Smoking is not addictive

by Carl V Phillips

Now that I have your attention, this long essay is my response to the frequent requests to summarize my analyses of the concept of addiction, particularly how it relates to tobacco product use. I should note that the headline is based on the most commonly-accepted definition of “addictive”. I will work my way through that to other senses of the word under which smoking might be considered addictive.

Since this is really a question of linguistics, not science per se, we need to start there. Sometimes a word is doing work and sometimes not. By “doing work” I mean that the meaning of a sentence hinges importantly on what the word means. So in the previous sentence, “meaning” and “what the word means” are doing the fundamental work, while “importantly” is doing some work (offering a caveat that there are fuzzy areas at the border). The word “sentence” is just there for structure, and its exact meaning does not matter; indeed, it should be obvious that the point is not restricted to series of words that form proper sentences. The “I mean that” phrase is purely stylistic and “hinges” is merely necessary for the sentence structure, and is basically an ornamental choice of verbs.

How well as word needs to be defined depends on how much work it is doing. If a narrative mentions someone is tall, but makes no further use of that information, it does not matter what the author considers “tall” to mean. If someone says “you should be able to find him there; he is tall and wears glasses”, then you need to have a rough idea of what the speaker would consider tall, and of the context (whether the room is full of Guatemalans, Swedes, or kindergarteners), which you can probably work out adequately. When you read “tall men have 18% greater income on average”, however, the word is doing a huge amount of work and you have little idea what it means. The precise outcome measure implies a bright-line definition, and a particular context (population), neither of which you know without further digging. You might think you learned something from reading that sentence, but you have not really. (Also missing is a definition of “average”; is it the mean — i.e., the statistic merely says that millionaires are more often tall  — or median?)

The use of “addiction” in serious discourse about tobacco products is even worse than that last example. Some of those using the word are using it as a meaningless narrative throw-away, but that is not acceptable. The word is fraught with weighty connotation. The word always does a lot of work, invoking a relationship with the product that is more than just consumption or even habitual consumption. It also implies “this is a Bad Thing” with innuendo about the way in which it is bad. But the word is not well-defined. There is not even an adequate vague definition like the one that lets you look for the “tall” guy in the room. Anyone who says that smoking is addictive thus needs to explain what they are claiming, especially in the many cases where the authors lean heavily on the assertion. They never do.

When a word is doing work in serious analysis and its definition is not generally agreed upon or exists only as a vague colloquial concept, authors need to define how they are using it. As another example, the phrase “tobacco product” is not clearly defined, so if it were doing any important work in this essay (it is not), I would define it. (I discuss the need to do this further here.) We do not know what someone means when they say tobacco products are additive, and it is clear the authors also seldom have any idea. This is true even when they go so far as to make quantitative claims (“highly addictive”; “more addictive than heroin”), even though they certainly cannot quantify something they cannot even define. All we usually learn from the use of the word is that the author is not a very good scientific thinker, and is probably trying to create propaganda rather than knowledge. (Not useless bits of information, of course, but presumably not what the author wanted to communicate.) A large portion of usage of the word seem to just mean “someone gets pleasure from it” or merely “someone uses it” even though “I personally disapprove of it.” That, obviously, is the exercise of propaganda and not honest communication.

Over the last fifteen years, I have made a practice of responding to statements about tobacco products being addictive by asking the author to define the word or explain what they mean. I have probably done that a thousand times (not counting times I reply to a tweet from some PR drone who I expect could not even define “prevalence”, let alone “addictive”). Usually I get no reply, which presumably represents a complete inability to answer (just how often does someone resist a chance to reply along the lines of “hah, you think you are so clever, well your question is actually easy to answer….”?). I have gotten at least a hundred responses, however, and not one of them has been (1) a legitimate definition that (2) included some or all tobacco products and (3) did not also including almost every other consumption choice. That alone, absent any analysis or further thinking, is sufficient to establish that those who make the claim have no idea what they are claiming.

Many of those I have challenged about the use of the word self-identify as “addiction experts”. The fact that even the self-styled experts cannot respond to “what did you mean when you said smoking is addictive?” with anything other than sustained stammering or nonsense makes me confident that my failure to identify a viable definition was not due to overlooking common knowledge.

Five years ago, I wrote a post in which I laid out what such a definition must and must not do, and presented the challenge to offer any definition under which tobacco product use is addictive. Five years, 59 comments on that post, and a hundred subsequent conversations have done nothing to shake my conviction that no one who makes that statement really knows what they are claiming.

