Dissolving the Placebo

Franklin Miller at the NIH and I are working on a new approach to the placebo and nocebo effects.  The eventual goal is to argue that the guidelines for use of placebo in both research and treatment context should be rethought, but that’ll come further down the line.  In this paper, we argue that the concept of placebo should be dissolved because thinking of it as a unified phenomenon, rather than a loose collection based on family resemblance, fails to appreciate the diversity of things referred to as placebo.

As always, comments, questions, and criticisms are most welcome.


1 Introduction


The patient visits his doctor with a complaint of fatigue.  After routine examination, she judges that there is no serious health problem.  She recommends taking multivitamins and calling back in two weeks.  When the patient calls back, he reports feeling less fatigue and more energy.  What are we to make of such a sequence of events?


First of all, we should be unsurprised.  Primary care providers only rarely prescribe pure placebos, such as sugar pills, but they often prescribe impure placebos, such as vitamins, homeopathic remedies, and antibiotics for (what they take to be) viral infections (Fässler et al. 2009, 2010; Nitzan & Lichtenberg 2004).


Perhaps we should also be disappointed: prescribing a placebo can undermine trust, as well as patient autonomy, especially if doing so involves deception.  Perhaps this case constitutes a gray area, but the vast majority of patients in palliative chemotherapy for metastatic lung or colorectal cancer believe that they are receiving curative treatment, indicating that care providers routinely fail to disabuse patients’ of misconceptions about their treatment and perhaps even actively engage in deceit (Weeks et al. 2012).  But placebos can produce their characteristic effects even without deceit Kaptchuck et al. 2010). And even if the physician doesn’t say outright, “This is a placebo,” there are ways of prescribing such interventions that are not clearly examples of deceit.  She could say, for instance, “I think this is a viral infection [true], but you can never be 100% certain [also true], so I’m prescribing antibiotics, just in case.”  Or she could say, “It’s not clear what will alleviate your symptoms [true], but a little extra vitamin C never hurts [also true], so try these vitamin supplements and see what happens.”


Perhaps we should cast a skeptical eye on the patient’s self-report.  He says that his fatigue has been alleviated, but is he sincere, or is he just saying what he thinks the physician wants to hear (response bias)?  Even more troubling, perhaps he is sincere but self-deceived.  He thinks that his symptoms have gotten better, but they haven’t (another type of response bias).  Or perhaps his symptoms have been tempered, but that would have happened anyway in the natural course of the illness.  Extremes tend towards the average over time; to pretend otherwise is to ignore regression to the mean.


Maybe, though, his self-report can be trusted, and the partial alleviation of his symptoms can be attributed to the placebo effect.  Its efficacy has been documented for a number of illnesses, including pain, Parkinson’s disease, depression, anxiety, and addiction, among others (Benedetti 2008).  Many of these illnesses involve a psychological component, and seem to be maladies, in a sense, of affect and attentional focusing.


But what does it mean to attribute the mitigation of symptoms to the placebo effect?  Here we enter thorny conceptual territory, as there is no consensus about what the placebo effect is or how it works.  It’s far too easy to fall into self-contradictory nonsense when trying to define the effect as, for example, what happens when an intervention with no causal power causes healing.  But other, more sophisticated, definitions likewise codify conceptual confusion, and the lack of adequate conceptualization stymies research in a variety of ways (Vallance 2006).


In this paper, we argue that the conceptualization of the placebo effect needs to undergo a paradigm shift.  We begin by distinguishing epistemic from metaphysical conceptions of the effect.  We then show that epistemic conceptions are fruitless and confused, but argue that unified metaphysical conceptions are explanatorily inadequate.  This dilemma is dissolved by dissolving the placebo effect itself: there is no such thing as the placebo effect because the phenomena we refer to under this rubric do not constitute a unified kind.  Talk of the placebo effect belies the heterogeneity of its mechanisms, which we group together only because of a weak family resemblance and our own positive evaluative attitude towards their characteristic outcomes.  If this perspective is sound, then the debate over whether or not placebo is powerful is fundamentally misguided (Hrobjartson & Gotzsche 2001, 2004, 2010; Vase, Riley, & Price 2002); it lumps together phenomena that should be considered and are distinct.  In the final section of the paper, we examine some of these phenomena and explore one – attentional focusing that triggers feedback loops and cascades – in more detail.  In so doing, we show how a pluralistic conception of placebo opens up space for the formulation and testing of new hypotheses and the reinterpretation of what previously looked like inconclusive or mixed results.

