4. I should point out that there is an alternative, more sensoryexplanation to the case of B. W. without his prosthesis. This is that thetactile stimulation around the stump with his prosthesis on couldpossibly have eliminated the phantom pain. While there is no evidencethat such mild tactile stimulation can function this way, there is evi-dence that vigorous vibration (Russell & Spalding 1950), or poundingthe stuinp many times as was done a few decades ago (on the theory that"the nerves would wear out"), or electrical shock to the stump (Wall &Sweet 1967) bring relief from phantom limb pain in some cases. Perhapssuch violence provides an insistent reminder to some deep level ofconsciousness that the limb stops there. In any case, this phenomenon ofrelief of phantom limb pain by a functional prosthesis has not, to myknowledge, been reported before and should be researched.

Chronic sensory pain

Patricia KitcherDepartment of Philosophy, University of Minnesota, Minneapolis, Minn.55455

The target article raises four different considerations that mightsupport a behavioral theory of pain. Rachlin offers a possiblefunction of pain - the "fire alarm " theory - that would make itnatural to think of pain as overt behavior. He reports some fairlyimpressive results from behavioral therapies for long-term pain.The fact that infants naturally and invariably respond to injury,hunger, and scary stimuli by crying or whimpering lends cre-dence to the idea that pain involves reflex elements. And thereis a pretty good fit between the properties of sensory pain andrespondents. Like respondents, sensory pain is fairly constantacross subjects and across times.

The function argument. I do not think that Rachlin intends thisloose discussion to bear much weight. My point is that it cannotbear any weight. The suggestion is that pain has a certainfunction as a fire which demands immediate response and onlybelated criticism. If this function story is accepted, the identifi-cation of pain with behavior seems quite natural. Covert firealarms would be an evolutionary mistake. But this kind ofstorytelling has no probative value. The data Rachlin explains byreference to this supposed function admit of obvious alternative

THE BEHAVIORAL AND BRAIN SCIENCES (1985) 8 1 63

Commentanj/RacMin: Pain and behavior

explanations; for example, in eases such as the sports injuriescited by Rachlin, what compels immediate attention is thepresence of an injured player. Pain is not the "fire alarm ; theinjury itself is what compels the aid of observers. (Think of caseswhere the player is unconscious.)

The argument from therapy. Rachlin reports that behavioraltherapy has a good success rate with psychological pain. As hepresents this material, however, it does not provide support fora behavioral theory against its rivals. Physiologists will point outthat morphine docs a very nice job of suppressing pain, so thatshould be evidence that pain is some physiological condition.(That morphine is addictive and progressively less effectiveundermines its utility as a long-term solution to chronic pain. Itdoes not impugn the evidence that pain may be a physiologicalcondition because it can be affected by chemical intervention.)A cognitivist will reply that the dramatic changes in surround-ings will affect the cognitive economy of the pain sufferer, so thatone does not really know whether changing the contingencies ofreinforcement alters the cognitive structure, which in turnalleviates the pain. Raehlin's presentation of Fordyee's methodslends some credence to this move. One of the crucial steps isrecord keeping so that the relation between pain behavior andreinforcement, of which the patient is usually unaware, is mademore "vivid to him.

Infants and respondents. Skinner distinguished respondentsfrom operants by two criteria, temporal order and the flexibilityof the behavioral response. Respondent behavior is fixed and"natural." Respondent conditioning is just a matter of .stimulussubstitution. Skinner appealed to these features of respondentsto motivate the need for operants, which are extremely plastic(Skinner 1953, chaps. 4-6). Unlike psychological pain, sensorypain does not seem to be very plastic. Infantile responses tonoxious stimuli have a reflexive character. So, separately, theseconsiderations lend some support to the behaviorist identifica-tion of sensory pain with respondent behavior. When con-joined, however, these considerations cut against each other.

Rachlin tries to defend the behaviorist theory of sensory painby taking on a favorite philosophical foil, the super-Spartans. Ithink he should be more worried about an extraordinary Vien-nese. Unlike the patients whose recoveries from chronic psy-chological pain Rachlin describes, Freud's chronic pain wassensory. So on the behavioral theory, we should be dealing witha respondent behavior. But Freud's reaction to the pain in hisjaw was nothing like the tearing reflex in the presence of onions,to use one of Skinner's paradigms. Even if Ernest Jones exagge-rates his hero's stoicism, it is probably true that Freud spentlittle of his sixteen-year battle against cancer crying and whim-pering (Jones 1961, e.g., p. 529). The behaviorist can deal withthis case in one of two ways (or by a combination of the two). Theclaim could be that besides crying and whimpering, babiesmake other overt responses to injury, hunger, and so on. Thenthe claim might be that these respondent elements are presentin the behavior of an adult stoic dealing with painful stimuli. Theother move would be to suggest that, with maturation, onereflex response to painful stimuli is replaced by another. So,with infants, the respondent behavior is crying and whimpering;in adults, the respondent behavior is some more subtle but stilldetectable pattern.

Rachlin's discussion implies that he would take the lattercourse. This is fine, but now one wants some account of the adultrespondent behavior. Departing from Skinner's molecular be-haviorism, Rachlin would claim that Freud might be respondingto a painful stimulus even though he is not doing anything rightthis minute. Still, whatever the adult reflex is to painful stimuli,Freud should have engaged in this behavior with increasingfrequency as his cancer grew worse. What could it be? To matchthe characteristics of sensory pain, there should be some unifor-mity across subjects and across time. This does mean thateveryone must make exactly the same response to painfulstimuli, but something should be constant, perhaps the rate of

responding. What I find unpersuasive about Rachlin's presenta-tion of this theory is that there is no attempt to specify what therespondent is, even though it must be overt behavior. Overtbehavior can still be quite subtle, something like gazing intomiddle distance, changing fixation, fidgeting mildly, or engag-ing in a subtle repetitive motion. Behaviors of these sorts mightwell have escaped the notice of Freud's daughter and friends.Such activities of patients with sensory pain are certainly avail-able for study, however. Could Rachlin and his colleagues showsome solid correlations between the intensity of the painfulstimuli and some parameter of overt behavior, then the case foridentifying sensory pain with that respondent would really havesome force.

Rachlin could reply with a tu quoque: There is no completephysiological account of pain, and no complete cognitive ac-count of pain either. But physiological processes and cognitiveprocesses are both inner, hence harder to detect. By contrast,the behavioral theory identifies pain with something overt.Hence the lack of substantive theory is more worrisome.

Pain behavior: How to define the operant

Hugh LaceyPhilosophy Department, Swarthmore College, Swarthmore, Pa. 19081

Among psychological research programs committed to compre-hensiveness but not necessarily to physiological reductionism,only cognitive science, many believe, remains viable. Againstthis trend Rachlin argues that radical behaviorism, interpretedas resting upon correlations rather than contiguities betweenthe environment and behavior, still remains a competitor. I willcriticize an important feature of Rachlin s account of pain, butthe criticism will leave open his contention that behavioristtheories of pain have greater merit in certain respects thanphysiological and cognitive alternatives.

Rachlin maintains that pain has an operant component be-cause Fordyce's (1976) procedures for the treatment of painshow that there is a class of behaviors, pain behaviors, whosefrequency of occurence is a function of contingencies of rein-forcement. It is also implicit in Rachlin's analysis that all behav-iors ordinarily taken to be expressions or effects of pain belong tothis class (except for those that are explicable in terms ofrespondent conditioning.) It follows that all pain behaviors arelawful, the relevant laws relating the behaviors only to environ-mental variables, with no need for recourse to internal vari-ables, whether physiological, cognitive, or phenomenal. This isin accordance with what Rachlin takes to be a central tenet ofradical behaviorism. Whether or not this would constitutesufficient grounds for identifying pain with pain behavior, itwould - if certain conditions were fulfilled - be sufficient tojustify a theory ofbehavior eschewing all reference to pain as aninner cause, unreduced state, mental event, or conscioussensation.

One of these conditions would be that the class of painbehaviors should be definable in terms that can be appliedindependently of such characterizations as "behavior caused bypain or "behavior that expresses pain, and hence that mem-bers of the class should be identifiable without being parasitic onascriptions of pain to the sufferer. According to the usual radicalbehaviorist tenet, the terms should presumably also be ex-clusively physicalist (interpreted broadly to include any termsordinarily ascribable to physical objects) and not occur in inten-sional contexts. (A context is intensional if the truth value of asentence within it can vary depending on the expression used todescribe a relevant object. For example, "taking his medica-tion and "drinking a light brown liquid" may both adequatelydescribe a particular event, but "He went to the kitchen in orderto take his medication ' mav be true and "He went to the kitchen


in order to drink a light brown liquid" false. Expressions thatinvolve reference to a person's mental states characteristicallyoccur in intensional contexts; see also Place 1981.) Fordyce'streatment procedures might work without this condition beingsatisfied, but the dispensability of the "mentalist" idiom of painin a behavioral theory requires it.

How is the class of pain behaviors defined? Rachlin does notsay. Clearly the class is large and varied, including an assort-ment of verbal behaviors. It is unlikely that there is a list ofnecessary or sufficient (physicalist) conditions for a behavior tobe a pain behavior. But, it will be argued, in behavior theory,classification is based on functional, not topographical, consid-erations. The topographical description of a rat's behavior is notpart of the definition of the operant "bar press." Any behavior ofthe rat belongs to this operant provided that it produces theeffect of pressing the bar, and it is upon the production of thiseffect that reinforcement is contingent. However, pain behaviorcannot be defined by any common effect since there is no suchrelevant effect, and looser formulations (e.g., whatever is rein-forced under specific conditions) are always on the edge oftautology.

I cannot see how the class of pain behaviors can be definedunless its instances arc characterized as expressions or effects ofpain. Moreover, it seems to me that certain behaviors will beidentifiable as pain behaviors only under intensional description(a description based on the intention, not the effects, of theaction, e. g., taking medication), and that many verbal behaviorscan be classified together as pain behaviors only in virtue of theirexpressing similar meanings or intentions. A similar point aboutthe classification of verbal behaviors arose when Rachlin and Iattempted to apply a behaviorist theory of choice to contextswhere speech played a role in the generation of choices (Lacey &Rachlin 1978). This, of course, is not a conclusive refutation ofRachlin's thesis, but it draws attention to a serious weakness inthe present supporting argument.

The thrust of my critique is not necessarily to advocatemoving toward theories that deal with inner causes. My argu-ment is that the classifications required by radical behaviorismcannot be made without relying on an intensional idiom. It is notthat the lawfulness of behavior cannot be expressed withoutrecourse to internal events. Radical behaviorism maintains thatan intelligible order can be found at the molar level, repre-sented in laws correlating behavior and the environment. Cog-nitive science denies that there is a general intelligible order atthe molar level because the laws of behavior involve internal,cognitive states as well as environmental variables.

An alternative position is that there is intelligible order to befound at the molar level, not represented in terms of laws thatexpress regularities, but in terms of teleological and otherintensional categories. On this alternative, when we say that X isin pain, we attribute to X a state that makes intelligible a widerange of behavior; it shows that certain behavior is apt, appropri-ate, and efficacious in the circumstances, that it is conducive tocertain goals, that X needs to be treated in a certain way, and soon. The part of pain talk that Rachlin wants to absorb underoperant conditioning then appears as part of intensional andteleological discourse; perhaps that is why it has been so elusiveto those who have tried to contain it within the quest for laws.From this perspective, the mistake of radical behaviorism is toattempt to reduce teleological explanations to lawful correla-tions between behavior and environment. At the same time, theperspective is compatible with there being regular correlationsbetween some behavior (classified on the basis of intensionaldescriptions) and the environment. Hence it is compatible withthe success of Fordyce's treatment procedures. This success,therefore, does not uniquely support the radical behavioristaccount of pain.

ACKNOWLEDGMENTThis commentary was prepared with support from NSF (SES-8308604).

Commentary /Rachlin: Pain and behavior

Against dichotomizing pain

John D. LoeserDepartment of Neurological Surgery, University of Washington, Seattle,Wash. 98195

As angels dance on the head of a pin, so psychologists cavort onthe meaning of pain, gyrating about the theoretical machina-tions of their cultural heritage. The rhythms of Aristotle, Des-cartes, and Wittgenstein are sure to be invoked, not to overlookthe Melzack-Wall (1965) gate hypothesis, which has become derigueur. How have these attempts helped the practitioner? Domedicine and psychology offer better therapies today than in thepast? If not, why has this area of health care delivery lagged sofar behind treatment of infectious diseases or sanitation?

The answer, I believe, lies in the creation of the problem asproposed by Rachlin. The dichotomy of pain, whether in thecognitive, physiological or behavioral model, simply does notlead the practitioner to rational conclusions; super-super-Spar-tans are irrelevant. Experimental pain, even that carried out onhuman volunteers, looks at only one, often trivial, aspect of theclinical phenomenon; the relationship between the stimulus asidentified by the experimenter and the response of the subjectdefined as relevant by the experimental paradigm. What is sooften omitted is the intervening human being.

There is ample clinical evidence to discriminate betweenacute pain and chronic pain due to benign diseases. Chronicpain due to cancer is really acute pain occurring on a long-termbasis. Modern medicine has dramatically improved the manage-ment of acute pain: local, regional, and general anesthesia aswell as narcotics have permitted painless surgery, childbirth,repair of trauma, and so on. Chronic pain, and the disabilityassociated with it, appears to have reached epidemic propor-tions. Has the physiologic or cognitive or behavioral basis ofmankind changed in the industrialized Western world? Why ischronic pain unknown in nonindustrialized parts of this planet?Why has chronic pain never been recognized in infrahumanspecies? In short, there is increasing evidence that the strategiesmounted by physicians to treat chronic pain, which are basedupon the successful treatment of acute pain, not only do notalleviate but may actually exacerbate chronic pain.

As Lewis Carroll illustrated, the first step in effective commu-nication is some agreement as to the meaning of words. In asociety that uses the phrase "a pain in the neck," Rachlin'sdichotomy is clearly inadequate. Many patients seek health carebecause of chronic "pain"; few have evidence of tissue damage.Many are depressed, often in part owing to the inappropriateprescription of medications and inactivity. Like it or not, we canmake a statement about another individual only if that indi-vidual does or does not do something. That event is behavior -in this case, pain behavior. Our diagnostic question is not thevalidity of such an event (which is what underlies almost allphysiological-psychological dichotomies) but what are the fac-tors which may contribute to such behavior. Tissue damage(nociception) is one, but it is usually operative only in cancerpain or acute pain. Injury to the nervous system is another:There is no tissue damage where it hurts. Instead, alteredfunction in peripheral or central nervous system appears togenerate the behavior (consider phantom limb pain). Sufferinginduced by depression, anxiety, fear, or isolation can alsogenerate pain behaviors. Finally, as Fordyce (1983) has soclearly shown, once pain behaviors are generated the environ-ment can play a major role in their perpetuation. If the brain isthe organ of behavior, all pain behaviors are real and involvephysicochemical processes. The distinctions between sensoryand psychological, cognitive, physiological, and behavioral arenot conducive to an increase in our understanding of the prob-lem of pain. In fact, they have created our dilemma.


Commciitary/Raclihn: Pain and behavior

Functional behaviorism: Where the pain isdoes not matter

A. W. LogueDepartment of Psychology, State University of New York at Stony Brook,Stony Brook, N.Y. 11794

Rachlin s attempt to answer some of the criticisms that cognitivepsychologists and philosophers have made of behaviorism, par-ticularly with regard to the characteristics of pain, differs in twoways from traditional behaviorism (see Logue, in press a, inpress b). One of these differences, molar behaviorism, greatlystrengthens the position of the behaviorists, while the other,identifying pain aversiveness with operant pain behavior, turnsaway from the original purpose of behaviorism.

The first change is the formal statement of what Rachlin callsmolar behaviorism. According to molar behaviorism, correla-tions between stimuli and responses are critical factors in deter-mining what an organism does. Thus, events in the distant pastcan be seen as affecting an organism's present behavior. Such aconceptual scheme is extremely helpful in studying pain behav-ior, for the events that caused the pain and behavior that signalsthe occurrence of those events can be widely separated in timeyet still be described within a behaviorist framework.

Molar behaviorism may be particularly appropriate forhuman behavior because, owing to humans' highly sophisti-cated language, events that are temporally distant, or evenevents that have occurred to other people, can easily be commu-nicated and can thereby influence a particular person's behavior(Catania 1983; Lowe 1983). Like someone's own experiencewith regard to the aversiveness of a particular stimulus, covertlanguage is unobservable, yet both the presence of pain aver-siveness and the presence of covert language sometimes appearto influence the effects that various current environmentalstimuli have on responses. For example, owing to what appearsto be their covert language behavior, humans are sometimesable to wait for the larger, more delayed reinforcer instead ofchoosing the smaller, less delayed reinforcer in a self-controlparadigm (Mischel 1981; Logue, Pena-Correal, Rodriguez, &Kabela 1984). In the case of pain aversiveness, someone maycomplain of a headache although there is no evidence of tissuedamage. Both pain aversiveness and language appear to beresponsible for the frequently observed lack of a preciseequivalence between the current environment and overt be-havior.

