PERSONS AND PLACEBOS= PHILOSOPHICAL DIMENSIONS OF THE PLACEBO EFFECT A Dissertaflon for the Degree of DI1. D. MICHIGAN STATE UNIVERSITY Howard Brody I977 iffi’.’ I ‘3‘!” CL,“ I lll/Illll/f/l l/I/ Ill/ILA ' 11“???“er w “4.4” file; if ‘3‘ %- y This is to certify that the thesis entitled Persons and Placebos: Philosophical Implications of the Placebo Effect presented by Howard Brody has been accepted towards fulfillment of the requirements for Ph . D . degree in Philosophy Major professor Date 4/21/77 0-7639 ’_1"’ \ 7%.?! 9 Ive/Iv“ 9 V7 SEPM 21306 "'1" M; £17,, MAI? 232005" ._..,_ 1 mama ABSTRACT PERSONS AND PLACEBOS: PHILOSOPHICAL DIMENSIONS OF THE PLACEBO EFFECT By Howard Brody The placebo effect occurs within medicine when a patient improves after being given a chemically inert substance, under the impression that he has been given an active drug. Although this phenomenon has been studied extensively by medical scientists recently, little atten- tion has been paid to it by philosophers. Three philosophical issues deserve exploration: 1) the meaning and breadth of the term 'placebo effect'; 2) implications of the placebo effect for the mind-body rela— tion; and 3) ethical problems regarding the clinical use of placebos. Investigation of these issues will proceed in an interlocking fashion, so that each sheds additional light on the others. A formal definition of the placebo effect may be approached by first reviewing the medical literature, and then analyzing the empiri- cal data through the use of illustrative case examples designed to in- dicate the boundaries of the term's applicability. This procedure yields a definition with several important features. First, the placebo effect may be present even where no placebo has been used-- in particular, we may wish to speak of a placebo component accompanying the administration of biomedically active therapy. Also, two important references must be made to belief states. One is the belief state of the individual, as the placebo effect can meaningfully be said to occur only if the indi- vidual believes that he is in a healing setting receiving some sort of Howard Brody therapy. The other is the belief state of medical science, as economy of explanation requires that we do not attribute a symptom change to the placebo effect if it is explainable on the basis of some other well-accepted medical theory. The importance of the individual's belief state points out the mind-body implications of placebo phenomena. Many theories falling within the framework of Cartesian dualism, including causal interaction- ism, behaviorism, identity theory, and eliminative materialism, can give plausible accounts of the placebo effect; but weaknesses in all of these theories can be found on other grounds. The first three the- ories have serious flaws that have been well characterized by tradi— tional philosophical arguments; eliminative materialism becomes less plausible when it is seen to involve a radical change in our form of life, one that would involve among other things the loss of any basis for moral reasoning. An alternative to the Cartesian tradition, however, is to take the concept of person rather than the concepts of mind and body as the primitive term. A theory can be developed which treats being a per- son as being an animal with the capacity to use symbols in such a way that the symbols acquire meaning through the use. The capacity theory of person gives an illuminating account of the placebo effect, squares well with considered judgments about other mind-body issues, and in par- ticular emphasizes that to be a person is necessarily to be a dweller within culture and language, a facet of human existence not adequately dealt with by other mind-body theories. Howard Brody The ethical problems, the third issue area to be discussed, re- duce in large part to the justification of deception in medical prac- tice. Historically, many physicians have attempted to justify placebo use, but both deontological and utilitarian arguments can be used to establish a prima facie presumption against the deceptive use of pla- cebos. The formal definition, by showing that one need not give a placebo in order to elicit a beneficial placebo effect, points the way toward alternative, nondeceptive ways of securing the benefits of the placebo effect for patients. These philosophical discussions suggest the utility of the model of the placebo effect which takes its crucial feature to be the imposing of meaning upon the patient's illness experience; this model in turn suggests several interesting lines of empirical inquiry. Finally, by touching in a mutually supportive way upon the empirical-conceptual, the metaphysical, and the normative realms, all of which are bound up in the theory and practice of medicine, this investigation of the pla- cebo effect provides a possible model for the discipline of philosophy of medicine. PERSONS AND PLACEBOS: PHILOSOPHICAL DIMENSIONS OF THE PLACEBO EFFECT By ar“ Howard Brody A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Philosophy 1977 Copyright by HOWARD ALLAN BRODY l977 ACKNOWLEDGMENTS The research and preparation of this dissertation was supported by a Fellowship from the Institute on Human Values in Medicine, Phila- delphia, financed by National Endowment for the Humanities Grant # EH- l0973-74-365. While I appreciate this financial support, I am especially grateful for the advice and encouragement from Institute directors and staff, including Thomas K. McElhinney, Ph.D., Edmund D. Pellegrino, M.D., and Ronald w. McNeur, Ph.D. The most gratifying feature of completing this work was the oppor- tunity to engage the intellects of a spirited and enthusiastic disser- tation committee-- Joseph F. Hanna, Ph.D., and Martin Benjamin, Ph.D., of the Department of Philosophy; and William 8. Neil, Jr., M.D., and James E. Trosko, Ph.D., of the Department of Human Development, Col- lege of Human Medicine. Helpful ideas were contributed by many people too numerous to men- tion, at Michigan State University and elsewhere; among those kind enough to suggest references or lines of inquiry were Arthur F. Kohrman, David S. Sobel, Joseph Margolis, Arthur Kleinman, Peter Vinten-Johansen, Donald A. Kennedy, and Sumer Verma. ii TABLE OF CONTENTS Introduction. Philosophical Dimensions of the Placebo Effect ........ l Chapter l. The Placebo Effect: A Review of the Medical Literature..ll 1.1. History and Definition ................................... 12 1.2. Nature and Scope of Response ............................. 14 1.3. Agents Acting as Placebos ................................ 16 1.4. Factors Influencing the Placebo Effect ................... 18 1.5. Theories of Placebo Mechanisms ........................... 23 Chapter 2. A Definition of the Placebo Effect ...................... 31 2.1. The Placebo Effect as Medical Anomaly .................... 31 2.2. Boundaries of the Placebo Effect ......................... 37 2.3. Formal Definition of 'Placebo Effect' .................... 46 Chapter 3. Traditional Mind-Body Views and the Placebo Effect ...... 56 3.1. A Reflective—Equilibrium Approach to Mind ................ 57 3.2. Overview of Alternative Mind-Body Theories ............... 60 3.3. Some Initially Plausible Theories ........................ 65 3.4. Rejection of Commonly Held Theories ...................... 68 Chapter 4. Eliminative Materialism ................................. 75 4.1. Features of Eliminative Materialism ...................... 75 4.2. Objections to Eliminative Materialism .................... 81 4.3. The Concept of Person .................................... 93 Chapter 5. A Theory of the Person .................................. 97 5.1. The Capacity Theory of Person ............................ 97 5.2. The Capacity Theory and the Placebo Effect .............. 102 5.3. The Capacity Theory and Considered Judgments ............ 106 5.4. Problems with the Capacity Theory ....................... 112 Chapter 6. Ethical Problems in Placebo Use ........................ 120 6.1. Historical Background ................................... 121 6.2. Arguments For and Against Placebo Use ................... 125 6.3. The Arguments Summarized ................................ 129 6.4. Limited Placebo Use ..................................... 133 6.5. Alternatives to Placebos ................................ 136 Conclusion. Recapitulation and Research Implications .............. 144 Bibliography ............ . ......................................... 156 iii Introduction. Philosophical Dimensions of the Placebo Effect Physicians have known for at least several centuries that patients often display marked improvement of symptoms when given a sugar pill, or another substance having no known medicinal properties, under the impression that it is an active drug. A biomedically inert substance given in such a manner to produce relief is known as a placebo, and the resulting influence upon the patient may be called the placebo ef- fect. With the advent of large-scale clinical trials of drugs and therapeutics within the last three decades, placebos have become an important way of eliminating investigator bias in medical research de- sign. As a result, a good deal has been learned about the placebo ef- fect in the course of studying other therapies, and this in turn has stimulated study of the placebo effect directly. Investigators, as they have come to learn more about psychosomatic medicine and about psychological and social determinants of disease, have made some at- tempts to develop a comprehensive psychophysiological theory capable of explaining the placebo effect. These developments in medicine give rise to three major philosoph- ical issues. First, physicians have not devoted much attention to de- fining 'placebo effect' rigorously, and to delineating sharply the sorts of phenomena to which it is intended to apply. This task is a diffi- cult one because much remains to be learned about the placebo effect. 1 2 0n the one hand. one wants a definition specific enough to serve as a helpful guide in laying out research strategies. On the other hand, one wants a definition general enough so that it does not treat mat- ters which ought to be settled empirically as conceptual issues to be settled a priori. Several questions in philosophy of science, inclu- ding what is to count as an adequate explanation and how one estab- lishes the truth of scientific theories, must be taken into account in seeking this balance. Second, what is known about the placebo effect suggests that, in some way, a patient's beliefs or expectations can influence his bodily states. This appears to have implications for the relationship between mind and body; some philosophical views of the mind-body relation may allow for such a connection while others may not. Third, the actual use of placebos by physicians in therapeutic en- counters raises ethical issues-- specifically, issues within that area of normative ethics which is becoming known as the subspecialty of "medical ethics.“ Each of these three sets of issues can be dealt with in straight- forward fashion by the philosophical approaches already mentioned. We might arrive at a formal definition of 'placebo effect,‘ for example, by looking at how the words are actually used by investigators in the field, or by looking at the operational methods employed to measure it. We might adapt an existing mind-body view, such as Cartesian causal in- teractionism, to give an account of the placebo effect in mind-body terms. And we can handle the ethical issues within either the utili- tarian or the deontological framework, as a subcategory of cases in- volving deception. In this way all the issues might be settled, each in isolation from the others. However, it is both more challenging and more satisfying to aim for a more comprehensive approach. We would like to arrive at a for- mal definition of "placebo effect' which not only takes care of nar- rowly empirical issues, but also illuminates the tasks of investiga- ting the mind-body relation and of framing an ethical argument. And we would like to find a mind-body theory that not only accounts for the existence of the placebo effect, but also helps us to understand it further in terms of its definition and its ethical import. That is, the philosophical accounts given in response to the three sets of is- sues ought to be not only plausible when taken singly, but also mutu- ally consistent and illuminating when taken together. But we also demand more. In each of these three issue areas we already have what we may call considered judgments, many of which have little or nothing to do with the placebo effect. In the empirical realm we know, for instance, that a sugar pill has no medicinal proper- ties according to existing pharmacologic theories, insofar as its chemical structure is concerned. In the area of mind and body, we know that we have minds-- that we are conscious of ourselves and our surroundings in a way that trees and rocks are not, and in a way that animals may share to some degree. And in the normative realm we know that it is wrong to torture others for our own amusement. These are the sorts of things we are certain of if we can claim to be certain of anything at all; we are confident that we are not led to believe these things simply because we are confused, or because we are biased by sel- fish interests (Rawls 1971, pp. 47-48). Therefore, we would be reluc- tant to adapt a mind-body theory or an ethical stance which conflicted 4 with these considered judgments, just because this seemed a neat, ad hoc way out of a specific dilemma raised by the placebo case. Thus the degree of overall "fit" we ideally aim for among our three philo- sophical theories and our three types of considered judgments is quite extensive, and the chances are that we will never get a perfect "fit" but only a workable approximation. And even this approximation may be upset by new discoveries, or by new types of cases which cause us to rethink our ethical positions.1 We can summarize this approach schematically. Let C and C' be two alternative conceptual theories for organizing the empirical data about the placebo effect. Let M and M' be two alternative metaphysi- cal theories about the mind-body relation, and let N and N' be two al- ternative normative theories (of moral obligation). Also, let c1, c2, ...cn, m], m2,...mn, and n1, n2,...nn represent our considered judg- ments about conceptual-empirical, metaphysical mind-body, and normative matters, respectively, all of which are independent of our beliefs about the placebo effect. Now, if all of the following are true: 1. C and C', M and M', and N and N' are each equally plausible insofar as they are able to explain the relevant features of the placebo effect, taken in isolation; 1. The model for philosophical inquiry that I am employing is an expansion of Rawls' "reflective equilibrium" for determining principles of justice most compatible with our basic moral judgments (1971, pp. 18-22). Rawls in turn cites Nelson Goodman on the justification of principles of scientific inference, suggesting that some related con- cept of "best overall fit" may be applicable within philosophy of science as well. I take Lakatos' (1970) description of "research programmes" in science to involve a similar equilibrium model; sci- entists have characterized this model as a cybernetic or a negative- feedback one, as contrasted to the hypothetico-deductive model (Med- awar 1967, p. 154). 5 2. Taken together, (C,M,N) is internally more consistent than (C',M,N), (C,M',N), (C,M,N'), (C',M',N), or any other combi- nation; 3. C is more consistent with C], c2....cn than is C', M is more consistent with m], m2,...mn than is M', and N is more consis- tent with n], n2,...nn than is N'; then we would have the strongest possible grounds for preferring C, M, and N over C', M', and N', respectively. Furthermore, even if C were slightly less plausible than C' when applied to the placebo effect in isolation, we might be willing to trade this off against the much greater degree of overall "fit" offered by the set of theories (C,M,N), and the agreement with existing considered judgments on other matters of philosophical importance.2 The approach just described arises from a particular preconception of philosophy in general, which also turns out to have important appli- cations to the new subdiscipline of philosophy which is becoming known as philosophy of medicine.3 On this view, we engage in philosophy in order to find a more intelligible and coherent view of the world, in- cluding our own places in the world. Technical precision in philosophy is desirable and advantageous for this and other reasons, but by itself 2. The "equilibrium" must involve changes in both directions, since if our existing considered judgments were never altered to fit at- tractive general principles, the system would offer little opportunity for growth. We are rather looking for general principles which, if they do not match our most basic considered judgments, "extend them in an acceptable way" (Rawls 1971, p. 19). 3. 0n the scope of this new subdiscipline see (Pellegrino 1976a). An illuminating debate on the possibility and the nature of a philos- ophy of medicine is found in (Engelhardt and Spicker 1975, pp. 211- 234 . 6 precision does not satisfy this basic need for comprehensive understan- ding which leads us to philosophy initially. And so, where a conflict arises, we may be willing to dispense with a certain amount of techni- cal precision in order to satisfy this need. Furthermore, since know— ing our own places in the world requires us to see ourselves as agents rather than always as spectators, the moral as well as the metaphysi- cal calls for a consistent place within the world view we are con- structing. We are not content to know about other people and situa- tions in a merely descriptive way; we want to know how we ought to act towards others and in certain situations. In particular, we want to avoid any metaphysical theory such as a crude determinism, however em- pirically intriguing such a theory might be, if it seems to leave no room for free actions and moral thinking.4 This preconception of philosophy seems especially applicable to philosophy of medicine. An increased emphasis on technical precision has taken medicine a long way, but precision has been shown to have its limits, and what reflective physicians have always referred to as the "art" of medicine continues to defy precise analysis. Furthermore, medicine necessarily crosses all of the boundaries that we have tried to draw between the empirical, the metaphysical, and the moral realms. Medicine, above all other fields of study, refuses to let any of us remain spectators for long. The physician cannot merely observe and describe the course of disease; he must intervene actively, in a way 4. This preconception of philosophy is neither new nor original. On the importance of a subjective sense of satisfaction, and of taking moral as well as conceptual elements into account when seeking "fit," see (James 1927, pp. 146-148). The importance of including both meta- physical and moral considerations will be stressed in the discussion of eliminative materialism in Chapter 4. 7 that has dramatic impact on the rights and interests of other people. And the physician himself is liable at any moment to switch roles and become the patient. When medicine is viewed in this way, the variety of issues raised by the placebo effect is seen to represent in a micro- cosm the larger sweep of philosophy of medicine. For this reason, our investigation of the placebo effect will touch upon a number of points of importance to philosophy of medicine generally. The view of philosophy stated above requires some additional com- ment. The notions of "overall fit" and "equal plausibility," for in- stance, require considerable amplification if the account is to be de- fended against possible criticisms. I certainly do not intend the no- tion of "fit" to be so strong as to suggest mutual logical entailment among C, M, and N. However, I think that the notions of fit and plau- sibility, as they will be employed below, can be understood intuitively; this at least will allow the account to serve as a rough guide for the investigation. The best way to amplify this intuitive level of under- standing is actually to carry out the investigation, and then to enumer- ate the points of carry-over and cross-fertilization that have arisen. The account can serve as a rough guide to what follows even though all of the steps in the schematic will not be carried out explicitly. Even with these problems aside, anyone adhering to the present Anglo-American analytic tradition is likely to look suspiciously on any search for a comprehensive philosophical overview in which inter- nal "fit" is stressed. And, indeed, one could cite examples of imag- inative and internally consistent philosophical accounts which are no more than meaningless exercises in fantasy. But the serious flaw in such accounts is not the attempt at comprehensiveness, or the value 8 placed on internal fit, but rather the failure to be grounded at any point on an acceptable base, empirical or otherwise. So long as our attempt takes into account our empirical, metaphysical, and moral con- sidered judgments, there seems little reason to fear that any perni- cious castle-in-the-air-bui1ding will occur. Some, of course, would insist that the empirical corner of the (C,M,N) triangle ought to be given priority over considered judgments of other types; and they should be happy to note that the emphasis on internal fit can be ex- pected to augment the empirical content of the other two corners. The body of this dissertation will deal sequentially with the three sets of issues arising from the placebo effect. Chapter 1 will review the empirical findings about the placebo effect from the med- ical literature. It will list, as a point of departure, various defi- nitions of 'placebo' offered by medical authors. Theories explaining placebo phenomena will also be reviewed. While in no way exhaustively reviewing the available literature, Chapter 1 will summarize the major empirical points which any philosophical investigation will have to take into account. Chapter 2 will consider these data from a critical standpoint. It will relate the placebo phenomenon more explicitly to the currently dominant theories and assumptions of medical science. Following con- sideration of a series of examples designed to mark out the boundaries of applicability of the term 'placebo effect,‘ a formal definition will be proposed. Certain features of this definition, in turn, will serve as a starting point for the other two sets of issues, the mind- body implications and the ethical questions. 9 The next three chapters will be devoted to the mind-body issue. Chapter 3 will give an overview of the Cartesian tradition in philos- ophy of mind and will consider three theories arising within this framework, behaviorism, causal interactionism, and identity theory. Chapter 4 will go on to deal at some length with eliminative materi- alism, which is in some ways the strongest mind-body theory; but while all of these theories are able to give some account of the placebo ef- fect, all turn out to give rise to other philosophical problems when tested against other considered judgments. At the conclusion of Chap- ter 4, an alternative approach will be suggested, which involves fo- cusing on the concept of "person" instead of the concepts of mind and body. Chapter 5 will then develop such a theory of the person, which attempts to integrate man's biological nature, man's mental states and activities, and man's participation in language and culture.5 This view, it will be argued, seems more compatible with what we know about the placebo effect than do mind-body theories lacking this scope. In addition, the person theory will be shown to fit in well with other considered judgments about the nature of mind. Chapter 6 will then address the ethical issues, beginning with a historical review of physicians' use of placebos and the arguments tra- ditionally offered for and against such use. These arguments will be critically reviewed to establish a prima facie presumption against de- ceptive use of placebos. It follows from the formal definition given in Chapter 2 that the placebo effect can occur without administration 5. Throughout this dissertation I will be using the undesirable masculine noun and pronoun forms for purposes of brevity, and also in order to reserve the term 'person' to designate the particular philo— sophical stance described in Chapters 4-5. 