2 Epistemic versus metaphysical conceptions of placebo


In this section, we introduce a distinction between epistemic and metaphysical conceptions of the placebo effect.  We then argue that epistemic conceptions are fundamentally misguided.  While this suggests that metaphysical conceptions are preferable, they suffer from a related problem.  We think that a pluralistic approach dissolves this problem, however, and attempt to show how in the next section.


Epistemic and metaphysical conceptions of placebo are differentiated by their answer to the question, “What distinguishes standard medical healing from mere placebo effect?”  According to epistemic conceptions of placebo, the answer is that, in the case of mere placebo, we do not understand or know or have a roughly correct theory of how the intervention works, but in the case of standard medical healing we do.  Whether a given remedy counts as placebo is thus indexed to the contemporary level of medical understanding.  According to metaphysical conceptions of placebo, by contrast, in the case of mere placebo, a particular mechanism is implicated, but in the case of standard medical healing that mechanism is not implicated.[1]  What the mechanism is depends on the metaphysical conception.  The two most popular in the literature are patient expectations (sometimes supplemented by caregiver expectations) and classical conditioning, but others, such as anxiety reduction, signal detection, the ritual of care, and the caregiver-patient relationship, have been proposed.


2.1 The trouble with epistemic conceptions


Perhaps the most straightforward statement of the epistemic conception of placebo is due to Szawarski (2004; see also Grünbaum 1994), whom we quote at length:


If we can, by referring to scientific evidence, logically explain what the specific activity of a drug or therapy is in the case of a particular condition, then it is redundant to bring into the explanation any mysterious psychological or psychoimmunological factors. […] On the other hand, if we cannot scientifically explain the results of a treatment, then it is natural to refer to placebo.


One problem with this way of conceptualizing placebo is the scope of the ‘we’.  Who, exactly, needs to be able to explain how the remedy works?  Certainly the knowledge or understanding needn’t be common.  Most people have little or no medical training and do not regularly read the Lancet or the New England Journal of Medicine.  Neither need the patient understand how the remedy works.  Informed consent relies on providing the patient with some level of information, but rarely extends to “logically explain[ing]” by referring to “scientific evidence” the “specific activity” of the remedy.  Nor yet, though this may be more surprising, need the physician herself understand how the remedy works.  There’s presumably a division of labor when it comes to such knowledge, in which both patients and many physicians pass the buck to researchers and other experts, whose own understanding might be rather murky.


So is the referent of ‘we’ simply at least one expert?  This is better, but it still doesn’t work, since the physician who prescribes the remedy might not be connected to the expert by a suitable chain of reliable testimony or communication.  The ‘we’ must refer – if at all – to at least one expert in a suitably connected chain of reliable testimony or communication.


Once this point is made, it should be clear why epistemic conceptions of placebo inevitably lead to confusion, for they are covertly relativized.  Of two doctors who prescribe the same remedy at the same time, one might be prescribing placebo and the other a drug, if the latter but not the former is suitably connected to a knowledgeable expert.  And the same doctor could prescribe placebo today and – after getting connected to a knowledgeable expert – the same remedy as a drug tomorrow.  This stretches the bounds of sense.


It also adumbrates a problem that will arise as well for metaphysical conceptions of placebo: disunity.  Ignorance is a veil that covers manifold cognitive sins.  There’s no reason to suppose that what we don’t understand forms a unified class.  Hence, there’s no reason, in the epistemic paradigm, to talk about the placebo effect as if it were a unified phenomenon.  This spells trouble for one of the hallmarks of evidence-based medicine: the randomized controlled trial, in which a treatment is judged by whether it produces better outcomes than placebo.  If there’s no such thing as the placebo effect but rather a congeries of effects that we happen to lump together, then determining that a given treatment outperforms one placebo effect is no guarantee that it outperforms all of them, or even many of them.  Presumably what matters is that it outperforms all placebogenic treatments that are relevantly similar.  It seems inappropriate to pit a new analgesic pill against sham surgery, or to pit acupuncture against a sugar pill, especially now that we know that sham acupuncture produces more of a placebo effect than inert pills, and that high-interaction sham acupuncture produces more of a placebo effect than low-interaction sham acupuncture (Kaptchuck et al. 2000, 2006).  But what makes a placebogenic comparison relevantly similar has nothing to do with our ignorance; instead, this notion smuggles in a metaphysical conception of placebo.