Yet this lack of precise equivalence between the currentenvironment and behavior need not result in the abandonmentof behaviorism. Both language and pain aversiveness can beconceived of as hypothetical constructs that represent the effectof distant events on behavior, defined as the functions thatexpress the relationship between the environment and behav-ior. Psychologists must determine the characteristics of thesefunctions and the rules that govern them. In this way a scientificapproach can be retained while studying what some may wish todescribe as internal mechanisms.

While Rachlin has expanded the scope of behaviorism byformalizing molar behaviorism, he has not construed pain aver-siveness as a hypothetical construct. He has answered therepeated queries of the cognitivists and philosophers regardingthe location of pain aversiveness for behaviorists by placingpain aversiveness squarely in operant pain behavior. This ishis second change in the perspective of traditional behav-iorism.

A few summary definitions may be helpful at this point. Painin its popular usage refers to the subjective experience of boththe sensory and the aversive, bothering qualities of pain.

Rachlin s definition of pain consists of two parts: respondentbehaviors that he identifies with the sensory aspects of pain, andoperant behaviors that he identifies with the aversive, botheringqualities of pain. Note that Rachlin repeatedly uses the actualwords sensory, aversive, and bother. These words are not givenoperational definitions; they are simply "identified with" re-spondent and operant pain behavior. For Rachlin, someone whonever shows even the smallest amount of operant pain behaviorhas no pain aversiveness; pain does not bother that person. Thisis not merely a semantic distinction; for Rachlin, operant painbehavior is synonymous with the subjective experience of aver-siveness. However, Rachlin himself states that whether thisassertion is true has not yet been settled. In fact, it is a questionabout which philosophers and psychologists could argue atlength without ever reaching resolution. Some of these argu-ments are detailed in Rachlin's article.

In resolving the issue of the location of pain aversiveness it ishelpful to recall the original purpose of behaviorism; to objec-tively examine and be able to predict behavior (Watson 1913).Toward this end Skinner (1979) defined operant behaviors interms of their function on the environment. For example, alever press was defined as all those behaviors that resulted in thedepression of a lever.

Therefore, in terms of traditional behaviorism, the issue is notwhere pain aversiveness actually is or what it acutally is, but howvarious behaviors can best be organized to predict future behav-iors. Simply because people s statements regarding pain aver-siveness are modifiable by contingencies of reinforcement doesnot prove that the subjective experience of pain aversiveness ismodifiable behavior or that pain aversiveness consists of behav-ior that is observable given our present techniques. It is impos-sible to prove that an unobservable feeling (aversiveness) hasbeen modified, or of what it consists, or even that it exists. Theoriginal behaviorists, Watson and Skinner, knew that this effortwas scientifically useless. By trying to locate pain aversive-ness Rachlin is himself making the error that they sought toavoid.

To summarize, if someone posed the question of whether aperson who never demonstrated any observable operant painbehavior was experiencing pain aversiveness, Rachlin would sayno, but a traditional behaviorist would say that the person hadnot been behaving in the way that was likely given that tissuedamage had occurred. In other words, the traditional behav-iorist would say that the person was not demonstrating the set ofbehaviors consistent with the organizing concept, the hypo-thetical construct, which has been named here pain aver-siveness. The traditional behaviorist would not discuss whetherthe person was actually experiencing pain aversiveness becausethat could not be known. All that can be known is what a persondoes and the extent to which those behaviors are usually corre-lated with actual tissue damage.

It may be helpful to call this version ofbehaviorism functionalbehaviorism (see also Roback 1923) because it focuses on:

1. organizing behavior according to its function2. the variations that occur in the functions relating environ-

mental events to behavior3. behaviorism as scientific psychology that gets the job

done, that fulfills the function of predicting behavior.Functional behaviorism retains the scientific character ofbehaviorism while encouraging investigation of behaviorsthat have been identified with so-called intei ial, unobser-vable events such as pain aversiveness and covert lan-guage. It is therefore likely to succeed in predictingbehavior where other versions of behaviorism are not.

ACKNOWLEDGMENTComments by H. Rachlin, M. Rodriguez, and M. Smith on a previousversion of this paper are appreciated.


One pain is enough

Wallace I. MatsonDepartment of Philosophy, University of California, Berkeley, Calif. 94720

Rachlin holds that a pain is really two pains - one the sensorykind (or component), which is respondent behavior, the otherthe "psychological" kind, which is operant behavior. But theevidence cited suggests that, on the contrary, there is onlysensory pain, to which the sufferer reacts - makes a judgment,adopts an attitude - in a manner that is not always a linearfunction of the intensity of the painful stimulus. This reaction isoperant behavior, no doubt, but it is not another pain or anothercomponent of the pain.

The dual pain theory, Rachlin claims, is supported by theseparability of the intensity and reaction scales: Since subjectstend to assign the same intensity-scale positions to the samestimulus but differ about degrees of aversiveness, the scalesmust be measuring different pains or pain components. But, ofcourse, the opposite inference should be drawn. There cannotbe a pain or pain component that is simply severe but neitherbearable, uncomfortable, . . . , nor agonizing; or a pain that isintolerable but neither just noticeable, mild, . . . , nor ex-cruciating. The intensity and reaction scales are well named;they measure intensity of and reaction to the same pain, andthey would be unintelligible if they did not.

This kind of distinction has wide application. One's children'splaying of rock music on the stereo varies in intensity, and one'sannoyance thereat has its separate degree scale. One can to acertain extent decide whether to erupt or to tune it out, whichby no means makes it literally inaudible. And we distinguishpeople whose loves and hates are insipid from those who arepassionate but self-controlled.

But what, then, are we to make of Fordyee s successfultreatment of chronic pain? (Here and below "Fordyee" is shortfor "Rachlin's report of Fordyee'; I have not checked thepublications referred to.) Fordyee accepts only patients inwhom he fails to find any persisting stimulus (external or inter-nal) causing pain behavior; nevertheless, the pain behavior ischronic. Granted that the patients are not malingering, andgranted that no stimulus is found because there really isn't any,does it not follow that the patients are suffering - really suffering- from purely "psychological" pain, the "operant component"by itself, which therefore is a second pain after all?

A more conventional and economical hypothesis is available:The patients are deceiving themselves. They believe - reallybelieve - that they are in pain, but they aren't; they do notactually feel any pain. There was a time when they really were inpain. They adopted an (operant) attitude toward it that wasmanifested in typically aversive behavior. The behavior broughtits rewards of narcotics and sympathy. Now the pain is (really)gone, but in order to continue to qualify (in their own view ofthemselves) for the rewards, the patients continue the operantbehavior that brings them. Fordyee s treatment consists inremoving the motivation for the self-deception - in a gradualand gingerly way avoiding martyrdom.

This hypothesis presupposes the possibility of sincerely be-lieving oneself to be in pain without actually being in pain.Behaviorists will accordingly reject it because to them believingthat one is in pain, behaving in all respects as if one were in pain,and being in pain are all the same thing. Many nonbehavioristphilosophers will reject it on the ground that a sincere report ofpain is incorrigible.

But the ineorrigibility of the pain protocol is a myth. Insuitable circumstances (e.g., fraternity initiations in the bad olddays), nonpainful stimuli (e.g., drawing an icicle across the skin)may produce shrieks of real anguish. This is the converse of thesoldier who doesn't notice for a while that he is missing an arm.And the behaviorists respondent/operant distinction - cor-rected to eliminate the redundant second pain - explains them

Co»»ne»t«rt//Rachlin: Pain and behavior

both: The stimulus (the pain) is one thing; the attitude we take toit is another. And the attitude, the set to action, with a view to itsconsequences, can continue after the respondent stimulus iswithdrawn.

Pain and parallel processing

Ronald MelzackDepartment of Psychology, McGill University, Montreal, Que., Canada H3A1B1

Dr. Rachlin has critically evaluated a wide literature that en-compasses all the major approaches to the problem of pain -physiological, psychological, clinical, and behavioral. I agreewith many of his arguments and disagree with some. Since heusually uses the words "sensory" and "psychological" in quota-tion marks, it is evident that he does not imply that theseprocesses exclude each other; rather, they are complementaryto one another. This is the sense in which Kenneth Casey and Iproposed that pain experience comprises a number of dimen-sions that reflect activities in parallel processing systems(Melzack & Casey 1968).

Clearly, all theories of pain are ultimately "physiological"theories. Cognitive theories implicitly assume that the neuralprocesses that underlie cognitive activities occur in parallel withsensory transmission and are able to influence it. Similarly,affective and motivational processes occur in parallel with bothsensory and cognitive activities, so that pain is not simply astring of sequential events but the product of transactions thatoccur among neural processes that go on at multiple levels.

This concept, which I consider to be crucial to understandingpain, is not presented with sufficient clarity in Rachlin's paper,although it is evident in several places in the paper that he isfully in agreement with it. I will therefore not enter into anyneedless dispute but rather will amplify a few points that, heseems to believe, are understood by most investigators in thefield. I have learned, often with astonishment and dismay, thateven seemingly obvious features of pain processes must bespelled out at virtually every opportunity.

Let us consider the first dichotomy that has led to confusion inthe field: Pain is a sensory experience, and everything else -emotion, motivation, thought, evaluation, coping strategy - isthe reaction to the sensory experience. This idea, given thestamp of approval by Beecher in 1959, has been the favoredhaven of many psychologists. However, this approach leads toconfusion. It is the basis of the frequent failure to understandsome of the most basic features of pain - such as the absence ofpain after severe, life-threatening injury. If sensation occursinevitably after injury and is in proportion to it, how can weunderstand the statement made by about 65% of soldiers whowere wounded in battle, that they felt no pain after their injury(Beecher 1959)? As Patrick Wall and I (Melzack & Wall 1983)have pointed out on many occasions, this leads to the paradox of"painless pain " - obviously a meaningless term. It is much morereasonable to recognize that neural processing permits cog-nitive activities to act on inputs - to open or close the "gates " -before those inputs give rise to sensory experience.

If we recognize that sensory, motivational, and cognitiveprocesses occur in parallel rather than in series, we also realizethat we cannot equate the word "sensory" with "intensity" andall other activities as "reaction." The cognitive-affective-moti-vational processes interact with sensory input beginning at theearliest synapses and therefore determine the eventual inten-sity of an input produced by injury. This is the basis of thevariable link between injury and pain (Melzack & Wall 1983).

The second major dichotomy that has led to confusion - whichRachlin fully recognizes - is the variable link between experi-ence and behavior. As long as we recognize that pain is not


Commentanj/RachUn: Pain and behavior

simply a reflex response to injury but that a complex brainintervenes between input and output, we can realistically facethe problem of the "super-Spartan" discussed in the paper. Ourbehavior in the face of injury or pathology ranging from akidneystone to cancer depends on the intervening activity ofbrain. The brain is the repository of our fears and anxieties, ourunderstanding of the situation, our concern for the future - notjust our own personal future, but the future of those whose livestouch on our own. It is for this reason that the link between painexperience and behavior is so variable. Once we recognize theneural basis of variability, we can avoid needless dichotomiesand instead look to multiple parallel processing systems as thebasis for the richly complex experience and behavior that char-acterizes pain.

A mentalistic view of "Pain and behavior"

H. MerskeyUniversity of Western Ontario, Department of Education and Research,London Psychiatric Hospital, London, Ontario, Canada N6A 4H1

Rachlin's target article reached me soon after I had interviewedmy first patient of the day. This lady, whose English is imper-fect, began as follows: "Everybody thinks is my imagination. Isno my imagination - is my back." She also complained of pain inthe right leg that was worse on walking, some headaches, and adeterioration in symptoms of her hiatus hernia. She added: "Isnot the worry that makes me pain, is the pain that makes meworry. Only the stupid people is not worry."

Her views raise the question of whether we can treat herremarks as "behavior." Some part of them certainly is im-pressive behavior, such as the vigor with which she spoke. Yet tosee it only as behavior and not as an important subjectiveexperience seems perverse.

I will deal first with some matters of evidence. No one hasshown a way to split the supposed experience of pain neatlyaccording to its presumed etiology. Leavitt and Garron (1979)distinguished between the words used by psychological andorganic groups, but usually it is not possible to tell from clinicaldescriptions how pain originates. Psychiatric patients with painand no lesion most often describe the same sort of experience aspeople who have recognized lesions that are thought to causepain. Beecher (1959) argued that pain had both a primarysensory component and a reaction component that could not beseparated completely from each other. A striking effort to provethat these elements could be separated came from signal detec-tion analysis. Rollman effectively disposed of this view (Rollman1977, 1979, 1980; Chapman 1977; Jones 1979, 1980).

Popular methods for the measurement of pain such as verbalscales, visual analogue scales, signal detection, and pain reliefmeasures are subjective. Direct behavioral studies are rela-tively rare, and less efficient. There is no satisfactory scientificevidence that a strict operant procedure works any better thanany other technique (Sternbach 1983). Other methods haveequal success (Pinsky et al. 1979; Catchlove & Cohen 1982).Fordyce (1983b) claims that some of these methods, includingpsychodynamic approaches, have operant characteristics. Butthe operant method does not have unique effects.

The theoretical problem relates to confusion between eventswithin the body and the mind. The use of the word pain is welldefined by practice and is psychological. Its meaning tends to belost to physicians since they are aware of physical events withinthe body that form part of the mechanism of production of pain.Thus doctors make contradictory remarks like, "Severe painsneed not be felt" (Jaspers 1963). Patients with vivid experiencesare told that they do not have pain because a lesion has not beenfound. Walters (1963) described this difficulty well as follows:

Physical pain is a psychic event and not a physical event. The physicalside is the physiological mechanism. . . .The pain is . . . the percep-tual experience of discomfort in a spot in the body. . . . This fact isoften ignored . . . . you will hear yourself or your colleagues say that"pain travels in the spinothalainic tracts" or that "the end organs pickup pain and transmit it up the pain pathways." But these impulsescentrally bent to excite further mechanisms . . . are no more the painthan the visual impulses from the retina are the perceptual fields ofcolour and pattern that present to us when our eyes are open.

Szasz (1957) developed an identical view. No one to my know-ledge has offered a worthwhile alternative to this position thatpain is a word for a psychological condition and not for specificphysical events.

My original definition of pain (Merskey 1964; Merskey &Spear 1967) served as the basis for an international definition(I.A.S.P. 1979). The latter is as follows: "an unpleasant sensoryand emotional experience associated with actual or potentialtissue damage, or described in terms of such damage."

Rachlin neglects this viewpoint, which implies that causalfactors promote or produce pain although they cannot be sepa-rated in subjective experience. The causes should be identified,must be studied, and, at times, can be quantified. Fordyceincorporates the subjective state in a behavioral statement byrelating the latter to the patient's words and saying that he hasno concern witli what the individual actually feels - merely withwhat he does verbally. That still amounts to denying part of theother persons experience. In a curious fashion, the behavioristof today resembles the solipsist of the eighteenth century.

My argument is mentalistic and monistic. As Rachlin recog-nizes, cognitive theory and the Melzack-Wall approach aresimilar. They do not require mysterious interactions betweenbody and mind. Rather, they say that the events which occur in ahuman being are a unity but may be described in differentlanguages. This approach is correct in theory, and is importantin the direct relationship with the pain patient. Any system likea pure behavioral one which produces an artificial circumlocu-tion around the topic, neglects experience, or refuses to take itinto consideration is incomplete. It also leads to frequent errorsin management, compounding theoretical mistakes with prac-tical failures.

Behavior, cognition, and physiology: Threehorses or two?

T. R. MilesDepartment of Psychology, University College of North Wales, Bangor,Gwynedd LL57 2DG, Wales, U.K.

This is an important paper. What follows represents not somuch disagreement as an attempt at reformulation. Rachlin'sposition as he has stated it is open to a number of theoreticalobjections, but these can be removed without any radicalchange to the substance of what he has to say.

In the first place the distinction he draws between "sensory"pain and "psychological" pain is potentially misleading. Ac-tions can have painful psychological consequences if they affectwhat people think and believe; these are different from "phys-ical" consequences such as injury or bleeding. By analogy,therefore, Rachlin seems to be suggesting that what peoplethink or believe about a situation may affect their reactions topainful stimuli. But it is not clear that this entitles him to saythat there are two kinds of pain or that there are two compo-nents of pain, nor is it self-evident that these two claims areequivalent. He does indeed call attention to the distinctionresearchers have drawn between an "intensity" scale and a"reaction" scale, but to say that the one measures "sensory"pain and the other "psychological pain does not seem to addanything.

