The relativity of ‘placebos’: defending a modified version of Grünbaum’s definition

This article argues that health care systems that are grounded in solidarity have the right to penalise some users who are responsible for their poor health.

Paper 8. The relativity of ‘placebos’: defending a modified version of Grünbaum’s definition by Jeremy Howick


Jack Freer August 3, 2020 at 7:00 AM

Jeremy Howick (JH) has attempted to clarify a perceived confusion about placebos, placebo effect and placebo controls. His approach is to refine the definition and descriptions developed by Adolf Grünbaum. The result is a complex scheme which attempts to characterize the placebo effect. Along the way, JH also attempts to define 'placebo' and to relate this to the notion of placebo controls in clinical trials.

In essence, JH is attempting to create a unified theory of placebos. I would argue that a single comprehensive conceptual framework is unnecessary. Instead, a description of the placebo effect [1] is the primary task. From that, we can derive a definition of placebo [2]. Finally, I would suggest that placebo controls [3] in research represent a distinct category that is independent of the mechanism of the placebo effect.

[1] [This description of placebo effect relies heavily on Howard Brody's 1977 monograph on placebos]:

The placebo effect occurs when a person, who receives an intervention in a healing context, undergoes a change in a medical condition or symptom (either better or worse), and that change is not due to any known therapeutic, physiologic, or pharmacologic effect of the intervention (either in whole or in part).

[2] A placebo is simply an intervention that produces a placebo effect.

[3] A placebo control in a clinical study is an inactive substance or intervention which is designed to be similar to the study intervention.

Ideally, the study intervention and the placebo intervention are administered to different groups that are allocated by random selection. The identity of the intervention in each group remains unknown to all involved (research subject, staff etc). Some placebo interventions are more difficult to mask (as in studies of psychoactive substances such as cannabis, or those involving surgery). No matter, the goal is to simulate the study intervention as closely as possible. The bare difference between the study and placebo interventions should only be the treatment modality being evaluated.

There are obviously features of the placebo control that take advantage of the placebo effect in its experimental function as described. A control arm of the study which results in either beneficial or adverse effects will help delineate the effects of the study intervention that are solely due to the properties of that treatment (since "real treatments" are known to produce placebo effects as well). The theoretical details of that effect, however, are irrelevant to the choice and design of the control. The only thing that matters is the control's similarity to the treatment (and, of course, its inertness).


Pat D August 3, 2020 at 11:17 AM

I find your account based on Harold Brody much clearer than JH’s account. I will post separately about his account, but my question for you is about mental health interventions. Do you or Brody consider talk therapy or cognitive behavioral theory to be “really” therapeutic? If so, how do you distinguish its therapeutic effect from a placebo effect?

Jack Freer August 3, 2020 at 2:02 PM


Actually, Brody's thinking on this evolved over time. He is an MD/PhD (Family Medicine/Philosophy). I believe the 1977 book was actually his dissertation at Michigan State. He was troubled by the conflict between placebo efficacy and ethical barriers. Searching for a way to harness the beneficial effects of placebos (without lying to patients), he decided it lay in the doctor-patient relationship. Specifically, the *power* of the physician in the relationship produced positive outcomes, not attributable to pharmacologic or physiologic mechanisms. This is then a means of eliciting the placebo effect without being dishonest with the patient.

Your question about mental health treatment really gets at our ambivalence about talk therapy. CBT is *real* medical treatment, and like every medication or surgical procedure, acquires a portion of its therapeutic response from the placebo effect. It seems impossible to actually estimate the proportion of each.

That actually raises an incidental question: what does does the research placebo for CBT look like? I seem to recall that it is a kind of psychobabble which superficially passes for therapy.


Stephen Kershnar August 4, 2020 at 6:26 PM


I am worried about these statements.

(1) A placebo is simply an intervention that produces a placebo effect.

(2) A placebo control [effect] in a clinical study is an inactive substance or intervention which is designed to be similar to the study intervention.

I tried to restate them as follows.

(1) A thing is a placebo if and only if it is an intervention that produces a placebo effect.

(2) A thing is an intervention that produces a placebo effect if and only if it is an inactive substance or intervention that is similar to the study intervention.

If we combine these statements, we end up with the following.

(3) A thing is a placebo if and only if it is an inactive substance or intervention that is similar to the study intervention.

The problem is that a placebo is an active substance or intervention. Specifically, it is prescribed for the psychological benefit to the patient.