10 of a placebo; this will suggest a line of argument which places in- creased stress on alternatives to the deceptive use of placebos. The Conclusion, finally, will briefly summarize the results of these inquiries in order to show how the "reflective equilibrium" strategy described above has guided the investigation. Some further comments on medical ethics will illustrate the internal consistency of the philosophical framework that has been erected. Finally, some lines of empirical research which are suggested by these investigations will be discussed. Chapter 1. The Placebo Effect: A Review of the Medical Literature Physicians have only recently approached the placebo effect as a subject for formal investigation and speculation. Pepper, in a 1945 paper sometimes considered a classic in this field, admitted that he was unable to find any articles on placebos listed in two major medi- cal bibliographic indices. Shapiro, a prolific reviewer, states that the recent interest in placebos dates from 1953 and was stimulated by the desire to design adequate double-blind therapeutic trials (1968).1 Probably the bulk of the medical literature on placebos treats the placebo effect as a nuisance variable, worthy of notice only for the havoc it can wreak upon inadequately designed experiments. Thus one reads, for instance, that psychotherapy will become a more potent tool when it is isolated from the concomitant placebo effect, just as fox- glove became a more useful medicine when the active ingredient, digi- talis, was extracted (Shapiro 1964). But other writers emphasize the positive therapeutic potential of the placebo effect, and the insight it may offer into psychosomatic disease and healing. This review of the medical literature will consider in turn the history of the term 'placebo' and definitions offered for it; the nature and scope of placebo responses; agents that can act as placebos; factors 1. In a double-blind trial, neither investigator nor experimental subject knows whether the subject is in the control or the experimental group. Thus if the experimental group is to receive a drug or other treatment, the control group must get a dummy treatment that outwardly resembles the experimental one but which lacks the ingredient under study. 11 12 influencing the placebo effect; and explanatory hypotheses that have been proposed. This will provide material for a more critical analysis in the next chapter. 1.1. History and Definition The word 'placebo' entered the English language in the 14th cen- tury as the name for the vespers sung for the dead (Shapiro 1968). The word was derived from the Latin version of Psalm 116:9: "Placebo Domino in regione vivorum..." (Pepper 1945), usually translated, "I shall walk before the Lord in the land of the living,” although the literal translation of placebo is "I shall please." From this original meaning, the word acquired both its medical application and its nega- tive connotation. In the former instance, doing something purely sym- bolic for patient and relatives when nothing curative can be done, something both soothing and inexpensive, could be compared to singing a hymn (Osmond 1974). In the latter instance, 'placebo' came to be used in Chaucer's time to mean a sycophant or servile flatterer, de- rived from the practice of singing vespers on behalf of strangers for pay (Pepper 1945). Mbtherby's New Medical Dictionary (1785) defined 'placebo' in neutral and uninformative terms as "a commonplace method or medicine" (Shapiro 1968). By contrast, Hooper's Medical Dictionary (1811) de- rided 'placebo' as "an epithet given to any medicine adopted to please rather than to benefit the patient" (Pepper 1945), as if the two were mutually exclusive goals. Contemporary definitions offered by investigators in the field tend to avoid judgmental terms, but still show significant differences 13 of opinion. Pepper represents the restrictive end of the spectrum by defining 'placebo' as an agent which is totally inert (1945). Pre- sumably he means inert in a pharmacologic sense only; if an agent pro- duced no effect whatsoever one would hardly want to label it a placebo. Wolf clarifies this point by defining 'placebo effect' as "any effect attributable to a pill, potion, or procedure, but not to its pharma- codynamic or specific properties“ (1959). And a very broad and inclu- sive definition is suggested by Modell's comment that the placebo reac- tion is "the only single action which all drugs have in common" (1955, p. 55). Probably the most detailed definition is Shapiro's: A placebo is defined as any therapy (or that component of any therapy) that is deliberately used for its nonspecific psychologic or psychophysiologic effect, or that is used for its presumed ef- fect on a patient, symptom, or illness, but which, unknown to pa- tient and therapist, is without specific activity for the condi- tion being treated. A placebo, when used as a control in experimental studies, is defined as a substance or procedure that is without specific ac- tivity for the condition being evaluated. The placebo effect is defined as the nonspecific psychologic or psychophysiologic effect produced by placebos (Shapiro 1968, p. 599). Shapiro notes that by his definition, a placebo 1) may be pharmacologi- cally inert or active; 2) may or may not produce the placebo effect in any given instance; 3) may produce an effect which is either positive or negative (i.e., placebo side effects). The reference to "presumed effect" allows for the notion of unwitting placebo use by physicians, and is the basis for the often quoted statement, "The history of medi- cal treatment can be characterized largely as the history of the pla- cebo effect" (Shapiro 1968, p. 597). The differences among these definitions, as well as specific 14 problems created by each definition, form the major issues requiring analysis in the next chapter. 1.2. Nature and Scope of Response Shapiro succinctly summarizes the importance and the breadth of the placebo effect: Many papers have demonstrated the importance and magnitude of the placebo effect in every therapeutic area. Placebos can be more powerful than, and reverse the action of, potent active drugs. The incidence of placebo reactions approaches 100 per cent in some studies. Placebos can have profound effects on or- ganic illnesses, including incurable malignancies. Placebos can often mimic the effects of active drugs. Uncontrolled studies of drug efficacy are reported effective four to five times more frequently than controlled studies. Placebo effects are so omni- present that if they are not reported in controlled studies it is commonly accepted that the studies are unreliable. Increased ap- preciation of placebo effects is reflected in the speculation that the major medical achievement of the last decade will be recorded by future medical historians as the development of methodology and controlled experiments (Shapiro 1968, p. 598). The symptom most often thought of in association with placebos is pain; but placebos modify both subjectively reported and objectively observable symptoms. One reviewer gives the following list of condi- tions in which placebos have been shown to produce relief: cough, mood changes, angina pectoris, headache, seasickness, anxiety, hypertension, status asthmaticus, depression, and the common cold (Bourne 1971). Placebos can lower blood sugar levels in diabetics (Singer and Hurwitz 1967) and can shrink tumors in patients with malignant lymphosarcoma (Klopfer 1957). When a subjective symptom and its physiological 2. The use of the word "mimic" might unintentionally suggest that the placebo effect is somehow less real than the pharmacologic effect of drugs. The problem of using neutral language in describing the pla- cebo effect, so as not unwittingly to beg the interesting questions, must be kept in mind. 15 concomitant (e.g., nausea and disturbed gastric motility) can be ob- served simultaneously, placebos can be shown to affect both (Wolf 1950).3 Placebos can also produce toxic side effects like those of "ac- tive" drugs. One typical study reported the following side effects among 25 patients experiencing negative placebo reactions (Honzak, Horackova and Culik 1972): Somnolence (10 cases) Palpitations (9 cases) Irritability and insomnia (8 cases) Weakness, with drop in blood pressure of more than 20 mm mer- cury (5 cases) Temporal headache (4 cases) Diarrhea (3 cases) Collapse (2 cases) Itching (2 cases) In addition, three of these patients developed dependence to the placebo and demonstrated withdrawal symptoms when the pill was stopped. In another study, one patient repeatedly responded to placebo administra- tion by developing a florid rash, diagnosed as classic drug-induced dermatitis by a consulting dermatologist; it ceased immediately upon discontinuance of the placebo, which in this case was plain lactose (Wolf and Pinsky 1954). Placebo reactions may resemble those of "active" drugs not only in the end results but also in the patterns of activity. These pat- terns include a peak effect a certain number of hours after administra- tion of the drug; a cumulative effect of increasing symptom relief as 3. These findings should put a lie to the myth, still prevalent among physicians, that if a patient responds to placebo his symptoms must be either imaginary or feigned, and that a placebo can be used in the differential diagnosis of psychic symptoms from "organic" ones. 16 the drug is continued over time, with a carry-over effect after the drug is stopped; and a decrease in efficacy as the severity of the symptom increases. These "pharmacologic" patterns occur with placebos as well as with "active" agents (Lasagna, Laties and Dohan 1958). Some investigators have reported that placebo effects are more tran- sitory than "real" drug effects (Lasagna et al. 1954), but there is enough contrary evidence to question this (Rosenthal and Frank 1956). 1.3. Agents Acting as Placebos Essentially any treatment modality can act as a placebo, and pa- tient reactivity will vary according to the supposed potency of the treatment one thinks one is getting. A placebo capsule, in general, is more powerful than a placebo pill; an injection works better than either; and an injection that stings is better than a painless one (Evans 1974).} In one study a white or yellow capsule produced the maximal therapeutic effect, while side effects occurred most frequent- ly with a reddish-gray capsule (Honzak, Horackova and Culik 1972). Surgery is an especially powerful placebo stimulus (Beecher 1961). It is not always easy to distinguish a placebo stimulus from ac- tive therapy. A recent study tried to compare true acupuncture ther- apy with a sham acupuncture procedure for chronic shoulder pain. The "placebo" treatment consisted of pricking the skin with acupuncture needles without actually inserting them, and then tapping them on the skin (Moore and Berk 1976). But since cutaneous stimulation of any type may promote pain relief (Melzack and Wall 1965), the sham proce- dure cannot be considered to be physiologically inactive. An especially intriguing study was Park and Covi's (1965) l7 "non-blind" placebo trial. These researchers gave sugar capsules to 15 outpatients with neurotic complaints, telling them that the pills were sugar and contained no medicine; that such pills had helped other patients in the past; and that the doctors were convinced that the patient would get relief also. Fourteen patients completed a week's trial of therapy, and all but one showed improvement of symptoms by a standard symptom inventory (the remaining patient's husband had made a suicidal gesture during that week). The patients could be divided into three groups: those certain that the capsule was a placebo; those certain that the capsule was an active drug; and those not certain. The two groups feeling certain either way showed the most improvement. 0f the first group, half at- tributed their improvement to the placebo, and half to their own abil- ities to cope-- one stated that the pill served as a constant reminder that she could do something to improve her own condition. Also, some of these patients were glad to be avoiding the addiction and overdose potentials of active medication. Among those sure that the placebo was really an active drug, most reasoned that this must be the case since they had improved. They either ignored the sugar-pill explanation, or dismissed it as a therapeutic gimmick of the physician to encourage patient self-sufficiency. Half of those certain that the pill was an active drug reported side effects, while none of those believing that it was a sugar pill did so. A major flaw in the Park and Covi study is that the investigators were initially unwilling to offer placebo as the only treatment modal- ity to these new patients seeking help. Therefore they explained that the placebo would be given only for one week and that subsequently other 18 treatment could be considered. This may have had the effect of putting the patients "on probation" for the week and thereby creating a great desire to please the doctors-- possibly accounting for the nearly 100 per cent placebo response as compared to the more usual 30 to 50 per cent response (Beecher 1955). It has also been shown that being on a waiting list to be seen at a psychiatric facility exerts a placebo ef- fect of its own and hastens recovery (Sloane et al. 1975).4 As more is written about the placebo effect, however, patients are more likely to conclude that they may have received placebos, and this may not necessarily hamper their response to therapy (Cousins 1976). 1.4. Factors Influencing the Placebo Effect As soon as the importance of the placebo effect began to be under- stood, investigators began to search for personality factors that would identify the "placebo reactor," in the hopes that eliminating such subjects from controlled studies would produce clearer data. An early study claimed that the placebo reactor displayed the following characteristics: more outgoing; more anxious; less emotionally mature; more concerned about visceral complaints such as constipation; and more satisfied overall with their hospital experience. But the non- reactors in this study responded less well to analgesics as well as to placebos for pain relief, raising the question of whether the reactors simply had less severe pain (Lasagna et al. 1954). A later study, done in a laboratory and thus perhaps not comparable to Lasagna's hos- pital data, held that reactors by psychological testing were more 4. For more on this waiting-list study see the Conclusion, es- pecially Footnote 3. 19 enthusiastic, outgoing, and verbal, and better adjusted, than nonreac- tors (Muller 1965); this picture seems to conflict with Lasagna's at several points. Yet another study found no difference between reactors and nonreactors when the same psychological test instrument was used, but a separate personality inventory showed reactors to be more neu- rotic and extroverted (Gartner 1961). In sum, there are so many in- consistencies among these and many other studies that one may reason- ably conclude that there is no single personality type characterizing placebo reactors (Kurland 1960; Shapiro 1968). There may, however, be some evidence to suggest that patients who develop worsening of symp- toms on placebo may be distinguishable on some personality measures either from positive reactors or from nonreactors (Shapiro et al. 1973). In more cases than not, an individual who responds to placebo un- der one set of circumstances will fail to respond under other circum- stances, even in the course of the same study; these inconsistent reac- tors generally outnumber consistent reactors and consistent nonreactors combined (Lasagna et al. 1954; Beecher 1955). The only study to show a nearly 100 per cent constancy of reaction or nonreaction, a 1946 study of headache, has not been replicated (Jellinek 1946). A large number of other patient variables have shown either no cor- relation or contradictory correlation with the placebo response. These include age, sex, intelligence, findings on Rorschach and other psy- chological tests, and presence of neurosis or psychosis (Shapiro 1968). One finding in patients that has been rather consistently corre- lated with placebo reactivity is stress or anxiety. Even here there are questions, however. Beecher (1955) claimed that patients with more severe pain were more likely to get relief from placebos, and suggested 20 that the psychic stress accompanying pain contributed to the placebo effect; but as already noted, Lasagna and co-workers found the opposite correlation between pain severity and placebo response (Lasagna et al. 1954; Lasagna, Laties and Dohan 1958). In treatment of anxiety neu- rosis, Rickels and Downing (1967) found that patients with less pre- treatment anxiety responded better to placebos. Other patient factors correlated with placebo reactivity are harder to measure, and include positive expectations, faith in the physician, motivation, and the need for emotional catharsis or for psychological defense mechanisms (such as the ritual of taking medi- cine as a means for reducing anxiety). A study of patients with para- noid symptoms found that those who exhibited readiness to enter into personal relationships with the therapists were good placebo reactors, while those holding back from such relationships were not (Freedman et al. I967). Expectations are commonly cited as an important factor in produ- cing the placebo response. In a study of placebo to improve short- term memory in elderly patients, patients' expectations were highly correlated with subjective improvement, and were correlated somewhat less well with objective improvement (Nash and Zimring 1969). One de- tailed attempt to study the role of expectations occurred in a study of how biofeedback could increase the frequency of alpha rhythm ("re- 1axation") on the subjects' encephalograms (Stoebel and Glueck 1973). The investigators designed an index to measure the combination of actual learning of alpha control and placebo effect. Using this index, they showed that patients did best in the long run when expectations and active learning were kept in relative balance. For instance, 21 subjects with very high initial expectations tended to be discouraged by the actual results first obtained, and thus performed less well on subsequent training. However, this index is rather speculative in na- ture, and perhaps ought to be viewed as a predictor of long-term out- come rather than as a measure of placebo effect.5 Also, these inves- tigators seem not to have distinguished carefully enough between ex- pectation and motivation; each may well contribute to a positive out- come, but through different mechanisms (Rosenthal and Frank 1956). However, if accepted, the results of this study would suggest that either too high or too low expectations could hamper placebo response. Clearly, factors such as expectations and motivation are not pa- tient variables strictly speaking, but could be expected to depend at least in part on the physician, the situation in which the placebo is administered, and other external factors: [Expectations] vary widely among patients, depending on such fac- tors as the patient's previous experiences with physicians and medications, his personal knowledge of the physician, the repu- tation of his physician in the community, the community belief in the recent achievements of medical science, various relevant prop- erties of the institution or the setting in which the physician operates, and the physician's personality and behavior and his own expectancies as to what he can accomplish (Whitehorn 1958, p. 662). While investigators for the most part have been reluctant to switch their attention from the placebo and the patient to the entire placebo context, the accumulating data have forced this change in focus (Wolf 5. It would have been interesting to employ as part of this study a true placebo or "dummy" biofeedback (i.e., the biofeedback signal being given to the subject randomly, instead of only when the subject was exhibiting alpha rhythm); unfortunately this was not done. This omission casts even more doubt on whether the index formulated by the investigators was measuring a "placebo" response. 22 1959, Shapiro 1968).5 In the absence of good objective data on the contribution of the physician to the placebo effect, a good deal is based on generaliza- tions from other types of studies, such as studies of experimenter bias in research and of the influence of therapists' behaviors and at- titudes on outcome in psychotherapy (Shapiro 1968). A classic study of experimenter bias had the experimenters being told that their rats had been especially bred either for brightness or for dullness, al- though all rats were in fact from the same genetic strain. The experi- menters then performed learning experiments on their rats and obtained the data that conformed to whatever their expectations of the rats' behavior had been (Rosenthal 1963). If scientists can somehow com- municate their own expectations and attitudes to rats, it seems reason- able to assume that physicians can unknowingly communicate expectations and attitudes to patients, altering the patients' therapeutic outcomes as a result. One study that did document the physician's attitude as a factor compared relief of anxiety by two sedatives and by placebo. When the drugs were administered by one doctor, who anticipated that there would be no difference among the two active drugs and who was viewed by pa- tients as more neutral and matter-of-fact in manner, there were no differences in relief among the three agents. When administered by a 6. This reluctance may stem from the desire to try simpler hypoth- eses with more readily measurable variables, and also from the trend in psychosomatic research in the 1940's and 1950's to define "personality types" associated with specific diseases. One might speculate that had watchmakers conducted the first experiments on hypnosis, they would have tried to correlate the trance state with the type of watch being swung before the subjects' eyes, and would have been chagrined when such a correlation failed to appear. 23 second doctor, who anticipated greater efficacy of the active drugs and whom the patients viewed as more optimistic and supportive, the two sedatives showed superiority to placebo. In addition, patients showed greater overall relief of symptoms when treated by the second doctor. One way of explaining these results would be that the two physicians had in fact found a way to guess correctly which pills were sedatives and which were placebo, so that the double-blind ex- perimental design had broken down; but this was checked for and found not to be the case (Uhlenhuth et al. 1959). Even less is known about the role played by other factors in the healing environment. We have already mentioned differences produced by changing the color or the route of administration of the placebo (1.3). 