Almost everyone knows that smoking harms them, that vegetables are good, and that exercise matters. Health behaviour barely moves on that knowledge. The applied science of health is largely the study of the gap between what people intend and what they do, and of why the biggest lever is usually not the person's mind but the world around them.
The first few decades produced a family of cognitive models of health decisions. The Health Belief Model framed action as a function of perceived susceptibility, severity, benefits, and barriers (Rosenstock, 1974). The Stages of Change model described movement through precontemplation, contemplation, preparation, action, and maintenance (Prochaska and DiClemente, 1983). The Theory of Planned Behaviour tied behaviour to attitudes, social norms, and perceived control (Ajzen, 1991). Underneath all of them sits Bandura's self-efficacy: the belief that you are capable of performing the behaviour, which turns out to be one of the most consistent predictors of whether people act (Bandura, 1977). The shared message is that health behaviour is not driven by information alone but by a bundle of beliefs, and especially by whether people feel able to act.
The modern synthesis collapsed this proliferation into something usable. The COM-B model holds that any behaviour requires three things at once: Capability (the physical and mental capacity), Opportunity (the social and physical environment allowing it), and Motivation (wanting to, reflectively and automatically). Its Behaviour Change Wheel maps intervention types and policies onto each component, giving designers a systematic way to ask which of the three is actually missing (Michie, van Stralen and West, 2011). A companion taxonomy catalogued 93 distinct behaviour-change techniques, giving the field a common language for describing and reporting what an intervention actually did (Michie et al., 2013).
The single most important applied finding, though, is the intention-behaviour gap. It is intuitive to assume that if you change what people believe and intend, behaviour will follow. It largely does not. A meta-analysis of experiments that successfully shifted intentions found they produced only a modest change in actual behaviour, a fraction of the change in intention (Webb and Sheeran, 2006), and the broader literature confirms that a large share of people who intend to act, to quit, to exercise, to eat better, simply do not (Sheeran and Webb, 2016). This is why health education, the leaflet, the warning, the awareness campaign, so reliably disappoints: it may move knowledge and even intention while leaving behaviour where it was.
The cognitive models are better at explaining after the fact than predicting in advance. They can account for a decision once it is made, but leave much of the variance in real behaviour unexplained, which is part of why interventions built on them so often underperform.
The most instructive case is the field's proudest success: the decades-long fall in smoking in many countries. It is tempting to read that as psychology winning, but the honest account is that it was structural and multi-channel, sustained taxation, smoke-free legislation, advertising bans, plain packaging, mass-media campaigns, and accessible cessation services acting together over many years. Persuasion and individual "nudges" were a small part of a largely regulatory and environmental achievement, and treating tobacco as proof that clever messaging changes behaviour badly misreads it. By contrast, interventions targeting diet and physical activity, where the environment keeps pushing the other way, have mostly produced small, short-term effects that decay once the programme ends.
The deepest and most uncomfortable finding concerns the social gradient. Across populations, income, education, working conditions, housing, and the sense of control over one's own life explain more of the variation in health outcomes than any psychological intervention does (Marmot, 2015). This matters ethically as well as scientifically, because a behaviour-change programme that requires time, money, headspace, and stable circumstances to act on will be taken up most easily by people who already have those things, so a poorly-designed intervention can widen the very inequalities it was meant to reduce. None of this means individuals lack agency. It means that treating a structural problem as a personal one is both bad science and, applied carelessly, unfair.
Much of the evidence rests on self-report, short follow-ups, and the gap between efficacy in a controlled trial and effectiveness in messy real life, with effect decay the norm rather than the exception. And the whole field is context-bound: food environments, health systems, and the determinants of health differ so much between countries that an intervention validated in one setting is a hypothesis, not a solution, in another.
How do you close the intention-behaviour gap durably rather than for the length of a study? How can interventions be designed so they help the least advantaged at least as much as the most advantaged, rather than widening the gradient? And how much can any psychological programme add once the structural determinants are accounted for?
The usable core: knowing is not doing, so stop spending on information alone; find which of Capability, Opportunity, or Motivation is actually missing and act on that; expect small, decaying effects from individual-level programmes; and design so you help the disadvantaged at least as much as the advantaged, because the alternative widens the gap.