You do not need to read the previous post to understand this one (though it offers a deeper dive on some points and triangulation if some things are not clear). To summarize: A candidate definition needs to mean something other than repeated consumption; the word is clearly meant to imply more than that. It cannot just mean serious dependence (i.e., that some consumers suffer serious withdrawal symptoms when they cease consumption). Dependence alone might keep someone using, but that is its own concept and, in any case, is seldom a good description of tobacco product use. It definitely cannot just mean tolerance (the effects diminish with extended use); there is already a word for that too. If something is addictive it must be bad, since the connotation is clearly negative, thus eliminating the large portion of candidate definitions that are merely about “needing” to keep doing something, which also encompass breathing, sleeping, or spending time with your children.

More generally, it must be some special concept and so cannot describe a large portion of normal consumption choices. This eliminates one of the favorite knee-jerk faux-definitions that gets offered, “persistent consumption despite the costs.” Every consumption choice has costs. When those offering this phrase desperately amend it to “…health costs”, it narrows a bit, but not much. Most consumption choices have health costs: reading a book is sedentary; hanging out with friends exposes you to infections. A further desperate attempt to amend that to something like “…major health risks” also fails, because it is clearly not what people mean. No serious commentators argue “smoking is addictive but snusing or vaping cannot be addictive — even if the consumer’s behaviors and feelings are exactly the same — because of the lack of major health risks.” No one says that an extant disease that dooms someone to die soon creates immunity to addiction, even though it means they face no further health risk. It is often argued that someone might as well let themselves become addicted to opioid pain killers when facing their imminent death, but the stated reasons for that do not include saying that is that they are immune to the scourge of addiction. Rather, they often include the observation that she might become addicted but it does not matter.

Similarly, the definition cannot merely be just that something is an acquired preference — that someone wants it more because he has consumed it before. That also covers a large portion of consumption choices. Genres of art (everything from classical music to a gripping television series), for example, often exhibit very strong increasing marginal benefits (the technical way of saying “the more you have consumed, the more you like the next bit of consumption”). Indeed, we are generally more interested in consuming something we tried and liked than we were interested in trying it the first time (ever heard the phrase “you don’t know what you’re missing”?).

One thing that is clear about what people seem to want the word to mean is that it is about behaviors and/or preferences (desires). Thus the definition must be about these, and so any proposed definition that is based on physiological or biological measures is necessarily wrong. This is simple to demonstrate with a thought experiment: Imagine someone can take-or-leave a drug and does not bother to consume much of it, suffering no problems from it, but she exhibits the brain-scan activity that is the proposed physiological definition of “addiction”. Since she obviously does not fit what people mean by addiction, it cannot possibly be that a physiological state can defines the condition (though it might be associated with it, with this individual representing an oddity). One favorite in this neighborhood, using the word to refer to mere dopamine production, is even worse; that error combines the failure of thinking it is possible to create a physiological definition and the failure of implying that “addiction” refers to any derivation of pleasure from consuming.

The only existing candidates for an “official” definition — something that is a reasonably proper definition and is widely endorsed — exclude tobacco product use. There is a family of definitions that describe a syndrome that sometimes includes some of the above concepts — dependence, tolerance, persistence, health effects — but whose most concrete condition is that the consumption and/or pursuit of the consumption substantially disrupts someone’s social and economic functioning.

This condition makes for a reasonably valid as a definition of a meaningful phenomenon, unlike most of those that are proposed. Syndrome descriptions like this (and like basically every other definition of a psychological condition) are never fully-valid definitions. They almost always merely replace the purely intuitive notion of the meaning — “I know it when I see it” — with a list of more specific “I know it when I see it” conditions. They almost never quantify or otherwise define how much you have to “see” of one those conditions before it counts as being present. Indeed, it is difficult to imagine doing so in most cases. Moreover, these syndrome quasi-definitions almost never hint at a method of aggregation — i.e., how strong one condition must be to make up for another being weak (this is largely moot in the present case, since it is really only about the one condition). Nevertheless, such definitions can feel generally right, and that is the best we can hope for.