Second, he suggests that "'sensory' pain is . . . a re-


Commentary/'Rach\\n: Pain and behavior

spondent while 'psychological' pain is an opcrant. " But there issomething logically very strange here. As Ryle (1949, Chap. 7)has pointed out, sensation words do not stand for events. Ifthey did, one could take a peep at an ache or tingle in the sameway as one takes a peep at the things that are going on in onesneighbor's garden. If a sensation of pain is not an event, how-ever, it makes no sense to classify it either as an operant or as arespondent. Perhaps what Rachlin might have said is that cer-tain stimuli are invariably followed by feelings of pain but thatthe extent to which one is troubled by the pain may be de-creased if one interacts with the environment in certain ways.He is thus (quite properly) commending those research pro-grams which investigate such interactions.

Now no one disputes that there are physiological proceduresby means of which pain can be lessened, for example, the useof injections or chloroform. In a science of behavior, however,one is interested in the ways in which expressions of pain aretreated by the community. Rachlin here introduces the veryinteresting example of "super-super-Spartans" who do not ex-press their pain either verbally or by wincing but who are fittedwith a red light that flashes when their physiological painmechanisms are activated. This is an informative analogy, sincethe real-life situation is different only in that instead of a redlight that anyone can see there is a bodily sensation felt by onlya single person. Indeed, the analogy could with advantage havebeen exploited further. If the word pain does not stand for anevent then it cannot stand for any of those allegedly privateevents that behaviorists are so often accused of ignoring; andindeed, if it did, it is hard to see how anyone could ever havelearned its meaning. In fact, however, cries, winces, verbalreports of pain, and so on are behaviors to which the communi-ty responds, just as they might in principle have responded tothe flashing of a red light. If a young child says that he can see acat it is likely that others will say that they also can see it; but ifhe says, "My leg is hurting, " he learns that others do not saythat they can feel the pain too. As Rachlin points out, theyeither try to comfort him or, in some circumstances, accusehim of shamming or of making too much fuss.

In general, then, Rachlin is right in saying that there can be"external study of pain behavior and internal study of pain

mechanisms." However, in including in the same paper a cri-tique of cognitive psychology it is arguable that he has at-tempted too much. His discussion under "Cognitive theoryand treatment' calls to mind the words of Bolingbroke inShakespeare's Richard the Second:

Oh, who can hold a fire in his handBy thinking on the frosty Caucasus?

But he implies that the correctness or otherwise of a cognitiveapproach to the study of pain is something that can be deter-mined empirically, since the answer depends on whether par-ticular strategies such as imagining numbness at the site of thepain are effective in reducing discomfort. The objections to acognitive approach, however, are in fact conceptual: It is acategory mistake to suppose that words such as image andbelief stand for any kind of entity that could be part of a causalchain between stimulus and response. As a matter of logic,therefore, there cannot be a distinctive role for the cognitivepsychologist: either his postulated mechanisms are physiologi-cal ones, in which case he is a physiologist, or he is studyingthe influence of verbal instructions (such as "Try to imag-ine. . . ") on behavior, in which case he is a behavioral scien-tist. Rachlin speaks as though there are three horses contend-ing in this particular race, namely, a physiological horse, acognitive one, and a behavioral one. He has said some veryhelpful things about the first and third of these horses (which,indeed, as he would agree, are not so much rivals as partners).If the above argument is right, however, the "cognitive" horsedoes not belong in this race at all and should not have beenadmitted to the starting gate.

Is there always a neurochemical linkbetween pain and behavior?

G. PepeuDepartment of Pharmacology, Florence University, 50134 Florence, Italy

Pharmacologists, aware of the limitations and side effects of drugtherapy for pain control, must favor all attempts to use nonphar-macological means for relieving pain. Rachlin s accurate andcritical review outlines the theoretical basis of a psychological orbehavioral approach to pain control. If a general criticism can beexpressed, some of the examples quoted in the review may berather obscure to a European reader. Cultural differences apart,the impression that a biologist gets from Raehlin's analysis of thephysiological, cognitive, and behavioral theories of pain is thatof a semantic dispute, as the author admits, concerning the twoaspects of pain, the nociceptive stimulus and its perception.Even in the rat it is possible to recognize a "sensory" and an"aversive' or "psychological" pain. Electrical stimulation of thetail according to the method of Paalzow and Paalzow (1973)elicits a motor response characterized by tail withdrawal andhindquarter movement, an immediate vocalization response,and a delayed vocalization response that occurs after the cessa-tion of the nociceptive stimulus. Paalzow and Paalzow call thislast response vocalization afterdiseharge. It involves the rostralregions of the brain and is considered the affective component ofpain. Morphine and anxiolytics are more effective on the vo-calization afterdiseharge than on the two immediate responses(Morichi & Pepeu 1979).

If we examine the rat's behavior in light of the theories asdescribed by Rachlin, the motor response and the immediatevocalization could, according to physiological and cognitivetheories, be considered the "real pain," whereas the vocaliza-tion afterdiseharge would be the "pain behavior." On the otherhand, for the behavioral theory the vocalization afterdisehargewould be the "real pain."

Apparently, only pharmacological or elcctrophysiologicalways can affect the rat's response to nociceptive stimulation. Inman, however, pain can be modified by "psychological" factors,as Rachlin emphasizes in the beginning of his review. Psycho-logical factors have traditionally been exploited in order tocontrol pain. An example, for which Rachlin forgot to find aplace within the framework of the pain theories, is offered by thepain control proposed by religions. The Catholic religion invitesthe sufferer to accept physical pain in order to obtain a reward inthe future life after death. Real pain is not obtunded by thisapproach; rather, pain behavior is changed. According to thebehavioral theory, however, real pain is reduced.

Perhaps the expectation of a reward after death as a means ofmaking pain less "painful" is the only truly "psychological" paincontrol. Other religious practices such as collective prayers,singing, and dancing appear to reduce pain perception. As anexample, we may recall the medieval flagellants. However, inthese cases conditions favorable to the release of endogenouschemicals modulating nociceptive information may occur. Thesymposium "Shamans and endorphins" mentioned by Henry(1982) is an example of the current tendency to explain behav-iors in term of neurochemical changes in the central nervoussystem. Shamanism is a trancelike state including both euphoriaand analgesia. Papers such as that published by Coid, Allolio,and Rees (1983) showing that mean plasma metenkephalinconcentration was increased in ten severely disturbed patientswho habitually mutilated themselves offer strong support to thephysiological theory of pain. In Coid et al.'s study, patientsreported that the mutilation was carried out without causingpain.

It is therefore very tempting to equate changes in painperception with alterations of a central mechanism controllingnociceptive input. In this physiological approach the psycholog-


Commentary/Rdchlin: Pain and behavior

ical and behavioral theories would find justification, so long asthey offered a way of modifing the central neurochemieal controlmechanism of the nociceptive input.

This assumption is tenable only if it can be experimentallyshown that behavioral procedures or mental attitudes relievingpain always affect brain neurochemistry. Some conditions in-ducing analgesia associated with a release of endorphins havebeen listed by Rachlin. The best known are stress and acu-puncture. Long-distance running and sexual activity are men-tioned by Henry (1982). Fear conditioning increases en-kephalin-mediated analgesia in rats (DeVries, Chance, Payne &Rosencrans 1979). Can cognitions, as in Rachlin's example("Cognitive theory and treatment," paragraph 5), do the same inman? Endorphins need not be the only neuroregulatorinvolved.

However, chronic pain increases metenkephalin in the spinalcord segment receiving a direct projection from the painful area(Faccini, Uzumaki, Covoni, Missalc, Spano, Covelli &Trabuc-chi 1984). It has been shown that chronic pain induces ahypoalgesia to acute nociceptive stimulation (Colpaert 1979).Could the metenkephalin increase in the spinal cord be modu-lated by nonnociceptive peripheral stimuli, for instance, byreinforcing the well behavior of Fordyce?

It must be stated that all this has a slight flavor of Aristotelianhumoral theory. Nevertheless, many other circumstantial datacould be marshaled to support the contention that there is arelationship between pain perception (aversive pain) and brainneurochemistry. These appear to give a neurochemieal founda-tion to the gate control theory of Melzack and Wall (1965).According to this theory, "pain," a state of perception in whichthe organism as a whole reacts to apparent injury, is triggered ifthe initial gate and subsequent stages of the transmission path-way are favorably set for the exhibition of this state (Wall 1979).The concept of "favorably set again leaves unanswered themain question of what it is that makes pain felt, the questionaround which the theories examined by Rachlin are built.Nevertheless the gate control theory has through the yearsexerted a strong heuristic influence in stimulating experimentaland clinical investigations that have also led to applications.

Semicovert behavior and the concept of pain

Ullin T. PlaceDepartment of Philosophy, University of Leeds, Leeds LS2 9JT, England

If I have understood him correctly, pain, according to Rachlin,consists of three elements, a pain stimulus, and two forms ofovert pain behavior: the "respondent" behavior, which occursas a reflex response to the pain stimulus and the "operant'behavior, which is reinforced insofar as its emission by theorganism is followed by an alleviation or termination of the painstimulus.

I take it that although he doesn't use the term in this paper, itis part of Rachlin's view that pain stimuli are normally "aversive"in the sense that they constitute "an establishing condition," touse Michael's (1982) term, whereby any operant that is followedby the alleviation or termination of the aversive stimulus isthereby reinforced. Rachlin then suggests, following Fordyce(1978), that the distinction between what he calls "sensory" and"psychological" pain can be accounted for in terms of thedistinction between respondent and operant pain behavior.

For the purposes of this discussion I shall assume that thedistinction Rachlin has in mind when he distinguishes between"sensory" and "psychological" pain is the same distinction asthat which is drawn in commonsense terms between what issometimes called "physical pain," where pain is a bodily sensa-tion that is usually extremely unpleasant and distressing, andpain in the sense of the emotional reaction of acute distress when

that reaction is evoked, not by pain qua bodily sensation, butrather by a thought, such as the thought that this pain is perhapsa symptom of some fatal illness.

If I am right in thinking that this is the distinction Rachlin hasin mind, then it is difficult to resist the conclusion that hisbehaviorist theory of pain simply does not contain sufficientconceptual resources to enable him to do justice to the fullcomplexity of the commonsense distinction.

I suggest that in order to do that, in addition to the painstimulus and the overt and predominantly operant pain behav-ior which, as we ordinarily understand the matter, is an effect or"expression" of the pain, rather than part of it, we need torecognize the existence of three distinct varieties of behavioreach of which is predominantly but not exclusively covert andpredominantly but, with one exception, not exclusively re-spondent. These three varieties of semicovert behavior, as wemay call them, are (1) attending behavior, (2) emotional reac-tions, and (3) self-directed verbal behavior or thinking.

Behaviorists of Rachlin's persuasion have traditionally beenreluctant to acknowledge the occurrence of these types ofsemicovert behavior. This, I presume, is because the so-calledintrospective reports of human subjects, of which behavioristshave always been suspicious, deal for the most part with thecovert and, hence, otherwise inaccessible aspects of such behav-ior. It seems to me that this reluctance is misplaced for tworeasons.

In the first place, since this behavior is only partly covert, itfollows that there are many occasions on which it consists partly,if not wholly, in publicly observable overt molar behavior whoseoccurrence even the most hardened behaviorist must acknow-ledge. Thus visual attending behavior normally consists in acomplex pattern of head and eye movements, such as tracking,accommodation, and convergence, whose effect is to keep theretinal image of the object attended to in focus. Similarly,auditory attending behavior may consist in controlling the noiseone might otherwise make oneself so as not to obscure the soundone is trying to catch. Olfactory attending frequently consists insniffing, gustatory attending in savoring movements of the lipsand tongue, and tactile attending in moving one's fingers overthe surface of the object of attention.

The occurrence of an emotional reaction, in contrast to theovert operant behavior, like pain behavior, for which the emo-tional reaction creates the establishing condition, is much lesseasily detected at the level of molar observation than is attend-ing behavior. Nevertheless, blushing in shame and embarrass-ment, weeping in joy and sadness, and the enlargement of thepupils in excitement and interest are overt, publicly observableaspects of such reactions.

In the case of thinking, most thoughts are uttered privately tooneself, without any actual movement of the voice musculature.People nevertheless often think out loud, not only on occasionswhen thinkers intend to share their thoughts svith others, butalso when the thought is entirely self-directed.

The second point that needs making in this connection is that,even in those cases where the occurrence of the behavior is anentirely covert event, taking place presumably within the cen-tral nervous system, it is usually possible to determine objec-tively that a covert response has occurred by observing thechange that has thereby come about in the establishing andother conditions controlling subsequent operant behavior.

Thus the effect of attending behavior is to increase what wemay call the "salience of the stimulus or stimuli to whichattention is paid and hence the vigor and accuracy of thediscriminative control exercised by those stimuli over the subse-quent operant behavior. The effect of an emotional reaction, likethe distress involved in both senses of the word pain, is to set upan establishing condition whereby the alleviation or disap-pearance of the stimulus or state of affairs that evokes it acts as areinforcer witli respect to any behavior that is followed by it.

The effect of thought on subsequent operant behavior is much


Commentary/Rachlin: Pain and behavior

less easy to pin down than is that of attending behavior andemotional reactions. However, recent work on the contrastbetween human and animal responding on fixed-interval sched-ules of reinforcement (Lippman & Meyer 1967; Lowe 1979;1983) is beginning to throw some empirical light on the dif-ference between what Skinner (1969) calls "contingencyshaped" and "rule (i.e., thought) governed" behavior.

If we try to relate these three types of semicovert behavior toSkinner's (1938) "respondent" and "operant" distinction, itappears that emotional reactions, as contrasted with the oper-ants that "express" the emotion thereby generated, are invari-ably respondents. Attending behavior and thinking on the otherhand both function in part as operants obeying the Law ofEffect. However, both these forms of behavior appear to besubject to what we may call "a respondent override mechanism"that ensures that attention and thought are directed as muchtoward stimuli that are highly aversive as they are toward stimulithat are reinforcing with respect to operant behavior. Withoutsuch a respondent override mechanism the operation of the Lawof Effect would have the maladaptive consequences predictedby the now long discredited theory of "perceptual defense,"whereby the organism would systematically ignore aversivestimuli such as pain, as well as those discriminative stimuliwhich act as danger signals with respect to such aversivecontingencies.

Pain without behavior: Inhibition of reactionsto sensation

Kelly G. Shaver and Jana J. HerrmanDepartment of Psychology, College of William and Mary, Williamsburg, Va.23185

The fundamental difference between a behaviorist approach topain - or to any other psychological phenomenon - and that ofcognitive theory is not that one examines actions of the wholeorganism while the other limits its inquiry to a fraction of theorganism. The difference lies rather in what adherents of thetwo approaches are willing to infer on the basis of the behaviorthey both observe. Can there be emotions, or are there merelyemotional displays? Can there be thought, or is there merelyverbal behavior? Can there be pain, or is there only painbehavior? True to the behaviorist tradition, Rachlin takes thelatter position, but there are both logical and empirical reasonsto question his argument.

Let us begin with Rachlin's closing assertion that the "bestargument for a behavioral definition of pain would be a trulyeffective behavioral treatment of pain." Were a truly effectivebehavioral treatment for pain to be discovered, even one thatchanged only the individual's overt expression of pain, the forceof Rachlin's paper suggests that his final claims would not benearly so modest as the closing assertion would lead us tobelieve. On the contrary, despite the present limits to theevidence, Rachlin concludes from Fordyce's (1976) successfulbehavioral treatments of individuals suffering from chronic painfor which no concomitant tissue damage can be found that allpain is nothing more than "pain behavior." Ignoring for themoment that chronic pain may not be the same as acute pain,and that pain without apparent tissue damage may not be thesame as pain with tissue damage, all that can logically beconcluded from a successful behavioral treatment of pain is thatthe method provides control over the disruptive manifestationsof pain. No psychologist familiar with social desirability con-straints and demand characteristics (Orne 1962) should accept atface value the idea that a change in a verbal report is isomorphicto a change simultaneously occurring in an internal state (unless,of course, one has defined the internal states out of existence).Similarly, no psychologist familiar with a falsificationist view of

scientific discovery (Popper 1959) should erroneously infer thatbecause one theoretical viewpoint has been confirmed all com-peting formulations have necessarily been ruled out. Demon-strating that effective behavioral therapy can provide controldoes not conclusively demonstrate that behavior theory pro-vides sufficient explanation.