Perhaps you do not want to count psychological activity as the relevant activity. First, this will not work unless you stipulate that psychological activity is not the relevant type of activity, even if it has a therapeutic effect. This strikes me as misleading. Psychiatric treatment that has physiological effects is active and therapeutic.

Second, this could be inquiry-specific so that what we want is a psychology-independent therapeutic treatment. But then why not just label the placebo as ‘psychology-dependent treatment’ and the active treatment ‘psychology-independent treatment’. We avoid the morass that accompanies the placebo concept and allows us to speak with greater clarity.


Steve K

Jack Freer August 4, 2020 at 10:03 PM


Certainly, placebos have activity, and the definition should contain the qualifiers "pharmacologic or physiologic" when stating they lack activity. The mechanism for the placebo effect is psychological because one must be aware of it if it is to work. Brody presents the hypothetical situation in which one person has read that placebos are effective (and relatively safe). He believes his friend's symptoms would improve with a placebo, but his friend is a Christian Scientist and will not take ANY medications, so he slips the placebo into his friend's coffee when he isn't looking. I don't like to emphasize the psychological angle because I believe it carries a lot of baggage (such as the belief that some people or some conditions are more susceptible to the placebo effects).


Stephen Kershnar August 5, 2020 at 3:54 PM


I guess I think that the word 'Placebo' is too vague to be useful.

In particular, I am not sure that what matters is whether it has pharmacologic or physiologic effect or whether the person prescribing it intends that they have these effects.

Also, I am not sure how this distinction matters in that anything that affects someone's thinking is physiological in some sense (assuming thinking occurs in the brain).

Perhaps the best approach is to stipulate a meaning for 'placebo' and avoid vague or circular accounts.

I hope your summer is going well,

Steve K

Neil Feit August 7, 2020 at 12:11 PM

Jack, does your account way up above take the statement "Placebos bring about the placebo effect" to be true as a matter of definition. It seems like it does. I haven't thought too carefully about this but it seems to me that the statement I quoted isn't true by definition but is capable of empirical test.

David H August 7, 2020 at 1:04 PM


Regarding (2) I thought Howick was right that a placebo effect should be distentangled from placebo ("placebo effect" not defined in virtue of "placebo") because non-placebos could have placebo effects (p.1380). In his language, a non-placebo that has a remedial characteristic effect could also have placebo effects due to its incidental features (pill's expense, bad taste, high tech delivery etc). If having a placebo effect was enough to make something a placebo then the same substance could be a placebo and not a placebo if the latter means it has known physiologic, pharmaceutical, therapeutic effects.

Maybe you are using "placebo effect" to include nocebo effects but it would otherwise seem to go against ordinary usage to have harmful placebos. You allow in (1) above that placebo effects could be harmful or beneficial. So, if you then follow Brody and define a placebo as something that has placebo effects, then there could be harmful placebos which is counterintuitive (p 1388). Of course, if the gains in explication are worth disregarding common usage in a refined definition, then so be it. Howick doubts it is worth it (pp. 1388-1389) and so introduces "harmful interventions" for characteristic factors that are harmful and relies upon "nocebo effects" for harms from incidental factors.

I think the key work that needs to be done in defining placebo -though my knowledge of the debate issue and literature is just Howick's article - is to distinguish therapeutic psychological effects from psychological placebo effects. I wish his characteristic/incidental distinction worked but I fear it doesn't. I offer a first pass below in one of my comments about the dependency upon a pro-attitude that could be undermined by awareness of that dependency as distinguishing psychological placebo from therapeutic psychological mechanisms

This assumes there are no placebos that don't work through mental mediation. Howick mentions placebo effects on unconscious bodies because of conditioning (p. 1381) in which even unconscious bodies respond to stimulation. He also mentions confounding influence of needle insertion, and bulking agents and experimental biases. He seems to include experimenter enthusiasm... (as) part of what we mean when we talk about placebo effects."

Jack Freer August 7, 2020 at 8:28 PM

Neil & David,

I believe the current terminology is unfortunate because there are fixed popular notions of placebo (not unlike brain death). Since we are stuck with placebo talk, I would think a more coherent system would look something like this:

Placebo Effect (PE) —The large body of knowledge describing the various phenomena, including nocebo, ubiquitous presence of PE in *all* medical treatment to some degree, and occurrence in 100% of the he population (although, to varying degrees).

Remedial Placebo (RP) — Since all “effective” remedies Include some PE, I would suggest a remedy be described as placebo (RP) only if there is intentional deception.

Placebo Control (PC) — Research strategy intended to administer a physiologically inert, but superficially similar intervention to the control arm of a clinical study.