Furthermore, epistemic conceptions of placebo have a pernicious institutional effect by discouraging research on the mechanisms that underlie placebo effects.  Imagine designing a research program aimed at explaining placebo.  It would be necessarily self-defeating: once you explain it, it’s no longer placebo.  Epistemic conceptions of placebo entail that the statement “I’ve explained the placebo effect” is necessarily false, and that it’s impossible to reflectively endorse the intention to explain the placebo effect (since you cannot intend to do what you recognize is impossible, i.e., to explain the inexplicable).  Indeed, not only is it not necessarily false, it’s actually true, at least of some placebo phenomena.  We now know which neurobiological interventions produce certain placebo effects; for instance, placebo analgesia is a product of expectation-induced activation of endogenous opioids and cholecystokinin (Levin, Gordon, & Fields 1978; Benedetti 2008).


We conclude that epistemic conceptions of placebo are, if not themselves confused, then confusion-inducing, as well as theoretically barren and institutionally constipated.  In the next section, we turn our attention to metaphysical conceptions of placebo.


2.2 The trouble with metaphysical conceptions


Recall that a conception of placebo counts as metaphysical if it identifies a specific mechanism as producing placebo effects.  Probably the best-attested such conception is the expectation theory.  Humphrey (2002, p. 256; see also Miller 2005) is a recent exemplar.  He defines placebo as:


a treatment which, while not being effective through its direct action on the body, works when and because:

  • the patient is aware that the treatment is being given;
  • the patient has a certain belief in the treatment, based, for example, on prior experience or on the treatment’s reputation;
  • the patient’s belief leads her to expect that, following the treatment, she is likely to get better;
  • the expectation influences her capacity for self-cure, so as to hasten the very results that she expects.


The difference between standard treatment and mere placebo, on this version of the metaphysical view, is that the patient’s expectations are causally implicated in their own satisfaction.  An appendectomy will end your appendicitis regardless of whether you think it will, but placebo analgesia works only because you believe it will.  One might wonder exactly what is meant by ‘expectation’ in this context.  There is a narrow sense of the term according to which someone only counts as expecting that x will occur if she occurrently – perhaps even consciously – anticipates x.  Another, natural enough, sense of the term is much broader, corresponding roughly to dispositional subjective probability: to expect x is to be disposed to regard x as a near certainty, or perhaps to be disposed to consider x more likely than not-x, or perhaps just to be disposed to regard x as at least a live possibility.


The expectation mechanism has been investigated for decades and is now well-substantiated.  People report more pain relief from branded aspirin than from unbranded aspirin, more relief from unbranded aspirin than from branded placebo, and more relief from branded placebo than from unbranded placebo (Branthwaite & Cooper 1981).  Patients who expect pain-relief from their saline drip because they’ve been told that it is infused with a painkiller request acute analgesia less frequently than patients who are less inclined to expect continuous analgesia (Pollo et al. 2001).  Positron emission tomography (PET) reveals that raising expectations of pain relief leads to a large release of dopamine in the brains of patients with Parkinson’s disease (de la Fuente-Fernandez et al. 2001).  In a functional magnetic resonance imaging (fMRI) experiment on both placebo analgesia and nocebo hyperalgesia, Wager et al. (2004) found that when expectations of pain relief are induced, there is a corresponding decrease in activity in brain regions associated with pain processing (the thalamus, insula, and anterior cingulate cortex), but when an increase in pain is anticipated, there is a corresponding increase in activity in the prefrontal cortex.


Thus the expectation theory has been corroborated as an explanation of not only some placebo phenomena but also some nocebo phenomena.  And there are many further examples, which we only touch on here.  For instance, the incidence of the listed side-effects of a given drug spikes even in the placebo arm of trials studying that drug (Amanzio et al. 2009).  The same treatment (influenza vaccine) leads to worse outcomes on both subjective and objective measures (reported side effects and absenteeism, respectively) when framed negatively (mentioning that side effects occur in N% of cases) than when framed positively (mentioning that side effects are not observed in (100 – N)% of cases), even though the facts expressed by the different framings are identical (O’Connor, Pennie, & Dales 1996).  Even the mere verbal suggestion of potential negative outcomes can generate nocebo (Benedetti et al. 2007).  For instance, when male patients treated with finasteride are told that it may cause erectile dysfunction, decreased libido, and problems with ejaculation, but that these are uncommon, their reports of these symptoms jump from 9.6%, 7.7%, and 5.7% to 30.9%, 23.6%, and 16.3% respectively (Mondaini et al. 2007).