The ethics come first here, because health is where individualising a structural problem does real harm. The evidence is clear that circumstances, income, control, environment, shape health more than willpower does (Marmot, 2015), so any programme that frames poor health as simply a failure of knowledge or motivation is both inaccurate and liable to punish the people with the least room to act. The responsible test for any health intervention is whether it reaches and helps the people facing the steepest barriers, not just the motivated and well-resourced who would probably have managed anyway.
The first discipline is to stop paying for awareness as if it were a solution. Knowing is not doing (Sheeran and Webb, 2016), so a campaign that raises knowledge or even intention while leaving the behaviour unchanged has failed, however good its engagement numbers look. Use COM-B as a diagnostic before designing anything: ask which of Capability, Opportunity, or Motivation is the actual bottleneck, because the answer dictates the intervention (Michie, van Stralen and West, 2011). If people already want to act (motivation is fine) but the environment blocks them (no time, no facility, the unhealthy option is the default), then more motivational messaging is wasted effort and the fix is opportunity: change the environment, the defaults, the availability. The behaviour-change-technique taxonomy gives a shared vocabulary for specifying exactly what an intervention does, which also makes honest evaluation possible (Michie et al., 2013). The most reliable individual-level techniques are the ones that attack the gap directly, making the behaviour easy and concrete, building specific plans for when and where, and reducing friction, rather than adding more reasons to care.
Two honesty disciplines protect you and your people. Set realistic effect sizes: outside tobacco-style structural programmes, most behaviour-change effects are small and fade, so a workplace wellness scheme promising transformation is overselling (and the board will eventually notice). And watch the gradient inside your own population: a programme joined mainly by the already-fit, already-healthy staff has not improved health so much as rewarded it, and the honest version is designed to reach the shift-worker, the carer, and the stressed rather than the enthusiast. The applied field has serious vehicles worth drawing on here: Susan Michie and UCL's Centre for Behaviour Change built COM-B and the Behaviour Change Wheel as practical design tools, and Robert West's work turned smoking cessation into a well-specified, evidence-led practice rather than an exhortation.
The transferable lesson is to be honest about what health messaging can and cannot do. Awareness and fear campaigns move knowledge more reliably than behaviour, so promising that an information drive will shift a population's diet or activity levels is a claim the evidence does not support. Where campaigners can genuinely help is in building the political will for the structural measures that actually work, since those (taxation, environmental change, service provision) are decided in the political arena, not the doctor's office. And the framing matters: presenting health as a matter of individual virtue invites blaming the sick, while the evidence points at conditions, which is both more accurate and, handled with care, more unifying.
The tobacco story is the template: the wins come from structural, multi-channel, sustained action, taxes, regulation, environmental change, and accessible services, far more than from asking individuals to try harder. Build interventions on a COM-B diagnosis of where the behaviour actually breaks for the target population, and expect that for most health behaviours the binding constraint is Opportunity, the environment, not a deficit of Motivation (Michie, van Stralen and West, 2011). Above all, design against the social gradient. Marmot's own applied answer, developed through the WHO Commission on the Social Determinants of Health and the "Marmot Review", is proportionate universalism: universal measures delivered with an intensity proportionate to need, so that the steepest barriers get the most help and the intervention narrows rather than widens the gap (Marmot, 2015). His books, The Health Gap (2015) and Status Syndrome, are the accessible applied statements of that programme. The global caveat is decisive: food environments, health systems, and the shape of the gradient itself differ enormously between countries, so a measure that worked in one system is a starting hypothesis elsewhere, to be tested against local conditions rather than imported wholesale.
Three habits. Treat information as necessary but rarely sufficient, and put the effort into Capability and Opportunity rather than assuming a knowledge or motivation deficit. Diagnose before designing with COM-B: find which of the three is actually missing, because the wrong intervention on the wrong component does nothing. And judge every health intervention by whether it narrows the gradient, reaching the people with the steepest barriers, rather than by whether the motivated few who joined got a bit healthier.
.png)