Something along the lines of these proposed definitions seems to be what most authors are trying to evoke when using the word. Someone is addicted if a consumption choices is badly screwing up his life, yet he keeps at it and he exhibits some or all of the “hooked on drugs” conditions like dependence. This describes some (but far from all) use of “hard drugs”, as well as some unhealthy relationships with video games, gambling, and television. In some cases it might describe an unhealthy pattern of something that is normally a healthy activity, like sex (or the pursuit thereof), exercise, work, or owning cats. But one thing it clearly does not describe is tobacco product use, except perhaps under some very bizarre and rare circumstances. Tobacco product use does not interfere with functioning, except to the trivial degree that all consumption choices do (i.e., it takes away a bit of time and money from other pursuits), and indeed enhances functioning in many ways.

(As an aside, it is worth emphasizing a critical point that is often misunderstood. If this or anything similar defines addiction, then it is not a property of the drug or other consumption choice itself. It is a property of how a particular individual relates to the consumption, which is often heavily influenced by a person’s circumstances. If two people who use heroin have the exact same physical and preference responses to it, but one leads lead a perfectly functional life and buys the drug with his disposable income, while the other is unable to keep a job because of his habit and steals from his friends to score, only one of them is addicted. One of the many implications of this is that the “opioid addiction crisis” is as much a description of unfortunate social circumstances as it is about drug use.)

Some versions of this dysfunction-based family of syndromic quasi-definitions qualify as “official” because they are endorsed by government agencies and self-proclaimed arbiters of psychology, like professional societies. For example, one agency of the US government endorses such a definition, despite other agencies persisting in calling tobacco product use addictive (without ever offering a viable definition for what they mean). In my experience, when you ask an “addiction expert” what she means by smoking being addictive, she will point to one of these definitions, failing to recognize the obvious problem that they do not describe smoking.

Thus the title of this essay: According to these — the most widely-endorsed family of candidate definitions — smoking is not addictive, nor is any other tobacco product use. Indeed, Michael McFadden informs me, based on his research on the history of anti-smoking literature, that up until sometime in the 1980s, the word was not used by serious anti-tobacco commentators. They often took pains to avoid it or even explicitly said that what they were describing was different from what we thought of as drug addiction. But as an early move in their practice of misappropriating words to create irresponsible rhetoric (see also the more recent case of doing that with “epidemic”), tobacco controllers started using the word as a trick, to evoke the specter of a scary phenomenon that is not actually an accurate description.

Because these “substantial dysfunction” definitions are the ones endorsed by governments and medical organizations, anyone using “addiction” in a government or health science context must either explain what they mean by the word or adhere to those definitions. Thus, the use of “addiction” to refer to tobacco product use in those contexts is simply wrong, absent an explanation of what is meant (which, again, is never offered, and those using the word are apparently not capable of explaining what they mean). It is purely misleading rhetoric and a pretty strong indication that the author should not be taken seriously.

However, this does not mean that this is the only valid use of the word in any context, nor that an explicitly-defined alternative use of the word would be wrong. Words mean somewhat different things in different contexts, and official definitions frequently do a poor job of capturing what most people mean by a word. Smokers say they are addicted, after all, and they presumably mean something by this.

The starting point for the common language definition pretty clearly can be found in the consumption choice having an unusually strong tendency for past consumption to motivate future consumption. In the 1970s, some Chicago School economists offered a quasi-definition based on this. Like most of their ilk, they were too clever by half, and what they came up with (which – again, true to form — was mostly about impressing their colleagues rather than advancing human knowledge) was wrong because it was both too broad and based on inaccurate assumptions. But they were on to something. Their biggest failure was anchoring the definition on the presence of any departure from one assumption in the standard welfare economics models, an assumption that happens to be badly incorrect. The assumption is that consumption choices exhibit diminishing marginal returns across all quantities of consumption — i.e., that each additional bit of whatever is consumed is less rewarding, and thus less wanted, than the previous bit. The Chicago economists ineptly “defined” addiction to be any departure from that assumption, such that there are increasing marginal returns.

This is obviously flawed because almost every consumption choice has increasing returns (increasing desire to consume) as a result of previous consumption over some range of consumption. Returns eventually decrease in most cases (there is nothing you want an unlimited amount of), including with “addictive” drugs. But before they get there, there is a lot of complication and variation.