If positive expectations systematically lead to placebo effects and negative expectations systematically lead to nocebo effects, why not just declare victory for the expectation-based version of the metaphysical conception and be done?  The problem is that, while the expectation theory helps us to explain some placebo and nocebo effects, it fails to explain all placebo and nocebo effects.  And of course, a condition on an adequate theory of X is that the theory apply to all and only the cases in which X occurs.  As we know from the philosophy of language, there are two main ways in which a theoretical construct can refer: the descriptivist route and the direct reference route.  When a kind is picked out descriptively, the theorist lays out necessary and sufficient conditions for membership, then turns to the world to find out what satisfies those conditions.  Whatever fits the description belongs to the kind; whatever doesn’t, doesn’t.  When a kind is picked out through direct reference, by contrast, the theorist points to examples of the kind and says, as it were, “That kind of thing – whatever it is.”  The theorist will typically have some notion of what properties the kind has, but she needn’t have in mind properties that all and only members of the kind possess, and she might even be mistaken, as were the biologists who assumed that archaea were evolutionarily more ancient than bacteria.  The expectation theory of placebo is a descriptivist theory, which means that it needs to pick out all and only the phenomena that really are placebo effects.  Other metaphysical conceptions of placebo (and nocebo) explain phenomena that the expectations theory does not cover, and so it cannot be considered an adequate theory of placebo (and nocebo).


By way of illustration, consider the conditioning theory, according to which placebo effects are generated by classical conditioning: the body learns to react to the conditioned stimulus (e.g., a pill of a certain shape and size) in the same way it responds to the unconditioned stimulus (e.g., the drug hitherto contained in pills of that shape and size).  Ader (2000), for instance, showed that starting a patient off on a course of treatment with pills that contain 100% of the normal dosage and then weaning them to sugar pills that look and taste exactly like the real thing leads to outcomes indistinguishable from normal treatment, but that using pills with 50% of the normal dosage throughout treatment is not as effective.  Perhaps even more telling is the fact that inducing expectations of increase or decrease in growth hormone and cortisol has no effect on their secretion, but preconditioning with sumatriptan (a 5-HT1B/1D agonist that inhibits cortisol secretion while stimulating growth hormone secretion) enables placebogenic enhancement of growth hormone production and placebogenic reductions in cortisol production (Benedetti et al. 2003).  Preconditioning was effective even when patients were led to expect the otherwise.  Moreover, the expectations and conditioning mechanisms can sometimes be harnessed in tandem to strengthen the effect that either would produce on its own (Benedetti et al. 2003, Hanour 2005, Stewart-Williams & Podd 2004).


It would be tedious to walk through further examples.  The point is that any given descriptivist metaphysical conception of placebo, no matter how well it explains some phenomena, is inadequate unless it explains all placebo phenomena.  And we have no a priori reason to think that a single conception will do this, and plenty of a posteriori evidence to the contrary.  Now, it’s possible to respond to this problem by switching to a direct reference theory (Kripke 1972, Putnam 1975), and defining placebo as whatever-it-is-that’s-responsible-for-that, using ostensive reference to the phenomenon in question.  Such a strategy seems to work when dealing with natural kinds, such as gold and hydrogen, and biological kinds, such as tigers and whales.  But the direct reference strategy fails when the object of investigation is not such a kind, and the partial successes of the expectation and conditioning conceptions of placebo suggest that the metaphysical approach to placebo is unamenable to this solution.


What unifies placebo phenomena (if anything) is a weak family-resemblance relation, and our own positive evaluative attitude towards their outcomes.  What unifies nocebo phenomena (if anything) is a weak family-resemblance relation, and our own negative evaluative attitude towards their outcomes.  Just as it would be futile to construct a theory of weeds or pests using the direct reference strategy, so, we contend, it is futile to construct a theory of placebo and nocebo using the direct reference strategy.  The phenomena referred to under this heading are just too diverse.  Expectation-induced release of dopamine in the striatum is a placebogenic treatment for Parkinson’s disease; preconditioning with immunosuppressive drugs such as cyclophosphamide and cyclosporine A modulates immune mediators such as IL-2, IFN-gamma, and lymphocytes (Benedetti 2008).  It seems incredible to argue that these phenomena fall under a single natural, biological, or psychological kind.


It should be clear at this point that diverse placebo phenomena bear only a tenuous family resemblance to each other, and as Wittgenstein (1953/2009) and others have argued, family-resemblance concepts are unamenable to analysis by means of necessary and sufficient conditions.  The same goes, of course, for nocebo phenomena.  The epistemic conception of placebo fails, but so do unified metaphysical conceptions, regardless of whether they are meant to refer in a descriptivist or a direct reference way.  We contend that the best response to this problem is to dissolve the heterogeneous placebo phenomena into their coherent parts, which should be theorized and investigated both independently and in their interactions.  Doing so will not only help to resolve the placebo-is-powerful debate, but also open up new avenues of inquiry.