1.5. Theories of Placebo Mechanisms All medical authorities speculating on how placebos might exert their influence agree on one point-- that a placebo "cannot possibly act" through a pharmacologic or physiologic route (Beecher 1955). Im- plicit or explicit in most investigators' definitions is that if a substance now held to be a placebo, such as lactose, turns out to have a biochemical effect, this datum will prompt the reclassification of the substance as an active drug and will not be accepted as empirical evidence to explain the placebo effect. Also implicit in most views is the assumption that as we learn more about the specific physiologic and psychologic mechanisms of drugs and other treatments, the realm of effects now attributed to placebos will shrink (Shapiro 1964)-- that is, that to call something a "placebo effect" is now as much an admission 24 of ignorance as a potential explanation. Byerly (1976)7 has classified possible placebo theories as men- talistic, conditioning, or mixed. Mentalistic theories presumably are those that make reference to the subjective states of awareness of the patient, while conditioning theories are types of behavioristic accounts, which make reference only to outwardly observable behavior. The most commonly encountered mentalistic theories are those re- ferring to patient expectation (e.g., Rosenthal and Frank 1956; Nash and Zimring 1969); such theories are also sometimes referred to as self-fulfilling prophecy (Beecher 1955) or response-bias theory (Mor- ris and O'Neal 1974). By all these theories the patient's expectation of symptom change is held to be causally connected to the change that occurs. Since the central nervous system, the autonomic nervous sys- tem, and the endocrine system all exhibit predictable changes in re- sponse to the person's emotional state, these are frequently suggested as the intervening psychophysiologic mechanisms (Wolf 1959). Theories that are almost purely mentalistic hold that the placebo effect works solely through alteration of the patient's subjective reac- tion to illness. In pain relief, the placebo is said to act strictly by relieving anxiety, which in turn produces relief of pain (Evans 1974); or pain itself is said to consist of a sensory component and a subjective-processing component, with the placebo affecting the lat- ter and not the former (Beecher 1955).8 But these theories ignore 7. To my knowledge this paper by Byerly is the only attempt to date to approach the placebo effect from the philosophical standpoint. 8. Beecher's two-stage pain theory derives from studies of narcot- ics in treating war wounds; soldiers given morphine claimed still to feel pain but no longer to be bothered by it. Beecher concluded that 25 both the ability of placebos to relieve many other symptoms besides pain, and the documented impact of placebos on objectively observable bodily function; such data seem to render any purely mentalistic the- ory untenable and to require some sort of psychophysiologic view. Another form of mentalistic theory, arising from the Freudian tradition, is transference theory (Forrer 1964). Transference is de- fined in psychoanalysis as the unconscious projection of feelings, at- titudes, and wishes properly displayed toward a significant figure in early development (usually the parent) onto another person in the in- dividual's current life (the doctor or therapist) (Freedman, Kaplan and Sadock 1972, p. 798). A satisfactory doctor-patient relationship invites the patient unconsciously to trust in the doctor, to submit to his wishes, and to expect him to "make it better" in a way similar to the parent-child relationship (Shapiro 1968). Transference may be seen as an adjunct to expectation theory, if positive transference en- courages optimistic expectations; or it may be seen as an independent mechanism, acting through an unconscious release of psychic tensions, for example. Some theorists have attempted to reduce placebo responsiveness to suggestibility, which may be defined as a state of compliant respon- siveness to ideas or influences (Freedman, Kaplan and Sadock 1972, p. 795); susceptibility to hypnosis is a commonly cited example of sugges- tibility. However, Shapiro notes that patients experiencing hysterical conversion reactions (psychic symptoms), who are commonly considered pain conSiSts of two phases, the sensation and the emotional reaction to it. I cannot fully evaluate the tenability of this view here; but cf: comments on "Pain and Suffering" by Jerome Shaffer in (Spicker and Engelhardt 1976. pp. 221-233). 26 to be extremely suggestible in the way most psychologists use the term, are very poor placebo reactors; he feels that this casts doubt on the suggestibility theory (Shapiro 1968). But other investigators question whether the case of the conversion hysteric is a true case of sugges- tibility as defined above (Kurland 1960). In one study, the more sug- gestible patients showed significantly greater placebo reactivity in the first week of therapy, but later showed less reactivity than other patients. The authors proposed that the more suggestible patients might "overrespond" to placebo at first, producing an apparent, rela- tive worsening of symptoms later on; at any rate, suggestibility alone could not account for the extent and duration of the placebo effect (Steinbrook, Jones and Ainslie 1965). In contrast to these mentalistic theories, conditioning theory takes a stimulus-response form which makes no reference to the inter- nal, mental states of the individual. Past instances of active ther- apy in medical settings are seen as the stimulus, while relief of symp- toms is the original, unconditioned response. As conditioning occurs, the medical setting itself becomes a sufficient stimulus and the thera- peutic response becomes conditioned, so that it occurs even without ac- tive treatment (just as, after repeated presentation of food together with a bell sound to dogs, the bell alone will produce salivation) (Gliedman, Gantt and Teitelbaum 1957). Difficulty in choosing among competing theories is illustrated by Bourne (1971), who argues that transference and conditioning suffice equally well to explain some of the commonly observed placebo phenomena: Finding: Placebo response is maximized by anxiety. Transference account: Anxiety produces a "set" for transference, by encouraging regressive behaviors harking back to an earlier 27 stage of psychic development. Conditioning account: Stress causes the organism to fall back on conditioned responses instead of trying new adaptive beha- viors. Finding: Placebos often work best in diseases characterized by quiescent periods broken by periodic flare-ups. Transference account: Recognition of experiences undergone in the past, such as a disease flare-up occuring as part of a recog— nized pattern, increases transference potential. Conditioning account: Repetition of the stimulus is essential for conditioning to occur. Finding: Placebos work best on symptoms under central nervous sys- tem, autonomic, or hormonal control. Transference account: Such symptoms are most susceptible to changes resulting from increase or decrease in psychic tension. Conditioning account: Such bodily changes are most accessible to conditioning, being physiologically most closely connected with sensory inputs. An additional mechanism that has been proposed is attribution theory, which is not a conditioning theory but which does not seem to be clearly mentalistic either. It holds that placebo reactors are sim- ply highly sensitive to subtle changes in their internal states. If a symptom lessens very slightly in severity following placebo administra- tion, the individual will detect this and will attribute the change to the placebo. One study of placebo response attempted to measure this internal sensitivity or "openness"; they found it not to be correlated with placebo reactivity, although patients' expectations were (Nash and Zimring 1969). Another study set out directly to test alternative theories by giving placebos labeled with either familiar or unfamiliar drug names. By conditioning theory, they reasoned, familiarity would enhance the placebo effect by providing a stronger conditioned stimulus. By attri- bution theory, unfamiliarity might be expected to enhance the placebo effect, as the patient would be familiar with and sensitive to the 28 pharmacologic effects of drugs that he had previously taken. By expec- tation theory, the physician's suggestion and attitude should be the controlling variable, with familiarity playing only a minor role. These investigators found no correlation of placebo response with familiarity or unfamiliarity, concluding that their results were most consistent with expectation theory (Morris and O'Neal 1974). Finally, Byerly (1976) suggests the possibility of other theories which avoid the rigid distinction between mental and bodily phenomena; as an example he cites a view of the "symbolic reality" of medicine which treats disease as inherently a cultural construct (Kleinman 1973). Earlier placebo writers mention the symbolic aspects without making clear whether they are construing symbolic import in strictly mentalistic terms, or whether they hold, with Kleinman, that symbolic significance influences bodily health and disease: [T]he physician is a vastly more important institution than the drug store. The reasons for this are deeply rooted in the main- springs of human behavior, for man in distress wants action-— rational action if possible, of course, but irrational action, if necessary, rather than none at all.... [T]he pill the patient swallows, no matter what its nature, acquires potency as a symbol of faith, wisdom, and support (Findley 1953, pp. 1822-1823). The physician's ability to relieve the emotional, reactive as- pects of a patient's illness through symbolic operations is there- fore an important aspect of his healing function.... Hence the prescription, pill or injection symbolizes the physician's healing function. The prescribing of a pharmacologically inert substance may thus, through its symbolic significance, produce favorable effects (Whitehorn 1958, p. 662). The clearest theoretical statement of a symbolic-cultural basis for the placebo effect is given by Adler and Hammett (1973) in what I shall be calling the "meaning model" of the placebo effect. Adler and Hammett identify two invariant features of healing practices in all cul- tures: l) a shared cognitive system which explains illness in terms 29 (whether of natural phenomena or of supernatural occurrences) readily understandable to those sharing the background of cultural beliefs ("system formation"); and 2) a relationship with a socially sanctioned healer occupying a role of parental power and influence, which in turn stimulates caring responses from family and community ("group forma- tion"): It is suggested here that these two factors-— group formation and system formation-- are as essential to psychic functioning as nourishment is to physical functioning, are the basic factors composing what is subjectively experienced as a feeling of "mean- ing," are invariably used in all successful interpersonal ther- apies, and are the necessary and sufficient components of the placebo effect (Adler and Hamnett 1973, p. 597). The data now available do not seem to be sufficient to exclude with certainty any of the theories that have been listed above.9 More research needs to be done, especially research like the Morris and O'Neal (1974) study cited above which sets out to compare different theories. Future research strategies suggested by some of these the- ories, especially research into the healing situation as a whole and into its symbolic and cultural aspects, will be mentioned in the Con- clusion. But to a great extent, interpretation of any future data will depend on getting clearer about precisely what is meant by 'pla- cebo effect,’ and about which phenomena are or are not applicable to its study; discussion of these matters will occupy most of Chapter 2. In addition, a study of the implications of some of the placebo the- ories for the nature of the mind-body relation might provide additional 9. It could, of course, be argued that there is not one mechanism but several responsible for the placebo effect. Beecher took the repro- ducibility of placebo response rates from study to study (an average of 35.2 per cent, with a standard deviation of 2.2 per cent, in 15 studies covering a variet of symptoms) as evidence supporting a single mechan- ism (Beecher 1955 . 30 grounds for accepting or dismissing some of the theories; this will be considered in Chapters 3, 4 and 5. In summary, while the word 'placebo' has been in the medical lexi- con for at least several centuries, contemporary definitions still disagree on the scope of effects attributable to it. It is widely agreed, however, that the placebo is very powerful and can accomplish in some instances the majority of effects of which "active" drugs and other therapies are capable. Placebos influence both subjective and objective symptoms and can produce toxic side effects. The proportion- ate incidence of placebo reactions is roughly predictable from study to study, but the search for specific personality traits that will pre- dictably pick out the "placebo reactor" has mostly failed. It appears that the same individual may or may not react to placebo depending on a complex set of internal and environmental factors, including the re- lationship with the physician and the nature of the healing situation as a whole. However, specific factors that do seem to be reliably re- lated to positive placebo effects include positive expectations and perhaps some degree of stress or anxiety in the patient. A variety of theories have been put forth to account for the placebo effect; these include mentalistic theories (expectation or response-bias, transference, and suggestion theories); conditioning theories; and "mixed" theories such as those calling attention to the "meaning" of the illness exper- ience for the patient. Chapter 2. A Definition of the Placebo Effect In 1.1 we reviewed several definitions of 'placebo' that had been suggested by medical authors. But these definitions lacked rigorous analysis, and were indeed mutually contradictory in some ways. Since they were formulated primarily to introduce and organize various par- ticular medical findings, they cannot be expected to bear much philo- sophical weight. In this chapter I shall try to formulate a defini- tion of 'placebo effect' that can serve as a basis for further philo- sophical investigation. The first section applies Thomas Kuhn's (1970) notion of a scien- tific paradigm to the placebo effect; this will allow us to ask later the extent to which a definition is dependent upon a particular explan- atory context. The next section gives a series of illustrative exam- ples to determine the range of phenomena that a definition must cover. The third section begins by offering formal definitions of 'therapy' and 'specific,‘ and then uses these to arrive at definitions of 'pla- cebo effect' and 'placebo.‘ 2.1. The Placebo Effect as Medical Anomaly Some concepts that have utility when applied to the placebo effect arise from Kuhn's (1970) reconstruction of the history of science. Scientists conducting research rely heavily, not only on the explicitly stated laws and theories of their science, but also on a set of assump- tions and explanatory presuppositions which remains implicit but which 31 32 uniquely characterizes the science that they are engaged in. These pre- suppositions create expectations about the world and suggest both what sorts of phenomena are most usefully studied, and how observations or experiments are best carried out. The presuppositions are thus very useful in guiding scientific research and in steering scientists away from troublesome areas not accessible to the scientific tools at hand. But invariably data will be collected which are at odds with this set of presuppositions, and which are unexpected according to the accepted set of laws and theories. Scientists will first attempt to account for these findings by making slight modifications in the existing the- ories; but over time more and more unexpected findings accumulate. At some point a few "revolutionary" scientists put forth totally new laws and theories, which are based on a different set of presuppositions. If this new set of theories both explains the previously unaccounted- for data, and embraces the accumulated knowledge of the old science while opening up new avenues for further research, scientists will come to adopt it, and a scientific revolution will have occurred. Kuhn terms the set of basic presuppositions and assumptions a paradigm,1 and calls the unexpected findings, that can lead to overturning para- digms, anomalies. An example from physics may illustrate how paradigms and anomalies 1. "Paradigm" may be used to refer either to the body of shared beliefs of a scientific community, or specifically to that community's puzzle-solving examples which have the most direct impact on research design (Kuhn 1970, pp. 174—191). I use the term here more in the for- mer sense, referring especially to the elements of heuristic models (e.g. of disease causation) and values (e.g. what counts as a "good" explanation). However Kuhn's latter sense of 'paradigm' cannot be com- pletely separated from this usage-- the place of Koch's postulates in contemporary medical science shows how values, heuristic models, and puzzle-solving examples are mutually bound up. 33 are related. The paradigm dominant in physics in 1895 led scientists to expect to find various sorts of rays, but rays that could cause a plate to glow across the room from a cathode ray tube were not among these. Thus, when Roentgen noticed such a glow, he was observing a phenomenon which had previously been created in many other laboratories, but which had not been observed because the theories and the presup- positions of physics did not tell anyone where to look for it. (By contrast, a totally expected finding might be the discovery of a new element whose properties had already been predicted by the periodic table.) Roentgen's announcement of his discovery, therefore, stirred immediate controversy. At the very least, accepting his data would re- quire that many accepted experiments be done over, since there was now this new variable that had to be controlled. The clash with existing assumptions was so strong that some eminent physicists, such as Lord Kelvin, refused to believe Roentgen's data. About the same time, how- ever, physics was accumulating other anomalies, including black-body radiation and the constancy of the speed of light; and so when the new paradigms of quantum mechanics and relativity appeared, which could account for these anomalies better than the old Newtonian paradigm, physics was ready to accept them (Kuhn 1970, pp. 57-61). In medicine, underlying paradigms include theories and assumptions about the nature of disease and therapy, and about laws and regularities in human pathophysiology. The present-day, Western medical paradigm emphasizes causal mechanisms affecting organs, tissues, cells, chem- ical factors, and physical phenomena. Theories relating psychological and sociological factors to disease and therapy are generally less well developed and held in lower esteem-- one might feel that they 34 will have to do until "real" explanations in physical-chemical terms become available through further research. Within such a paradigm, the fact that a chemically inert pill can change symptoms and "organic" bodily states constitutes a significant anomaly. As was the case with x-rays, accepting the placebo phenomenon entails rejecting a major body of previously accepted data, since until recently most of what was known about therapeutics came from uncontrolled trials. A dis- covery such as the placebo effect is likely to be greeted with con- sternation among medical scientists, unlike, for instance, the dis- covery of a new antibiotic to treat tuberculosis; the dominant para- digm leads the scientist to expect the latter but not the former. In the absence of an attractive alternative paradigm that can totally replace the existing medical paradigm, we see different at- tempts to deal with the placebo discovery. The serious physician to- day cannot deny the placebo data; but he can instead adopt an attitude of exclusion towards it-- he may be content merely to label the pla- cebo effect so that it can be readily recognized and therefore exclu- ded from research. The early attempts to determine a placebo-reactor personality type so that such subjects could be excluded from clini- cal trials (1.4) is an example of this approach; the scientist reasons that he might as well focus his attention on those phenomena which are most readily explainable by accepted theories, and put any anomalies he finds "on the shelf." The suggestion that the placebo effect is an impurity which ought to be removed from psychotherapeutic modalities (Shapiro 1964) is another example of exclusionary thinking. This sort of thinking may influence and may implicitly occur in the definition one adopts for 'placebo' and 'placebo effect.‘ 35 By contrast, an inclusive approach would seek new laws or causal factors, to expand or modify the existing paradigm so as to bring the placebo effect within it. The fact that the dominant paradigm has grudgingly admitted phenomena such as psychosomatic disease might lead one to think that this expansion or modification need not be a drastic one, and that the paradigm will emerge stronger for the change. Re- search studies such as many cited in Chapter 1, which look at the pla- cebo effect as a phenomenon to be studied on its own grounds rather than as a variable to be controlled, exemplify the inclusionary ap- proach. X-rays were anomalous from the viewpoint of the Newtonian para- digm, but not from the viewpoint of modern physics. Similarly, how one construes the phenomena we have been calling the placebo effect depends on the paradigm of reference. Consider an African native vil- lage with two witch doctors who use essentially identical healing rit- uals; an anthropologist discovers that one is viewed by the villagers as more highly expert at his craft, and that that one achieves a sig- nificantly higher cure rate than the other. The anthropologist might conclude that 1) all healing accomplished by either is due either to the placebo effect, or to the normal vicissitudes of disease; and 2) the greater healing rate of the one is due to a differential placebo effect, produced by greater expectations on the part of the patients. But this is to view the matter from the Western paradigm, which holds treatment not explainable in our accepted theoretical terms to be bio- medically inefficacious. The disease paradigm operating in that vil— lage, however, may hold that a witch doctor's cure always works unless the patient fails to follow directions exactly, or thinks impure 36 thoughts while involved in the ritual; all treatment failures may be explained in these terms. The villagers might then postulate that the more respected witch doctor is better able to banish impure thoughts and to command compliance from his patients. Not only does this para- digm explain the phenomenon that we would attribute to the placebo ef- fect in totally different terms; but this paradigm indeed seems to leave no room within its explanatory model for anything like the pla- cebo effect at all. Thus when Shapiro claims that the history of medicine before the present century is the history of the placebo effect (1968, p. 597), he is saying that therapies then in use are deemed worthless by mod- ern medical science, and that nevertheless patients got better at a rate not attributable entirely to the natural recuperative powers of the body.2 But this is again to apply our present paradigm uncriti- cally; a serious medical historian would seek rather to determine what paradigms dominated the thinking of those earlier physicians. (I will suggest later that the use of the term 'placebo effect' in Shapiro's statement can be understood only in a deriVative or metaphorical sense.) Since the placebo effect is already a rather slippery concept, as the next section will show, one might want to begin the task of defi- ning it by accepting at least one firm reference point; and for my dis- cussion this will be our currently accepted medical paradigm. I will, however, try to indicate explicitly the paradigm-dependent elements in 2. Respect for the body's self-healing potential is justified by such classic treatments as (Cannon 1963). However, whether modern med- icine is so clearly superior to past practices is cogently called into question by (Powles 1973). 37 the discussion, so that they will not escape critical scrutiny. 