Recall the previous observation that addiction obviously does just mean acquired appreciation. The increasing returns for consuming your favorite music — you had little or no desire to listen to it before you first did, then you wanted to hear it again sometimes, and now it is on your daily playlist — are clearly not addiction. If you know chocolate cake is on the dessert menu, you might want it to order it, but if you your companion say “no dessert, just the check” you do not really care. But if you ordered it, took one bite and loved it, and then your companion said “pack that up now so we can leave” you would be disappointed and insist on enjoying another bite. If you already have a functioning car, your desire to acquire a car is hugely diminished (the favorite simplistic Econ 101 example of diminishing returns). But someone living in a city might be content to not own a car for years, get one, and then find they desperately want to replace it when it dies, demonstrating a highly increased return resulting from past consumption.

There are all sorts of variations on consumption choices having non-monotonicities or discontinuities in terms of returns to scale. Many people find that when they are playing a computer game or watching television that they cannot pull themselves away. Once they do, however, within a few minutes the urge is gone and they might not even bother to return to it. You might find yourself just dying to see the next episode of a good television series after watching the latest one, but then a month later you notice you have four new episodes on your DVR and think “meh” and delete them. Or if you prefer a more literary example, there is the book that you lost sleep to keep reading, but when you noticed it on your nightstand a week later you just pulled out the bookmark and returned it to the library.

An interesting one is cheese: Many people report that if they eat a lot of cheese, reaching the point where they do not want any more at the time, they find they have a strong desire to eat more a few hours later. But he next day they are back to their normal level of desire for cheese. That example gets closer to the drug use experience — it is clearly not about momentary captivation (as with turning off the television), nor about having acquired habits and made investments that depend on continuing the consumption (as with replacing the car).

In short, there is no consistent pattern of returns to scale as a function of how much of something you have consumed before. Economists’ simplistic assumption of diminishing returns seems to be wrong more often than it is right. The change in returns varies across what is being consumed, quantity, and recentness of past consumption, as well as across different consumers. Still, it is possible to identify particularly extreme patterns of increasing returns and make those the anchor for a non-dysfunction-based definition of “addiction”. It is not going to be a bright-line definition like the economists’ hubris led them to believe they had. But it should be clear by now that finding a bright-line definition is out of the question.

To their credit, those economists quietly amended their “definition” to add that accumulated past consumption must lower someone’s welfare. Thus there has to be some “running fast just to stand still” quality: After smoking for a while, the marginal benefit of having a cigarette is increased (you want one more than you did before you smoked), but part of why it is increased is because being bereft of a smoke has you feeling worse than you would have if you had never smoked. Thus, part of that greater benefit is used up just digging yourself out of the hole. I suspect most people who say “I am addicted to smoking” would find that this description rings true.

With this, we have a reasonably concrete concept that seems to capture what smokers are getting at when they use the word. “Addiction”, in this sense, means something with a unusually huge tendency for past consumption to increase the marginal value of future consumption. But, alas, this is not that cause for celebration that it would be by itself (greater returns means greater welfare gains), because it also means a lot of that increased marginal value is used up just getting back up to baseline.

There is still plenty of fuzziness: Marginal value is still sometimes decreased by past consumption, depending on quantity and timing (immediately after having one cigarette, someone wants another much less). The marginal value in the television, computer game, or chocolate cake examples might be huge while currently consuming, but lowered dramatically five minutes later, and very low after binging for a day and thinking, “geez, that was too much!” Thus, the increased marginal value condition needs some clarification like, “whatever happens today, which can be complicated and non-monotonic, consumption over the past year increases the marginal value of consumption tomorrow.” I will not attempt to make that any more precise, trusting that the reader just gets it and knows it when s/he sees it.

The “running to stand still” effect requires additional nuance. That effect might be so great that someone’s current welfare, even after having just indulged, is lower than it would have been had they never started consuming it in the first place. Believing this to be the case presumably explains why many smokers lament, “I wish I had never started.” This is the specter that prohibitionists are trying to invoke with rhetoric like, “a generation addicted to vaping.” But the “digging oneself out of the hole created by past consumption” effect will vary quantitatively. (Moreover, it is impossible to measure or even subjectively assess very well: Who can know how she would be feeling and functioning today had she never smoked?)

It is ambiguous whether this condition — that the hole is (or will become) so deep that the consumer is always below her counterfactual never-started baseline — is being invoked when it is asserted that smoking/gambling/junk food is addictive. But it is built into the innuendo, which goes something like: “perhaps addiction in this sense is not at all as bad as the dysfunction from the official definitions, but it is still unambiguously bad because the net welfare effects are negative.” If that is part of the implicit definition, the word is clearly used far too expansively. For many consumers, not all of the marginal benefit is used up to get out of the hole. If it is not part of the definition, however, it is difficult to see any reason that addiction is necessarily bad.