3 A case for pluralism


Suppose, for the sake of explanatory clarity, that only two distinct mechanisms – expectation and conditioning – led to what is now called the placebo effect.  Suppose also that some illnesses were susceptible to placebogenic treatment via expectations, some via conditioning, some via their combination, and some via neither.


Illnesses susceptible to treatment by… Expectations No expectations
Conditioning A B
No conditioning C D

Table 1: Dissolving placebo mechanisms


Now imagine how things will look to researchers working in the expectation paradigm: only illnesses of type C will appear to be a promising target for placebogenic treatment.  By contrast, from the point of view of someone working in the conditioning paradigm, only illnesses of type B will look promising.  And now imagine a meta-analysis of placebogenic treatment that lumps all of this research together: expectation-based investigations of A, B, C, and D, as well as conditioning-based investigations of A, B, C, and D.  Only two out of these eight categories of studies (expectation-based investigations of C and conditioning-based investigations of B) will show any effects, and they could be partially or even fully drowned out by the noise from the other six categories.  If, as seems likely, there aren’t just two, but many placebo pathways, things get even messier.


This is not an idle speculation.  We suspect that the meta-analyses of Hrobjartson & Gotzsche (2001, 2004, 2010), who urge that there is no evidence of clinically significant placebo effects, may suffer from exactly this methodological drawback.  This point aligns with the critique of Vase, Riley, & Price (2002), who found that if the meta-analysis is restricted to well-designed, theory-driven studies, effects in categories B and C are detectable, as well as even larger effects in category A.  Meta-analysis is a terrifically important statistical tool, but it is best at making apples-to-apples comparisons.  Thus, we do not mean to suggest that Hrobjartson & Gotzsche’s work has been misguided or that it delivered the wrong results.  In fact, we take it to be not only consistent with but even mutually confirmatory with our own view: what they found is exactly what should be found if researchers into placebo have overzealously conflated dissimilar mechanisms.


When the mechanism of placebo is left completely unspecified (as it must be in the epistemic conception) or only vaguely specified (as it often is in the metaphysical conception), it’s difficult if not impossible to draw the kinds of distinctions made in the previous paragraph.  Consider again the expectation theory.  Above, we distinguished between expectations as occurrent, conscious mental states, on the one hand, and dispositional subjective probabilities on the other.  We also distinguished several ways of spelling out the dispositional subjective probability view: it could involve being certain or nearly certain of an outcome, or being disposed to think the outcome more likely than not, or even just not being disposed to reject the outcome as a live possibility.  It could also involve an affective component, as the attitude of hope is taken to comprise both cognitive and affective components (McGeer 2008).  This latter possibility receives some support from the work of Ted Kaptchuck and his colleages, who found that placebo interventions predicted not only some clinical outcomes but also “dramatic behavioral and psychosocial changes,” such as no longer fearing eating and feeling more positive affect (Kaptchuck et al. 2006), especially hope (Kaptchuck et al. 2009).


Distinct placebogenic mechanisms may interact not only with each other, but also with traditional treatment pathways, such as drugs.  In a typical randomized controlled trial of a pharmacological intervention, the effect of a drug is compared with the effect of an indistinguishable placebo pill.  (Suppose for the sake of simplicity that the only placebo mechanism is expectation-based.)  In more sophisticated trials, there is a third no-treatment arm, which helps researchers distinguish the natural course of illness from any placebogenic effects.  An assumption built into this methodology is that that the effects of drug and placebo are linear and independent: effect of(drug + placebo) = effect of(drug) + effect of(placebo).  The extent to which the combination of drug and placebo outperforms placebo alone is assumed to equal the independent, specific efficacy of the drug.  This assumption is dubious.  In some cases, drug and placebo may interact, so that effect of(drug + placebo) > effect of(drug) + effect of(placebo).


Treatment with… Placebo No Placebo
Drug W X
No Drug Y Z

Table 2: Interaction between drug and placebo


This point becomes clear when one attends to the fact that what is being measured in the active treatment arm of a typical randomized controlled trial – even one with a no-treatment arm – is denoted by W, not X, in the table above.  W is not the specific efficacy of the drug, but its “placebo-enhanced” efficacy (Biller-Andorno 2004).  (The placebo arm is Y, and the no-treatment arm is Z.)  The standard assumption is that X + Y = W, which would justify subtracting Y from W to determine X.  There is good reason to think this assumption is false.  Hidden injections of analgesics are significantly less effective than open injections (Amanzio et al. 2001).  Overt administration of opioids is more effective than covert administration in treating pain, anxiety, and Parkinson’s disease (Colloca et al. 2004).  Patients’ knowing that a drug has good (bad) effects even increases (decreases) the rate at which it is absorbed by the body (Flaten, Simonsen, & Olsen 1999).