2.2. Boundaries of the Placebo Effect The term 'placebo effect' can be construed very narrowly so as to refer to only a few sorts of phenomena, or very broadly so as to include much of medical practice and many nonmedical occurrences as well. There are some "core" uses of the term that almost all medical scientists would agree to; and there are uses of the term that fall near the "boundaries" of its applicability that might engender con- siderable debate. A good way to get clear on these boundary condi- tions is to consider a series of illustrative examples, which show what is at stake if we draw the boundary lines at various points. This will provide a basis for the formal definition in the next sec- tion. Case 1. A patient suffers from pain due to periodic flareups of rheumatoid arthritis. During one such episode the physician administers sugar capsules, telling the patient that this is a new analgesic drug. The patient subsequently reports dramatic relief. Case 1 seems to be a straightforward and uncomplicated instance of the “core" sense of the term 'placebo effect.‘3 None of the medi- cal authors cited in Chapter 1 would hesitate to apply the term in such a case. Case 2. A and B both contract a cold at the same time, under similar circumstances. A is administered a sugar pill, being told that it is a potent cold remedy; B gets no treatment. Both A and 8 recover from their symptoms at the same rate, with the 3. Recall Shapiro's (1968) contention that the placebo effect can be either positive or negative; for simplicity the case illustrations will deal only with positive placebo effects except where noted. 38 same level of discomfort until their colds subside. Would one want to say in Case 2 that A experienced a placebo ef- fect? By current medical thinking, the recovery of both A and B can be explained on the basis of the self-limiting aspects of viral infec- tions, immune defense mechanisms, restoration of homeostatic proces- ses, and so forth. There are thus good grounds to regard the taking of the sugar pill as irrelevant to A's course-- all things being equal, he would have gotten better in an identical fashion without the pill. It would sound paradoxical to attribute an "effect" to an intervention which played no role in the outcome; and on this analysis we would not regard Case 2 as an example of the placebo effect. We would rather say that both A and B got better as a result of the body's natural re- storative processes. Suppose on the other hand that one wanted to argue for a possible role for the placebo effect in Case 2. Taking seriously the ceteris paribus assumption, this would entail the presence of the placebo ef- fect in such a way that we could not find out about it in terms of ob— servable outcome. I will assume that one purpose of defining 'placebo effect' is to stimulate and guide empirical research into its workings, and that adequate understanding of it will involve empirical issues as well as conceptual ones. Given that purpose, there is nothing to be gained, and some measure of clarity to be lost, in taking Case 2 to involve the placebo effect. Case 3. A large number of individuals are suffering from a wide variety of diseases. Half of these individuals are fed an especially nutritious diet while the other half are fed a nutri- tionally inadequate diet. A larger percentage of the first group recover as compared to the second, although a number of individ- uals in the first group do have worsening of their disease despite the diet.4 39 The effect of diet on disease resistance has some features in com- mon with the placebo effect. The same basic diet will be effective for a large number of diseases. There is a measurable positive response to diet therapy, but it generally falls well below 100 per cent (except in cases of specific nutritional deficiencies). And the diet presuma- bly does nothing directly to alter the basic causative mechanism of the disease (microbes, cellular malignancy, or whatever). However, as with the so-called natural restorative powers, we can explain the results of nutritional therapy by pathophysiologic theories which appear to be independent of the placebo effect; and so postulating a placebo effect in Case 3 as it stands would appear need- lessly to multiply explanations (assuming that the improvement of the first group is not "over and above" the amount that can be explained on the basis of the theories alluded to ). The features that nutri- tion and the placebo effect have in common suggest what medical authors have in mind when they refer to "nonspecific therapies." Exercise, and modalities which enhance the efficacy of the body's immune system, might be cited as other examples of "nonspecific therapy" (which it- self stands in need of formal definition). Thus the placebo effect would be one type of nonspecific therapy, but is not coextensive with that class. Case 4. Imipramine is the drug of choice for treating certain types of depression. Both Dr. A and Dr. B use this drug, with the same dosage schedules, on large numbers of depressed pa- tients. Dr. A is a surly fellow while Dr. B is encouraging and supportive. Of Dr. A's patients, 75 per cent are significantly improved in three weeks while 90 per cent of Dr. B's patients are improved in that time. 4. I will, of course, not attempt to defend the ethics of such an experiment, were it to be done deliberately. 40 Imipramine is certainly not an inert substance; it is both active and specific for the condition being treated. But to explain the dif- ferent results (again assuming the patient populations otherwise equal) we are inclined to view the total therapy as consisting of the drug plus the emotional-psychological features of the doctor's interaction with the patient. Like many investigators, we have been forced by the data to turn our attention away from the drug itself and to look in- stead at the total context (1.4); and on this basis we might attri- bute Dr. B's increased success rate to a placebo effect. If we do so, we are using 'placebo effect' to designate the results of one component of the therapy-- a component which in the actual setting might be so intermingled with other features of the doctor-patient exchange as to be practically indistinguishable.5 This is different from the simple sugar-pill case; but on balance it seems a reasonable extension of the term. Some medical authors (Houston 1938; Wolf 1959; Shapiro 1968) define 'placebo effect' so as to allow for this use while others (Pep- per 1945) do not. Case 5. A person who never goes to doctors decides to improve his health by undertaking an exercise program. He develops strength and endurance, as well as a more general sense of fit- ness and well-being.6 In Case 4 we explained the total result in terms of both patho- physiological and psychological features; and we attributed the latter component to the placebo effect. The increase in strength and endur- ance in Case 5 can be explained as specific outcomes of exercise. On 5. The objection raised in Case 2 above does not apply here. "Prac- tically indistinguishable" does not mean empirically undetectable in principle; it merely challenges the ingenuity of the investigator for controlling for subtle variables. 6. I am indebted to David S. Sobel for this example. 41 the face of it, the psychological sense of well-being one gets in ad- dition does not seem dissimilar from the added boost that a supportive doctor-patient relationship can give to an active medication. 15 there any reason not to attribute this result to a placebo effect of the ex- ercise? To argue for such a reason I must introduce the notion of the "healing context." This is derived from the concept of the "sick role" first introduced by medical sociologists, which has proven very useful in cross-cultural studies of response to illness.7 One feature of the sick role is that the sick person must submit to the authority of the socially-designated healer for the purposes of attempting a cure. While the healer may be a medical doctor, an herbalist, a sha- man, or whatever, such socially-designated healing roles exist in vir- tually every culture that has been studied. Furthermore, it is usual if not universal for a particular setting-- hospital, cave, temple, etc.—- to be identified with the healing activity, and for certain rit- ual behaviors (often including behaviors not tolerated by that culture anywhere else) to become associated with that setting and with the purpose of healing. This combination of the designated healer, desig- nated setting, and designated rituals I refer to as the "healing con- text." It refers to something that may be present in all cultures, without referring to the specific healing practices of any particular culture or the specific beliefs of any particular medical paradigm. As a rule, of course, the psychological reaction of the patient in the 7. For discussion of the sick role from the sociological perspec- tive see (Parsons 1951; Parsons 1961; Siegler and Osmond 1973; Friedson 1970, pp. 205—243). 42 healing context can be elicited only if the culture-bound features of the healing context are those of the patient's own culture.8 Consider- ing the universality of patterns of social response to sickness, I think it is reasonable to assume that there are important similarities, say, between the native's psychological reaction to being in the sha- man's temple and the Western individual's reaction to hospitalization. The question posed by Case 5, then, is whether we wish to impose as a boundary condition on the term 'placebo effect' that it apply only to events occurring within a healing context. Even this condi- tion may be too weak, as Case 6 illustrates. Case 6. A patient who is scheduled to undergo open heart sur- gery, and who is in acceptable physical condition, becomes very depressed and insists, despite support and reassurance from the medical staff, that he is sure that he is going to die during surgery. The operation is begun and all is going well until, for no apparent reason, there is a sudden drop in blood pressure. All attempts to correct this fail and the patient dies. The ability of persons to "think" themselves into otherwise unex- plainable deaths is well documented (Frank 1974, pp. 50-55; Engel 1976).9 If we follow Shapiro's reasonable convention of referring to placebo effects as being potentially either positive or negative, could we at- tribute the death in Case 6 to a negative placebo effect? Unlike the situation in Case 5, the events in question occur within the healing context. But the psychological effect, depression, is neither the intended therapeutic intervention, nor a concomitant of the intended therapeutic intervention (as in Case 4); indeed the doctors in Case 6 8. Hence, treatment problems arise for Western medicine when the patient adheres to folk medical beliefs of his own subculture which are not understood by his physicians (Snow 1974). 9. Psychological risk factors in open heart surgery are currently being studied by Dr. Sumer Verma (personal communication). 43 try deliberately although unsuccessfully to counteract the depression. Cases 5 and 6 suggest a conceptual "slippery-slope" problem with the boundaries of 'placebo effect.' There are an almost endless num- ber of instances where suggestion or auto-suggestion, or other psycholog- ical states, influence persons' bodily processes or their perceptions of bodily processes (Frank 1974; Kiritz and Moos 1974). The psycho- physiological mechanisms by which these occur require empirical elu- cidation. While it would be surprising if the mechanisms by which a sugar pill can ameliorate symptoms turned out to be totally different from the mechanisms involved in these other instances, the precise degree of similarity or dissimilarity needs to be investigated; it does not seem to be a matter to be decided by definitional fiat. I have suggested already that the task of defining 'placebo ef- fect' can be viewed as a preparatory step toward this needed research. In what ways, then, can the choice of definition either help or hinder research? It might help if the definition called attention to simi- larities between the defined phenomenon and an already-known class of events, where investigators had not already discerned the possible con- nection. But the placebo literature shows no reluctance to view the placebo effect in light of what is known about other psychophysiologic correlations-— we have already reviewed attempts to apply such stan- dard psychophysiologic theories as conditioning and transference to the placebo problem (1.5). Alternatively, a definition might hinder research if it was too inclusive, tempting the investigator to pass over important differences among classes of phenomena. For instance, so long as respiratory dis- eases caused by bacteria, mycoplasma, and viruses were all lumped 44 together as "pneumonia," the investigation of the worth of penicillin in treatment was bound to be impeded. We saw, in 1.3, how much re- mains to be learned about the psychophysiologic phenomena that occur within and as part of the healing context. I assume that a reasonable research strategy would be to get clearer on these instances before trying to generalize the findings to other aspects of human existence. If one accepts this empirical bias and this strategic assumption, it makes good sense to exclude the phenomena described in Cases 5 and 6 from the definition of 'placebo effect.'10 Case 7. A is a Christian Scientist and, despite being severely ill with rheumatoid arthritis, refuses to take any sort of drug or other medical therapy. 8, who is concerned about A's welfare, knows of studies showing that arthritic patients improve when given a placebo such as lactose. B obtains a supply of pink lactose tablets; but, knowing A's aversion to medication, con- trives secretly to slip the tablets into A's coffee, without A being aware of this.ll Case 7 points out another feature of the healing context as it re- lates to the placebo effect. It makes sense to say that B has slipped a sugar pill into A's coffee, but does it make sense to say that B has slipped a placebo into A's coffee? The latter terminology seems to involve a conceptual absurdity, regardless of whether A's condition in fact changes or not.12 The lesson of Case 7 is that it is not enough 10. It might eventually turn out that the similarities among the different classes of phenomena were so striking that 'placebo effect' would be dropped altogether in favor of a more general term such as 'autosuggestion.‘ Still, the term 'placebo' would remain in use to designate a dummy medication or treatment. 11. I am indebted to Martin Benjamin for this example. 12. The absurdity here is similar to that in a story Abraham Lin- coln liked to tell, about an Irishman who had taken an abstinence Pledge and was forced to order lemonade at a bar on a hot day; he fi- nally leaned confidentially toward the barkeeper and asked, "Couldn't you put a wee drop 0' the creetur into it unbeknownst to me?" (Sandburg 1939, 1:572, IV:158). 45 for the subject to be in a healing context in order to allow applica- tion of the term 'placebo effect'; the subject must believe that he is in a healing context. Generalizing from numerous studies, it seems that the subject need not believe that the treatment being given is efficacious;13 it appears sufficient that the subject believe that it is treatment, that it is a deliberate intervention being given in response to his illness condition with the intent of benefit. Suppose that we inform B of his conceptual error, and he now has to decide what to do with his large supply of pink tablets. If he uses them in his own coffee, as a sweetener, we would not want to say that he is using placebos on himself. We have already noted the em- pirical findings that have led placebo investigators to fbcus on the context of placebo use, not on the dummy treatment itself; our anal- ysis of Case 7 adds to this empirical observation the stronger con- ceptual point that the meaning of 'placebo effect' is context-depen- dent in the way that we have noted. If belief in the healing context is a necessary condition, is it also sufficient? We could imagine an elaborate sham in which an indi- vidual was made to believe wrongly that he was in a clinic receiving treatment from a doctor when in fact he was getting dummy pills and shots from actors on a movie set. If the victim of this subterfuge experienced a relief of symptoms attributable to this experience, we could say without contradiction that a placebo effect had occurred. We might even want to go so far as to say that the belief itself was 13. In the nonblind placebo trial (Park and Covi 1965), several of the patients responding positively had initially expressed doubts that placebos would work; if placebos can work in the face of doubt they ought to work also in the weaker case of nonbelief. 46 sufficient to make the movie set a healing context for that individual in his present belief state, in the same way that a witch doctor's thatched hut may be a healing context for an African native but not for a Wall Street stockbroker. However, for our present purposes we need not debate this latter point. From a practical standpoint exam- ples such as the movie-set sham do not pose any significant problem for defining 'placebo effect,‘ however interesting they might be in terms of isolating the key features of healing contexts. Before turning finally to the matter of formal definitions, it is important to emphasize the difference between the boundary condi- tions indicated by Cases 5 and 6 and by Case 7. The former, requir- ing the healing context as a necessary condition, is a stipulative device suggested because of its probable utility for research. But the latter, requiring belief in the existence of a healing context, is conceptual and points out an essential feature of the word 'pla- cebo.‘14 2.3. Formal Definition of 'Placebo Effect' The considerations from the preceding section can now be used to evaluate critically the formal and informal definitions offered by medical authors, and to make suitable changes. Four major definitions, already cited in 1.1, may be summarized as follows: 1) Pepper (1945): The placebo effect is a therapeutic effect pro— duced by a biomedically inert substance. 14. However, see the formal definitions and discussion in 2.3, below, on why a placebo need not be present for the placebo effect to be said to occur. 47 2) Wolf (1959): The placebo effect is a therapeutic effect or side effect attributable to a treatment, but not to its pharmaco- logic properties. 3) Shapiro (1968, p. 599): The placebo effect is the nonspecific effect of a therapy, which may or may not have a specific ef- fect in addition. 4) Modell (1955, p. 55): The placebo effect is what all treat- ments have in common. These definitions are listed in order of increasing breadth, and in- creasing range of phenomena that fall under them. For instance, Pep- per's definition would hold that if a specific pharmacologic effect is present, the placebo effect cannot be present; Wolf's and Shapiro's, that if a specific pharmacologic effect is present, the placebo effect may also be present; and Modell's, that if a specific pharmacologic effect is present, the placebo effect must be present. On grounds already discussed we can eliminate the most narrow and the most broad of these four proposals. Pepper's approach is ruled out by our willingness to look at different components of a total therapeutic encounter, and to ascribe a placebo effect to a nonspecific component which may accompany administration of an active treatment. Modell's all-inclusive statement is refuted by an example mirroring Case 7. In The Moonstone a physician, angered by statements from the hero on the total worthlessness of medicine, secretly places some lau- danum into the hero's coffee; and the hero, who had previously been troubled by insomnia, slept unusually soundly that night (Collins 1868). This is a clear case of pharmacologic potency without any accompanying placebo effect.15 48 The remaining proposals are substantially similar and are roughly consistent with our previous discussions. But they make use uncriti- cally of the terms 'therapy' and 'nonspecific,‘ which seem to require some elucidation. I begin by offering a definition of 'therapy': T is a therapy for condition C if and only if it is believed that administration of T to a person with C increases the empiri- cal probability that C will be cured, relieved, or ameliorated, as compared to the probability of this occurring without T. This definition of 'therapy' is intended to be as general as pos- sible, embracing drugs, surgery, physical therapy, psychotherapy, and so on, even though it does not include measures aimed solely at pre- vention of disease. "Administration" should be interpreted to in- clude acts of omission, such as salt restriction, and self-administra- tion by the person himself; but it is intended to restrict 'therapy' to acts of deliberate intervention or human agency. The definition does not explicitly require that condition C be a disease or a symptom of disease; this interesting issue is not pertinent to the matter at hand. The phrase "it is believed that" is included to allow one to speak of ineffective therapies; if this were omitted the definitions for 'therapy' and 'effective therapy' would be the same, contrary to general usage.16 The definition also indicates implicitly when one is justified in believing that T will relieve C, by including the reference 15. Modell elsewhere seems aware of this problem when he states that the placebo effect invariably accompanies every prescription of a drug (1955, p. 54). 16. Who believes this is deliberately left vague; it might be the person with C, the person administering T, or some third party as out- side observer. The importance of specifying which of these hold for a specific case is illustrated by the witch-doctor example in 2.1 above. 49 to empirical probability-- either a randomized controlled study must show that T is more likely than no treatment to relieve C; or other theories of pathophysiology, which are themselves supported by empiri- cal data, must predict T's efficacy based on known causal mechanisms for C. Anecdotal evidence or personal experience justify the belief only in a derivative sense; one must be willing to assume that a future controlled trial, if carried out, would confirm this evidence. There- fore, this definition of 'therapy' is dependent upon our present med- ical paradigm, which holds up the standard of the randomized, con- trolled trial over any other form of investigation. By the defini- tion, we might say that physicians of other historical periods, or in other cultures, used therapies; but by our present paradigm we would not be willing to say that they were justified in considering these measures to be therapies. We could still note that these physicians could have been considered justified by reference to the paradigm un- der which they were operating; the problem of cross-paradigm criticism and justification is a general problem in history and philosophy of science, and is not peculiar to this definition or to the placebo prob- lem. By contrast, we might envision a culture which related the cause of all disease to transgressions against basic social mores, and for whom therapy was seen in terms of atonement or expiation. This disease- therapy paradigm could be internally consistent, and could have social- cohesiveness value as well, even to the extent that whether a particu- lar therapy ever did any good for the individual patient in an empiri- cally verifiable way might be irrelevant. This culture might offer a definition of 'therapy' which would be radically different from ours, 50 but which in its own way would be equally paradigm-dependent.l7 The acceptability criterion implicit in the definition of 'ther- apy' has an important implication-- that one is never justified in con- sidering an intervention to be a therapy. in the absence of an accep- ted theory of pathophysiologic mechanism, based on observation of only one patient. This attitude is consistent with the present unwilling- ness to accept anecdotal evidence in medicine. If, then, we are later to define 'placebo effect' in terms of a sort of therapy, it would fol- low that to ascribe the placebo effect to one patient is implicitly to formulate a hypothesis about a class of patients. Turning to the next problematic term: T is a specific therapy for condition C if and only if: 1) T is a therapy for C 2) There is a class A of conditions such that C is a subclass of A and that for all members of A, T is a therapy 3) There is a class B of conditions such that for all members of B, T is not a therapy; and class B is much larger than class A. An example might be penicillin used for pneumococcal pneumonia. Peni- cillin is a therapy for this disease, since it increases the empirical probability of recovery. Pneumococcal pneumonia is one of a class of diseases (infectious diseases caused by penicillin-sensitive organisms) for all of which penicillin is a therapy; but there is a much larger class of diseases (noninfectious diseases, and infectious diseases caused by viruses, rickettsiae, etc.) for which penicillin is not a therapy. Therefore penicillin is a specific therapy for pneumococcal 17. I am suggesting here that the notion of "therapy" is connec- ted very intimately with that of "disease," a point I cannot argue for here. Unfortunately recent philosophical inquiries into the concepts of health and disease have almost totally neglected this point. 