Rachlin's inclination to overstate the theoretical import ofFordyce's (1976) therapeutic success is abetted by his extremecharacterizat'on of the cognitive alternative as "mentalism,"roughly the psychological equivalent of "alchemy." The argu-ment can be stated more constructively by pointing out that acognitive interpretation of pain need only suggest that a con-scious process might intervene between the sensation of painproduced by identifiable tissue damage and the reaction to thatsensation that would be obvious to an external observer. Al-though Rachlin describes the psychophysical measurement pro-cedures that Tursky, Jamner, and Friedman (1982) have devel-oped to distinguish the intensity of the sensation of pain from thereaction to the pain, he does not seem aware of the problemsthat this distinction presents for a behaviorist theory. If painwere nothing more than "pain behavior, " then an effectivebehavioral treatment that produced a diminution in the reactionscale would of necessity produce a corresponding reduction inthe intensity scale. Only if one supposes that some consciousprocess can intervene between sensation and reaction is itpossible to imagine a therapy (like hypnotherapy) that couldalter how bearable pain might be without also changing (toexactly the same degree) the felt intensity of pain.

In his discussion "What is pain?" Rachlin compares a person"in pain . . . though not evincing pain behavior" to a rat press-ing a bar at a certain rate although not pressing the bar "rightnow." What distinguishes the two cases may well be somethingabout the very meaning of pain (we would agree with the"antibehaviorist philosophers" who would argue for the pos-sibility of inhibiting an overt response to a sensation of pain). Butin a larger sense this argument about the meaning of pain ismerely the most recent skirmish in a metatheoretical controver-sy that predates the emergence of psychology as a separatediscipline. At this broader level the question has to do with aparticular resolution of the classical dilemma of determinism:The conscious volition required to inhibit the overt expressionof pain is in principle no different from the conscious volitionrequired to refrain from "immoral" though immediately re-warding behavior. It may be possible for behaviorists to excludeconscious processes from the expression of pain more success-fully than they have been able to exclude them from morality.But the research Rachlin reviews should give them pause, notcomfort.

Molar behaviorism, positivism, and pain

Charles P. ShimpDepartment of Psychology, University of Utah, Salt Lake City, Utah 84112

Rachlin's paper is another effort by a radical behaviorist to showhow a purely behavioral analysis can adequately deal withprivate events, in this case, pain. My comments focus on two ofRachlin's assumptions. I wish merely to point out that there arealternatives to these assumptions. The alternatives may lead toimportantly different views on the nature of behaviorism andtherefore on behavioral interpretations of private events such aspain.

Rachlin assumes there are two kinds of pain. One is "funda-mental," is "sensory" in nature, and is correlated with anantecedent stimulus such as a wound. The second is derivedfrom the first and is correlated with the way various pain-relatedbehaviors are controlled by their reinforcing consequences.This distinction might be called a "foundationist" approach to


Commentary/RachWn: Pain and behavior

the analysis of the conceptual status of pain. This distinctionbetween two kinds of pain, one sensory and fundamental andthe other derived, closely parallels more general, positivistie,analyses of the nature of seeing and the nature of knowledge.Such analyses hold that there is a fundamental, sensory know-ledge that provides the bedrock or foundation for scientificknowledge and that this fundamental knowledge is separablefrom other, derived, kinds of knowledge.

This "foundationist" position that discovers empirical bed-rock in perceptual metaphors has not gone unchallenged (Han-son 1969; Rorty 1979; Ryle 1949; Wittgenstein 1958). Oddly,one of the works Rachlin cites, that by Ryle (1949), gives a goodanalysis of conceptual problems inherent in the foundationistepistemology. Rachlin references Ryle for various behavioralideas that, taken out of context, resemble those of radicalbehaviorism. But he does not comment on the complex relationRyle assumes between knowing and seeing, and as a resultRachlin seems to adopt a relatively unsophisticated position onthe nature of scientific observation. As a result, he places greaterfaith in the fundamental scientific status of overt behavior than isjustified. Rachlin, like Skinner (1950) and many other radicalbehaviorists, seems to feel that by dealing with overt behaviorhe can deal with something that provides a privileged, funda-mental kind of knowledge.

A related, positivistie claim appears at the end of the section"Behavioral theory and treatment" and elsewhere. It is saidthere that pain is behavior. This kind of operationism seems toomit much of the grammar of the concept of pain: It does notadequately take into account the relations among the individualbehaviors (crying, kicking, etc.) said to define pain and es-pecially the way their meanings depend on ordinary language.This contextual material is very much part of the meaning of theconcept of pain, but it does not seem to appear anywhere inRachlin s analysis. This omission is characteristic of positivistieapproaches in general. The idea that there are little knowledgeunits, in Rachlin's case deriving from Pavlovian and operantconditioning, that together adequately redefine a concept suchas pain, is today an anachronism; it resembles Watsonian behav-iorism and in some ways the logical atomism of the earlyWittgenstein (1922). It seems not to take into account contem-porary thinking in epistemology, the philosophy of science, orcognitive psychology.

Rachlin seems fully to appreciate that if a theory of pain is tobe confined to overt behaviors, one needs to know somethingabout how to deal scientifically with overt behavior. Rachlinoffers "molar behaviorism" as the solution to the problems ofwhat behavior is and of how to deal with it scientifically. Heasserts that this kind of behaviorism has never to his knowledgebeen refuted. The defining exemplar of molar behaviorism is thegeneralized matching law (for example, see Baum, 1973). Long-term readers of the Journal of the Experimental Analysis ofBehavior will recall that Rachlin (1971) published an articlecalled "On the tautology of the matching law," the point ofwhich was that the matching law, being a fundamental causallaw, cannot be refuted. Data that might on the surface appearinconsistent with it simply indicate the presence of some inap-propriately scaled variable or some methodological flaw. Thus,we have Rachlin asserting here that molar behaviorism hasnever been refuted and asserting previously that its definingexemplar in principle cannot be refuted. Yet molar behaviorismmight not be as frighteningly impregnable as this awesome logic-suggests, and there may yet be a possibility for a molecularanalysis. Thus, of the eight or so examples of molar "currentbehavioral theory listed in the sixth paragraph of "Behavioraltheory and treatment, at least half were developed by theoristswho subsequently have gone on to develop more molecularapproaches.

Here is not the place even to list the reasons why a molecularapproach to behavioral analysis is continuing to prove fruitful.Suffice it to sav that the extremelv radical version of molar

behaviorism that Rachlin advocates confronts an evergrowinglist of conceptual issues and empirical phenomena that appar-ently are beyond its capacity to deal with successfully (Peele,Casey & Silberberg 1984; Platt 1979; Shimp 1975; 1976; Hinson& Staddon 1983). Thus, it seems doubtful that molar behav-iorism can support the kind of philosophical burden Rachlin ishere placing on it.

If behaviorism is to provide a satisfactory philosophy ofprivate events in general, and of pain in particular, behavioralaccounts will have to abandon their commitment to a classicpositivistie position on the relation between observation andknowledge. And if molar theory is to provide a comprehensivebehavioral theory, it will have to find some means of assimilatingmolecular phenomena.

The reign of pain falls mainly in the brain

Dennis C. Turk and Peter SaloveyDepartment of Psychology, Yale University, New Haven, Conn. 06520

At times it is difficult to understand the relationship betweenthe two parts of Rachlin's target article, the overview of currentpain research and the behavioral theory that relies heavily onthe analogy of the "super-Spartan." We will focus on the firstpart of the paper since, in our opinion, the available evidencedoes not support the model presented. First, Rachlin suggeststhat there is a "cognitive theory" of pain that can be contrastedwith a distinct "behavioral theory" of pain. He cites Turk,Meichenbaum, and Genest (1983) to show that cognitive copingstrategies have not proven to be effective in enhancing paintolerance. He then infers that the data do not support a cognitivetheory of pain. The leap from a conclusion regarding the efficacyof various cognitive coping strategies (e.g., imaging that you areon the beach in Hawaii) to the efficacy of "cognitive" treatmentfor chronic pain is specious.

The cited quotation from Turk et al. (1983) does not supportthe conclusion that cognitive strategies are ineffective. Rather,laboratory subjects have already developed some coping strat-egies, and they do not suffer from a coping skills deficiency.Teaching new coping techniques to subjects who have existingcoping repertoires will not necessarily lead to any improve-ment. Furthermore, Rachlin fails to consider the literature onacute clinical pain that supports the efficacy of both cognitiveand behavioral coping strategies (e.g., Langer, Janis & Wolfer1975).

Few consider a purely cognitive treatment to be a satisfactoryapproach to chronic intractable pain (we will return to this pointshortly). Yet Rachlin fails to cite the one study that most closelyresembles a "cognitive" treatment, that of Rybstein-Blinchik(1979), who demonstrated the efficacy of an approach employingrational restructuring and relabeling. She reported that thistreatment resulted in a reduction in both self-reports of pain anilpain behaviors observed by the nursing staff.

The application of what we call a cognitive-behavioral ap-proach to the treatment of pain involves a wide range of treat-ment strategies. Cognitive-behavioral interventions are active,time-limited, structured treatments designed to help patientsidentify, reality test, and correct maladaptive beliefs concerningtheir problems. Patients are encouraged to monitor the impactthat negative pain-engendering thoughts and feelings play inthe exacerbation and maintenance of maladaptive behaviors. Inaddition, patients are taught to recognize the connectionsamong cognitions, affect, physiology, and behavior. The thera-pist is concerned with both the role that patients cognitions playin contributing to the disorder and the nature and adequacy olpatients behavioral repertoire.

We have already noted that there is a body of literature thatsupports the efficacy of cognitive approaches with acute pain.

72 THE BEHAVIORAL AND BRAIN SCIENCES (1985) 8.1

Response/Rachhn: Pain and behavior

There is also some evidence supporting the efficacy of cognitiveand cognitive-behavioral approaches to chronic recurrent pain.For example, Holroyd, Andrasik, and Westbrook (1977) andBakal, Demjen, and Kaganov (1981) have reported on theeffectiveness of stress-management techniques for tensionheadaches and vascular headaches, respectively. In fact,Mitchell and White (1977) reported that the greatest amount ofimprovement in migraine headaches was demonstrated follow-ing the addition of cognitive strategies to more behavioral ones.It is possible that Rachlin did not review these studies becausethe operant theory has given these types of pain little attention.

Even in the treatment of chronic intractable benign pain,there is a growing body of evidence supporting the utility of acognitive-behavioral approach. For example, Turner (1982) re-ported on the efficacy of a cognitive-behavioral treatment versusrelaxation with low-back patients. She noted that the formerresulted in significant pain reduction and increased activitylevels. These results were maintained from one and one-half totwo years following the intervention. Several related studieshave supported the importance of cognitive factors in the effec-tive use of biofeedback with chronic-back-pain patients (e.g.,Flor, Haag, Turk & Koehler 1983; Nouwen & Solinger 1979).

We are not suggesting that the efficacy of cognitive or cog-nitive-behavioral approaches has been demonstrated unequivo-cally. In fact, we agree with Rachlin s conclusion that the jury isstill out. We believe, however, that there are a large number ofstudies that should be considered before one subscribes to apurely behavioral theory of pain.

Rachlin contends that there are two psychological theories ofpain with two distinct treatment protocols. However, as wehave suggested, there is actually no "cognitive" treatment(Rybstein-Blinchik, 1979, notwithstanding). Rather, most treat-ment studies have adopted a cognitive perspective that ac-knowledges the importance of patients' processing informationabout treatment and their plight and employ some behavioralstrategies to bring about cognitive ami behavioral changes.

We would suggest that even those who adopt an operantapproach incorporate important cognitive components in treat-ment. Fordyce (1976) spends a good deal of time attempting tochange patients views of their problem as part of an operanttreatment. He provides them with an alternative conceptualiza-tion of pain that needs to be interpreted, appraised, and accept-ed if his treatment is likely to have any beneficial effects.

Rachlin has failed to consider data that are relevant to histhesis. He has misrepresented the cognitive-behavioral modelas being a purely cognitive one, and, moreover, he has present-ed the operant model as if it were exclusively behavioral. Pain isa perception, not a sensation, and thus, the experience iscomprised of cognitive, affective, and sensory as well as environ-mental factors. Although we agree that behavior is an importantcomponent of the experience of pain, it cannot be viewed assynonymous with it.

Not "pain and behavior" but pain inbehavior

Patrick D. WallCerebral Functions Group, Department of Anatomy, University CollegeLondon, London WC1E 6BT, England

This paper perpetuates a Cartesian dualism by its subdivision ofpain into "sensory" and "psychological." This is an intellectualartifice invented to preserve a concept of divided brain andmind. Rachlin proposes a two-stage process. First, a mindlessmechanism reports to the brain the state of the periphery as bestit can. Second, cognitive processes get to work to decide what tomake of the news. Attractive as this sequential process may be tothe classical intuitive philosopher, there is not a scrap of physio-

logical or psychological data to support the dualism (Mel/ack &Wall 1982; Wall 1979, 1984; Wall & McMahon 1984). What isobserved in others and felt in oneself is an integrated wholeorganism which is on occasions in a state of sensory-psychologi-cal pain.

Author's Response

Ghostbusting

Howard RachlinDepartment of Psychology, State University of New York at Stony Brook,Stony Brook, N.Y. 11794

First, I will deal with some common objections to thetarget article under various specific topics. Then I willdiscuss each commentary individually. This organizationnecessarily involves some repetition but will prove con-venient for the reader interested only in certain topics orin certain commentaries.

Specific topics

Introspection. Psychological theories (like those of othersciences) may be evaluated by standards of coherence,parsimony, experimental success, and meaningful ap-plication to practical concerns. They cannot be evaluated,I claim, by agreement with our intuitions or introspec-tions any more than a theory in physics can be so evalu-ated. If a psychological theory that is found to be best bythe above standards goes against our intuitions and intro-spections, then our intuitions and introspections must beillusions. The moon illusion is an illusion because itcontradicts accepted physical theory. It is conceivable,however unlikely, that a theory of physics that sees themoon to be shrinking as it moves from the horizon to thezenith will someday be generally accepted. In that eventthe moon illusion will not be an illusion but a veridicalperception.

The reasons for accepting or rejecting a physical theoryhave nothing to do with perceptions of the subject matterand everything to do with the other standards mentionedabove (which may constitute a perception of the theoriesthemselves). As Skinner (1953), Kantor (1963), and otherbehaviorists have argued, psychological theories shouldbe judged by the same standards as other scientifictheories.

Introspection was never rejected by Watson or anyother behaviorist as an object of study - as behavior likeany other. Introspection was rejected rather as a priv-ileged sort of behavior - as a substitute for the scientificstudy of behavior itself. Faced with the choice of regard-ing our mental states as given by introspections or asdetermined by whatever is the best theory we currentlyhave, it is always tempting to rely on introspections. Therewards for elevating them into scientific truth are imme-diate. Introspection requires little work and little intel-ligence. And no one may contradict us. Ainslic believesthat these benefits enable us to trust what introspectiontells us as the truth. But the penalty for so doing, although

THE BEHAVIORAL AND BRAIN SCIENCES (1985) 8.1 73

Response/RacMin: Pain and behavior

delayed, may be severe. It is the possibility of sacrificingcoherence, parsimony, experimental success, and mean-ingful application - all the criteria of a scientific theoryother than introspection. So we ourselves in our theoriz-ing, like the pigeons in Ainslie's own experiments on self-control (Ainslie 1975), are faced with a choice between asmall immediate gain (a temptation) and a much largerbut delayed gain. We do not succumb to temptation whenwe introspect. We succumb to temptation when wesubstitute introspection for scientific reasoning.

Theory and application. Most of my colleagues will proba-bly agree with me that psychology, relative to some othersciences, is far from ready to apply theory directly to theproblems of everyday life. The best we seem able to do atpresent is to develop theory and practice in parallel.Psychological theory can (all too easily) explain post hocwhy certain techniques work and why others fail. Anytheory may act as a heuristic, guiding practice in a vagueway, but none is precise enough to predict in advancewhich techniques will work and which will fail. At pre-sent, in psychology, theory can benefit much more by astudy of practice than practice by a study of theory. (Untilrecently physics was only slightly better off. The develop-ment of steam engines probably contributed more tophysical theory than vice versa.) This is why the targetarticle so strongly emphasized the importance of clinicaltechniques in the study of pain.

As several of the commentators pointed out, the radicalbehavioristic definition of pain is, given current intui-tions, counterintuitive. I do not argue that all the evi-dence is in. I do not ask the reader to accept behavioraltheory and reject all others. I do claim, however, that thebehavioral theory of pain is no less viable (howevercounterintuitive it may be) than any other current theory.Thus, contrary to the assertions of Clark and Shaver &Herrman, it is significant that behavioral techniques(techniques vaguely guided by behavioral theory andeasily describable in its terms) are no worse than others incontrolling pain.