Of the two instances of placebo use (RP & PC), one is grounded in deception (RP), and the other explicitly explained in excruciating detail in a written informed consent document (PC).

PE is a conceptual explanatory system with little direct practical significance. RP is a real life situation and since it entails intentionally deceiving a patient, it is virtually always ethically impermissible (I can't think of any analogies to lying to the Nazi at your door in wartime Netherlands when he asks if Jews are hiding in the attic). PC is also a real life practical situation. Its ethical permissibility turns on the quality of the informed consent (the conversation, not the document).


Bob Kelly August 10, 2020 at 10:54 AM

Jack, sorry for jumping in late on this one. I agree with Neil above that the definition of 'placebo', while nice and clear with a good logical structure, probably shouldn't make placebos successful (producing a placebo effect) by definition. It seems like we could give someone a placebo but it not have the intended effect. Though, this of course doesn't amount to a rebuttal of your view since we have to basically beg the question against your definition for the case to work. But it raises a question of whether placebos are necessarily effective.

I am also curious about your modified definitions directly above. 'Placebo Effect' is defined as a 'body of knowledge'. Neither placebos nor their effects are bodies of knowledge. And 'placebo effect' seems to show up in the definition. The 'placebo effect literature' is a body of knowledge, which is about the phenomenon we call 'placebo effect'. The latter should not be defined in terms of the body of knowledge which discusses it. And I also don't see 'placebo control' as a research strategy. 'Use/Implementation of a Placebo Control' seems like a research strategy (perhaps we need to add 'for the purpose of goal X' at the end), but 'placebo control' seems like the thing that is used as a control, or in the control group, not the strategy itself. 'Remedial Placebo' seems to be defined as 'a remedy that involves intentional deception' (is there another kind of deception?). This seems much too permissive. Suppose Dr. Smith's patient Anne is really scared of taking prescription medicine, but she has a dangerous and likely fatal infection. Dr. Smith gives her what appears to be a bottle of an herbal, homeopathic substance, and tells her that's what it is. Really, it's an antibiotic. Anne, feeling fine about herbal remedies, takes the medication and the antibiotic works perfectly normally to cure the infection. This seems like a normal remedy with (in my view, appropriate) deception, but without a placebo effect. Of course, what a placebo and its effect actually are is what is at issue, so this is just my intuition.

Bob Kelly August 10, 2020 at 11:02 AM

Also, maybe you don't think the case of Dr. Smith and Anne is sufficiently described to make the deception permissible. But let's just add that Anne is bed-ridden in the hospital, she cannot be moved anywhere else, doctors only have a few days to treat the infection before she dies, she explicitly wants to be treated and to survive but won't knowingly take non-herbal remedies (let's say it's due to an irrational fear of them having dangerous side effects), and doctors know the antibiotic is 99% effective and Dr. Smith knows that if he offers Anne an herbal remedy she will take it. Dr. Smith notices he happens to have an old empty bottle of herbal medicine. The antibiotic happens to look a lot like an herbal pill. He puts the antibiotic into the herbal medicine bottle and tells Anne that if she takes this, it may increase her chances of fighting off the infection. She takes it and the infection clears up.

What else should he have done?

Pat D August 3, 2020 at 11:18 AM

I find JH’s account of placebos muddled. Without going into detail, here are several reasons.

First. I am not sympathetic to his behaviorist presuppositions. For instance, what is the difference between expectations and beliefs? Or, what distinguishes cognitive behavioral therapy from behavioral therapy? Or, what does philosophy bring to medicine that medicine lacks? I don’t think that these questions can be answered without distinguishing biological and psychic & cognitive functioning with respect to human (or animal) health.

Second, JH’s definition of characteristic features (p. 1392) fails for several reasons. We do expect a proven treatment – say, antibiotics for bacterial pneumonia – to be characteristically effective. We learn about effective treatments by experimentation and reasoning, not be association or nominal definition. There is also a circularity in defining placebos as lacking characteristic features and defining characteristic features as adding something to placebos.

Third, as I shared with Phil Reed in an earlier email exchange, I am sympathetic to Pekka Louhiala and Raimo Puustinen’s argument to limit placebo terminology to research contexts and to speak instead about care effect in a clinical (and research) context with respect to beneficial therapeutic effects that remain unexplained by a medical treatment.