51 pneumonia. 'Specific' is used in several ways in medical discourse, and this definition is consistent with what one might call the loose sense-- the sense in which "specific therapy" might be contrasted with "general therapy." It should be noted that (2) does not require that C be a proper subclass of A, so C may be identical with A (i.e., the therapy is specific for one condition only, such as iron for iron deficiency anemia). There is also a stronger sense, however, in which "specific ther- apy" is roughly equivalent to "best therapy." By the definition given above, penicillin would be a specific therapy for Escherichia coli in- fection, since penicillin is better than no therapy. But in practice physicians would not refer to penicillin as a specific therapy for E. coli, since the organism is four to five times more likely to be re- sistant to penicillin than it is to ampicillin, for example. To deal with this usage we might introduce an additional definition for 'pre- ferred specific therapy.‘ For a therapy to be a preferred specific therapy it would have to be a specific therapy as defined above; and there would have to be no therapy T' which offered a better risk-bene- fit ratio than T for C, taking into account both therapeutic efficacy and absence of significant side effects. However, for the placebo context we will not need this additional definition. (Nor will we need the still stronger sense of 'Specific' in which a therapy is not only known to be best empirically, but could also have been predicted to be best based on an established pathophysiologic theory-- for exam- Ple, vitamin 812 as a specific therapy for pernicious anemia.) We can now combine these elements into a definition of the placebo 52 effect: A placebo effect occurs for person x if and only if: 1) x has condition C 2) x believes that he is within a healing context 3) x is administered intervention I as part of that context, where I is either the total active intervention or some component of that intervention 4) C is changed 5) The change in C is attributable to I, but not to any spe- cific therapeutic effect of I or to any known pharmacologic or physiologic property of I. The mention of belief in the healing context, and the possibility that I may be only one component of the total healing intervention, reflect the boundary conditions discussed in 2.2.18 The definition, like Wolf's and Shapiro's, allows for both positive and negative changes in C. "Not attributable to any pharmacologic or physiologic property of I" excludes changes due to diet or other nonspecific therapies. To whatever extent psychotherapy can be shown empirically to be effica- cious, it is also a specific therapy and so is excluded even though it has no "pharmacologic of physiologic" effect. The word "attributable" may be interpreted in light of our discussion of 'therapy' and the ac- ceptability criteria implicit in the current medical paradigm; it also refers to the present state of medical knowledge, and leaves open the possibility that newly-discovered properties of I may cause us to change our view that C was modified by the placebo effect.19 It is even conceivable, from the form of the definition, that everything we 18. Since what counts as a healing context depends on the culture of the individual, inclusion of this term in the definition means that the placebo effect is inherently culture-dependent (Riley 1976). 19. Note that I may not be known specifically-- in the sugar-pill case, the cause of the symptom change is assumed not to be the chemical content of the pill, and no other medication is known to have been used, so some other element of the total episode is assumed to be responsible. 53 now attribute to the placebo effect will someday be attributed to new laws of medicine, leaving 'placebo effect' without reference. There are thus two very different epistemic elements in our definition-- the belief state of the individual subject regarding the healing context (which is culture-dependent), and the belief state of medical science regarding what can be explained by existing laws and theories (which is paradigm-dependent). It does seem unsatisfactory that 'placebo effect' has been defined by exclusion, as something not attributable to other things. Why not an inclusive definition, such as one attributing the placebo effect directly to the psychological component of the healing intervention? Certainly in practice "not attributable to known pharmacologic or physio- logic properties" could amount simply to "attributable to psychological properties"; but it could also mean attributable to presently unknown pharmacologic properties, or to some completely different sort of prop- erty. This seems to be a matter best left for empirical research. Further, if one framed an inclusive definition but left out mention of what sort of property to which the placebo effect was to be attributed, it would reduce to a definition of a nonspecific therapy, and one would be unable to distinguish the placebo effect and the effects of diet or exercise. Significantly, I have offered a definition of 'placebo effect' without having given a definition of 'placebo.‘ This is in keeping with the trend we have already noted several times, of looking at the total context instead of at the inert medication. Thus, as Case 4 showed, we can apply the term 'placebo effect' to instances where no placebo is in evidence. In such cases one can label the purported causative 54 component of the intervention the "placebo stimulus" to emphasize this distinction. We can then be satisfied with a more traditional, restric- ted definition of 'placebo' proper: A placebo is: l) a form of medical therapy. or an intervention designed to simulate medical therapy. which at the time of use is be- lieved not to be a specific therapy for the condition for which it is offered, and which is used either for its psy- chological effect, or to eliminate observer bias in an ex- perimental setting. 2) (by extension from 1) a form of medical therapy now believed not to be efficacious, though believed efficacious at the time of use. Clause (2) is added to make sense of a sentence such as, "Most of the medications used by physicians one hundred years ago were actually pla- cebos." One of the epistemic elements from the definition of 'placebo effect' reappears, the mention of the present belief state of medical science. Where a placebo is used for therapy, we can assume that the second epistemic element is present also, since to have a "psychologi- cal effect" the therapy must be believed to be such by the recipient (as Case 7 illustrates). But this element of belief may be lacking in the experimental setting, which is another important use of 'placebo.‘ I have argued that a definition of 'placebo effect' ought to aid and stimulate research. The definition given above fills the bill. It asks the question: if the change in symptomatology is not attribu- table to known pharmacologic or physiologic properties of the interven- tion, to what is it attributable? At the same time, it avoids closing off lines of research by a priori stipulations of what sorts of proper- ties to consider. But beyond the empirical questions, I am concerned with the philosophical significance of the definition. One important line of investigation is suggested by the possibility that psychological 55 mechanisms might be producing bodily changes, and that the belief state of the subject is a necessary condition for this to occur; this would seem to have important consequences for theories of the mind-body re- lationship, which will be taken up in the next three chapters. A dif- ferent line of investigation is the ethical question of the use of pla- cebos as therapy; and Chapter 6 will show that the formal definition given above has significance for that issue also. Chapter 3. Traditional Mind-Body Views and the Placebo Effect Having reviewed the empirical data on the placebo effect and hav- ing formulated a definition of this phenomenon, we may now ask what im— --»plications this line of inquiry has for the mind-body relationship. By itself, the placebo effect raises interesting questions about philosophy of mind; but in addition, to the extent that mens sana in corpore sano is a goal of medical practice, these questions are central to any phil- osophy of medicine. This chapter and the next two will be devoted to mind-body issues. This may seem to be a disproportionate amount of attention, especially since much of this present chapter will be devoted to listing possible theories only to reveal later their weaknesses and defects. But in fact, although all proposed mind-body theories have flaws, very few of them are outright nonsense; almost all theories capture some portion of the complex of intuitions that we hold about our bodies and our minds. In general, theories fail, not by failing to capture and to illuminate the intuitions to which they are addressed, but rather by failing to take into account other, equally basic intuitions. Thus, reviewing many alternative theories will place us in a much better position to examine critically the theory we will, in the end, find most satisfac- tory, even if in the process we are led somewhat far afield from the placebo effect itself. 56 57 Despite the large amount of space being devoted to the mind-body issues (as compared, say, to the ethical issues in Chapter 6), it will be necessary to skim rather lightly over many possible areas of con- troversy, and to summarize in rough—and-ready fashion philosophical ar- guments that are very complex in their full development. Thus, the following discussion may suggest a wider agreement on many philosophi- cal points than is actually the case, as examination of the references cited will readily show. This chapter begins by reviewing the "reflective equilibrium" strategy mentioned in the Introduction, as it relates specifically to the mind-body issue. The next section provides an overview of tradi- tionally accepted mind-body theories. In the third section, several of these theories which seem plausible will be applied to the placebo ef- fect, and the grounds for their plausibility will be explored; but in the final section, significant defects in each of the theories will be shown. It is by attempting to modify such theories to eliminate these defects that an even more plausible theory, eliminative materialism, emerges; that theory will be discussed in the next chapter. 3.1. A Reflective-Equilibrium Approach to Mind If we accept the notion that the accumulated data about the placebo effect require some sort of explanation in terms of how the mind and the body are related, and that this is a matter for philosophical analysis rather than for additional empirical research alone, we can approach the task of explanation in different ways. One is to seek out the view of the mind-body relationship which best explains the placebo effect as an isolated phenomenon, or which at least does not conflict with any of 58 the known empirical findings. Whether this view accounts for mind- body issues not directly raised by the placebo data (for instance, the question of whether minds can exist apart from bodies) would be con- sidered irrelevant by this approach. This approach is consistent with the pragmatic, task-oriented way in which physicians and medical scien- tists have approached the mind-body problem, when they have approached it at all.1 Thus, one finds in the medical literature proposals for "double-language theory" (Graham 1967), holist emergentism (Wolff 1962), and "methodological dualism" (Boss 1975). But as a rule, these accounts deal with medically-related issues only at the expense of other features of a comprehensive philosophy of mind-- whether minds can exist apart from bodies, whether we can know that there are minds other than our own, and so on.2 If a philosopher notes that a medical mind—body the- ory raises problems and conflicts in these other areas, it seems as if philosophy is simply raising impediments to medical research and prog- ress. Small wonder under these circumstances that medical people might come to regard the "mind-body problem as philosophically senseless and scientifically wasteful of time and effort" (Freedman, Kaplan and Sa- dock 1972, p. 432). An alternative approach is the "reflective equilibrium" strategy 1. See, for example, Engelhardt's analysis of the research-connec- ted motivations that led the 19th century neurologist John Hughlings Jackson to adopt the doctrine of parallelism (Engelhardt 1975a). 2. An exception is the approach taken by the philosopher-physician Tristram Engelhardt. His more sophisticated theory, in the Kantian- Hegelian tradition, takes mind and body to be two separate domains of significance, such that attempts to relate them causally constitute category mistakes. 0n matters such as psychosomatic medicine and the placebo effect, his views seem to be a type of epiphenomenalism; but this may be my misreading of his position (Engelhardt 1973). 59 described in the Introduction. On this view, the task is to find the overarching theory that best makes what we know about the placebo effect hang together, in a consistent and mutually illuminating way, with other conceptual considerations regarding mind and body. Our particu- lar concern with the placebo phenomenon should not make us forget that we have many basic considered judgments about mind and body. These might include, for example, our certainty that we need no grounds to ascribe a sensation such as pain to ourselves-- we simply are in pain, we do not infer that we are based on evidence-- while we do need grounds to ascribe it to others; yet we unhesitatingly treat another who is in pain as if he has the same sensation that we do when in pain ourselves. So we want a mind-body theory that deals adequately with the placebo effect, but we also require that our theory "fit" with basic considered judgments of the sort mentioned. We are willing to work from both ends, either giving up some fineness of grain regarding the placebo effect in exchange for better overall "fit," or sacrificing some degree of "fit" for a theory which promises to highlight the pla- cebo effect in a particularly illuminating way. If the search for this kind of broadly-based theory fails, we may then wish to accept a narrow, medically-oriented view. But since, in the course of searching for the best "fit," we might find our considered judgments about other matters giving us new insight into the placebo effect, and vice versa, it would be a methodological mistake to settle for the narrow theory without making some attempt to search for a more comprehensive one first. The mind-body theories that we will consider, then, will be looked at both from the standpoint of the placebo effect and from the stand- point of our basic considered judgments. It will turn out in many cases 60 that different theories give equally adequate, if equally vague, ac- counts of the placebo effect; thus the basic considered judgments will play the larger role in ranking the different theories according to their philosophical p1ausibility.3 3.2. Overview of Alternative Mind-Body Theories Almost all views of the mind-body relation assume that there is a significant and basic difference between statements about sensations, volitions, thoughts, memories, etc., and statements about the structure and function of physical bodies.4 Originally, Descartes characterized mind as thinking and unextended (i.e., neither occupying nor moving through physical space), and body as unthinking and extended; a human person was seen to consist of a mind plus a body.5 While mind was orig- inally thought of as nonmaterial substance, difficulties with this con- cept have been avoided by speaking instead of mental events. Mental events differ from physical events in that we have some sort of nonin- ferential access to some of them (i.e., our own), so that as a rule we cannot be mistaken about them; and in that mental events are not locali- zable in space in the precise way that physical events are. 3. That mind-body theories give us vague accounts of the placebo effect should not by itself count against them; we would not want phil- osophical theories to fill in details that can properly be provided only by further empirical research. 4. This section follows roughly in its organization (Shaffer 1967). For an overview of significant contemporary positions on mind-bod , with- in the Anglo-American tradition, see (Chappell 1962; Shaffer 19651. 5. For the original statement of this position see (Descartes 1927, pp. 145—165); (Spicker 1970, pp. 3-23) provides a summary of the prob- lems that it raises. 61 Clearly, speaking of mental events in general requires that we lump together such very different things as smelling an unpleasant odor and thinking about a differential equation. While much of the discussion that follows will be based on such a lumping together, it will be useful to distinguish two important types of mental events, sensations and intentional states. A rough way of making the distinc- tion is to note that sensations include events such as hearing a bell, feeling a pain, seeing a bright color, and so on; they often correspond to something "out there" but not necessarily, as in cases of hallucina- tions and after-images. Intentional states include believing that the Battle of Hastings was fought in 1066, thinking about the predictability of earthquakes, and fearing that you are going to hit me; these cannot be described completely without mentioning the object (often a proposi- tion introduced by the word 'that'), and the object need not be present or may not even exist-- I can think about Moses or about unicorns. Also, as a rule, sensations are a more primitive sort of event; all sentient animals can have them while only more complex organisms can have intentional states. As we might expect, some mind-body theories give good accounts of sensations while having difficulty with intentional states, while other theories have the opposite characteristics; in par- ticular, intentions seem more susceptible to behavioral analysis than sensations are. If we recognize the mental and the physical as distinct and primi- tive types of events, we can deal with their relationship in two ways. We may choose a monistic theory, which either recognizes the essential reality of only one type of event, or else tries to derive one type of event from the other, or both types from some third type. Or we may 62 select a dualistic theory which recognizes both types of events as equally basic and seeks to explain their relation without slighting either. One sort of monistic theory regards one type of event as totally dependent upon and arising secondarily from the other. Idealism at- tributes reality only to mental events and regards the physical world as totally dependent upon our mental images of it. A tree, for in- stance, would exist only as the object of someone's perception, and would not exist at all if someone were not at that moment perceiving it. Idealistic theories are seldom proposed today. Much more popular is materialism, which holds mental statements to be about certain physi- cal events which occur in the brain. For instance, our seeing a tree consists of photons of light striking our retinas, which then excite neurons to discharge, thus exciting other neurons, and so on. When we have described all these physical events, we have said all there is to say about "seeing a tree"; there is nothing mental "over and above" these physical events. Behaviorism, which we can regard for our pur- poses as a form of materialism, seeks to reduce all statements about mental events to statements about the publicly observable behavior of organisms. The recent successes in neurophysiology research and in operant-conditioning psychology have made materialism and behaviorism, respectively, seem especially credible. Some confusion is avoided if several forms of behaviorism are dis- tinguished. Methodological behaviorism is a statement of research strategy for scientists, which says essentially that one can discover interesting, lawlike regularities by investigating the behaviors of or- ganisms. It is fully consistent with methodological behaviorism that 63 there could exist mental events apart from any observable behavior; such events would simply be excluded from scientific inquiry. Thus, methodological behaviorism is of limited philosophical interest. We shall be concerned later in this chapter with logical behaviorism, which makes a metaphysical assertion which is held to be a general truth, namely, that mental events can be understood in a coherent way only if they are taken to refer somehow to publicly observable behavior. A still more sophisticated metaphysical thesis is radical behaviorism (Skinner 1974), which will be discussed in the next chapter. Other monistic theories attempt to slight the status of neither the mental nor the physical. Identity theory agrees that talk about mental events cannot be reduced to talk about brain events; rather it asserts that these two kinds of talk, though having different meanings, in fact refer to identical happenings—- that the claim that mental events are contingently identical with brain events of the appropriate type is a coherent and empirically testable hypothesis. Double-aspect theory holds that the mental and the physical are different aspects of some third kind of substance; this theory founders on what that third substance might be like, and how mind and body can be "aspects" of it, or of anything else. Dualistic theories are conveniently characterized by the types of causation that each admits. Parallelism holds that physical events can cause other physical events, and that mental events can cause other mental events, but neither cause the other, even though certain mental events seem to be constantly correlated with certain physical events. But on this theory, such a constant correlation is a highly mysterious coincidence; in the absence of causal connections, it is hard to see 64 why a broken bone might not be correlated with pain one time and with joy another time. Some philosophers have brought in divine interven- tion to explain the coincidence, but this is to offer an explanation which is even less understandable than the phenomenon being explained. Thus parallelism is usually rejected. Epiphenomenalism holds that physical events can cause other physi- cal events, and that some physical events (occurring in the brain) can cause mental events; but the mental events can cause nothing. Epiphe- nomenalism seems to acknowledge our considered judgment that our inner mental states are real occurrences, and that they are reliably correla- ted with certain physical events, while avoiding the sticky problem of how nonphysical, nonspatial mental events can cause physical events. But it ignores our equally basic considered judgment that our mental events, such as acts of will or of deliberation, do cause things to occur in the world. It also requires laws of physicopsychic causality to be of a strange sort, in that the postulated effects simply "dangle" and play no further role in any causal chain. Interactionism holds that physical events can cause both other physical events and mental events, and that mental events can cause both other mental events and physical events. This satisfies our considered judgment about the causal efficacy of our mental states, but demands that we face squarely the puzzle of psychophysical causality. We can see that some of the above theories can be dismissed more readily than others. Behaviorism, interactionism, and identity theory seem to have enough initial plausibility to warrant further consideration. 