When trying to determine the specific efficacy of a drug, what one wants is X, but it can be extremely difficult to investigate X directly.  Whenever you give people a drug, they know that they’ve been given a drug, so what you end up with is W, not X.  This leads to what Colloca & Benedetti (2005; see also Benedetti 2008 and Hrobjartson, Kaptchuck, & Miller 2011) call the uncertainty principle: “One can never be sure regarding the action of a pharmacological agent, as its pharmacodynamic action is perturbed by the act of administering it.”  As framed, this principle is too strong.  After all, you could in principle administer drugs to people without their knowing it; observing how they differed from people who knew they were receiving drugs, people who were only receiving placebo, and people who received no treatment at all would tell you the value of X for that drug.  But it would of course be morally and legally problematic to do so.  What the uncertainty principle should say is that, if one is investigating in a morally defensible way, then one can never be sure about the specific efficacy of a pharmacological agent.


The mechanisms of placebo phenomena are diverse.  They interact with each other.  They interact with traditional treatments.  For these reasons, we advocate dissolving the conception of placebo into a plurality of potential mechanisms that are spelled out in detail, and investigating these both independently and in their interactions.  This suggestion is in line with Hrobjartson, Kaptchuck, and Miller’s (2011) recommendation to design clinical trials “in which the presumed [placebogenic] causal factors are characterized a priori.”  In the next and final section, we show how this might be done for a potential mechanism that we find promising.


4 Somatic attention, feedback loops, and feedback cascades


As we mentioned above, many of the illnesses that seem susceptible to placebogenic treatment could be construed as maladies of affect and attention (e.g., pain, depression, anxiety, and addiction).  In this section, we outline a model of how these placebo effects might be generated.  We start with the example of pain, but other health problems may be similarly affected.  Although there is philosophical dispute about whether pain is necessarily conscious (Kripke 1972; Rosenthal 2005), certainly many pains are conscious and attended to.  The idea we want to explore here is that some cases of placebo analgesia could be due to a combination of somatic attention, construal, feedback loops, and feedback cascades.


Daily experience is a continuous deluge of stimuli.  We notice only some of them.  We attend to even fewer.  When we attend, we interpret or construe in particular ways.  How we construe depends both on what the stimuli are and our current state of mind – including our current beliefs, desires, moods, and emotions.  Construal then affects our state of mind, our attentional focus, our behavior, and through these channels the stimuli that we experience later.  If your doctor primes you to expect pain in your left big toe, you will pay more attention to the toe and be more disposed to interpret ambiguous sensations there as pain.  Thus, you will be more disposed to notice slight pains that would otherwise have slipped under the radar of attention, as well as to interpret ambiguous stimuli as pain (Allen & Siegel 2002).  If instead your doctor primes you to expect relief of your headache, you will pay more attention to sensations in and around your head and be more disposed to interpret ambiguous stimuli as the alleviation of pain.  Thus, you will be more disposed to notice slight lessening of the intensity of your headache, as well as to interpret ambiguous stimuli as pain relief.


One way in which placebogenic analgesia might occur, then, is simply by raising the threshold for both noticing pains and for counting a sensation as pain rather than heat, pressure, or what have you.  Even if it is possible to have a pain of which you are unconscious, still it seems evident that it’s worse to have a conscious than an unconscious pain.  And even if it is possible to misinterpret what really is a pain as some other kind of sensation, still it seems evident that it’s worse to have a pain and know it, than to have the same pain but misconstrue it as pressure.


So far, we have only discussed somatic attention from a punctate point of view.  What distinguishes the mechanism we propose is that it is essentially dynamic, drawing on the connectionist picture of the mind and brain (Tryon 2009).  When you attend to your head, expecting or hoping for pain relief, and you end up construing a sensation as alleviation, you may become less anxious, which would have (or perhaps even be constituted by) bodily changes such as decreased blood pressure and heart rate.  These bodily changes would then cause further relief of pain.  Thus, a feedback loop connects your sensations, somatic attention, construal, and bodily states.


When a feedback loop repeats over and over again, it can generate a feedback cascade.  The slight or ambiguous pain relief that you notice only because your doctor led you to expect or hope for pain relief leads to slight bodily changes.  These slight changes cause changes in sensations, in somatic attention, and in disposition to construe, which in turn lead to noticing further pain relief, which leads to further bodily changes, and so on.