Vaping or snusing might well (sometimes) be addictive in the sense currently under discussion, but the hole appears to usually be pretty shallow. Yes, some of the benefit of consumption is used up feeding neurological expectation or fulfilling a habitual urge. But many vapers and snusers express joie de vivre about their habit (as do some smokers, but a far lower portion), and are confident they could easily stop if they wanted to, but they do not want to. This makes it impossible to believe that the net welfare effect is negative. With this observation we have a departure from the condition that addiction is necessarily bad.

But there is a bigger problem here. Nicotine, heroin, television, computer games, gambling, and junk food are not the only consumption patterns that fit this proposed economic definition. Many traditions and rituals have high marginal returns due to doing them in the past, but with most of that due to the hole — i.e., you would feel bad if you did not do it. You might be totally sick of a particular social group or activity, but keep hanging out it because it would gnaw at you to no longer be a part of it (even though you would obviously not miss it at all had you never done it). You might be far happier at a tropical beach on December 25 but you cannot stand the thought of missing Christmas at the family homestead for the first time in your life.

The fact that you, dear reader, are (I hope) thinking “oh, yeah, those do fit the pattern, but it never would have crossed my mind to call them an addiction” means there is still a fundamental problem here. Cigarettes cannot be said to be addictive (in some sense) merely because they exhibit the same economics as increasingly unpleasant family reunions. It is something more.

This is the point in the story in which it, like many good stories, becomes about sex.

Yes, really.

In humanity’s literature about “morals”, sex is the most reviled aspect of the human experience. (It is also one of the most celebrated, but we are focusing on the negatives here.) Sexual desire is a threat to our concept of ourselves as rational and autonomous. It is a reminder that we are ultimately animals, a mere biological accident and still a part of the biological soup. Sex is a great equalizer, offering big welfare benefits to both the “worthy” and the “unworthy”. In various ways, it threatens the productivity and social orders that benefit the elite. The combination of these, consciously assessed or subconsciously felt, has resulted in millennia of reviling.

In the spectrum of human feelings, sexual desire is most similar to hunger (a word which, of course, is often used metaphorically to describe sexual desire). But unlike hunger, which is at its core a biological signal that you are going to start failing to function if you do not take particular actions, sexual desire is your biology demanding you pursue actions that are not necessary for your personal well-being. Much of the time it bubbles up, you would prefer to turn it off (well, I suppose, unless you have an exceptional lifestyle). But you cannot.

The drive to gamble or play a video game, even though it might qualify as an addiction under the official definitions, seldom feels like hunger. The desire to smoke (or snus or vape), however, often does. It bubbles up from somewhere deep in your brain rather than seeming like thought. You might have a conscious thought-based fantasy associated with the desire (envisioning what you are going to do with the cigarette or other human body), but that is not where the preference is centered, unlike when you contemplate what tactic you want to try next in your favorite video game.

In addition, and perhaps most important here, the desire to consume tobacco produces the illusion of physical sensation, one that like an itch demands action. You “feel” the back of your throat, lips, lungs, or other involved body parts “asking” to experience tobacco use, as occurs with sexual desire and a different list of body parts. This sensation is all in your brain, though. Of course, all physical feelings are technically in your brain, but most of them are your brain modeling something that is actually happening at the site. Not so these particular tingles.

Our fate is to suffer (and enjoy) that oft-reviled preference pattern of sexual desire, and its frequent inconvenience and pesky reminders that we are not the purely cognitive and spiritual beings that many would prefer. So be it. But to choose it? Voluntarily re-creating those horrors (as they are viewed by some) by getting hooked on tobacco products is different. That need not happen. The reviling of addiction (in the sense in which it describes smoking) is a reviling of voluntarily increasing feelings that are similar to sexual desire. Of course, that might mean some people’s motivation for hating addiction is that sex is a gift from the gods (or something like that) and it is blasphemy (or something like that) to presume to choose a profane shadow of its Platonic perfection. But given the type of people who condemn others’ choices to become addicted, it seems like a safe bet that reviling rather than reverence is at the heart of their views.

With that thought, I will abruptly end this, without trying to summarize or go out on a definitive note. I have been working on a piece focusing on how addiction, in this sense, is perhaps best understood a Platonic shadow of sex. Like this essay has been years in the making (which is to say, I have not been working on it). So hopefully there will be a sequel to this.