65 3.3. Some Initially Plausible Theories In determining which of the mind-body theories should be investi- gated in depth, we might ask how they would account for the placebo ef- fect, and which considered judgments seem to support them. To apply mind-body theories to the placebo case, we can return to the formal defi- nition from 2.3 and restate it in the form of antecedent and conse- quent events. The antecedent events are that the individual has a symptom, that he believes that he is in a healing context, and that he is administered an intervention. The consequent event is that the symp- tom is changed. An additional observation is that the change cannot be explained on the basis of specific properties of the intervention or of pathophysiologic laws as now known. The link between antecedent and consequent will generally be construed as causal, although this need not necessarily be $0.5 Attempts to apply classical conditioning theory to the placebo ef- fect (1.5) suggest the possibility of a behaviorist account. Such an account would have to construe all the antecedent and consequent events in terms of publicly observable behavior. The potential problem areas are giving behavioral accounts of subjective symptoms such as pain, and of believing that one is in a healing context. The usual method is to account for these in terms of dispositions to behave, such as, "I am in pain" means "I am disposed to yell, pull away, etc." If these 6. The view that all medical thinking is necessarily causal derives from our own dominant paradigm. The entire, complex system of ancient Chinese medicine was essentially noncausal (Porkert 1977). In the para- digm dominant in 18th century Europe, recognition of disease was based on the concept of a "motionless, simultaneous picture" (Foucault 1975, pp. 3-16, 188-189). 66 strategies are acceptable, then the behaviorist account can be comple- ted. Since behaviors occur within the realm of physical events, the causal connection between antecedent and consequent events presents no problem. Causal interactionism would view the placebo effect as a case of a mental event (believing that one is in a healing context) and some physical events (the intervention, the existing bodily state) causing another mental and/or physical event (the subjective and/or objective symptom change). This entails causality between mental and physical events, and we have seen that this needs at least some further explana- tion. Identity theory would essentially accept the account given for the interactionist view, but would add that the mental events referred to are in fact identical to certain physical events in the brain; see- ing this eliminates the puzzle over causality. For research purposes, we would presumably want to learn which brain states are identical to the mental states referred to, so that we could then study their connec- tions on a neurophysiologic basis; we could then learn the precise na- ture of the causal network. Indeed, if this research led to our adop- ting new "laws of pathophysiology," the placebo effect would cease to be unexplainable in terms of those laws and hence would cease to be the placebo effect as we have defined it. Given the methods of modern neuro- science, such a research program does not seem impossible. It is not only the case that each of these three theories manages to account for the placebo effect; in addition, each can point to basic considered judgments that support it. One such judgment is that we rely heavily on the behavior of others to determine what thoughts, beliefs, 67 and sensations they are having. Indeed, even though we generally feel that a person cannot be mistaken when candidly reporting his own men- tal states, we may on occasion reject a first-person report of another on behavioral grounds, as when a person, red in the face and with fists clenched, shouts, "I'm not angry!" This seems to support behaviorism. Another considered judgment is that our increasing knowledge of neuro- science does in fact tell us interesting and informative things about the mind; in particular, it tells us that certain mental events are in some way dependent upon certain brain events, as when an electrode im- planted in a selected brain site reliably stimulates a feeling of plea— sure or a specific memory trace. This considered judgment seems espe- cially compatible with identity theory. Finally, interactionism is supported by the two considered judgments referred to earlier-- that our inner mental states have undeniable reality and causal efficacy. Thus, the three theories are each prominently but not uniquely supported by certain considered judgments. For the committed proponents of one of these theories, the importance of the considered judgments is likely to be exaggerated. Instead of being merely a prominent feature of what we mean by mental events, the considered judgment that supports one's own pet theory is likely to be seen as the crucial feature of the mental realm. Thus, it is important to subject these three theories to more critical scrutiny, especially taking note of the problem areas that have been mentioned. 68 3.4. Rejection of Commonly Held Theories Each of the three theories considered in the previous section can be shown either to conflict with other considered judgments, or to give rise to troublesome conceptual puzzles. We will, it seems, have to look farther afield for a satisfactory theory to account for the pla- cebo effect. Behaviorism holds that descriptions of any psychological state can be reduced to descriptions of behaviors that are publicly observable in principle.7 Therefore, if we can find any psychological states which cannot be so reduced, we will have raised serious doubts about the doc- trine of logical behaviorism (however useful methodological behaviorism might remain as a working hypothesis in psychology). It is useful to focus on the mental-state report, "I believe that I am in a healing context," as our example (passing over for the moment the fact that "healing context" is an abstract concept unlikely to arise in daily conversation). Attempted behavioral accounts of belief states commonly take forms such as dispositions to behave or dispositions to make assertions. Such accounts of our mental-state report might be: 1. I am disposed to follow instructions given by the healer, to allow examination of my body, etc. 2. I am disposed, if asked, "Are you in a healing context?" or the equivalent, to answer affirmatively. 7. Such behavior need not be readily observable in practice for the behaviorist to make his metaphysical case. For example, some have at- tempted to analyze thought in terms of subvocal laryngeal contractions. 69 But these accounts as they stand are incorrect. I may believe that I am in a healing context but not be disposed to act in the appropriate ways if, for example, my fears of the medical procedures outweigh my desire to be cured. And I might believe that I am in a healing con- text but not be disposed to answer a question to that effect if, for instance, I have a desire to deceive the questioner. We could, it is true, expand our account to include such qualifiers: "...disposed to... if I have no desire to deceive, if I understand the question put to me, etc." But such an expanded account is no longer behavioristic, since mental terms such as ”deceive" and "understand" have crept into it. If in turn we try to give a behaviorist account of "deceive," we will have to add similar qualifiers which include mental terms of their own, and so on. Thus it would seem that any behaviorist analysis of this sort will either be incomplete, or will include unreduced mental terms in the analysis itself (Chisholm 1957, pp. 168-173). Further reflection suggests that this problem reflects a general feature of behaviorism, and is not the result of the particular exam- ples that we chose. For instance, "knowing that..." involves being disposed to answer certain questions correctly if'I want to, if’I am not confused, etc.; and "wanting to answer," in turn, involves being disposed to answer correctly if’I know the answer, if‘there is nothing else I want more, etc. It seems to be a necessary feature of psycho- logical states that they can be characterized completely only in terms of their relations to other psychological states, although they can and indeed must be characterized in part in terms of observable behavior. Thus, no psychological term can be characterized adequately in such a way as to eliminate all psychological terms from the explanation 7O (Putnam 1964)-- any more than we can describe the relation, "the tree stands to the right of the boulder," merely by describing the struc- ture or the behavior of the tree itself. If behaviorism must be rejected as an adequate account of belief states, we must also reject classical conditioning theory, with its simplisitc stimulus-response characterization, as an adequate placebo explanation. This runs counter to the assertion that experiments showing a "placebo effect" in animals provide empirical support for conditioning theory (Byerly 1976). Can an animal believe that it is in a healing context? We can attribute to animals concepts whose presence can be manifested by non-verbal recognition; a dog can show by his behavior that he believes that his master will be coming home soon (Kenny 1976, p. 51). But the concept of a healing context seems to be an abstract concept not open to this possibility. We must con- clude that what was seen in the animal experiments was not the "pla- cebo effect" as we have defined it. It may still be the case, however, that certain limited features of the healing setting can become condi- tioned stimuli, evoking responses in both animals and humans. The problems with behaviorism are avoided by interactionism, since the latter theory explicitly includes mental terms. But inter- actionism gives rise to two problems of its own. One, already alluded to, is the puzzle of causality between the mental and the physical. We are used to accounting for causation in terms of one body exerting a force on another, or in terms of transfer of energy. But if one event occurs in a body which has mass and can move through space, and another event occurs in something nonsubstantial and nonspatial, it is hard to see how any causal connection could exist. 71 The notion of 'cause' used here is essentially a Newtonian one; and Gasking (1955) has suggested that this is not the primitive or the root meaning of 'cause.’ This primitive meaning he takes to be that of a recipe for producing a certain effect-- A causes 8 when one can pro- duce a state or event of the A sort as a means to producing a state or event of the 8 sort. The sense of 'cause' that appears in the Newton- ian or scientific context is properly viewed as a special case of this root sense.8 But the price we pay for adopting a looser sense of 'cause' is to give up the powers of explanation and prediction that ac- company 'cause' in the stricter sense. Still, the causality puzzle might be tolerable if there were not another serious problem with interactionism. If I consist of a mind plus a body, and if thought and consciousness are properties only of the former, it is quite possible for me to conclude that my mind is the only one that exists. I do in fact see many other persons, but I see only their bodies and never their minds; for all I know they may be cleverly constructed automata which have no thoughts or consciousness. But certainly the possibility that I could have grounds for thinking this runs counter to our basic considered judgments. It has been ar- gued that I know others have minds by analogy from my own case; but such a use of analogy would be inappropriate. Having seen, for example, the internal wiring and gears in many railway semaphores, I could con- clude by analogy that the next semaphore I encounter will have such an internal structure. But since minds have no physical substance, I could never in principle check out my assumption about other minds existing, 8. But see 5.2, below, for a refinement of Gasking's position, sug- gesting that there is no one "root sense" of causation. 72 in the way that I can check out a railway semaphore (Ryle 1949, pp. 51- 56). Interactionism, then, seems to take a considered judgment about which we feel firmly convinced (i.e., that other people have minds like ours) and to relegate it to the status of something we must take purely on faith and can never in principle be certain about. Any reasonably plausible mind-body theory that avoids this troublesome other-minds problem would therefore be preferable to interactionism. Identity theory, in turn, avoids the problems that attend both be- haviorism and interactionism; but it avoids these by postulating an iden- tity relation of a sort that requires considerable scrutiny. An impor- tant feature of the identity relation is that anything that can truly be said of one term of the relation can truly be said of the other. We can say "the Morning Star is identical to the Evening Star" because any property of the Morning Star (size, position in space, etc.) can truly be predicated of the Evening Star, and vice versa. But the mind-body problem has arisen precisely because things that can truly be said of mental events (nonspatial, noninferential access to our own, etc.) can- not be said of physical events. We might try to reformulate our con- cepts of physical and mental events to remove some of these differences, but we would succeed only by either "mentalizing" physical events or by "materializing" mental events (for example, by adopting a linguistic convention that allows us to locate mental events precisely in space). In either case, how one sort of event could take on properties of the other would be as puzzling an issue as how the mind is related to the body; so we would not have succeeded in clarifying the mind-body problem. Another feature of identity relations is that two things can be said to be identical only if they are of the same sort. This follows 73 from the way that we define physical space-- two things of the same sort cannot be in the same place at the same time, unless they are identical. Two things of different sorts can occupy the same space at the same time-- a tree may be in the same place as an aggregate of cellulose molecules. But in this case we would say that the tree is constituted of cellulose molecules, not that the tree is identical to the aggregate of molecules. For one thing, we can truly ascribe properties to the tree that we cannot to the molecules, and vice versa; we can, for instance, talk of the mean kinetic energy of the molecules but not of the tree. Also, the tree and the aggregate of molecules have different conditions for survival through time. If the tree is cut up into logs the aggregate of molecules survives but the tree does not; if the tree is pruned and the clippings burnt, the tree survives but the aggregate of molecules does not (Wiggins 1968). Thus, for two things to be identical there must be some "sortal concept" that applies to both; for the Morning Star and the Evening Star it is the concept "planet." The sortal concept is important be- cause it tells us where to look to see if the identity statement is true or not.9 To see whether the Morning Star is identical to the Eve- ning Star, we first trace one planet through space, then the other, to see whether we have traced the same planet or two different ones. But what sortal concept could serve this function for mental events and physical events? It cannot be a very vague concept such as "event" or "phenomenon"; because then we have no clear idea what to trace. Some- thing that occurs, such as raising my arm, could be viewed equally well 9. I owe my understanding of this refinement of Wiggins' analysis to an unpublished paper by Martin Benjamin. 74 as one event or many events, depending on our purposes (it could be one arm movement, or the simultaneous contraction of many muscles). But if the sortal concept is made definite enough to trace through space or time, it would have to take on either physicalistic or mentalistic properties, and hence would not apply equally well to the two terms of the identity statement]0 We must conclude from this that the proposed mind-body identity assertion, despite its straightforward appearance, in fact conceals a number of sticky problems; it is not at all clear that the assertion is a coherent or meaningful one. We have thus found serious problems with all three of the mind-body theories that seemed initially plausible. But this does not rule out the possibility that one or more of them could be modified so as to avoid some of the criticisms. By making some major modifications in behaviorism, on the one hand, or identity theory, on the other, one can arrive at a position called eliminative materialism, which agrees well with the considered judgments noted above and which is immune to several of the criticisms we have listed. This will be the focus of discussion in the next chapter. 10. James Cornman, "The Identity of Mind and Body," in (Borst 1970, pp. 123-129) argues for "cross—category" identity, such as "the tempera- ture of the gas is identical to the mean kinetic energy of its molecules." But if the identity is truly cross—category, there can be no common sor- tal concept, and the identity statement is incoherent. Indeed, in Cornman's example, "identical to" seems strained at best; "directly proportional to" is much more natural. Chapter 4. Eliminative Materialism The refutations offered in the previous chapter for behaviorism and for identity theory will not satisfy many defenders of these the- ories, who might object that we have looked at these theories only in their weakest forms. More recent authors, it will be stated, have mod- ified these theories so as to make them immune to refutation on the grounds we have mentioned. This chapter will be largely devoted to an analysis of this claim. The first section looks at features of these modified theories, under the title of "eliminative materialism," once again borrowing from Kuhn's idea of paradigms as stated in 2.1. The second section attempts to provide arguments against eliminative materialism as a satisfactory mind-body theory. One argument in favor of eliminative materialism is that the only alternative to such a theory is one of the types of dual- ism that were found to be unsatisfactory in the previous chapter; so the final section will show that another alternative is available, by going outside of the Cartesian tradition. The alternative approach, Strawson's concept of 'person,' will provide a logical framework for more detailed development in the next chapter. 4.1. Features of Eliminative Materialism What I will be calling eliminative materialism has developed out of identity theory and behaviorism, in response to some of the criticisms mentioned in the previous chapter. Identity theorists, noting the 75 76 failure of attempts to translate mental-state talk into brain-state talk, and observing the sort of ad hoc reformulations of mental and physical characteristics that would be required to make the identity assertion appear coherent in its original form, have moved to a "dis- appearance form" of identity theory. On this view, as we learn more about the neurophysiology of the brain, we will simply adopt the lan- guage of science in talking about our own internal experiences, and traditional mentalistic talk will "disappear." Instead of saying, "I have a pain," we will say, "My C-fibers are firing"; talk about pains will drop out of our language in the same way that talk about demons has dropped out of our talk about disease. And the new language will offer greater economy, as the same terms which we will use to describe our everyday experiences will also function in scientific observation and theory-building.1 A similar advance has been made in behavioristic thinking. As operant-conditioning theories have become more sophisticated, views of what is to count as "behavior" have broadened to include various "inner" bodily states, and the past history of the organism has been taken into account along with present states. An example of such a sophisticated theory is the "radical behaviorism" of B. F. Skinner (1974). A follower of Skinner, for example, would argue that in refuting classical conditioning as a plausible placebo theory (3.4), we have in effect demolished a straw man, since psychologists have long since 1. Representative papers on the "disappearance form" are Richard Rorty, "Mind-Brain Identity, Privacy, and Categories," and Paul Feyer- abend, "Materialism and the Mind-Body Problem," (Borst 1970, pp. 187-213, 142—156). 77 abandoned classical conditioning for the more refined operant condi- tioning. An operant-conditioning account of the placebo effect might go something like this. Suppose that there is a certain bodily state (analogous to alpha rhythm, for example) such that the self-healing powers of the body work best when the body is in that internal state. Upon repeated exposures to the healing context, achievement of that state will be positively reinforced by quicker relief of symptoms. Over time, the individual might become conditioned to achieve that state upon being presented with the healing context as a stimulus; this will occasionally result in relief of symptoms even if no active in- tervention is given in the healing encounter. Further, instead of asking the circumstances needed for the individual to believe that he is in a healing context, we might ask about the degree of "stimulus generalization" present in this case of conditioning; the latter ques- tion is open to precise study and quantification.2 Despite important differences, it is useful to consider the "dis- appearance form” of identity theory and radical behaviorism together. First, it must be seen that the “disappearance form" is really no longer a form of identity theory at all. Using the demon analogy, replacing demons as the purported causal agents in disease with a germ theory is not to say that demons are identical to bacteria; it is to say that, in the past, when we talked about demons, we were hopelessly confused; and we should change our account so that it reflects the facts as we now know them. In both radical behaviorism and the disappearance form, the suggestion is made that we eliminate our traditional mentalistic 2. I am indebted to Joseph Hanna for pointing out to me the possi- bility of such an account. 78 talk in favor of language that (it is asserted) is more scientifically correct. (Radical behaviorism does attempt to give new meanings for our present mentalistic terms and advocates retaining such terms in the language; but the change in meaning is so drastic that it amounts prac- tically to eliminating the terms as we use them.)3 And this new lan— guage will be materialistic, in that it will make reference only to physical states and events and will seek to explain human behavior in terms of deterministic laws akin to the laws of physics and chemistry. Hence the title, "eliminative materialism," for the combination of both theories.4 Eliminative materialism must be understood as a radical reconstruc- tion of our notion of mental events, as contrasted to previously dis- cussed theories, which were rather attempts to explicate the notion. It is this radical-reconstruction feature that allows eliminative mate- rialism to escape the criticisms leveled at behaviorism and identity theory in 3.4. Behavioristic attempts to deal with the problem of be- lief states, for example, failed, because the behaviorist attempted to give an account that would capture all of what we presently mean when we talk of beliefs as mental states. And the identity theorist, in or- der to make his identity statement seem coherent, was tempted to try to 3. Obviously the radical behaviorist's use of mentalistic terms with radically modified meanings makes argument in this area especially difficult; and to some extent, as we will show below, the plausibility of the behaviorist's position depends on this ambiguity. This sort of language problem is a general feature of cross-paradigm debates in sci- ence (Kuhn 1970, pp. 198-204). 