Self-medication: Why the behaviour is not the problem

When people notice they are relying on a habit too much, the evening glass of wine, the late-night scrolling, the work that bleeds into the weekend, the usual move is to look at the habit and try to stop it. Drink less. Scroll less. Work less. The behaviour is named as the problem. Stopping it is named as the goal.

This is rarely how change actually works. The behaviour is not random, and it is not weakness. It is doing something for the person. It is reducing some internal state that the person does not currently have another way to reduce. That is what self-medication actually is: not a category of vices, but a category of function. Anything that reliably softens an unwanted feeling can serve the role.

Once a behaviour is doing this work, removing it without replacing the function tends to fail. The function is still there. The pressure is still there. Something else will fill the gap, often something with a similar shape. The useful question is not how to stop the behaviour. It is what the behaviour is achieving, and whether that same achievement can come from somewhere that costs less.

The clinical use of the term comes from work by Khantzian (1997), who described how people with substance use difficulties were often using their substance of choice to manage a specific affective state, not pursuing a substance for its own sake. The concept has since broadened. Self-medication is now understood as any behaviour that reliably reduces an unwanted internal experience, and the category extends well beyond drugs and alcohol.

The forms it takes in adult life are familiar. Alcohol after a difficult day. Cannabis to wind down. Food to take the edge off restlessness. Caffeine and sugar to push through fatigue that is really exhaustion. Work to outrun the discomfort of stillness. Exercise pushed past health into compulsion. Scrolling, gaming, or shopping to disappear from the room for a while. None of these is inherently a problem. The form does not determine the function. What turns any of them into self-medication is the role they are playing in the person's emotional life.

The clinical line is whether the behaviour is restoring capacity or replacing it. A coping behaviour that lets a person rest, recover, and re-engage with what they care about is doing something different from a coping behaviour that helps them avoid a feeling indefinitely. Both look similar from the outside. The internal function is what tells them apart.

Three mechanisms keep self-medication patterns running, and they reinforce each other. The first is timing. The relief these behaviours provide is immediate, while the cost is delayed. The brain learns from immediate outcomes more strongly than from later ones, because immediate outcomes are easier to attribute to the action that produced them. By the time the cost arrives, often hours or days later, the brain has already filed the behaviour as something that worked.

The second mechanism is what the research now describes as experiential avoidance: the persistent attempt to control or escape unwanted internal experiences, even when doing so creates problems elsewhere. A comprehensive meta-analysis of 441 studies and over 135,000 participants found moderate-to-large associations between experiential avoidance and every internalising condition examined, including anxiety, depression, generalised anxiety disorder, and post-traumatic stress disorder (Akbari et al., 2022). The strategy is universal because the underlying need is universal. People want unwanted feelings to be smaller. Self-medication is one of the most reliable short-term ways to make that happen.

The third mechanism is repetition. Behaviours that are performed in the same context, with the same trigger, and that reliably produce the same internal effect, become automatic over time. Habit research consistently finds that habit formation does not require months of conscious effort, but it does require the conditions that link a cue to a response. Once those conditions are met, the behaviour stops being a choice the person actively makes and starts being something that happens. This is why people often describe self-medication as something they did before they realised they were doing it.

This is also why willpower is the wrong frame. By the time a self-medication pattern is established, the behaviour is no longer being driven by a daily decision. It is being driven by an automatic response to a felt state. Asking the person to try harder is asking them to override an automaticity that is, by definition, not asking permission.

Recognising what a behaviour is achieving requires a different kind of looking. Not at the behaviour, but at what immediately precedes it and what immediately follows it. Three questions tend to be useful, in this order.

First: what feeling am I in just before I reach for this? Not the story about the day. The actual internal state. Tired. Restless. Anxious. Empty. Bored. Lonely. Angry. Numb. The feeling does not have to be dramatic to drive the behaviour. Mild discomfort that has been building for hours is more often the trigger than acute distress.

Second: what does the behaviour actually do to that feeling? Does it numb it, distract from it, soften it, postpone it, replace it with a different feeling? The honest answer is usually some combination, and the combination matters. A behaviour that numbs is doing something different from a behaviour that distracts.

Third: what would I have to feel if I did not do this? This is the question most worth sitting with. The answer is the function the behaviour is serving. It is also, often, the thing the person is most reluctant to look at, which is partly why the behaviour exists.

This is the point at which compassion becomes a clinical move, not a moral one. The behaviour was reasonable given what it was doing. Whatever it is now, it began as an attempt at self-care. Looking at it accurately, with the function visible, is what allows the next step. Looking at it with shame or judgement keeps the person stuck in the same loop, because shame and judgement are themselves unwanted feelings that the behaviour is well-positioned to soften.

Once the function is visible, the change strategy reorganises itself. The question is no longer how to stop the behaviour. It is which other behaviours can serve the same function, ideally with lower cost, and how to make the new behaviour available at the moment the old one would have been triggered.

The strongest evidence for behaviour change at this level is around what are called implementation intentions: specific, pre-decided plans of the form 'when X happens, I will do Y'. The plan binds a chosen response to a chosen cue, which lets the new behaviour run on something closer to automaticity rather than on willpower. Implementation intentions consistently produce small to moderate effects on behaviour change across health domains, and they are particularly useful when the existing behaviour is already cue-driven, which self-medication usually is.

In practical terms, this looks like specifying both halves of the loop. Not I want to drink less, which leaves the cue intact and the response unchanged. More like when I get home from work and feel the pull toward a drink, I will sit on the back step for ten minutes first. The new response does not have to be the perfect long-term replacement. It only has to occupy the same place in the loop, briefly enough to interrupt the automaticity and create space for a different choice.

The realistic timeline matters. Australian research synthesising twenty habit-formation studies found that new health behaviours took a median of fifty-nine to sixty-six days to feel automatic, with substantial individual variability ranging from four days to over three hundred (Singh et al., 2024). The implication is not that change is slow. The implication is that change is non-linear, and the first weeks usually feel worse, not better. The new behaviour has not yet been reinforced by the brain. The old one has been reinforced for years. The early period is where most attempts fail, not because the strategy is wrong but because the discomfort of doing the unfamiliar thing is louder than the eventual cost of the familiar one.

Some self-medication patterns are not workable as a self-help project. Substance dependence with physical withdrawal, restrictive eating that is reducing weight or function, depression that is severe enough to make engagement itself difficult, and trauma symptoms that drive the behaviour from below are all situations where the right next step is professional support, not a different cue-response plan.

The evidence for working with these patterns clinically is reasonably strong. A 2024 meta-analysis of psychosocial interventions for adults with co-occurring anxiety and substance use disorders found that integrated treatment approaches reduced symptoms of both conditions across post-treatment and follow-up, with the largest effects when anxiety and substance use were addressed together rather than sequentially (Nardi et al., 2024). The clinical point is that the pattern usually has more layers than a single behavioural plan can hold, and a longer relationship is what allows those layers to come into view.

A useful internal test is whether the behaviour is restoring something or hiding something. Restoration is sustainable. Hiding is not. If the answer is unclear, that uncertainty is itself a reasonable reason to bring another mind into the work.

Self-medication is what most people do, in some form, with internal states they have not yet found another way to meet. The pattern was rational at the time it was learned, and it has been doing useful work ever since. Seeing what that work actually is does not make the behaviour worse. It makes it possible to do it differently.