Psychophysics and behavior. In beginning the targetarticle with the separation in psychophysics betweenintensity and reaction scales I meant to illustrate thatpeople recognize two aspects of pain. I then went on toidentify those two aspects as operant and respondentbehavior. Some commentators, however, seem to havebeen so beguiled by my initial reference to the psycho-physical results that they did not read what I said after-ward. Genest, Wall, and Loeser all (to a degree) make thesame criticisms of current views of pain as I do. But theyerroneously attribute those views to me. As Melzackindicates, the fault is mine for not emphasizing stronglyenough that I disagree with the linear view of painimplied by the dichotomy between "intensity" and "reac-tion." In part, then, Genest, Wall, and Loeser criticize,not my view, but the linear pain theory as put forth in allits oversimplicity by Matson.

Words and pain. Where do the words that people use todescribe their pain come from - and why do they usethose words when they do? A behavioral theory (contraryto traditional psychophysics) assumes that these wordsare not descriptions of internal states but rather are

designed (like what Place calls "establishing conditions")to produce certain actions in listeners which the speakeror the general environment eventually reinforce. Thisnotion requires a common understanding between speak-er and listener. When migraine headache sufferers de-scribe their pain, are they attempting to produce certainbehavior in the listener or just reporting on an internalstate accessible only to them? Jamner and I (unpublished)asked a group of non-migraine sufferers to fill out theMcGill pain questionnaire (Melzack 1983) pretendingthey were suffering from migraine. We found no dif-ference between the words they used to describe theirimaginary pain and the words used by a group of actualmigraine sufferers. This constitutes some evidence (ofcourse, not proof) that pain descriptions are social actsaccessible to all people with a common culture ratherthan reports of the quality of inherently private states (asGenest would claim).

Cognitive behavior modification. We should not evenconsider cognitive theory as applied to pain, Miles be-lieves. Between a good behavioral theory and a goodphysiological theory there would indeed seem to be noroom for a cognitive theory. To slightly alter Miles'smetaphor: with two mature, healthy horses pulling thewagon we would not need a third. But are physiologicaland behavioral psychology mature or healthy? I don'tthink so. Theories labeled "cognitive" are, therefore,worth considering, because they aid in description andgeneralization of effective clinical practice.

The problem in applying cognitive theory to clinicalpractice, however, is that there seems to be no relation atall between the work of clinicians who call themselves"cognitive behavior modifiers" and experimental cog-nitive psychology (Rachlin 1977a; 1977b). It would be toomuch to ask for direct application of theory to practice.But in this case we do not even have post hoc explanation;there is simply no relation between theory and practice.Whatever elegance, symmetry, or experimental preci-sion is contained in current cognitive theory is lost incognitive behavior modification. All that remains in com-mon is the name "cognition" and a vague sense that"thoughts" are being dealt with.

But what about laboratory and clinical techniques de-veloped by the cognitive behavior modifiers? A littlesuccess in practice would compensate for a lot of failure intheoretical elegance. Turk & Salovey make the remark-able claim that we cannot look for success in laboratorystudies of cognitive behavior modification since laborato-ry subjects "do not suffer from a coping skills deficiency."The claim is remarkable because the only decent evi-dence we have that self-description is related in any wayto pain control comes from the study of laboratory sub-jects by Turk, Meichenbaum, and Genest (1983) men-tioned in the target article. In that study some subjects(with low pain thresholds) "catastrophized." If that studyis now to be repudiated we are left with only clinicalstudies for evidence of the effectiveness of cogntive be-havior modification. As indicated in the target article,clinical studies are difficult to conduct and, almost neces-sarily, difficult to interpret. That is why I made no specialclaims for Fordyce's results. I said his methods were noworse than any others. Cognitive behavior modifierscannot be satisfied with this. They must claim that their


Respou.se/Rachlin: Pain and behavior

cognitive-behavioral techniques are better than purelybehavioral techniques. Otherwise why use cognitivetreatments at all? Why not just stick to behavioraltreatments?

In support of the clinical efficacy of cognitive additionsto behavioral techniques Turk & Salovey cite a study byRybstein-Blinchik (1979). In that study 11 chronic painpatients were told to use "milder" words to describe theirpain and were given examples of mild words: "I feelnumbness" or "I feel itching" instead of "I feel pain."Then they were given practice in using these words by atherapist familiar with the purposes of the study, and theywere told that their pain would be helped if they persistedin using these mild words to describe it. It is not surpris-ing, therefore, that patients in this group obeyed instruc-tions and actually did describe their pain with mildwords, giving their pain lower ("milder") ratings thanpatients in control groups. They did also show less overtpain behavior when it was measured by an experimenterwho they knew was involved with the study. I do notintend to belittle these results, but Turk & Saloveysuccumb to wishful thinking when they say that thistreatment resulted in reduction of "pain behaviors ob-served by the nursing staff." Rybstein-Blinchik reportsthat the nurses observed no differences in pain behaviorbetween the treated patients and others.

Respondents and operants. The target article does notadequately distinguish between respondents and oper-ants. Since a behavioral theory of pain depends on it, thisdistinction needs to be clarified. I beg the reader's indul-gence, therefore, while I briefly discuss laboratory ex-periments with animals where operants and respondentshave been most extensively analyzed.

According to Skinner's (1938) original distinction, arespondent is an act controlled by its antecedents andalterable by classical (Pavlovian) conditioning, while anoperant is an act controlled by its consequences andalterable by operant (instrumental) conditioning. Skinnerspeculated that certain acts might be naturally and whollyrespondents (perhaps those mediated by the autonomicnervous system), while other acts might be naturally andwholly operants (perhaps those mediated by the skeletalnervous system). Dogs' salivation is a prototypical exam-ple of the former, and pigeons' key pecking is a pro-totypical example of the latter.

But it was soon shown that no act is wholly respondentor wholly operant. Salivation may be controlled by itsconsquences and key pecking by its antecedents. Thelatter process, known as "autoshaping," is so reliable and"law-abiding" that it has become one of the two or threestandard methods in the United States for studying classi-cal conditioning itself (see, for example, Locurto, Terrace& Gibbon 1981; Rescorla 1984). It is currently agreed thatno act can be a "pure" operant. According to Teitelbaum(1977), all operant behavior evolves in some way fromrespondent behavior no matter how different it mayappear to "be in its final form. Teitelbaum's notion ofoperants developing from respondents corresponds nice-ly to Melzack's notion of parallel processing in the phys-iology of pain. Teitelbaum has shown that the develop-ment of eating patterns as a rat's (or a human's) brainmatures corresponds to changes in eating patterns as thatanimal recovers from brain operations. At first eating is

stereotypic, reflexive, and relatively unmodifiable by itsconsequences. As the higher centers of the brain devel-op, eating becomes more idiosyncratic, relatively unre-flexive (not reliably elicitable by a given stimulus), andhighly modifiable by its consequences. In other words,eating in rats and humans starts out as mostly respondentand evolves into an operant. Yet buried within thatoperant, so to speak, the respondent still lurks. At timesof stress or extreme deprivation, respondent eating maytake the place of operant eating in adults. (See Breland &Breland, 1961, for an example of such "instinctive drift"in pigs.)

In the case of eating by mammals such as rats, pigs, andhumans, normal respondent and normal operant behav-ior seem to have different (but overlapping) topographies.But in the case of pecking (hence eating) by pigeons,normal respondent and normal operant behavior havealmost identical topographies (although there have beensome attempts to tease them apart; see Schwartz &Gamzu, 1977, for an example). The best way to tell thedifference between a respondent and an operant pigeonpeck is to do experiments to discover which of its aspectsare controlled by its antecedents and which by its conse-quences. As patterns of pecking become more compli-cated, the patterns themselves are more likely to becontrolled by their common consequences (the patternsare operants), while the elements, the individual keypecks, are controlled by their antecedents (the pecks arerespondents). Thus, respondent and operant behaviorare intertwined.

Returning to the subject at hand, pain, if it is indeedovert behavior, would seem to be at least as complex as apigeon's key peck. To what extent pain is like mammalianeating (and differs in topography in its respondent andoperant manifestations) or like avian eating (and is similarin topography in its respondent and operant manifesta-tions, differing only in complexity of pattern achievable)has yet to be explored.

I agree with Clark, Logue, and Matson that tissuedamage is the fundamental and most common cause ofpain. But tissue damage is still only the unconditionedstimulus for respondent pain. Beyond direct elicitationthere may be classical conditioning of (respondent) painand operant conditioning of pain. I should have said in thetarget article, not that there are two kinds of pain, but thatthere are two causes of pain and therefore two ways ofcontrolling pain. If pain is indeed overt behavior, then itmust be controlled by the same methods used to controlother kinds of overt behavior - namely, respondent andoperant conditioning. If I had said this I would haveavoided the criticisms of those commentators who claim-ed that I had gotten the number of kinds of pain wrong.(Campbell, Fordyce, and Genest say there are more thantwo kinds. Loeser, Logue, and Matson say there is onlyone kind.) But this is a pointless argument, and it is myfault for allowing it to arise.

Whole and part. Four commentators (Bernstein,Graham, Logue, and Place) disagree with my objection(based on molar behaviorism) to a behaviorism of theinner organism; one (Shimp) objects to molar behav-iorism in general; and one (Logue) explicitly supports it.The issue is too complicated to debate fully here. (Seespecial issue on the canonical papers of B. F. Skinner,

THE BEHAVIORAL AND BRAIN SCIENCES (1985) 8.1 75

Pain and behavior

BBS 7(4) 1984, and my commentary on Skinner's "TheOperational Analysis of Psychological Terms" for a discus-sion.) I will merely say that, to me, the expectation thatparts of an animal will be found to function like the animalas a whole (which is the essential assumption of inner-organism behaviorism) is like the expectation that thecarburetor of an automobile will function like an auto-mobile. Furthermore, it seems impossible to explainmental states such as pain with a molecular behavioristictheory unless that theory is extended to the innerorganism.

When asked for an explanation of mental states unac-companied by current overt behavior, molar behavioristspoint to a pattern of past overt behavior and identify thatpattern with the current mental state. But molecularbehaviorists do not have this option, because for thempatterns are not units. This forces (or at least encourages)them, when asked to explain mental states, to point insidethe organism. Then mentalists would have behavioristsjust where they want them. The behaviorists internalmythology is no better than that of the mentalist.

Curare. Two commentators, Campbell and Foss, cite the"fact" that curarized subjects feel pain even though theycannot behave. (Significantly, both Campbell and Fossare philosophers. None of the several commentators whoare physicians or physiologists or even clinical psychol-ogists cited this "fact.") Both Campbell and Foss fullyrealize that, even if it were true, the "fact" would notconstitute evidence against the behavioral theory of painas I presented it. Both commentators therefore ask us toimagine a person curarized from birth to death. Wouldthat poor person feel pain? That, of course, is a mean-ingless question unless one already has a behavioraltheory of pain, in which case the answer is no.

Let us now return to the "fact" that curarized subjectscan feel pain. Foss cites no references. Perhaps hisdentist told it to him, since he refers to curarized dentalpatients. But it is unusual for dental patients to be givencurare. Campbell cites two sources for the failure ofcurare as an anesthetic. I could find only one (Hutchinson1960). In that study 700 patients were interviewed afteroperations to discover whether they remembered any-thing about the operation. The question was put in termsof whether they remembered anything that had hap-pened or whether they had dreamt anything during theoperation. Nine patients reported dreams. Of these, onedream was rejected because it was unconvincing. Of theother eight subjects, six reported discomfort in theirdream, and of these only three reported pain. Two saidthey had not been dreaming but claimed to have beenawake during the operation. Since most of the patientswere anesthetized with a combination of several agents(including curarelike relaxants), this study might seem atbest irrelevant to Campbell's and Foss's point. But,fortunately, 216 of the patients were anesthetized by theLiverpool "pure" technique (uncommon in the UnitedStates) which consists only of a mixture of nitrous oxideand oxygen (which dentists do sometimes give) plus acurarelike relaxant. Of those 216 patients, 6 claimed laterto have had dreams. Of these, not more than 3 (possibly 2)claimed to have had a pain in the dream, and only 2(possibly only 1) described events that had actually oc-curred in the operating room. About these recollections

the author says, "There is very little proof that theexperiences described by the patients really occurredduring the operation." Those two "memories" (or thatone "memory") could have been due to unusually lightanesthesia, to something that occurred as the anestheticwas wearing off or before it took effect, to what thesurgeon might have told the patient before the operation,or to other unknown factors.

It is easy to understand why an anesthetist would bereluctant to take a chance that even one out of a hundredpatients will feel pain during an operation. My questionto Campbell and Foss is, what about those 210 patientswho not only felt no pain, not only felt no discomfort, butdid not remember what happened during the operation?It may well be that, as opposed to dental doses of nitrousoxide plus oxygen, surgical doses are effective anestheticsand do, by themselves, cause loss of consciousness. Inthat case the Hutchinson study is irrelevant to the argu-ment. On the other hand, if the doses of nitrous oxide plusoxygen used on the patients in this study would not havebeen enough by themselves to cause complete anesthesiaand loss of consciousness, the study constitutes strongevidence, not only/or a behavioral theory of pain but alsofor a behavioral theo-v of consciousness (see response toJaynes).

Super-Spartans and pain actors. Happily, not a singlecommentator specifically defended Putnam's super-Spar-tan refutation of behaviorism. This is surprising, since Ihave heard Putnam's argument referred to as "proof" ofthe falsity of behaviorism. However, Campbell did ask formore discussion of "crybabies" (Putnam's perfect actors),and Harman illustrates the super-Spartan and the per-fect-actor argument with homely incidents from his child-hood. The problem we have in relinquishing these argu-ments is our desire to retain the incorrigibility of painreports at all costs. But if you see pain as I do - as awidespread pattern of behavior - it must be possible foryou to be wrong about a specific instance of pain behavior.A simple act may prove not to be pain behavior eventhough it seems to be so at the time; or an act may prove tobe pain behavior even though it does not seem to be so atthe time. It depends on how the pattern develops. To usea familiar Gestalt example, one or two notes may seem tobe part of one melody but prove later to have been part ofanother. What is particularly hard to accept is that thisapplies to our own behavior as well as to that of otherpeople.

As I indicated before, insistence on the incorrigibilityof introspection in this regard is a temptation that weshould train ourselves to resist (as pilots must trainthemselves to resist the temptation to fly by the seats oftheir pants and instead to rely on instruments). Such atemptation exists in physics as well as in psychology (inour intuition that the world is as we immediately perceiveit to be). But the temptation is particularly strong (be-cause the need for immediate solutions is so great) inpsychology. Because we know that the temptation will beso strong, we should be still more on our guard against it.The long-term reward for not trusting to the seats of ourpants with regard to our own pains could be the effectivecontrol of chronic pain. It is particularly disappointing tonote that philosophers, who should be the first, areactually the last to give up their reliance on introspection.

76 THE BEHAVIORAL AND BRAIN SCIENCES (1985) 8 1

Psychologists are not for ahead of them. If we do not giveup this self-indulgence we will truly deserve the wrath ofpeople such as Loeser, who seems to blame us for thevery existence of chronic pain.

Individual commentaries

In this response I go from abstract to concrete: philoso-phers first, then psychologists, then physicians and phys-iologists. Within each division responses are alpha-betically organized. I apologize for any misclassifications.To save space, I emphasize disagreements (of which thereare a sufficiency).

Philosophers. It is quite rightly pointed out by Campbellthat if pains could change in quality while remaining thesame in all their behavioral manifestations, behavioristscould not account for all there is to pain. But a behavioraltheory does not see the words by which the quality of painis usually described (stinging, grinding, shooting, throb-bing) as qualities of pain although, as the names imply,they may be qualities of stimuli that cause pain (see priorsection "Words and pain"). A person who describes a painas "stinging" could be describing a stimulus, not the painitself. The qualities of pain itself would be topographies ofovert behavior (twisting, writhing, grimacing, yelling,crying, and patterns thereof). Campbell was (understand-ably) misled by my initial discussion of psychophysicalresearch (see section "Psychophysics and behavior"). Theproper categories of pain qua behavior are respondentand operant, not sensory and aversive (see "Respondentsand operants"). Of course, within these categories, be-havior varies in topography.