Neil Feit August 4, 2020 at 2:39 PM

Pat, regarding your second point, I was also struck by an apparent circularity when I first read Howick's definition of a characteristic feature (of a treatment/therapy) and then again on p. 1392 as you mention. His definition of what it is for a treatment to be a placebo control makes use of the idea of the characteristic features of the treatment. This notion of a characteristic feature, in turn, is defined by making reference to a placebo control. Right now, this seems like a circularity that can't be fixed up in any way, but I might be missing something. Also, his definition of a (generic) placebo makes reference to characteristic features of the treatment, so the circularity affects this analysis too. Still, I think Howick says a lot of interesting and sensible things about the issues.

Stephen Kershnar August 4, 2020 at 6:46 PM

Neil and Pat:

I agree with your comments. I do not know if this is correct, but here is my attempt to summarize JH’s final position (pages 1391-1392).

(1) An effect is a placebo effect if and only if it

(a) … or

(b) is it an effect of a (revised) generic placebo.

(2) A treatment (process) is a generic placebo if and only if none of the characteristic treatment factors, C, are effective in patient, S, for condition, D.

(3) Factors, C, are characteristic treatment factors if and only if C does not include the expectancy that a treatment is effective and has an incremental benefit on the target disorder over a legitimate placebo control in a well-controlled trial.

Hence, the following is true.

(4) Whether something is a placebo effect depends on whether it is a genetic placebo.

(5) Whether something is a genetic placebo depends on a fact about legitimate placebo control.

I’m wondering whether (a) if this is an accurate summary of JH’s position, (b) this overstates the circularity, and (c) independent of JH, whether ‘placebo’ must be analyzed in circular terms.


Steve K

Pat D August 5, 2020 at 2:56 PM


Thanks for your comments. I agree that JH says some interesting and sensible things, especially about how hard it is in many cases to design the control arm of a randomized study.


David H August 7, 2020 at 2:15 PM


About the circularity. I don't think Howick both defines placebo in terms of lacking characteristic features and defines characteristic features as adding something to placebos as you say in your second point above. I think he only does something like the first in defining placebos as having characteristic features that are not effective in benefiting or harming. A placebo is something whose characteristic features are not beneficial or harmful. So he does define characteristic feature in terms of placebo control. But he isn't committed to defining placebo effect and placebo control in terms of placebo (p. 1370). So it isn't a problem if "placebo" is defined in terms of "characteristic feature" and the latter is defined in terms of "placebo control". That might make "placebo" depend upon "placebo control" but he doesn't care to define "Placebo" first and "placebo control" in terms of placebo. He says to think that is needed is to be misled by linguistic features. (p. 1370)

The circular worry seems more along Neil's concerns of placebo control lacking benefits from characteristic features but placebo controls playing a role in determining what is characteristic features (condition 2 of a characteristic feature). There may be some sort of bootstrapping or back-to-back people raising each other by pushing against each other in the co-discovery of what is a placebo control and characteristic feature at the same time in an experiment. So a placebo control must lack a characteristic effect on the disorder. But a placebo controlled experiment is needed to discover what is a characteristic effect. What we thought was incidental turns out to be characteristic or vice versa (pp. 1379-1380). I don't think it is problematic if an experiment reveals what we thought was a placebo control isn't because we discover the alleged placebo had characteristic features that were beneficial. so it isn't vicious if placebo control depends upon characteristic features and characteristic features can be revised in a placebo controlled experiment. We can discover at the same time that something was or was not a characteristic feature and thus a treatment was or was not a placebo control

Pat D August 8, 2020 at 5:25 PM


Thanks for pointing out the distinction between placebo and placebo controls that JH makes. I'm not sure that it avoids the problem of circularity that Neil articulated more carefully than I did. JH's overall approach to analyzing and defining placebo control is problematic. First, he doesn't start with the basic idea of a control arm in an experimental study or what makes a placebo controlled study different from other randomized controlled studies. The quote from A.B. Hill (p.1368) about study and control groups being similar in all respects except for the treatment is not specific to placebo controlled studies. JH refers in passing to well-conducted non-placebo control studies of acupuncture and vertebroplasty (pp. 1386-87). These studies contribute to our knowledge of what JH calls the 'characteristic features' of a treatment, independent of 'placebo controls.' Second, JH does not differentiate the statistical nature of these treatment studies (placebo controlled or not) from the functional nature of pathophysiological (or basic pharmacological) studies. Treatment studies concern characteristic (or predictable) outcomes; they do not elucidate mechanisms, which is what I understand by characteristic 'features'. I would relate the discovery and bootstrapping process you mention to the complementary interaction of statistical and functional investigation. It's true that investigators often learn from their mistakes (or mistaken assumptions), but if a controlled study works as intended, you are not investigating the control arm of the study.