4. Further justification for combining the two theories is the fact that radical behaviorism seems especially strong in accounting for in- tentional states, while the "disappearance form" is most credible in dealing with sensations. 79 impose mentalistic features on physical events, or physical features on mental events. By being able to drop mentalistic talk completely, the eliminative materialist can avoid being backed into such corners. The objection, "But what you have just described doesn't include everything that is included when we talk about beliefs (or sensations, or thoughts, etc.)" is simply no longer relevant. What the eliminative materialist is proposing may usefully be com- pared to the idea of a paradigm shift in science (2.1). For example, to say that when chemistry adopted the oxygen theory of combustion in place of the older phlogiston theory, chemists adopted a new terminology, is to miss the actual extent of the revolution in thinking. There is an important sense in which the oxygen chemists were observing different data and studying different problems as contrasted with the older state of the science. Further, since one cannot work within a paradigm with- out accepting its set of basic presuppositions, cross-paradigm disputes are at least to some extent insoluble. Neither the oxygen chemists nor the phlogiston chemists could, in this sense, win over the other side by arguments as to the superiority of their theory, since they would in effect be arguing about two different things; each side could accuse the other of question-begging in the way that they have stated their theory. Replacing "I am in pain" with "my C-fibers are firing" represents a similarly radical paradigm shift. For instance, if we were to object that I can be mistaken about "my C-fibers are firing" while I cannot be mistaken about "I am in pain," the eliminative materialist would reply that the only reason we regard incorrigibility as an important feature of mental events is because we are totally immersed in our present men- talistic language. Our objection is analogous to one the phlogiston 80 chemist might raise: "Your oxygen theory is very nice as far as it goes, but it can't be correct because there is no room in it for the existence of phlogiston." However, despite the problems of cross-paradigm de- bates, we will see if some telling points against eliminative materi- alism cannot be raised. While eliminative materialism is a strong theory precisely because it engages in this radical reconstruction, its supporters may sometimes be tempted, as a debating tactic, to downplay this feature of their theory, and to talk as if eliminative materialism were, after all, nothing but a minor modification of identity theory. After all, if the Morning Star is identical to the Evening Star, we have nothing important to lose by agreeing to call the planet by one name whether it appears in the morning or the evening; we do not have to give especially strong reasons for making this shift in language. Similarly, the elimi- native materialist might play upon the confusion of the "disappearance form" with identity theory proper, to convince us that replacing "I am in pain“ with "my C-fibers are firing" is a similarly inocuous termino- logical shift. But, as we will be arguing, we cannot let the elimina- tive materialist off the hook so easily when a radical paradigm shift is at stake. It will not do to say that elimination of our present men- talistic language ought to be carried out simply because such an elimina- tion is conceivable. Another point in favor of eliminative materialism is that the con- sidered judgments listed in 3.3, which individually supported behavior- ism, interactionism, and identity theory, respectively, combine mutu- ally to support eliminative materialism. The theory accounts both for the emphasis on behavior in determining mental states, and the 81 importance of neurophysiological discovery in elucidating the "mind." And, assuming that the elimination of our mentalistic language can be carried out, we will be free to recognize the reality and the causal efficacy of the firing of our neurons. Furthermore, as already noted, eliminative materialism seems to be immune to the objections raised against the other three theories. Finally, and importantly from the medical standpoint, eliminative materialism, in calling for use of a more scientific language and for reduction of psychological explana- tions to deterministic and materialistic explanations, seems consis- tent with trends in contemporary medical science. Given its sophisticated nature and the problems of cross-paradigm debate, it would seem difficult to launch a strong attack against elim- inative materialism. The next section will take up this matter. 4.2. Objections to Eliminative Materialism Essentially two kinds of arguments can be raised against elimina- tive materialism. The first tries to make it seem less plausible that brain-state talk could replace mental-state talk as readily as we might think; this type of argument does not directly confront the radical paradigm-shift feature of the theory. The second does confront the paradigm shift, and asks directly what we stand to gain or lose by making it. In particular, we might challenge the materialist emphasis on the purported scientific advantages of making the shift; in line with our reflective-equilibrium approach, we might demand that moral considera- tions be taken into account as well. Replacing talk about mental states with brain-state talk seems most likely to succeed if there is a one-to-one correspondence between brain 82 states and what we now call mental states; at the very least, there ought to be a many-one correspondence, with any one of a set of brain states corresponding to a single mental state. But, at least with re- gard to intentional states, this "correspondence hypothesis" seems highly questionable. Goldberg (1968) takes the example "thinking about George Washington." Suppose a teacher asks three students to write down the name of the first President. A hears the question and immediately writes the words, "George Washington." 8 first has a men- tal image of the picture on the dollar bill, then recognizes it as the face of George Washington, and writes "George Washington." C has the same mental image as B, but fails to recognize the person by name, and so writes down nothing. If we are asked which students were thinking of George Washington, we must say that A and B were; although C had the right mental picture, we would not want to say that one can be thinking about George Washington and simultaneously not know that one is thinking about George Washington. At best, C was thinking about a picture of George Washington. But if we ask which students had the same thing go- ing on in their heads (and presumably in their brains), the most likely answer is B and C. Thus we might well doubt that the relation between brain states and mental states is of the sort that makes the eliminative- materialist program a likely venture.5 Another line of argument can be raised against the suggestion that descriptions of the structure of the brain can replace psychological 5. The materialist might weaken this example by suggesting that it does not make us doubt the existence of a one-one or a many-one correspondence; it merely makes us doubt our ability to articulate it in ordinary language. 83 explanations with mentalistic content. Fodor argues that a psycholog— ical explanation must consist of both an analysis of behavior in func- tional terms, and a description of the underlying structure or mechan- ism that makes the behavior possible. Describing only the mechanism will not suffice, because for any functional description, there are an indefinite number of mechanisms capable of producing it (Fodor 1965). It so happens, for example, that the firing of C-fibers is the mechanism that corresponds with pain sensation in all human beings studied to date. But we could easily imagine some elaborate series of switches and wires that could be implanted in a body to serve the same function; and the number of different mechanisms we could postulate would depend only on our ingenuity. There is no necessary connection between the functional description and any specific one of these functionally-equiv- alent mechanisms. Further, a description of one such mechanism only would be merely a description of the interactions among the parts of the mechanism, and "would fail to describe the role of these interac- tions in the production of behavior" (Fodor 1965, p. 177).6 With regard to this point a moral consideration arises as well. We might imagine making contact with creatures from outer space, who might turn out to have psychological states analogous to our own, inso- far as we could tell from their conversation and other behaviors, but whose bodily physiology might be completely different: Are we to re- gard these creatures as our moral equals, and accept moral duties not to kill them or to cheat them? Or do we regard them as wholly alien 6. Fodor here does not deny that functional explanations play a role within neurophysiology itself; but vis-a-vis the behavior of hu- man beings, all neurophysiological explanation, whatever its internal form, is mechanistic. 84 life forms, refusing to believe that they could have psychological states similar to ours because their structural form is so different? Eliminative materialism seems to steer us toward the latter course.7 But these lines of argument will be rejected as irrelevant by the committed eliminative materialist. Both the correspondence argument and the functional-explanation argument, he might insist, still mis— perceive the radical paradigm shift, and reply to the materialist as if he were trying to give an explication of our traditional mental- state talk. Once the task is seen as one of radical reconstruction rather than explication, it will be seen to make no difference what the brain states correspond with, or what form psychological explana- tions ought to take. Thus we have to face the materialist proposal head-on, and ask what it would be like to make the required paradigm shift. We might raise two problem areas-- what sort of attitude we would have to adopt towards ourselves, and what sort of attitude we would have to adopt towards others. Following Wittgenstein in taking a language-system to constitute a "form of life" (1958, I, 241), we could ask how the form of life under the materialist program would dif- fer from our present one in these two respects. Since talk in neurophysiology and in operant-conditioning psy- chology is essentially the talk of spectators witnessing an event with- out participating in it, the new life form would involve looking upon one's own inner states only in the role of spectator-- or, put another way, regarding our present and future behavior in the same way that we 7. This argument applies to the "disappearance form" but not to Skinner's radical behaviorism, to the extent that the latter includes functional explanations of behavior. 85 regard our past behavior. This new life form precludes the role of being an agent in the world, and certainly undermines our moral thinking; in what sense can we be said to be responsible for our future behavior if we have no more control over it than we have over our past behavior that is over and done with? And to the extent that being a scientific observer presupposes the subjective experience, and the agency, of the "I" who is doing the observing, this life form undercuts scientific thinking as well (Platt 1972).8 The eliminative materialist might reply that there is nothing new about any of this. We have already been forced to reexamine our moral thinking as we have become more knowledgeable about how we are con- ditioned by our environment and our early upbringing. But here the materialist is waffling between two positions-- on the one hand he is claiming that his position represents a radical paradigm shift; on the other he is claiming that his position is merely a logical extension of features of our present paradigm. But he cannot have it both ways. It is true, within our present paradigm, that we have had to reconsider the scope of our free agency in light of new knowledge of conditioning, unconscious impulses, and the like. But such a reconsideration still presupposes the possibility of free agency as a background condition; without this condition moral discussion would simply make no sense. The radical paradigm shift would remove the very possibility of free agency; it would not be merely an extension of our present moral think- ing. 8. Support for Platt's assertion comes from the trend in philos- ophy of science to view observation as inherently theory—laden, thus emphasizing the role of the scientist as an active participant in what he studies (e.g., Hanson 1958). 86 Skinner is guilty of this waffling when he advocates reforming language by eliminating mentalistic terms such as "freedom" and "dig- nity," and reforming life by more conscious use of behavior-modifica— tion techniques (Skinner 1971). This call for reform seems to suggest a role for choice and action within Skinnerls world view, when in fact Skinner's deterministic metaphysics makes such choice and action, in the sense that we speak of them, impossible. Skinner himself would say that we do not choose to accept his reform proposal, if we do so; we rather are caused to adopt it by a pattern of deterministic rein- forcers. One might claim that this is simply "choice" and "free ac- tion" as interpreted within the new paradigm; but if so it is not at all clear that the new paradigm leaves any room for moral thinking. These points are reinforced by looking at the attitudes towards others that the new paradigm would have us adopt. Strawson (1968) notes two different types of attitudes that we presently adopt towards others under our present form of life. First, there are what we might call participatory attitudes, which are responses to the atti- tudes that others have towards us. These include attitudes such as re- sentment and gratitude, which in turn are closely bound up with the more general attitudes of moral indignation and moral approval. We also have what might be called objective attitudes, which regard others as things to be manipulated rather than as persons. Towards some special classes of humans (e.g., small children and the insane) we have objec- tive attitudes all the time. We also on occasion have objective atti- tudes towards some normal individuals, for purposes of scientific in- quiry, or for furtherance of policy, or simply as a respite from the emotional involvement that accompanies participatory attitudes. But, 87 as these examples show, when we do adopt objective attitudes towards others we do so fbr particular reasons. Participatory attitudes, by contrast, are the norm for human encounters; when we have such attitudes it does not make sense to ask for the reasons why we do. (That is, we might ask for reasons why one has one participatory attitude and not another-- "Why did you have such strong resentment to such a silly in- sult?"-- but not fbr reasons why one has participatory attitudes, in general, instead of objective ones.) Strawson then argues that a deterministic thesis, of which elimina- tive materialism is an example, would require us rationally to adopt objective attitudes towards all people at all times, in effect giving up participatory attitudes completely. However, all interpersonal re- lationships as we know them, aside from purely instrumental relation— ships, are based on the context of participatory attitudes that makes up the norm for our form of life; and to suspend all participatory at- titudes as the deterministic thesis would require is to remove the pos- sibility for interpersonal relationships. To think that we could even have a choice in this matter is grossly to misperceive the nature of our human commitment to the form of life that we presently live. It is to think that somehow the universal context of participatory attitudes can come up for review, in the way that we can review specific instances of application of these attitudes. We will, it is true, revise our at- titudes toward a burglar once we learn that he was motivated by klepto- mania; but we cannot in the same way revise our views on whether we should have participatory attitudes at all. Strawson compares this commitment to participatory attitudes to our commitment to inductive reasoning (1968, p. 94). Could we give up 88 inductive reasoning? Induction pervades our form of life, influencing us every time we pick up the phone when it rings, confident that there will be a voice at the other end; and every time we turn the page of a book, confident that the printing will continue on the next page. We could speak of doubting whether induction is justified, but this would be mere verbal expression of such a doUbt; we have no idea how to live our lives except in a way that presupposes the validity of inductive reasoning. But none of this restricts us from questioning specific uses of induction, or from trying to revise and refine our rules for apply- ing inductive principles to specific cases. But if this statement of the nature of our commitment to our pres- ent life form does not impress the materialist, let us suppose that we have become somehow able to make the choice between our present way of life and life under the materialist's new paradigm-- the choice, that is, that we have just argued is outside of our ability to choose. We would then have to choose either to continue with our present life form, or to make the radical paradigm shift. Presumably we would want to ar- gue this choice based on the gains or losses involved in the change; and presumably the eliminative materialist would want to argue that the gains outweigh the losses. But what do we have to tell us what counts as a gain and what as a loss, except the background context of our in- terpersonal relationships? Our notion of benefit and loss presupposes that background context. The materialist, for example, tells us that it would be more “rational" if we were to make the shift and adopt ob- jective attitudes to the exclusion of participatory attitudes. But our concern must be not with what is "rational" in the abstract, but with what is to count as rational behavior towards others; and our 89 interpersonal behavior is rational or not depending on the nature of our interpersonal relationships-- it is rational to act towards my wife in ways that it would be irrational to act towards a supermarket clerk. But again, our relationships presuppose the background context of participatory attitudes, of our attitudes towards others and others' attitudes towards us. It seems that the materialist cannot even join in this debate over gains and losses without implicitly accepting the framework of participatory attitudes, and the life form, which he is urging us to dispense with. And this, in turn, lends further support to our previous conclusion, that giving up our commitment to this life form cannot be a matter for rational choice. The eliminative materialist still has a reply. It seems as if all this talk about background contexts and forms of life has created a smokescreen around what the materialist originally wished to claim. And that is simply that if we are confronted with an organism whose be- havior is determined in lawlike ways by its internal physiologic func- tions and the stimuli that it receives from the environment, our atti- tudes towards it is rationally what Strawson calls objective attitudes. This conclusion seems completely plausible. Furthermore, "rational" is a mentalistic term and also has to be radically reconstructed to fit the materialist paradigm. Under this paradigm, rational behavior would simply mean behavior that enhances the survival probability of the in- dividual or the group. But as soon as we investigate the plausibility of this reply more closely, we find that the plausibility rests on our implicitly assuming that our encounter with such an organism takes place as a special case against the background of our normal human relationships-— indeed, that 90 is the only way that we could conceive of such an encounter, given the human commitment we spoke of earlier. Thus in understanding how we would react to such a case we are dependent on the background context, just as we cannot understand kleptomania as a special case unless we first understand theft as a free action done for reasons and motives. Again, as Strawson says about our commitment to inductive reasoning, we can argue about the rationality or irrationality only of our judg- ments about specific cases. Our commitment to the universal background context of participatory attitudes is non-rationa1-- it precedes and underlies our criteria for determining rationality or irrationality. The eliminative materialist always has a final reply-- since what he is proposing is a radical paradigm shift, it is hopeless to argue with people who are so habituated to the old paradigm and its way of thinking that they can see no alternative. But, if we can force the materialist to adopt this as his final word in the matter, his position becomes much less plausible. For originally he seemed to be proposing not only that a radical paradigm shift ought to be made, but also, more importantly, that he could give good reasons for making the shift. If such reasons were to be both relevant and persuasive, they would have to be of such a nature as to bridge the gap between the two paradigms; they would have to show us, in effect, a way to make the transition in our thinking. And we have now seen that no such reasons are forthcom- ing. The reasons proposed are either completely foreign to our way of thinking and hence fail to persuade us; or else they are dependent upon our present paradigm, and hence give us no reason for making the shift. This conclusion can be illustrated by a diagram: 91 reasonsc present paradigm a material i st paradigm reasonsa reasonsb We can give many reasonsa to explain why we adhere to our present para- digm, such as pointing out how it allows for moral agency; and the materialist can give many reasonsb for his paradigm, such as its use- fulness in terms of psychological research. But reasonsa and reasonsb each show only the internal consistency of the respective paradigms; reasonsb can never convince us to make the radical paradigm shift, and reasonsa can never suffice to refute the materialist paradigm. If we could somehow imagine people living under a life form like that of the materialist paradigm, and who were considering the radical paradigm shift to our own present life form, our reasonsa could never suffice to persuade them to make the shift, any more than the materialist rea- sons can conclusively persuade us. The only reasons which speak directly to the paradigm shift are the "bridging" reasonsc. It is this sort of reason which, we have just been arguing, does not exist-- at least, none of the reasons pro- posed by the materialist have been found to qualify. Since reasonsc are the only reasons that could possibly show why we should make the paradigm shift, the materialist cannot speak to us of "should." All he can say is that his paradigm is "rational" in the sense that it has survival value, and that groups that adopt the new paradigm will sur— vive while those that adhere to our present paradigm will die out. But this is an empirical claim; all we can do is wait and see. 92 If the materialist, then, cannot give reasons for the paradigm shift, he can really say only two things about his proposal-- first, that it is not logically impossible; and second, that as science pro- gresses and subtly changes the way we look at and live in the world, such a shift may come to pass. But the same two things can be said about building a bridge to the moon. If this is all the materialist has to say, we may yet choose to make the shift he proposes, but it is then clear that we cannot be said to do so for sound reasons (i.e., for reasonsc if any existed). At most we would be doing it as a strange sort of leap of faith-- strange because, unlike religious faith, the change undercuts rather than supports many of our most basic con- sidered judgments. This, then, concludes the various arguments that may be raised against eliminative materialism. The arguments have raised doubts about the theory, but we can hardly claim to have refuted it-- indeed, being the sort of paradigm shift that it is, it seems immune from refutation in any ordinary sense. Therefore, eliminative materialism remains a strong theory which we would probably be willing to adopt by default in the absence of any attractive alternatives. In particular, many mate- rialists seem to defend their views with special vigor because they feel that theirs is the only realistic alternative to the troublesome Cartesian dualism. Therefore, if it is possible to formulate a theory which is consistent with all the considered judgments previously lis- ted, but which avoids some of the problems that eliminative materialism raises, an important reason for adopting eliminative materialism would disappear. To see what such an alternative theory would look like, it is necessary first to challenge some of the fundamental presuppositions 93 of the Cartesian approach to the mind-body problem. 