Campbell, after presenting the curare example (see"Curare"), asks for more discussion of fake pain. Let ustherefore consider a great actor in a play. Let us say he isso convincing in portraying pain that everyone believeshim to be actually in pain. Is he really in pain? If, beforeand after the play, he acts normally, then people wouldsay he was faking during the play. If before and after theplay he acts as if in pain, then people would say he was notfaking. They could be wrong. Being a great actor, hecould be acting pain-free outside the theater (say, he hascancer and wants to reassure his friends) but really inpain. Then we might have to observe his behavior overstill wider periods of time to discover the truth. But at nopoint could we settle the question by asking him orpeering inside him. Mis response to our questions wouldjust go along with his other behavior; no instrument forexamining his brain exists whereby a doctor could dis-cover that he was in pain even while he convincinglydenied it. Yet his immediate verbal denials could becontradicted by the pattern of other overt actions.

Campbell claims that a person can be in "comfort"even while "pretending" from birth to death to be inexcruciating pain. That person would be like the man whoclaims on his deathbed to have loved his wife "deepdown" even though he was consistently unfaithful to herand beat her. "Deep down," in these cases, means"seldom." Comfort is just as much overt behavior as painis. Campbell's super-super-crybabies are as inconceiva-ble as Putnam's super-super-Spartans, and for the samereason. They are self-contradictory conceptions.

Foss admits that behaviorism has the logical possibility

Response/Rach\in: Pain and behavior

of being true. Given this, and his apparent rejection ofbehaviorism, you would then expect him to cite theempirical evidence against it. But aside from the curareexample (see "Curare"), Foss provides us only with hisown intuitions and the argument that, because physiolog-ical mechanisms underlie behavior, behaviorism must befalse. He must then also believe that, because physicaland chemical mechanisms underlie neural behavior,physiology is false.

Graham agrees with me that the respondent-operantdistinction applies to pain. But he conceives the stimulusfor the respondent to be inside the organism. What is thedifference, then, between an operant and such a re-spondent? It cannot be that the operant is modifiable byits consequences while the respondent is not. Re-spondents may also be modifiable by their consequences(see "Respondents and operants"). Thus, given internalstimuli, the operant-respondent distinction, the distinc-tion that gives meaning to a behavioral theory of pain, islost.

Graham compares pain to an internal process likedigestion. He claims that digestion leads to movements ofthe whole organism. But it does so only as it interacts withother internal processes. Those movements are bestexplained not in terms of internal antecedents like diges-tion but in terms of deprivation and overt environment-behavior interactions. Digestion per se can be definedfunctionally wholly within the organism. This is not trueof pain.

Harman, when he was a child, faked having pain inorder to distract his parents. He claims now that he wasnot really in pain then. His parents probably had moresense than he gives them credit for. He might have fooledthem once or twice. But had they consistently reinforcedHarman's fakery he might well be a chronic pain patientnow and (looking back with some insight) might trace hispain to those events.

Harman also claims to have had headaches and to havelearned to wait them out. He says he never mentionedthem to anyone else and never acted any differently. If so(unless they have read the commentary), his parents, hisdoctor, his closest friends, and his spouse and children (ifany) must, to this day, still not know about those head-aches. Does anyone want to bet?

Kitcher rejects my pain as "fire-alarm" theory. Shesays that it is the player's injury, not the pain, that stops abaseball game. To carry the metaphor a step further, theinjury is like the fire and the pain is like the fire alarm.Kitcher would say that people make way for fire trucks,not because of the fire alarm, but because of the fire. Allright, but that still leaves the fire alarm with an importantfunction, analogous to the function of pain.

Next, she points out that behavior therapy for paincould be explained by cognitive theory too. I agree. It isalways possible to imagine cognitive mediation of behav-ioral events (just as one can imagine the hand of Godbehind physical laws). But it is behavioral theory thatencourages a concentration on the overt consequences ofpain behavior, and that concentration, I argue, is respon-sible for its success.

Finally, Kitcher makes an attempt to provide a behav-ioral account of Freud's response to his cancer. Sheassumes that operants and respondents must have rigidlydifferent topographies (see "Respondents and operants").


Res/jonse/Rachlin: Pain and behavior

But Freud's response to his cancer must be partly re-spondent and partly operant, the latter having beenshaped from the former.

The single point on which Lacey and I disagree iswhether a behavioral definition of pain depends on phys-ical description of pain behavior. He says it does andpoints to the physical description of a lever press. But apain is not like a single lever press; it is like a pattern oflever presses. With patterns of presses (say fixed-intervalscallops) there is no agreed-upon physical definition(Schneider 1969). Yet surely such patterns are behaviorand not internal states. Similarly, a gymnastic or ice-skating performance or a dive is an overt and not aninternal act. Yet there are no accepted physical criteria forsuch behavior. That is why we need panels of humanjudges to evaluate them.

For responses to Matson's commentary see Melzack'sand Wall's commentary.

Place's concept of "semicovert" behavior solves noneof the problems involved in the concept of covert behav-ior (see "Whole and part" and response to Graham).Place points out that a blush is not all there is to shame. Iagree. We disagree about where the rest of shame is to befound. He looks for it in covert behavior. I look for it inother overt behavior, which, with the blush, forms thepattern we call shame.

Psychologists. The argument of Bernstein for a behav-iorism of organs rather than organisms is really an argu-ment for physiology. I do not favor abandoning phys-iology. I just want to abandon physiological mythology asan explanation of behavior of whole organisms (see"Whole and part," response to Graham, and response toPlace).

Glark reports on signal-detection studies of pain. Thesignal-detection model was intended to distinguish be-tween a subject's ability to make discriminations and thesubject's criteria for reporting those discriminations. Thisis the linear process condemned by Melzack and by Wall.First a stimulus comes (tissue damage), then an "internalphysiological mechanism" is activated. (What or wherethat mechanism is is completely unknown to Clark, but itis exactly there that he locates pains.) Then one decideswhether to tell the world about the pain. If one's criterion(i.e., threshold) is low relative to the pain itself, onereports pain. If it is high, one does not. This picture isexactly what Wall says "there is not a scrap of physiologi-cal or psychological data to support. " Even ignoring theconsiderable problems of interpretation of the signal-detection model, Clark's redundant claim that "empiricalevidence' supports his assertions is gratuitous.

Clark and I agree that if a diseased tooth causes pain,getting rid of the diseased tooth generally gets rid of thepain. But for him that is the extent of pain treatment. Painwithout a diseased tooth or tissue damage of some kind isbeyond his comprehension. He would not label "as inpain people who display pain behavior without tissuedamage." That to him is a "labeling error." Let him tellthis to a person with a migraine headache.

Fordyce's commentary provides a definition of painbehavior that he equates with those behaviors that ob-servers identify as constituting a pain problem. They mayfollow a given stimulus (in which case they are respon-dents) or occur without any clear-cut stimulus (in which

case they are operants). I identify those behaviors withpain. Fordyce prefers to reserve the word "pain" forinternal events to which those behaviors are a response;then he claims to be indifferent about whether the "pain"is really gone when the pain behaviors are gone. Thattactic seems to me to be a waste of a perfectly good wordand a needless concession to his critics. Fordyce claims tocare about people's pain behavior; why should he denythat he cares about their real pain? That denial too easilyallows his significant contributions to the treatment ofpain to be dismissed as "mere " behavioral manipulationsby moralists of the inner man like Shaver & Herrman.And it encourages his identification as a "solipsist of theeighteenth century" by mentalists like Merskey witheighteenth-century models of the mind.

Much of my response to Genest is contained in thesection "Cognitive Behavior Modification." I would add,first, that Genest misses the point that pain is a pattern ofbehavior extended over time, not an individual act. Iagree with him that a person might be in pain nowwithout evincing pain behavior now. But I disagree withhim that a person can be in pain now without evincingpain behavior ever (see "Super-Spartans and painactors").

A second point raised by Genest is quite important. Hefinds it absurd that mental events may not be privateevents (see "Introspection"). This attitude must affect hisbehavior as a clinician. It must sometimes encourage him(contrary to the recommendation of Turk & Salovey) touse cognitive techniques, not as harmless additions tobehavioral ones, but instead of behavioral ones. Hethinks it is ridiculous to look in the mirror to determinewhether you are in pain. Instead, he seems to acceptLocke's metaphor of an internal mirror in which mentalstates are reflected. As Rorty (1979) has convincinglyargued, the metaphor of the internal mirror is not viable.The behaviorist view suggests that the mirror by whichwe see our own bodies is outside of ourselves, in theenvironment, particularly that part of the environmentsensitive to our behavior. This is one reason that behav-iorists stress the importance of what they call "feedbackfunctions" (Baum 1973). Feedback functions (includingschedules of reinforcement) are, for the behaviorist, thetrue mirrors of the mind.

Genest cites a study in which alcoholics reported accu-rately on their own impairment without knowing exactlyhow much they had drunk. My question is, were theyallowed to move? Would they have done still better ifthey could have tried to walk a straight line or look in amirror? What does Genest think he is doing when he tellshis clients to record their own behaviors and gives themfeedback? Who does he think can judge better whether aman is an alcoholic - the man or his family? Finally,Genest says that he can use "entirely internal cues" tomake claims about his myopia. This implies that he canmake such claims with his eyes closed!

Jaynes makes several interesting points that can onlybe dealt with schematically here. First, I do not say thatbecause chronic pain can be controlled by operant meth-ods it logically follows that pain is behavior (see "Theoryand application"). But if chronic pain could be controlledby operant methods, the burden of proof would be shiftedto other conceptions of pain. They would have to showtheir superiority in terms of explanatory power, experi-


Response/ Rachlin: Pain and behavior

mental validity, parsimony, elegance, and so on despitethe success in application of behavioral theory. I claimthat introspections are invalid alongside these other crite-ria (see "Introspection").

Second, it is not feelings of control but actual controlover pain stimuli that diminishes pain in the self-adminis-tered morphine procedure. This fact is entirely consistentwith molar behaviorism, which sees behavior as a func-tion of a relationship between behavior and the environ-ment. A schedule of reinforcement is just an example ofsuch a relationship. It is a mode of control.

Third, phantom limbs are a fascinating and compli-cated subject that I do not wish to oversimplify. But abehaviorist would put phantom limb pain in the samecategory as phantom limb walking. Walking and pain areboth highly practiced behaviors with respondent andoperant components. Without restraint, recent amputeesmust often forget they have only a single leg, get out ofbed, and begin to walk. Such phantom walking would bequickly punished by falling and might be expected todisappear. Phantom pain remains unpunished and, asJaynes points out, is perhaps even rewarded.

Finally, the notion of body consciousness is a usefulway to summarize phantom limb phenomena, especiallythe tendency of phantom limb pain to disappear with theattachment of a prosthesis. Rewards obtainable from painbehavior might be insignificant compared to rewardsobtainable by acting as a whole person. Then the substitu-tion of one set of rewards for the other would be easilymade. But there could be cases where the reverse mightbe true. I would guess that when an amputee is notquickly fitted with a prosthesis and instead adjusts to lifewithout it, rewards might be more evenly balanced andthe coming and going of pain not regulated so well by thegoing and coming of the prosthesis.

Jaynes (1976) is convincing that something significant,which can be characterized as the emergence of con-sciousness, happened in the first millennium B.C. I agreethat consciousness may be considered an operant (see"Curare"), but as such, I insist, it is a pattern of overtbehavior. What happened at that time might have hadmore to do with social relations than anatomy.

Logue agrees with me that molar behaviorism is a goodidea but disagrees about how it should deal with mentalevents. For her, mental events are "hypothetical con-structs" to be verified (presumably at some time in thefuture) by direct observation. In this respect, her com-mentary belongs with those of Bernstein, Graham, andPlace. Skinner (1984) also rejects a behaviorism based onwholly overt behavior. Since covert behaviorism is such acommon way of dealing with mental events (dating fromWatson, 1924) let me try to spell out its assumptions andwhy I object to it. In the case of operant conditioning weknow that when a response is reinforced, other responsesare also strengthened. For instance, if a key peck with aforce of 10 grams is reinforced, key pecks somewhatgreater and less than 10 grams will also be emitted athigher rates than before. Why (these theorists argue)should key pecks of zero grams - internal, covert keypecks - not also be strengthened? And, if covert keypecks can be reinforced, why not covert behavior of allkinds? Response generalization, however, has meaningonly in terms of observable behavior. The conception ofcovert generalization leans heavily on the potential un-

covering of these covert responses. How are they to beuncovered? As behaviorists, Bernstein, Graham, Place,Logue, and Skinner would, I assume, rule out introspec-tion. Another possibility is myographic measurement ofminute muscular movements. This route has been takenin the past and been found inadequate. Minute move-ments are conditionable (Hefferline & Keenan 1963) butseem to have little to do with the mental events that theywere originally supposed to explain. There is the hopethat Logue's "hypothetical constructs" will eventually beuncovered by physiological investigation, but as long asthe hypothetical construct is phrased in behavioral ratherthan physiological terms the hope seems to me a vain one.A "hypothetical construct" of hormonal secretion or ner-vous discharge might eventually be confirmed by phys-iology. A hypothetical construct that amounts to a condi-tionable homunculus seems unlikely to be found byphysiologists. What Logue fails to realize is that the molarbehaviorism that she endorses can explain mental eventswithout such homunculi.

Melzack, a great hero in the study of pain, has morefaith than I do in the future of physiological investigationof fear and anxiety as these states reside in the brain.Otherwise, we agree.

Miles rightly points out that a sensation is not an event.But that does not mean a sensation cannot be an operantor a respondent. Operants (and even respondents), likesensations, may be patterns of events. (See "Cognitivebehavior modification" for a further discussion of Miles'scommentary.)

Shaver & Herrman reveal their prejudices by con-trasting "mere emotional displays" with "emotions." Icannot speak for them, but there has been nothing"mere" about some of the emotional displays I havewitnessed. If I could have been exposed to the full-fledged emotion without the "mere display" I would haveliked it better, I think. Other criticisms raised by Shaver& Herrman have been dealt with in "Theory and applica-tion," "Respondents and operants," and, with more au-thority than I command, in the commentaries of Melzackand of Wall. Shaver & Herrman's final antibehavioristpoint is that it takes "conscious volition . . . to refrainfrom 'immoral' though immediately rewarding behav-ior." What do they say about the feet that pigeons can betrained to refrain from immediately rewarding behavior(Mazur & Logue 1978)?

Shimp classifies me as a positivist. Recently, Shimp(1982) called me a British associationist. I am happy to seehe has moved me up from the nineteenth to the twentiethcentury. His identifications of respondents with "founda-tionist" knowledge and operants with "derived" knowl-edge is strained and beside the point. He classifies Wat-son as a logical positivist, but it was Brunswik (1952) whoidentified himself with the logical positivists - not Wat-son, who (perhaps wisely) wanted nothing to do withphilosophers. Aside from this, Shimp defends molecularbehaviorism by arguments from authority and energeticassertion. But he does not indicate how molecular behav-iorism would account for pain.

My response to Turk & Salovey is mostly given in"Cognitive behavior modification." Their commentary isa good example of the reasoning prevalent in this field. Idefy the reader to discover from their commentary whatcognition is. Is it a coping skill? Is it verbal behavior? Is it


References/RachUn: Pain and behavior

reflection on sensation or behavior? Is it internal restruc-turing of the environment? Is it processing information?Is it perception? Or does it serve as a word to coverwhatever the authors cannot, at the moment, under-stand?

Physicians and physiologists. My answer to Ainslie'sthoughtful commentary is mostly contained in "Intro-spection" and "Words and pain." The difference betweenus lies in the distinction between person A's direct in-teraction with the environment and person B's observa-tion of As interaction with the environment. Ainslie and Idisagree about how to classify A's introspection in thisscheme of things. I say that A's introspection belongs inthe class of A's other behavior. It is part of A's interactionwith the world. Ainslie says that A's introspection isactually an observation - a B-like behavior. In what sensecan a person be A and B at the same time? In other words,how can you know yourself? Ainslie believes you canknow yourself by focusing inward - by taking B's attitudetoward an A that exists inside yourself - that is, byintrospection. I believe you can know yourself by focus-ing outward - by taking an observer's attitude toward theinteraction of your whole body with the environment.This characterizes the difference, not only betweenAinslie and me, but between me and several other com-mentators, for instance, Genest. It is not ridiculous tolook in the mirror to discover your mental state. Onefunction of a therapist, I believe, is to serve as a sort ofselective mirror and, in so doing, to give a person know-ledge of his own mental state.

Atkinson & Kremer cite evidence that different behav-ioral indices of pain are often reliably accompanied byspecific physiological and neurochemical events. Thebehavior that constitutes a given kind of pain may bedifficult to observe because it may extend over widetemporal intervals. The physiology that accompanies agiven kind of pain is difficult to observe because it isalmost always inaccessible. Nevertheless, there maycome a day when instruments will be developed tomeasure these internal events easily. If that occurs it islikely that we will all define pain (as we now define fever)in terms of these measurements rather than in terms ofthe still-difficult-to-observe overt behavior. Then a doc-tor will truly be able in the face of contrary behavioralevidence to tell a person that he is in pain. Until that dayconies, however, we had better define pain behaviorally.