One last comment: it appears that JH's definition of intentional placebo applies to a clinical and not a research context. It is not clear what this has to do with 'placebo controls'.


Harvey Berman August 4, 2020 at 6:40 PM

Jack’s 3-point-dissection of JH’s paper was welcome and helpful as was Pat’s direction to the work of Louhiala and Puustinen. I view placebos — no differently than nocebos — as an emotional response to a positive or negative expectation; it could occur in a doctor’s office, a classroom, a student-mentor relationship. (That was what Mr. Miyagi did for young Daniel-Son in the first Karate Kid.)


(1) Is the placebo response a tacit component of the doctor-patient relationship?

I question whether talk therapy is therapy/treatment as opposed to a process of self-exploration. It does not take place in an instant in time, a la giving a pill or laying on of the hands, but over an extended process. Maybe it shouldn’t be called therapy.

(2) Is it possible that the emotional/psychological benefit — placebo response ?? — is part of the comfort of human friendship? Why is it that some people in crisis, having friends and family, fare batter than the individual who is alone or suffering from loneliness?

(3) I have a question for the physicians — Freer and Daly — about vertebroplasty. It seems that the real treatment — surgery + glue, et al — was more harmful to the patient than the so-called “sham” surgery. So, sham surgery would be the preferred treatment?

Seriously, isn’t such sham surgery unethical and an unnecessary procedure on the patient?

(4) Finally, a question from Steven Wright: What is the side effect of a placebo? (No answer necessary.)


Jack Freer August 4, 2020 at 10:04 PM


The study design either compares the experimental treatment to placebo or "standard therapy" (if one is currently accepted). The sham surgery is simply a placebo control, and only used to mask the "real" intervention (and not an experimental arm of the study).


David H August 5, 2020 at 1:00 AM


The side effect of a placebo is a headache for an analytic philosopher trying to explain what it is.

Stephen Kershnar August 5, 2020 at 4:04 PM


I think the emotional/psychological benefit can occur without a placebo (e.g., people with a better friendship network might have better health outcomes) and a placebo could occur without the emotional/psychological benefit (e.g., one buys it off of the internet and decides it will work after reading WebMD). So I think this might be a pathway, but is not an essential condition of a placebo.

Ditto for it being part of the doctor-patient relationship.

However, we analyze placebo, it has to be different from the Frank Sinatra functionality test, "Basically, I'm for anything that gets you through the night - be it prayer, tranquilizers or a bottle of Jack Daniels."


Steve K

Pat D August 6, 2020 at 9:09 AM


I looked at the study of vertebroplasty for osteoporotic spinal compression fractures by Buchbinder et al. It was approved by the research ethics committees at the participating hospitals and required patient consent; it also met standards for publication in NEJM. So, from a research standpoint, I think that the placebo control arm of the study (= sham surgery) met reasonable ethical standards. The primary endpoint of the study was a comparison of pain relief (reduction on a 1-10 pain scale). There was considerable overlap between the treatment and control groups: "Mean reductions in the score for overall pain in the vertebroplasty and placebo groups were 2.6±2.9 and 1.9±3.3, respectively." So I wouldn't say that the sham surgery was better. The study also brings home the fact that we don't have very good discriminatory measures of pain. In any event, I think that sham surgery in a clinical (non-research) setting would be unethical, if for no other reason than the impossibility of truly informed consent.

Phil Reed August 7, 2020 at 5:11 PM

Pat, Your last sentence is a nice point that echoes your idea about limiting placebos to experimental contexts.

Isn't the sham surgery better in the sense that it was less risky?

Stephen Kershnar August 4, 2020 at 7:07 PM

Here is the Merriam-Webster definition of ‘placebo’.

Definition of placebo
1a: a usually pharmacologically inert preparation prescribed more for the mental relief of the patient than for its actual effect on a disorder

b: an inert or innocuous substance used especially in controlled experiments testing the efficacy of another substance (such as a drug)

These are two different concepts. The first is intention-dependent. It is unclear if has an implicit success condition (specifically, it succeeds or usually succeeds in providing mental relief but not disorder relief).

The second is intention-independent. It has a failure condition (inert or innocuous and, thus, not therapeutic).

I wonder if JH is trying to unify two different concepts. This is why he ends up with an unnecessarily complex and seemingly circular analysis of it.



Pat D August 5, 2020 at 2:26 PM


I think that you are right, that this is one source of JH's difficulties. The problem with the Miriam-Webster definition is that 'placebos' actually improve biological processes - e.g. fewer heart attacks, fewer strokes - not just a person's sense of well-being.