4.3. The Concept of Person All of the theories that we have investigated so far share the Cartesian assumption that "mind" and "body" are the two primitive terms by which other phenomena must be explained. Even though it represents a radical paradigm shift in other regards, eliminative ma- terialism rests on this assumption nonetheless-- it assumes that the realm of mind can be eliminated completely and that all that will be left will fall into the realm of body. One way to begin the search for alternative mind-body theories, then, is to challenge this assumption. This has been done very effectively by Strawson in his essay, "Persons" (1958). Essentially, Strawson's conclusion is that there are two kinds of predicates, mental and physical, and two kinds of entities, (mere) ma- terial bodies and persons. Material bodies can correctly have ascribed to them only physical predicates, while persons can have ascribed to them both physical and mental predicates. When one ascribes a mental predicate to another person, one does so in the same sense that one as- cribes it to oneself; and the "I" to which I ascribe mental predicates ("I am in pain") is the same "I" as the one to which I ascribe physical predicates ("I am six feet tall"). The important feature of Strawson's account is that the concept of person is logically required to be more basic than the mental and physi- cal predicates ascribed to it. To see why this is so, consider the posi- tion of the other-minds skeptic that we encountered in discussing inter- actionism (3.4). As a skeptic, I might want to speak of my pain, but I 94 am reluctant to admit that there exist any other minds that could have pains of their own; all I can actually observe are other bodies and never other minds. Thus it makes sense for me to doubt whether anyone besides myself has pains. But the coherence of this skeptical position presupposes that I have certain concepts. If I can speak meaningfully of my pains as dis- tinct from others' pains, my concept of mental predicates must be that of something ascribable in principle to a class of individual entities of the same logical type. But I can ascribe mental predicates to an entity only if I can identify that entity as an individual; and I can do this only if that entity exists in the physical world. Thus, for anything at all to be the subject of mental experiences in the sense that my skeptical argument requires, there must exist individuals of this unique type of having both mental and physical predicates ascriba- ble to them-- that is, persons (Strawson 1958, p. 342). Only by first having the concept of person can I move by abstraction to the concept of pure mind or pure consciousness (p. 341). I The power of Strawson's position lies in the fact that it is not an argument for the primitiveness of the concept of person-- rather, it is an explanation of why no such argument is needed. For the Cartesian dualist, in order to state his skeptical refutation of the existence of persons, has to presuppose the very concept he wishes to refute (p. 349).9 9. It will not do for the skeptic to claim that he accepts the no- tion of person, but doubts whether the other bodies that he observes are persons. Persons, which are not mind-plus-body in the Cartesian sense, can be adequately known and identified through their bodies. To take this position, the skeptic ceases to be merely an other-minds skeptic and becomes skeptical about the reality of the physical world. 95 Thus, by looking at the concept of person instead of at mind and body, we might avoid both a dualism in which two very dissimilar things must be brought together (interactionism), and a monism which seems to gloss over significant features of reality (behaviorism, identity theory, eliminative materialism). Moreover, if we accept Strawson's account as a very strong and a very basic way of refuting the Cartesian-dualist position, we have further grounds for refusing to adopt eliminative ma- terialism. The materialist, after all, feared that if we allowed men- talistic terms to remain in our language, we would inevitably slip back into the unsatisfactory position of dualism. But Strawson's power- ful argument rests precisely on the logical features of mental predi- cates; so that losing such predicates from our language would deprive us of this very strong argument against dualism. However, Strawson's account deals with the logic rather than the characteristics of the concept of person. The account is therefore of very limited use in elucidating the philosophical features of problems like the placebo effect. It can tell us, for instance, that the person who believes that he is in a healing context is the same person as the one whose body undergoes change-- that is, that the former is not men- tal substance while the latter is physical substance-- and that is cer- tainly something gained. But the fact that multiple predicates can be ascribed to the same individual tells us nothing about the relationships among those predicates, if any; and that is where the interesting ques- tions about the placebo effect lie. Furthermore, in looking at our other considered judgments, we find that the concept that Strawson has described does not fit well with our existing notion of 'person.’ We can truly ascribe at least some mental 96 predicates, such as sensations, to many animals as well as to human persons. It hardly seems consistent with our usual use of 'person' to call these animals persons merely because we can ascribe sense-con- sciousness predicates to them (Frankfurt 1971). The challenge for the next chapter, then, is to develop a concept of person that has the logical features noted by Strawson, but which avoids some of the shortcomings of his position and which is more il- luminating for the placebo effect. If this can be done, the new theory must then be tested against the sorts of considered judgments noted in the previous chapter. We can then compare the degree of overall "fit" for this new theory and for eliminative materialism. Chapter 5. A Theory of the Person Strawson's account of the concept of person indicates some logical features that a theory of the mind must possess to avoid both Cartesian dualism and reductionistic monism. The concept could be fleshed out in a number of ways, each yielding a "person theory" with distinguishing features. One such person theory, based on statements about mind by Anthony Kenny (1973a) as considerably amplified by Marjorie Grene (1976), will be discussed at length in this chapter.1 The first section offers a formal reconstruction of the Kenny- Grene position. The next section shows how this person theory can be applied to the placebo effect. The following section considers the theory in light of some of the basic considered judgments mentioned in the previous chapter; and the last section lists some of the problems that the theory raises. 5.1. The Capacity Theory of Person Essentially the Kenny-Grene theory holds that persons are animals possessing a particular capacity, the ability to use symbols in special ways. I will refer to the theory as the "capacity theory" for short. 1. Grene sees her amplification of Kenny's account as arising from the tradition of Continental philosophers Helmuth Plessner and Maurice Merleau-Ponty, and from the epistemology of Michael Polanyi. Since the Kenny-Grene account can stand on its own, I have not attempted to in- vestigate these sources or other matters relating properly to histor- ical background rather than to the theory itself. 97 98 The theory can be stated in terms of three major assertions, with some explanatory notes. (1) Animals have capacities; different sorts of capacities require different sorts of explanations. 'Capacity' here is left essentially undefined by both Grene and Kenny, although we shall later consider some distinctions Kenny makes between a capacity, its exercise, and its vehicle. Grene gives three basic types of principles which may be used to explain various capaci- ties of animals. (a) Animals, as bodies in physical space consisting of chemical substances, obey the laws of physics and chemistry; and some of their rudimentary behavior can be explained in terms of these laws. (b) Animals can also exhibit goal-directed behavior which must be explained in terms of teleological organizing principles. We have already reviewed arguments to show that psychological explanations can- not be reduced to explanations of structure without function (4.2). Moreover, even an inanimate machine, such as a clock, cannot be under- stood as a machine unless reference is made to such functional organi- zing principles-— the laws of physics and chemistry can explain it only as a collection of masses and substances, and cannot explain it as a time-keeping deviCe (Polanyi 1958, p. 330). Grene further distinguishes first-order goal-directed behavior, in which an animal pursues a goal which is in effect already given, from second-order goal-directed be- havior, in which the animal is able to choose among alternative goals. A similar distinction is made by Frankfurt (1971), who speaks of first- order desires, whose objects are to do or not to do certain things, and second—order desires, whose objects are to have or not to have certain 99 (first-order) desires. Many animals have the capacity for the former, while only persons have the capacity for the latter.2 (c) Another sort of organizing principle that can explain certain animal capacities is not teleological in form, but rather is normative or typological. A key example is the use made of species and species resemblance in biology. The explanations given are in terms of adher- ence to or deviation from certain descriptive norms or types. The basic point of this enumeration is the potential richness and variety of animal capacities. We come to think of animal capacities as a restricted class of phenomena by noting only capacities that can be explained by one sort of principle and forgetting the applicability of different sorts of principles. (2) Some animals have the capacity to acquire the ability to use symbols in such a way that the use confers meaning upon the symbol. This description leaves out computing machines, for example, which are able to process symbols but for which the meaning of the symbols de- pends on outside personal agency. Grene stresses that this capacity should not be viewed as different by some order of magnitude from other animal capacities-- that is, having this capacity is a special way of being an animal in the world, not something "extra" added onto the ani- mal nature such that the animal is no longer "only" an animal. Kenny 2. This observation and what follows leave open the possibility that non—humans, such as chimpanzees who have learned sign language, creatures from outer space, or highly sophisticated machines, could be persons. I accept this, but since our major concern is with hu- mans, I will use "human" and "person" interchangeably. The arguments that follow also suggest that these non-human persons would necessar- ily be tied to their bodies in a way analogous to humans. lOO emphasizes that the notions of 'symbol' and 'symbol-using' have fuzzy boundaries and in fact are infinitely open, which indeed corresponds to the way that we use 'mind.‘ What connects different instances of symbol-using is a "family resemblance" as Wittgenstein used the term.3 Being a symbol—user in this way has two important implications. First, an animal with this capacity is at least a potential dweller within culture and within a language system. Second, an animal with this capacity, as Kenny points out, also has a capacity for responsible behavior. In order to choose knowingly and responsibly among alterna- tive goals or actions, one must have ways of representing, through sym- bols, goals or actions not immediately present. Also, to use symbols in such a way that they acquire meaning is necessarily to have purposes and pursuits, since, as Wittgenstein observed, meaning (for at least a large class of cases) is use within a language system; and language sys- tems, in turn, constitute "forms of life" (Wittgenstein l958, I, 43, 139; I, 241).4 (3) Animals with the capacity to use symbols in this way are per- sons. To be an animal with this capacity is to have a mind.5 3. Wittgenstein contrasted the idea of multiple resemblances, which "overlap and criss—cross" among members of a class, with the idea that there must be one essential element as the common denominator among all members. "The strength of the thread does not reside in the fact that some one fibre runs through its whole length, but in the overlapping of many fibres" (l958, I, 67). See also (Kenny l973b, pp. 153-163). 4. Frankfurt (1971) argues that what distinguishes humans from other animals is the capacity to form second-order desires; if the Kenny-Grene account is correct, it would seem that being a symbol-user in the proper sense is both necessary and sufficient for having second- order desires. This point requires further exploration. 5. Specifically: "To have a mind is to have the capacity to ac- quire the ability to operate with symbols in such a way that it is one's own activity that makes them symbols and confers meaning on them" (Kenny 1973a, p. 47). l0l To summarize, then, having a mind is a special way of being an animal in the world-- a way that entails participating in language and culture. By the capacity theory, the notion of 'person' cannot be grasped without realizing that it stands, in effect, with one foot in the biological realm and the other foot within culture and sociality. Our Cartesian assumptions tempt us to see these two feet as two dif- ferent parts of personhood; but since Strawson has proved the primitive- ness of the concept of 'person,' we must resist this temptation to fall back into dualism and its attendant problems. It is not dualistic, however, to recognize that different capacities of the person require explanation in terms of different sorts of principles. The capacity theory might, indeed, be seen as a sort of revival of double-aspect theory (3.2). If "mindedness" refers to a certain capac- ity that an animal has, and "bodily" refers to its other capacities, then we have a way of making sense of a statement that mind and body are two different aspects of person. The difference, of course, is that originally double-aspect theory was proposed within the framework of Cartesian dualism-- while not itself dualistic, the theory implicitly accepted the Cartesian formulation of the mind-body problem. The capac- ity theory, on the other hand, is a theory of the person having the logi- cal characteristics proposed by Strawson's fundamental critique of dual- ism (4.3). The concepts of "mindedness" and "bodily" as they apply to the capacities of persons are both necessarily derivative from the con- cept of 'person' itself, and no longer indicate a fundamental dualism. The capacity theory, instead of trying to answer the Cartesian question, tries instead to show that the question is erroneously framed. 102 5.2. The Capacity Theory and the Placebo Effect As we have done with other mind-body theories, we must ask how the capacity theory of person accounts for and illuminates the placebo ef- fect. So long as we had to deal with mind and body, we had difficulties bringing together the belief state of the subject and the changes in bodily condition, which, by our definition, had to be linked in order for the placebo effect to occur. Either a mental state had to be made to impinge in some suspicious way upon a bodily state (interactionism), or a mental state had to be reconstrued in different terminology, thus denying some of its crucial features (behaviorism, eliminative materi- alism). Furthermore, since the mind-body relation orients us toward consideration of the individual mind-body link, the crucial social and cultural dimension of human existence tends to be lost sight of. At best, this sociocultural realm is seen as an extension of the function of individual minds, rather than as in itself a central aspect of the human condition. The capacity theory of the person changes the picture considerably. Our subject who experiences the placebo effect is no longer a mind and a body, but is a person. Being a person entails having all the capac- ities of a biological organism, and in addition the special capacity to be a symbol-user and necessarily to be a dweller within culture. If being a dweller within culture is a special way of being an animal, it should not be anomalous if this characteristic were found to influence other animal capacities-- including the capacities to undergo changes in bodily status and function. Experiencing symptom change due to the placebo effect is therefore the bodily expression of the person's l03 participation in the healing context as a culturally determined, sym- bolic phenomenon. Of course, the mechanisms by which this symbolic-cultural event finds its bodily expression need to be studied empirically; the capac- ity theory cannot answer such questions on an a priori basis. The theory does suggest, however, that what is to be studied is the rela— tionship between various capacities of the person, not relationships between two radically different substances, or between two categorially different domains of meaning. We are still likely to need different sorts of explanations for different capacities-- the physical-chemical laws which explain tissue damage, for instance, will not suffice for explaining how culture influences the person's belief states. Perhaps we will even have to develop new, bridging principles to connect dif- ferent sorts of explanations, before our account of the placebo effect will be complete. But this problem is quite different from that of re- lating "mind" and "body" as traditionally conceived. The notion of the person as symbol-user also suggests an additional sense of 'cause' which may be operating in the placebo effect, and which has very different characteristics from the scientific sense of 'cause‘ mentioned in our discussion of interactionism (3.4). Kenny states, "[T]o use something as a symbol and not as a tool is to use it in such a way that any effect which it may have on the environment lacks the immediacy and regularity characteristic of physical causality" (l973a, p. 47). Since using symbols in and on the world has definite effects even though 'cause' in the physical or scientific sense is not appli- cable, we might want to speak of a sociocultural sense of 'cause.’ Consider the way in which a "no parking" sign might be said to 104 "cause" certain behaviors of motorists. For some motorists, no-park- ing behavior will follow, and will be the result of, an inspection of the sign, even though in the vast majority of cases the inspection will take the form of immediate recognition and will not give rise to any train of reasoning. For other motorists, probably those most familiar with the neighborhood, no-parking behavior will occur without looking at the sign, and indeed without any overt or conscious awareness of the sign's existence. Some motorists will park at that spot, and of these, some will receive tickets and some will not. Some will engage in a sort of compromise no-parking behavior, perhaps parking for briefer periods than they would otherwise. Clearly the way in which the sign might be said to "cause" any or all of these is very different from the way in which the sign could be said to cause a shadow to be cast on a sunny day. Of the varied effects produced by the sign, none occur with the predictability or the regularity we expect of physical causality. And the sorts of things that would count as counterexamples for physical causality do not apply-- even if there were cars parked by the sign more often than not, we would not want to deny its significance or its import. But still, all of the varied behaviors that may occur, either conforming to the no-parking norm or deviating from it, are readily explainable. If more people park by the sign this week than last, and fewer are ticketed, we might explain this as the result of a police strike. Our explanation would draw heavily on the past histories and prior states of the individuals involved, and upon unforeseen present circumstances; and this fits well with the probabilistic nature of the behavior that we actually observe. 105 One might object to using the word 'cause' at all in such circum- stances. Don't these sociocultural cases (the sign "causing” no-park- ing behavior; an argument "causing" someone to change his mind) lack the constant, or at least statistical correlation between cause and effect which is a minimal necessary condition for ascribing causality? But it is not clear that there is some one central sense of 'cause' such that this condition applies. Feinberg notes that purely empiri- cal investigation normally yields an indefinite number of "causal fac- tors" connected with an event. In giving a causal explanation we are forced to select one or a few of these factors; and the grounds for selection depend on our purposes in seeking the explanation. These purposes may include satisfying our intellectual curiosity ("What causes the tides to rise and fall?"), making practical changes in the world ("What causes automobile fatalities?"), and ascribing moral re- sponsibility ("What caused the death of the innocent bystander?"). Our purposes will determine the criteria we use to judge the accepta- bility of a proposed causal explanation (Feinberg l970, pp. 201-207). Gasking (1955) claimed that the "recipe" sense of 'cause,‘ which corresponds to the second of Feinberg's three purposes, is the primi- tive or root sense. But none of the three purposes seems to be neces- sarily more basic than the others. The thought that there must be some root sense of 'cause,‘ either Gasking's recipe sense, or the Newton- ian sense, or some other, arises from looking at the causal ascription in isolation from the various human contexts in which it can arise. Thus, it is a mistake to assume that there must be some one common ele- ment, such as constant correlation, connecting all uses of 'cause'; again, a family resemblance is all that is needed (Note 3, above).6 106 The sense in which a culturally-designated healing context can cause changes in symptoms may be seen as the same sociocultural sense crf 'cause.‘ This is also the sense implied by the term "sociosomatic," vvhich Kleinman (1973) employs to describe medicine's "symbolic reality." [Elsewhere, in a study of native healers in Taiwan, this author states: But our argument is that providing effective treatment for dis- ease is not the chief reason why indigenous practitioners heal. To the extent that they provide culturally legitimated treatment of illness, they must heal (Kleinman and Sung 1976).7 I