The same story can be told about certain nocebo effects, such as nocebogenic erectile dysfunction.  If a man is led by his physician to expect or worry about erectile dysfunction, he will be more disposed to notice even slight or potential problems, and to construe ambiguous phenomena as evidence of dysfunction.  Noticing and construing in this way will lead to bodily changes and loss of confidence, which in turn may affect experienced stimuli, somatic attention, and processes of construal.  If this cascades – that is, if the loop repeats several times and builds on itself – it may lead to outright erectile dysfunction.


Few investigations of the mechanism of placebo and nocebo have theorized that somatic focus played a role, and, to our knowledge, none has been formulated in terms of feedback loops and cascades.  In a recent study, Geers et al. (2006) found that participants who were instructed to attend to their bodily states for potential negative side effects of an inert pill reported feeling more anxious and nauseated than participants who were not so instructed.  This study provides a small dose of support for our view.


A worry that might raised at this point is that our account so far is quite speculative.  Somatic attention combined with feedback loops and feedback cascades may sound promising, but it is far from empirically validated.  The more precise the model is, the less likely it is to be correct.  This is of course true, but we are not bothered by the fact that we might be wrong.  The point is not to guess right as to the mechanism the first time (though that would be nice) but to formulate precise hypotheses that can be operationalized and tested.





Ader, R. (2000). Classical conditioning in the treatment of psoriasis. Cutis, 66, 370-372.

Allen, L. & Siegel, S. (2002). A signal detection theory analysis of the placebo effect. Evaluation and the Health Professions, 25, 410-425.

Amanzio, M., Pollo, A., Maggi, G., & Benedetti. F. (2001). Response variability to analgesics: A role for non-specific activation of endogenous opioids.  Pain, 90, 205-2015.

Benedetti, F. (2008). Mechanisms of placebo and placebo-related effects across diseases and treatments. Annual Review of Pharmacology and Toxicology, 48, 33-60.

Benedetti, F., Lannote, M., Lopiano, L., & Colloca, L. (2007). When words are painful: Unraveling the mechanisms of the nocebo effect. Neuroscience, 147, 260-271.

Benedetti, F., Pollo, A., Lopiano, L, Lanotte, M., Vighetti, S., & Rainero, I. (2003). Conscious expectations and unconscious conditioning in analgesic, motor, and hormonal placebo/nocebo responses. Journal of Neuorscience, 24, 4215-4223.

Biller-Andorno, N. (2004). The use of the placebo effect in clinical medicine – ethical blunder or ethical imperative?  Science and Engineering Ethics, 10, 43-50.

Branthwaite, A. & Cooper, P. (1981). Analgesic effects of branding in treatment of headaches. British Medical Journal: Clinical Research Edition, 282, 1576-1578.

Colloca, L. & Benedetti, F. (2005). Placebos and painkillers: Is mind as real as matter? Nature Reviews Neuroscience, 6, 545-552.

Colloca, L., Lopiano, L., Lanotte, M., & Benedetti, F. (2004). Overt versus covert treatment for pain, anxiety, and Parkinson’s disease. Lancet Neurology, 3, 679-684.

De la Fuente-Fernandez, R., Ruth, T., Sossi, V., Schulzer, M., Calne, D., & Stoessl, A. (2001). Expectation and dopamine release: Mechanism of the placebo effect in Parkinson’s disease. Science 293, 1164-1166.

Fässler, M., Gnädinger, M., Rosemann, T., & Biller-Andorno, N. (2009). Use of placebo interventions among Swiss primary care providers. BMC Health Services Research, 9:144, 1-9.

Fässler, M., Meissner, K., Schneider, A., & Linde, K. (2010). Frequency and circumstances of placebo use in clinical practice – a systematic review of empirical studies. BMC Medicine, 8:15, 201-10.

Flaten, M., Simonsen, T., & Olsen, H. (1999). Drug-related information generates placebo and nocebo responses that modify the drug response. Psychosomatic Medicine, 61:2, 250-255.

Geers, A., Helfer, S., Weiland, P., & Kosbab, K. (2006). Expectations and placebo response: A laboratory investigation into the role of somatic focus. Journal of Behavioral Medicine, 29, 171-178.

Grünbaum, A. (1994). The placebo concept in medicine and psychiatry. In Graham & Stephens (eds.), Philosophical Psychopathology, 285-324. MIT Press

Hauour, F. (2005). Mechanisms of the placebo effect and of conditioning. Neuroimmunomodulation, 12:4, 195-200

Hrobjartson, A. & Gotzsche, P. (2001). Is the placebo powerless? An analysis of clinical trials comparing placebo with no treatment. New England Journal of Medicine, 344, 1594-1602.