Loeser has a simple way to cut through all the nonsenseabout pain. Pain must be caused by something. Get rid ofthe causes and you get rid of the pain. What are thecauses? Tissue damage and injury to the nervous systemare mentioned first. Other causes are "suffering inducedby depression, anxiety, fear, or isolation." Now all wehave to do is figure out how to get rid of suffering,depression, anxiety, fear, and isolation and we'll be allset.

For Merskey, pain is a "unity [that] may be describedin different languages" - cognitive and physiological.This is a sort of double-aspect theory. But as with all suchtheories, one is tempted to ask: aspects of what? ForMerskey they are two aspects of mind. This conceptionhelps him personally to empathize with his patients, buthe has no way of knowing whether it helps or hinders orhas any effect at all on their pain.

Pepeu's commentary is similar to that of Atkinson &Kremer, and my response is the same. One point raisedby Pepeu with which I would quibble is his statement thatwhile a person's response to painful stimuli can be af-fected by "psychological" factors, the response of rats topainful stimulation can be affected only by pharmacologi-cal or electrophysical manipulation. But we know thatsimply signalling an electric shock will strongly affect arat's response to it (Badia, Culbertson & Harsh 1973).[See also Dinsmoor: "Observing and Conditioned Rein-forcement" BBS 6(4) 1983.]

I agree with Wall's criticism, but it is not a criticism ofmy position. The fault is mine for not having stated itclearly enough.

NOTEJ. R. Kantor (1888-1984) has recently died. Yet he lives, as

behaviorism lives. This article is dedicated to him.

References

Ainslie, G. (1975) Specious reward: A behavioral theory of impulse control.Psychological Bulletin 82:463-96. |rHR]

{in press) Aversion with only one factor. In: Quantitative analyses ofbehavior, vol. 5, The effect of delay and of intervening events onreinforcement value, ed. M. Commons, R. J. Hcrrnstcin 6c H. Kachlin.Ballinger. [taHR]

Almay, B. G. L., Johansson, F., von Knorriug, L., Terenius, L. 6:Wahlstroni, A. (1978) Endorphins in chronic pain. 1. Differences in CSFendorphin levels between organic and psychogenic pain syndromes. Pain5:153-57. [JHA]

Alston, \ \ \ P. (1974) Can psychology do without private data? Behaviorism1:71-102. [CA]

Ainir, S. 6c Amit, Z. (1978) The pituitary gland mediates acute and chronicpain responsiveness in stressed and non-stressed rats. Life Sciences24:439-48. [JHA]

Atkinson, J. H., Kremer, E. F., Risch, S. C. & Janowskv, D. S. (1984) Pre-antl postdexamethasone cortisol concentrations in patients with chronicpain syndromes. Psychiatry Research. In press. 1J11A]

Atkinson, J. H., Kremer, E. F., Risch, S. C , Morgan, C. I)., Afiid, H. F..Ehlers, C. L. At Bloom, F. E. (1983) Plasma measures ol'bcta-endorphin/beta-lipotropin-tike iminunoreactivity in chronic painsyndrome and psychiatric subjects. Psychiatry Research 9:319-27. [JHA]

Badia, P., Culhertson, S. & Harsh, J. (1973) Choice of longer or strongersignalled shock over shorter or weaker unsignalled shock, journal of theExperimental Analysis of Behavior: 19 2 5 - 3 2 . [ iHH]

Bakal, D. A., Deinjen, S. & Kaganov, J. (1981) A cognitive behavioraltreatment of chronic headache. Headache 21:81-86. |I)CT|

Bauin, W. M. (1973) The correlation based law of elfect. Journal of theExperimental Analysis of Behavior 2(1:137-53. [tarHR, CPS]

Beecher, H. K. (1956) Relationship of significance of wound to the painexperienced. Journal of the American Medical Association161:1609. [taHR]

(1959) Measurement of subjective Responses: Quantitative effects of drugs.Oxford University Press. |HM, RM]

Blanchard, E. B., Andrasik, F. A., Arena. J. G. 6c Teders, S. J. (n.d.) Theeffects of differing chronic pain experience on the ps\choph\sical scalingof pain descriptors. Manuscript. State University ol New York atAlbany. ltaHR|

Bliimer, D., Zorick, F., Heilhronn, M. 6: Roth, T. (1982) Biological markersfor depression in chronic pain, journal of Xervous and Mental Disease47:713-15. [JHA]

Bodnar, R. J., Kelly, D. D., Spiaggia, A., Ehrcnheig, C. 6: Clusman, M.(1978) Dose dependent reductions by naloxonc of analgesia induced Incold-water stress. Pharmacology. Biochemistry <L~ Behavior 667-72. [JHA]

Bowers, K. S. (1968) Pain, anxietx, and perceived control. Journal of

Consulting and Clinical Psychology 32:569-602. |JJ|Breland. K. is Breland. M. (1961) The misbehavior of organisms. American

Psychologist 16:681-84. [rHR|Brunswik, E. (1952) The conceptual framework of ps\eholog\. In:


References/ Rachlin: Pain and behavior

International encyclopedia of unified science, vol. 1, no. 10. University ofChicago Press. [rHK]

Buxton Hopkin. D. A. (1980) Hazards and errors in anaesthesia. SpringerVerlag. [KC|

Bykov, K. (1959) The cerebral cortex and the internal organs. Moscow:Foreign Languages Publishing House. [DJB]

Catania, A. C. (1971) Elicitation, reinforcement and stimulus control. In: Thenature of reinforcement. e<l. R. Glazcr. Academic Press. |taHR]

(1983) Behavior analysis and behavior synthesis in the extrapolation fromanimal to human behavior, In: Animal models of human behavior, ed. G.C. L. Davey. Wiley. [AWLl

Catchlove, R. & Cohen, K. (1982) Effects of a directive return to workapproach in the treatment of workman's compensation patients withchronic pain. Pain 1-1:181-91. [HM]

Chapman, C. R. (1977) Sensory decision theory methods in pain research: Areply to Rollman. Priin 3:295-305. [HM]

Chesher, G, B. & Chan, B. (1977) Foot-shock induced analgesia: Its reversalby naloxone and cross tolerance with morphine. Life Sciences 21:1569-74. |JHA]

Chishohn, H. M. (1957) Perceiving: A philosophical study. Cornell UniversityPress. ItaHR]

Clark, W. C. (1984) Applications of multidimensional sealing to problems inexperimental and clinical pain: An introduction. In: New approaches topain measurement in man, ed. B. Bromm. Elsevier. IWCC]

Clark, W. C. & Yang, J. C. (1983) Applications of sensory decision theory toproblems in laboratory and clinical pain. In: Pain measurement andassessment, ed. R. Melzaek. Raven Press. IWCC]

Coid. J., Allolio, B. & Rees, H. (1983) Raised plasma metenkephalin inpatients who habitually mutilate themselves. Lancet 2:545-46. IGP]

Colpaert, F. C. (1979) Can chronic pain be suppressed despite purportedtolerance to narcotic analgesia? Life Sciences 24:1201-10. [GP]

Dallenbach, K. M. (1939) Pain: History and present status. American Journalof Psychology 52:331-47. [taHR]

Dennett, D. C. (1978) Brainstorms. Bradford Books. |taHR](1983) Intentional systems in cognitive ethology. Behavioral and Brain

Sciences 6:343-90. [GO]DeVries, G. H., Chance, W. T., Payne, \V. R. & Rosencrans, J. A. (1979)

Effect of autoanalgesia on CNS enkephalin receptors. Pharmacology,Biochemistry and Behavior 11:741-44. [GP]

Driver, D. (1984) Good nutrition and pain control vital with burns. MedicalPost. April 17 p. 64. |JJ]

Elmore, A. M. (1979) A comparison of the psychophy.sical and clinicalresponse to biofeedback for temporal pulse amplitude reduction andbiofeedback for increases in hand temperature in the treatment ofmigraine. Thesis, psychology department. State University of New Yorkat Stony Brook. [taHR]

Etzel, B. C. & Gewirtz, J. L. (1967) Experimental modification of caretaker-maintained high-rate operant crying in a 6- and a 20-weck-old infant(infans tyrannotearu.s): Extinction of crying with reinforcement of eye-contact and smiling. Journal of Experimental Child Psychology 5:303-17. ItaHR]

Faccini, E., Uzuniaki, H., Govoni, S., Missale, C , Spano, P. F., Covelli, V.& Trabncchi, M. (1984) Afferent fibers mediate the increase ofmetenkephalin elicited in rat spinal cord by localized pain. Pain 18:25-32. [CP]

Flor, H., Haag, C., Turk, D. C , & Koehler, H. (1983) Efficacy of EMG

biofeedback, psychotherapy, and conventional medical treatment forchronic rheumatic back pain. Pain 17:21-31. |DCT]

Fodor, J. A. (1981a) Rcprcscntationfs: Philosophical essays on the foundationsof cognitive science. MIT Press. [WCC]

(1981b) The mind-body problem. Scientific American 244:114-23. [\VCC[Fodor, J. A., Garrett, M. F. & Bever, T. G. (1974) The psychology of

language. McGraw-Hill. |taHR]Fordyce, W. E. (1976) Behavioral methods for chronic pain and illness. C. V.

Mosby. [WCC, HL. taHR, KCS, DCT](1978) Learning processes in pain. In: The psychology of pain, ed. R. A.

Sternbach. Raven Press. [UTP, taHR](1983a) The validity ol pain behavior measurement. In: Pain measurement

and assessment, ed. R. Melzack. Raven Press. [KC, taHR](1983b) Behavior is behavior is. . . . Contemporary Psychology

9:730. [HM]Fordyce, W. E., Fowler, R. S., Lehman. J. F. & DeLateur, B. J. (1968)

Some implications of learning in problems of chronic pain. Journal ofChronic Diseases 21:179-90. [JHA, JJ]

Geueli, P. (1957) Mental acts. Humanities Press. |taHR]Gibbon, J., Berryman, R. At Thompson, R. L. (1974) Contingency spaces and

measures in classical and instrumental conditioning. Journal of theExperimental Analysis of Behavior 21:585-605. [taHR]

Graham, G. (1982) Spartans and behaviorists. Behaviorism 10:137-49. [GA,GG]

Graham, G. & Stephens, G. L. (1984) Are qualia a pain in the neck forfunctionalists? American Philosophical Quarterly. [GG]

Hanson, N. R. (1969) Perception and discovery, Freeman, Cooper &Co. [CPS]

Hefferline, R. F. 6c Keenan, B. (1963) Amplitude-induction gradient of asmall-scale (covert) operant. Journal of the Experimental Analysis ofBehavior 6:307-15. [rHR]

Henry, J. L. (1982) Circulating opioids: Possible physiological roles in centralnervous function. Neuroscience and Bio-behavioral Reviews 6:229-45. [GP]

Herrnstein, R. J. (1970) On the law of effect. Journal of the ExperimentalAnalysis of Behavior 13:243-66. [taHR]

Hilgard, E. R. (1969) Pain as a puzzle for psychology and physiology.American Psychologist 24:103-13. IWCC]

Hilgard, E. R. & Hilgard, J. R. (1975) Hypnosis in the relief of pain. WilliamKaufman. [taHR]

Hinson, J. M. 6c Staddon, J. E. R. (1983) Hill-climbing by pigeons. Journal ofthe Experimental Analysis of Behavior 39:25-47. |CPS]

Holroyd, K. A., Andrasik, F. & Westbrook, T. (1977) Cognitive control oftension headache. Cognitive Therapy and Research 1:121-33. [DCT]

Hutchinson, R. (1960) Awareness during surgery. British Journal ofAnaesthesia 33:463-66. [KC, rHR]

Hull, C. L. (1943) Principles of behavior. Appleton-Century. ItaHR](1952) A behavior system. Yale University Press. [taHR]

International Association for the Study of Pain (1979) Pain terms: A list withdefinitions and notes on usage. (Recommended by the IASPSubcommittee on Taxonomy). Pain 6:249-52. [HM]

Jaspers, K. (1963) General psychopathology. 7th ed. Trans. J. Hoenig 6c M.W. Hamilton. Manchester University Press. [HM]

Jaynes, J. (1976) The origin of consciousness in the breakdown of thebicameral mind. Houghton Mifflin. [JJ1

(1982) A two-tiered theory of emotions. Behavioral and Brain Sciences5:434-35. [JJ]

Jones, B. (1979) Signal detection theory and pain research. Pain 7:305-12. [HM]

(1980) Professor Rollman's response. Pain 9:382. [HM]Jones, E. (1961) The life and work of Sigmund Freud. Basic Books. [PK]Kantor, J. R. (1963) The scientific evolution of psychology, vol. 1. Prineipia

Press. [rHR]Kelly, D. D. (1981) Somatic sensory system. 4. Central representations of

pain and analgesia. In: Principles of neural science, ed. E. R. Kandel &J. H. Schwartz. Elsevier. ItaHR]

Kleinknecht, R. A. 6c Donaldson, D. (1975) A review of the effects ofdiazepam on psychomotor performance. Journal of Nervous and MentalDisease 161:399-411. IJHA]

Koenig, P. (1973) The placebo effect in patent medicine. Psychology Today1:60. [taHR]

Kremer, E. F., Block, A. 6c Atkinson, J. H. (1983) Assessment of painbehavior: Factors that distort self-report. In: Pain measurement andassessment, ed. R. Melzack. Raven Press. [JHA[

Lacey, H. M. & Rachlin, H. (1978) Behavior, cognition and theories ofchoice. Behaviorism 6:177-202. ]HL, taHR]

Langer, E. J., Janis, I. L., 6c Wolfer, J. A. (1975) Reduction of psychologicalstress in surgical patients. Journal of Experimental Social Psychology1:155-65. [DCT]

Leavitt, F. 6c Garron, D. C. (1979) The detection of psychological disturbancein patients with low back pain. Journal of Psychosomatic Research 2:149-54. [HM]

Lenburg, C. B., Glass, H. P. 6c Davitz, L. J. (1970) Pain in relation to thestage of the patient's illness and occupation of the perceiver. NursingResearch 19:392-98. [MG[

Levine, J. D., Gordon, N. C. 6c Fields, H. L. (1979) The role of endorphinsin placebo analgesia. In: Advances in pain research and therapy, vol. 3,ed. J. J. Bonica, J. C. Liebeskind 8c D. Albe-Fessard. RavenPress. [taHR]

Lippman, G. & Meyer, M. E. (1967) Fixed interval performance as related toinstructions and to subject's verbalisations of the reinforcementcontingency. Psychonomic Science 8:135-36. [UTP]

Locurto, L. C , Terrace, H. S. fit Gibbon, J. (1981) Autoshaping andconditioning theory. Academic Press. [rHR]

Logue, A. W. (in press a) The growth of behaviorism. In: Points of view inthe modern history of psychology, ed. C. Buxton. Academic

Press. [AWL](in press b) The origins of behaviorism. In: Points of view in the modern

history of psychology, ed. C. Buxton. Academic Press. [AWL]Logue, A. W., Pena-Correal, T. E., Rodriguez, M. L. & Kabela, E. (1984)


fie/erences/Rachlin: Pain and behavior

Sensitivity to reinforcement in humans: Variations in reinforcer amountand delay. Submitted. [AWL]

Lowe, C. F. (1979) Determinants of human operant behavior. In: Advances inthe analysis of behavior, vol. 1: Reinforcement and the organization ofbehavior, ed. M. D. Zeiler & P. Harzem. Wiley. [UTP]

(1983) Radical behaviorism and human psychology. In: Animal models ofhuman behavior, ed. G. C. L. Davey. Wiley. [AWL, UTP]

McCaul, K. D. & Malott, J. M. (1984) Distraction and coping with pain.Psychological Bulletin 95:316-33. [WCC]

McCollam, J. B., Burish, T. C , Maisto, S. A. & Sobell, M. B. (1980)Alcohol's effects on physiological arousal and self-reported affect andsensations. Journal of Abnormal Psychology 89:224-33. [MG]

McGlashen, T. H., Evans, F. J. & Orne, M. T. (1969) The nature of hypnotic-analgesia and placebo response to experimental pain. PsychosomaticMedicine 31:227-46. [taHR]

Maier, S. F., Seligman, M. E. P. & Solomon, R. L. (1969) Pavlovian fearconditioning and learned helplessness. In: Punishment and aversivebehavior, ed. B. A. Campbell & R. M. Church. Appleton-Century-Crofts. [taHR]

Mann, R. E., Vogel-Sprott, M. & Cenest, M. (1983) Alcohol and cognition.Journal of Studies on Alcohol 44:900-905. [MG]