I think that there are other sources of JH's difficulties. One is the way he handles the mind-body problem, which seems ultimately to discount mind. From the standpoint of a nested functional hierarchy in living things, the problem relates to differentiating bottom-up and top-down causation. This applies not only to placebo effects, but more broadly to eco-social conditioning of disease - for instance, diseases of despair leading to shortened life expectancy in the rust belt. Or, per Louhiala, the beneficial effects of a caring relationship.

Pat D August 5, 2020 at 2:31 PM

Another source of JH's difficulties is the arbitrary distinction between characteristic and incidental effects.

Stephen Kershnar August 5, 2020 at 3:57 PM


Good point. However, if a drug were to significantly reduce a patient's chance of stroke or heart attack or speed her recovery from them, intuitively, I don't see it as being a placebo. This is true even if it were disguised skittles.

And as you point out, all thought-mediated effects are physiological, at least if one thinks that a person is in part or whole a brain.


Jack Freer August 5, 2020 at 7:40 PM

While the psychological mechanism of placebo effect is undeniable, I believe it leads to bias against patients who respond (more specifically, *when* they respond, since it is a universal phenomenon). Here is a true tale from my resident days. A young woman who was hospitalized for an unrelated problem, developed an asthma attack every evening. The resident on call would come and evaluate the patient and treat her with sub-cutaneous epinephrine (adrenalin) and she would improve. One night, a particularly frazzled (and cynical) resident realized that the nightly asthma attack occured immediately after visiting hours. On a hunch, he ordered a syringe of sub-cutaneous saline --and the patient improved. After that, each night when the resident was called (they had long hospitalizations in those days) the saline worked well. Then the nurses felt badly about bothering the resident and just gave the saline without an order (hey, it was just salt water).

What did the residents learn from this? What did they think they learned about asthma? "Neurotic" young women? Placebos?

Pat D August 6, 2020 at 9:17 AM


What would George Engel say about this patient? Would it have been worthwhile to explore relational conflicts (or perfume or...) as a trigger for asthma attacks following visiting hours?


Jack Freer August 6, 2020 at 9:06 PM


You raise an important point in bringing up George Engel’s perspective. The case is loaded with psychological, social and physical possibilities. A detailed history might well reveal an allergic etiology of the asthma. There might be major depression or generalized anxiety. And, who has been coming to see the patient during visiting hours?

I don’t know Engel’s thoughts on placebo, or if he even wrote about it. But as a conscientious physician, I assume he would have been appalled by the glib sloppy misuse of psychosomatics by the residents.



David H August 6, 2020 at 4:17 PM

Howick’s definition’s rule out placebos and non-placebos for non-diseases

All of Howicks and Grunbaum’s definitions involve treatments that are effective or ineffective for diseases. Obstetrical pain, non-pathological grief are not diseases and the “worried well “are not diseased and so according to Howick’s definitions could not be given a placebo or non-placebo as those are defined as presence or absence of characteristic factors having remedial effects on target disease D. Surely, we want there to be non-placebos and placebos that are effective for such non-pathological distress.


Pat D August 7, 2020 at 9:30 AM


JH doesn't specify what he means by disease. He might include disvalued states like the ones you mention. Is this what he means by relativizing placebo to the patient? But you are right to point out this source of ambiguity in his definition of placebo.


Phil Reed August 7, 2020 at 5:18 PM

David and Pat,

I think Howick wouldn't want to include disvalued states as part of his concept of disorder/disease, since that would give him a crappy concept of disease. But a similar easy solution, I think, would be just to change every instance of "disorder" to "disorder or unwanted life process."


David H August 6, 2020 at 4:17 PM

Too much theoretical relativization:

It is certainly correct to relativize placebos and non-placebos to the disease as antibiotics are placebos for flu but not bacterial pneumonia. But I think it is a mistake to relativize to treatment so , for instance, the same communicative interaction is incidental to one therapy and characteristic of another. That would make the same interactions a placebo and not. I think it is better not to define “placebo” and “non-placebo” in terms of characteristic and incidental factors. I am also worried that one can just construct a theory that makes certain pro-attitudes that are intuitively remedial placebo effects into characteristic features of a treatment. A positive psychology theory or cognitive behavioral therapy could just include an incidental factor as a characteristic given its theory.