Hrobjartson, A. & Gotzsche, P. (2004). Is the placebo powerless? Update of a systematic review with 52 new randomized trials comparing placebo with no treatment. Journal of Internal Medicine, 256, 91-100.

Hrobjartson, A. & Gotzsche, P. (2010). Placebo interventions for all clinical conditions. Cochrane Database of Systematic Reviews, 1, CD003974.

Humphrey, N. (2002). Great expectations: The evolutionary psychology of faith-healing and the placebo effect. In Humphrey (ed.), The Mind Made Flesh, pp. 255-285. Oxford University Press.

Kaptchuck. T., et al. (2000). Do medical devices have enhanced placebo effects? Journal of Clinical Epidemiology, 53, 786-792.

Kaptchuck, T., et al. (2006). Sham device versus inert pill: Randomized controlled trial of two placebo treatments. British Medical Journal, 332, 391-397.

Kaptchuck, T., et al. (2009). “Maybe I made up the whole thing”: Placebos and patients’ experiences in a randomized controlled trial. Cultural Medical Psychiatry, 33, 382-411.

Kaptchuk, T., Friedlander, E., Kelley, J., Sanchez, M., Kokkotou, E., Singer, J., Kowalczykowski, M., Miller, F., Kirsch, I., & Lembo, A. (2010). Placebos without deception: A randomized controlled trial in irritable bowel syndrome. PLoS ONE, 5:12, e155591.

Kripke, S. (1972). Naming and Necessity. Oxford: Blackwell.

Levine, J. Gordon, N., & Fields, H. (1978). The mechanisms of placebo analgesia. Lancet, 2, 654-657.

McGeer, V. (2008). Trust, hope and empowerment. Australasian Journal of Philosophy, 86:2, 237-254.

Miller, F. (2005). William James, faith, and the placebo effect. Perspectives in Biology and Medicine, 48:2, 273-281.

Mondaini, N., Gontero, P., Giubelei, G., et al. (2007). Finasteride 5 mg and sexual side effects: How many of these are related to a nocebo phenomenon? Journal of Sexual Medicine, 4, 1708-1712.

Nitzan, U. & Lichtenberg, P. (2004). British Medical Journal, 239:7472, 944-946.

Pollo, A., Amanzio, M., Arslanian, A., Casadio, C., Maggi, G., & Benedetti, F. (2001). Response expectancies in placebo analgesia and their clinical relevance. Pain, 93:1, 77-84.

Putnam, H. (1975). The meaning of meaning. Philosophical Papers, Vol. 2: Mind, Language, and Reality. Cambridge: Cambridge University Press.

Rosenthal, D. (2005), Consciousness and Mind. Oxford: Clarendon Press.

Stewart-Williams, S. & Podd, J. (2004). The placebo effect: Dissolving the expectancy versus conditioning debate. Psychological Bulletin, 130:2, 324-340.

Szawarski, Z. (2004). The concept of placebo. Science and Engineering Ethics, 10, 57-64.

Tryon, W. (2009). Cognitive processes in cognitive and pharmacological therapies. Cognitive Therapy Resources, 33, 570-584.

Vallance, A. (2006). Something out of nothing: The placebo effect. Advances in Psychiatric Treatment, 12:4, 287-306.

Vase, L. Riley, J., & Price, D. (2002). A comparison of placebo effects in clinical analgesic trials versus studies of placebo analgesia. Pain, 99, 443-452.

Wager, T., Rilling, J., Smith, E., Sokolik, A., Casey, K., & Davidson, R., et al. (2004). Placebo-induced changes in fMRI in the anticipation and experience of pain. Science, 303:5661, 1162-1167.

Weeks, J., Catalano, P., Cronin, A., Finkelman, M., Mack, J., Keating, N., & Schrag, D. (2012). Patients’ expectations about effects of chemotherapy for advanced cancer. New England Journal of Medicine, 367, 1616-1625.

Wittgenstein, L. (1953/2009). Philosophical Investigations, fourth edition. P. M. S. Hacker & J. Schulte (eds.). Oxford: Blackwell.

[1] An even weaker formulation of the distinction is this: in according to epistemic conceptions, not only do we fail to know how healing works, but we also fail to know that it works because we cannot predict when it will work and when it will fail; in contrast, according to metaphysical conceptions, we at least know that the intervention works, even if we don’t know how.

One thought on “Dissolving the Placebo

  1. Quite interesting. Hit upon your post by Googling oxford nocebo felon experiment. I chose this list from referring to Science and Health by Mary Baker Eddy and wanting more background information on the nocebo bleeding to death of a felon referenced on pg 379.

Leave a Reply