Mason, J. W. (1975) Emotions as reflected in patterns of endocrineintegration. In: Emotions, their parameters and measurement, ed. L.Levi. [JHA]

Mason, J. W., Brady, J. V. & Tolliver, G. A. (1968) Plasma and urinary 17-hydroxycorticosteroid responses to 72-hour avoidance sessions in themonkey. Psychosomatic Medicine 30:608-30. [JHA]

Mazur, J. E. & Logue, A. W. (1978) Choice in a "self-control" paradigm:Effects of a fading procedure. Journal of the Experimental Analysis ofBehavior 30:11-17. [rHR]

Meites, J., Bruni, J. F., Van Vugt, D. A. & Smith, A. F. (1979) Relation ofendogenous opioid peptides and morphine to neuroendocrine functions.life Science 24:1325-36. [JHA]

Melzack, R. (1973a) How acupuncture can block pain. Impact of Science onSociety 23:1-8. [taHR]

(1973b) Thepuzde of pain. Penguin. [taHR]Melzack, R., ed. (1983) Pain measurement and assessment. Raven

Press. [tarHR]Melzack, R. & Casey, K. L. (1968) Sensory, motivational and central control

determinants of pain: A new conceptual model. In: The skin senses, ed.D. Kenshalo. Charles C. Thomas. [RM]

Melzack, R. & Wall, P. D. (1965) Pain mechanisms: A new theory. Science150:971. [GP, taHR]

(1970) Psychophysiology of pain. International Anesthesiology Clinics 8:3-34. [taHR]

(1982) The challenge of pain. Penguin Education. |PDW](1983) The challenge of pain. Basic Books. [JJ, RM]

Meltzer, H. Y., Fang, V. S., Tricon, B. J., Robertson, A. & Piyaka, S. J.(1982) Effect of dexamethasone on plasma prolactin and cortisol levels inpsychiatric patients. American Journal of Psychiatry 139:763—66. [JHA]

Merskey, H. (1964) An investigation of pain in psychological illness. D. M.

thesis, Oxford. [HM](1968) Psychological aspects of pain. Postgraduate Medical Journal 44:297-

306. [taHR](1983) Development of a universal language of pain syndromes. In:

Advances in pain research and therapy, ed. J. J. Boniea, V. Lindblom &

A. Iggo. Raven Press. [JHA]Merskey, H. & Spear, F. G. (1967) Pain: Psychological and psychiatric

aspects. London: Bailliere, Tindall & Cassell. [HM]Michael, J. (1982) Distinguishing between discriminative and motivational

functions of stimuli. Journal of the Experimental Analysis of Behavior37:149-55. [UTP]

Millan, M. J., Przewlocki, R. & Hera, A. (1981) A non-B-endorphinergicadenohypophyseal mechanism is essential for an analgetic response tostress. Pain 8:343-53. [JHA]

Mischel, H. N., Fuhr, R. & McDonald, M. (1984) Children's dental pain: Theeffects of cognitive coping training in a clinical setting. Manuscript,

Stanford University. [WCC]Mischel, W. (1981) Objective and subjective rules for delay of gratification.

In: Cognition in human motivation and learning, ed. G. d'Ydewalle & W.Lens. Lawrence Erlbaum. [AWL]

Mitchell, K. R., & White, R. G. (1977) Behavioral self-management: Anapplication to the problem of migraine headaches. Behavior Therapy

8:213-22. [DCT]Morichi, R. & Pepeu, G. (1979) A study of the influence of hydroxyzinc and

diazepam on morphine antinociception in the rat. Pain 7:173-80.[GP|

Nisbett, R. E., & Wilson, T. D. (1977) Telling more than we can know:

Verbal reports on mental processes. Psychological Review 84:231-59. (MC, taHR]

Nouwen, A. & Solinger, J. W. (1979) The effectiveness of EMC biofeedbacktraining in low back pain. Biofeedback and Self-Regulation 4:103-11. |DCT]

Orne, M. T. (1962) On the social psychology of the psychological experiment:With particular reference to demand characteristics and theirimplications. American Psychologist 17:776-83. [KCS]

Paalzow, G. & Paalzow, L. (1973) The effect of caffeine and theophylline onnociceptive stimulation in the rat. Acta Phannacologica et Toxicologicu32:22-32. [CP]

Pavlov, I. P. (1927) Conditioned reflexes. Trans. C. V. Anrcp. OxfordUniversity Press. |taHR]

Peele, D. B., Casey, J. fit Silberbcrg, A. (1984) Primacy of interresponse timereinforcement in accounting for rate differences under variable-ratio andvariable-interval schedules. Journal of Experimental Psychology: AnimalBehavior Processes 10:149-67. ]CPS]

Pinsky, J. J., Griffin, S. E., Agnew, D. C , Kamdar, M. D., Crue. B. L. &Pinsky, L. H. (1979) Aspects of long-term evaluation of pain unittreatment program for patients with chronic intractable benign painsyndrome: Treatable outcome. Bulletin of the Los Angeles ScurologicalSociety 44:53-69. [HM]

Place, U. T. (1981) Skinner's Verbal behavior. 1. Why we need it.Behaviorism 9:1-24. |HL]

Platt, J. R. (1979) Interresponse time shaping by variable-interval-likeinterresponse-time reinforcement contingencies. Journal of theExperimental Analysis of Behavior 31:3-14. [CPS]

Popper, K. R. (1959) The logic of scientific discovery. Basic Books. |KGS|Putnam, H. (1980) Brains and behavior. In: Readings in philosophy of

psychology, vol. 1, ed. N. Block. Harvard University Press. [taHR]Raehlin, H. (1971) On the tautology of the matching law. Journal of the

Experimental Analysis of Behavior 15:249-52. [CPS](1976) Behavior and learning. W. H. Freeman. |taHR)(1977a) Reinforcing and punishing thoughts. Behavior Therapy 8:659-

65. [rHR](1977b) Review of Cognition and Behavior Modification by M. J. Mahoney.

Journal of Applied Behavioral Analysis. [rHR](1978) A molar theory of reinforcement schedules. Journal of the

Experimental Analysis of Behavior 30:345-60. ItaHR]

(1984) Mental yes. Private, no. Behavioral and Brain Sciences 7:566-67. [rHR]

Razran, C. (1961) The observable unconscious and the inferable conscious incurrent Soviet psychophysiology: Interoceptive conditioning, semanticconditioning, and the orienting reflex. Psychological Review 68:81-147. [DJB]

Reseorla, R. A. (1967) Pavlovian conditioning and its proper controlprocedures. Psychological Review 74:71-80. [taHB]

(1984) Associations between Pavlovian CS's and context. Journal ofExperimental Psychology: Animal Behavior Processes 10:195-204. |rHR|

Roback, A. A. (1923) Behaviorism and psychology. Cambridge UniversityBookstore. [AWL]

Rollman, G. B. (1977) Signal detection theory measurement of pain: A reviewand critique. Pain 3:187-211. [HM]

(1979) Signal detection theory pain measures: Empirical validation studiesand adaptation-level effects. Pain 6:9-21. [HM|

(1980) On the utility of signal detection theory pain measures. Pain 9:375-79. [HM]

Rorty, R. (1979) Philosophy and the mirror of nature. Princeton UniversityPress. [rHR, CPS]

Russell, W. R. & Spalding, J. M. K. (1950) Treatment of painful amputationstumps. British Medical Journal 2:68-73. [JJ|

Rybstein-Blinchik, E. (1979) Effects of different cognitive strategies on chronicpain experience. Journal of Behavioral Medicine 2:93-101. |rHR, D(.T|

Ryle, G. (1949) The concept of mind. Hutchinson. |TRM, taHR. CPS]Sanders, S. (1980) Assessment of uptime in chronic low back pain patients:

Comparison between self-report and automation. Paper presented atAssociation for Advancement of Behavior Therapy, New York,N.Y. [JHA]

Savin, H. (1980) Introduction: Behaviorism. In: Readings in philosophy ofpsychology, ed. N. Block. Harvard University Press. [taHR]

Schelling, T. C. (1960) The strategy of conflict. Harvard UniversityPress. [GA]

Schneider, B. A. (1969) A two-state analysis of fixed-interval responding in the

pigeon. Journal of the Experimental Analysis of Behavior 12:677-89. [rHR]

Schwartz, B. & Gamzu, E. (1977) Pavlovian control of operant behavior. In:

Handbook of operant behavior, ed. W. K. Honig & J. E. R. Staddou.Prentice-Hall. |tarHR]


References/ Rachlin: Pain and behavior

Senile, J. H. (1980) Minds, bruins and programs. Behavioral and BrainSciences 3:417'-57. ICA, DJB, laHR]

Shimp, C. P. (1975) Perspectives on the behavioral unit: Choice behavior inanimals. In: Handbook of learning and cognitive processes, vol. 2, ed. W.K. Estes. Lawrence Erlhaum.

[CPS| (1970) Organization in meinory and behavior. Journal of theExperimental Analysis of Behavior 26:113-30. [CPS]

(1982) Heinforcemcnt and the local organization of behavior. In:Quantitative analyses of behavior 2. Matching and maximizing accounts,ed. M. L. Commons. R. J. Herrnstein & H. Rachlin. Ballinger. [rHR]

Siegel, S., Hinson, R. E. & Kmnfc, M. D. (1981) Morphine-inducedattenuation of morphine tolerance. Science 212:1533-34. [taHR]

Siminel, M. (1962) Phantom experiences following amputation in childhood.Journal of Neurosurgery and Psychiatry 25:69-72. [JJ]

Skinner, B. F. (1935) The generic nature of the concepts of stimulus andresponse. Journal of General Psychology 12:40-65. [taRH]

(1938) The behavior of organisms. Appleton-Century-Crofts. [C.A, UTP,tarHR]

(1945) The operational analysis of some psychological terms. PsychologicalReview 52:270-77. [DJB]

(1950) Are theories of learning necessary? Psychological Review 57:193-216. [CPS]

(1953) Science and human behavior. Macinillan. (PK, rHR](1969) Contingencies of reinforcement. Appleton-Century-Crofts. [UTP](1979) The shaping of a behaviorist. Knopf. [AWL](1984) The operational analysis of psychological terms. Behavioral and Brain

Sciences 7:547-53. |rHH]Studdon, J. E. R. (1973) On the notion of cause with application to

behaviorism. Behaviorism 1:25-64. [taHR]Staddon, J. E. R., ed. (1980) Limits to action: The allocation of individual

behavior. Academic Press. [taHR]Sternbach, R. C. (1983) Fundamentals of psychological methods in chronic

pain. In: Advances in pain research and therapy, ed. J. J. Bonica ct al.Raven Press. [HM]

Szasz, T, S. (1957) Pain and pleasure; A study of bodily feelings. TavistockPublications. |HM]

Teitelhauin, P. (1977) Levels of integration of the opcrant. In: Handbook ofoperant behavior, ed. J. E. R. Staddon & W. K. Honig. Prentice-Hall. [tarHR]

Teske, K., Daut, R. L. & Cleelaml, C. S. (1983) Relationships betweennurses' observations and patients' self-reports of pain. Pain 16:289-96. [MC|

Thomas, S. N. (1978) The formal mechanics of mind. Cornell UniversityPress. [GG]

Tohnan, E. E. (1948) Cognitive maps in rats and men. Psychological Review55:189-208. |taHR]

Turk, D. C , Meichenbaum, D. H. fit Genest, M. (1983) Pain and behavioralmedicine: A cognitive-behavioral perspective. Guilford Press. IMG,tarHR, DCT]

Turner, J. A. (1982) Comparison of group progressive relaxation training andcognitive-behavioral group therapy for chronic low back pain. Journal ofConsulting and Clinical Psychology 50:757-65. |DCT]

Tursky, B-, Jamner, L. fit Friedman, R. (1982) The pain perception profile: Apsychophysical approach to the assessment of pain report. BehaviorTherapy 13:376-94. [GG, MG, taHR, KGS]

Tursky, B. & Stcmbach, R. A. (1967) Further physiological correlates ofethnic differences in responses to shock. Psychophysiology 4:67-74. itaHR]

Wall, P. D. (1976) Modulation of pain by nonpainful events. In: Advances inpain research and therapy, vol. 1, ed. J. J. Bonica fit D. Albc-Fessard.Raven Press. [GP]

(1979) On the relation of injury to pain. Pain 6:253-64. [PDW](1984) Introduction. In: Textbook of pain, ed. P. D. Wall & R. Melzack.

Churchill Livingstone. |PDW]Wall, P. D. & McMahon, S. B. (1984) Microncuronography and its relation to

perceived sensation. Pain. In press. [PDW]Wall, P. D. & Sweet, W. H. (1967) Temporary abolition of pain. Science

155:108-9. [JJ]Walters, A. (1963) The psychological aspects of bodily pain. Applied

Therapeutics 5:853-56. [HM]Watson, J. B. (1913) Psychology as the behaviorist views it. Psychological

Review 20:158-77. [AWL, taHR](1970) Behaviorism. W. W. Norton & Company (Orig. publ. 1924). [GA,

tarHR]Weiner, H. (1977) Psychobiology and human disease. Elsevier. [JHA]Weisenberg, M. (1975) Editor's comment: Section 7, surgical intervention to

pain. In: Pain, clinical and experimental perspectives, ed. M.Weisenberg. C. V. Mosby. [taHR]

(1977) Pain and pain control. Psychological Bulletin 84:1008-44. iGA]Winokur, A., Amsterdam, J., Caroff, S., Snyder, P. J. fit Brunswick, D.

(1982) Variability of hormonal responses to a scries of neuroendocrinechallenges in depressed patients. American Journal of Psychiatry 139:39-42. [JHA]

Wittgenstein, L. (1922) Tractatus logico-philosophicus. Routledge and KeganPaul. [CPS]

(1958) Philosophical investigations. 3rd ed. Trans. G. E. M. Anseombe.Macinillan. [taHR, CPS]

Wolitzer, H. (1983) In the Palomar Arms. Farrar, Straus and Giroux. [taHR]Zuriff, G. E. (1979) Ten inner causes. Behaviorism 7:1-8. IGA, taHR]


New from Bradford BooksQualitative Reasoning about Physical Systemsedited by Daniel G. BobrowQualitative or Naive Physics is a rapidly developing area of cognitive science. The contributions in thisbook present the most recent work on qualitative reasoning about the real (physical) world. A themecommon to all of them is explaining how physical systems work-from heat flow to transistors to digitalcomputation. These explanations are so detailed and exact that they can be used by computer programsto reason about physical work in the same kinds of ways that people do. The articles in QualitativeReasoning about Physical Systems constituted a special issue of the Journal of ArtificialIntelligence.$22.50 (softcover)

Gavagai!The Future History of the Animal Language Controversy

by David PremackHow do aliens communicate? Suppose one says "Gavagai!" What does it mean? In this book, DavidPremack questions our fascination with language as a tool to our understanding of the species. Heclearly and wittily untangles the complex skein of arguments, put forth from Darwin's time to ours,over whether man is unique because he can talk. We would do well, he asserts, to rid ourselves of ourinfatuation with language as the major human specialization and focus instead on other evolutionarypuzzles.$12.50

Now available in paperbackSituations and Attitudesby Jon Barwise and John PerrySituations and Attitudes tackles the slippery subject of "meaning," providing the basics of a realisticmodel-theoretic semantics of natural language, explaining the main ideas of the theory, and contrastingthem with those of competing theories.

"Situations and Attitudes is a godsend to many a serious but overworked student of cognition, espe-cially in the areas of psychology and artificial intelligence... [It] is not only a new theory in semantics, itis the theory of the future."-John Macnamara, McGill University$9.95

From Folk Psychology to Cognitive Science:The Case Against Belief

Stephen P. StichIn this book, Stich puts forth the radical thesis that the notions of believing, desiring, thinking, pre-ferring, feeling, imagining, fearing, remembering, and many other common-sense concepts thatcomprise the folk psychological foundations of cognitive psychology should not-and do not-play asignificant role in the scientific study of the mind.

"A common view of cognitive science is that it is succeeding in making honest scientific theories of oureveryday mentalistic lore-doing for folk psychology more or less what Galileo did for folk physics. Thereare a variety of avenues to this optimistic and comfortable view: Stich claims to have blocked them all...This is the best sort of expert research and authorship. It is vivid, informed, up-to-the-minute, andvigorously argued."-Daniel C. Dennett, Tufts University.$8.95

The MIT Press28 Carleton Street, Cambridge, MA 02142

Pain and Behaviour (Rachlin, 1985)

Documents

psychological pain

kinds of pain

severe pain

kind of pain

study of pain

behavioral theories

fundamental sensory

subjective experience