David H August 6, 2020 at 4:18 PM

Mischaracterizing “characteristic feature”

Howick defines a characteristic feature as a feature which:

(1) is not expectancy that a treatment is effective

(2) has an incremental benefit on the target disorder over a legitimate placebo-control in a well-controlled trial

I would think a patient could be relaxed by the practitioner without doing so on the basis of an expectancy that a treatment is effective. For example, one can relax someone with a terminal disease and do so without getting the patient to believe his disease is not terminal. So, returning to the non-terminal patient, imagine that relaxation improves a treatment outcome. So, the benefits of relaxation will satisfy condition 2 and condition 1 is satisfied because it is not an expectation that the treatment is effective. But surely getting the patient to relax is more of a placebo than non-placebo


Pat D August 7, 2020 at 9:45 AM


These are examples why Louhiala and Puustinen prefer to call these care effects rather than placebos. The challenge in a research setting is to differentiate therapeutic effects due to an intervention from those due to 'care effects' and to clearly define/design the placebo arm. The placebo arm aims to control for known 'incidental' factors, but tries to control for unknown 'incidental' factors by making control and study groups resemble one another as closely as possible. I think that it is a mistake to think that researchers presume to know "all" incidental factors of an intervention.


Neil Feit August 7, 2020 at 12:07 PM

This is interesting. I don't have a clear intuition that, in David's case of relaxing the non-terminal patient, the treatment is a placebo. If a theory counts it as a non-placebo that doesn't strike me as a count against it. But I am beginning to think that defining "placebo" in terms of characteristic features or factors is a mistake.


David H August 6, 2020 at 4:19 PM

Tentative proposal of an alternative account of placebos and nonplacebos:

There is something dependent, deceptive, and potentially self-defeating about remedial placebos. Given the above problems with distinguishing and too easily relativizing characteristic and incidental factors, I would like to substitute a role in the definition of placebo for a precarious dependent effect upon a pro-attitude. A placebo produces a remedial effect by some sort of pro-attitude to a treatment/context/doctor etc. but in ignorance of the dependency of the remedial effect on the pro-attitude. The efficacious pro-attitude to a treatment is due to a belief that something is remedial independent of the pro-attitude and so awareness of the beneficial work being done by the pro-attitude may render the placebo inefficacious. Awareness of the remedial effects of a placebo can be self-defeating and eradicate the remedial effect. It is important to stress that a placebo’s remedial effects are precariously dependent upon the pro-attitude towards something else to distinguish a psychological placebo from a non-placebo psychological treatment. When one learns that a psychoanalytic or positive psychology confidence building/relaxing/distracting/ etc. cognitive device worked, it doesn’t undermine its efficacy. Awareness of their efficacy is not potentially self-defeating. But awareness that is it the pro-attitude e.g. expectation of a treatment’s success/doctor’s knowledge and abilities etc. that is effective is what makes the remedial effect possible. So my first pass at a definition of generic placebo is something that is not remedial (or harmful) except in a way that is dependent upon the pro-attitude towards the treatment (broadly construed to include the interactions/setting etc.) The definition does away with characteristic and incidental as well as disease. A non-placebo would have a beneficial effect independent of the pro-attitude to the treatment.


Pat D August 7, 2020 at 9:55 AM

This is close to what I say above in terms of care effect, but I think 'care' avoids the negative connotations of 'placebo' terminology. To be clear, I am not referring to something special about the doctor-patient relationship. Any caring relationship will do. For instance, in palliative care, nursing assistants are among the most valuable caregivers.


Neil Feit August 7, 2020 at 2:14 PM

David, your first pass and the ideas leading up to it seem promising to me. Would the account below be an accurate Howick-style definition of generic placebo along your lines?

A treatment process t is a placebo for D iff either (i) t is not effective (remedial or harmful) in patients X for D, or (ii) t is remedial for D in a way that essentially depends on X's pro-attitudes toward t and X's not believing [or perhaps not being justified in believing] that t's efficacy depends on such pro-attitudes.

Notes: The last clause is to rule out non-placebo psychological treatment. For nocebo, replace "remedial" with something like "harmful" in (ii). D might be a disease but might also be some other disvalued condition like normal grief. This seems on the right track to me...

David H August 7, 2020 at 2:38 PM

I like it


David H August 6, 2020 at 4:19 PM

Too Little relativization:

Howick discusses Waring’s paradoxical effects in which there is an exacerbating response on the target disorder produced by a drug that is normally remedial. He then suggests relativizing to the patient. But one might have to relativize to a patient at a time since the same patient may have their say depression (or other illness) lifted at one time by a drug that makes it worse at another time

David H August 6, 2020 at 4:20 PM

Small problem with Grunbaum/Howick’s definition of “intentional placebo”:

(b) “the practitioner believes that the characteristic factors C all fail to be remedial for D (the practitioner believes that t is generic placebo)”

(c) “The practitioner believes that some patients will benefit from the treatment due to one or more of its incidental features.”

I would think something could be an intentional placebo even if the practitioner is uncertain rather than believes the C will be remedial as long as they think the patient will benefit from the incidental features. That uncertainty in the remedial effects of the characteristic features doesn’t prevent t from being an intentional placebo if t is a generic placebo and the practitioner confidently believes that the incidental features will be beneficial

David H August 6, 2020 at 4:20 PM

Small problem with Grunbaum/Howick’s definition of “inadvertent placebo”:

(b) the practitioner believes that some of the characteristic features C are remedial for D.”

I would think the practitioner could be uncertain about the benefits of the characteristic features. It strikes me that still counts as an inadvertent placebo understanding ‘inadvertent” as unintentional or unplanned. The MD is certainly not planning or intending the treatment to be ineffective, it just turns out to be a generic placebo. I think hoping when unsure rather than believing the treatment may be effective should work as a condition (b) for an inadvertent placebo.

Phil Reed August 8, 2020 at 7:29 AM

Objection One: what he says here seems inadequate to show that placebo has a meaningful definition.

I think there is a real question whether, for some concepts, there is a meaningful definition that we can provide or properly analyze. Placebo is a good candidate for such. ‘Placebo’ might be like ‘game.’ Howick wants to deny this. But the way he does this is to show that placebo is a meaningful concept, which is not sufficient to show that we can give a meaningful definition of it.

For example, he thinks that because using penicillin as a placebo in a trial for pneumonia is obviously absurd, this shows “that we work with some concept that involves judgments about what can and cannot count as ‘appropriate’ or ‘legitimate’ placebos and placebo controls” (1368). Fair enough, ‘placebo’ isn’t completely meaningless. Just like ‘game’ isn’t completely meaningless. But the burden is to show that we can give necessary and sufficient conditions for placebo, not to deny that it is a meaningless concept. Specifically, he is responding to Gotzsche, who claims “the placebo concept as presently used cannot be defined in a logically consistent way and leads to paradoxes.” The fact that we can’t call my pencil a “placebo” does not show that Gotzsche is wrong about this.

Phil Reed August 8, 2020 at 7:30 AM

Objection Two: his own account is not that far removed from those who deny that placebo has a meaningful definition.

As we saw, Gotzsche says “the placebo concept as presently used cannot be defined in a logically consistent way and leads to paradoxes.” Howick objects to this. Howick particularly chides Gotzsche for adopting a “practical” approach and characterizing placebos “practically as an [any!] intervention labelled as such in the report of a clinical trial.” Howick inserts the “any!” to poke fun at Gotzsche and Howick comments “it hardly needs remarking that this approach is untenable” (1367).

In fact, it seems that Howick’s own account depends heavily on practice in the same way. Is prozac a placebo? It depends. It depends on the therapeutic theory, on the individual, on expectations, on what we identify as characteristic features. Part of Gotzsche’s point (and Nunn and Turner and so on) is that we can’t say, of any intervention T, that T is (or is not) a placebo. Thus whether an intervention [any intervention!] is a placebo is relativized to practice.

Moreover, Howick’s account explicitly includes paradoxes. As he tells us, whether an SSRI is a placebo depends on the patient, so an SSRI is both a placebo (for some) and not a placebo (for others). As Gotzsche says: the definition leads to paradoxes.

Howick no doubt thinks this is a virtue of his account, but my point is simply that he is not so far from those at the beginning of the paper whom he sets up as his opponents.

Phil Reed August 8, 2020 at 7:32 AM

Clarificatory comment/question:

Howick insists on relativizing the definition of placebos to individual patients.e.g. SSRIs help depression for some patients and worsen it in others. I have two questions about this:

(1) Couldn’t an intervention for the same patient at the same time (to bring in David’s objection) both worsen and remediate a disorder? Just in different ways? So for example, maybe an SSRI causes less gloominess but also increases suicidal thoughts. If so, then a treatment for a single patient would be both a nonplacebo and a harmful intervention. Is that a problem for Howick?

(2) Isn’t the issue of relativizing placebos to individual patients problematic for clinical trials. Suppose I wanted to use SSRI as a placebo in a clinical could I do this knowing it depends on how individual patients respond? And since the concept of placebos, according to Howick, is by definition relativized to patients, then calling something a placebo in a clinical trial seems problematic, no?