What the science says

Consensus

COM-B is quickly stated. Any behaviour sits at the meeting point of Capability (the physical and psychological capacity to do it, from strength to skill to self-control), Opportunity (a physical and social environment that permits and prompts it), and Motivation (the reflective wanting of deliberate goals and the automatic pull of habit and impulse). Take one away and the behaviour does not happen. Around this core, the Behaviour Change Wheel adds an outer ring of nine intervention functions, from education and training to environmental restructuring, restriction, and coercion, and a further ring of policy categories, so that a diagnosis on the inside points to a family of actions on the outside (Michie, van Stralen and West, 2011). The Behaviour Change Technique Taxonomy then supplies ninety-three named, standardised techniques that map onto those functions, giving the field a shared vocabulary for describing what an intervention actually contains (Michie et al., 2013), and the Theoretical Domains Framework offers a finer, fourteen-domain elaboration for research use (Cane, O'Connor and Michie, 2012).

‍

The distinctive fact about COM-B, and the one that explains its standing, is how it was made. It was not another investigator's pet theory. Michie and colleagues ran a systematic review of nineteen existing frameworks of behaviour change and assessed each against three criteria: whether it was comprehensive, whether it was coherent, and whether it linked to a defensible overarching model of behaviour. COM-B is what remained when those nineteen were distilled to their common core (Michie, van Stralen and West, 2011). That pedigree is why UK bodies including NICE and Public Health England took it up and applied it to public-health guidance (Michie, Atkins and West, 2014): it reads less as a rival claim than as a summary of the field's shared knowledge.

It is a map, and that is its real use

Here is the part that matters most once you have met the other frameworks in this series, and the part the applied treatment in L5-14 does not dwell on. Because COM-B was distilled from the field rather than added to it, nearly every other tool we have covered turns out to be a region of the same map. The Theory of Planned Behaviour (L6-01) lives almost entirely inside reflective motivation, with a corner of opportunity in its notion of perceived control, which is exactly why it predicts intention well and stumbles on behaviour: it is working one province of the map and calling it the whole country. Fogg's model (L6-02) is a coarser cut of the identical territory, folding capability and opportunity together into "ability." Nudges and defaults (L5-13, L5-16) operate on physical opportunity. Social proof and norms (L4-03, L5-18) act on social opportunity. Habit (L1-04) is automatic motivation. Messenger and authority effects (L5-17) work through motivation and social opportunity together. None of these is wrong; each is a specialist working one region. COM-B's job is to hand you the whole map so you can see which region your problem is actually in, and therefore which specialist to call.

Controversies

The serious debate about COM-B is not whether it is useful but what kind of thing it is, and it is a genuine, unresolved dispute worth presenting from both sides. Jane Ogden (2016) put the sharp end of the critique in a paper pointedly titled "Celebrating variability and a call to limit systematisation." Her argument is that the Behaviour Change Wheel and its taxonomy impose a tidy classification on a domain that is irreducibly variable and context-bound, and that a scheme like this describes and organises behaviour change without explaining or predicting it. It tells you where to look; it does not tell you why an intervention worked here and failed there. The framework's proponents replied, in the same exchange, that a shared, systematic language is itself a real scientific advance, because a field where everyone names their interventions differently cannot accumulate evidence at all. Both are right about something, and the honest resolution is not to pick a winner but to hold the distinction: a comprehensive map that every team can read is genuinely valuable, and a map is still not a predictive theory.

‍

That distinction has teeth, because comprehensiveness has a cost. A framework broad enough to place every behaviour and every intervention somewhere is also, by construction, hard to falsify: if any result can be slotted in after the fact, the framework has not stuck its neck out. And when COM-B has been tested head to head against an older model, its edge is modest rather than decisive. Howlett and colleagues (2021) built a COM-B model of a real behaviour, weekly sitting time, and compared how well it predicted behaviour against the Theory of Planned Behaviour. The COM-B constructs did predict, explaining about 27% of the variance in sitting time, but the older Theory of Planned Behaviour explained about 23%, a broadly comparable result rather than the newer framework sweeping the older one aside. Wide adoption, in other words, is not the same as demonstrated predictive superiority.

Limitations

The limits follow from what COM-B is. Because it is a classification and design framework rather than a causal theory, the actual effect sizes live in the techniques, not in the map: the moderate, context-dependent success rates of any intervention come from the specific behaviour-change techniques used, and those vary by population and setting. The three-way cut is also coarse in places. "Automatic motivation" quietly bundles habit, raw affect, and identity, which behave differently and can pull in opposite directions. Assigning a given barrier to capability versus opportunity versus motivation is often a judgement call rather than a reading off an instrument. And the wheel's intervention functions are not culturally neutral: coercion and restriction, which sit on the outer ring as ordinary options, are routine public-health levers in some systems and politically unacceptable in others, so a design that is defensible in one country is a live question in the next.

Open questions

The deepest open question is whether a fully comprehensive framework can ever be predictive, or whether comprehensiveness and prediction are simply different jobs that no single scheme does at once. Below that sit practical ones: does the finer fourteen-domain elaboration predict better than the three-way cut, or just cost more to measure? And how well does COM-B combine with implementation-science frameworks and system-level modelling when the target is a whole population rather than one person's behaviour?

So what

The mistake with an integrative framework is to treat it as if it were a productive theory that will generate the answer. It will not, and expecting it to is what Ogden was warning against. Its value is orientation. Used well, COM-B is the index to the whole toolkit this series has built, and the discipline of consulting the index before acting is worth more than any single tool. To see how, take one concrete goal and carry it all the way through: you want far more people to cycle to work for short trips.

‍

1. Locate the problem on the map first. Before reaching for a favourite tactic, place the stuck behaviour on the map. For cycling that means asking, for your specific target group, which region is actually binding. Some people cannot cycle confidently in traffic (a capability gap). Many have no safe route, nowhere to lock a bike, and nowhere to change when they arrive (a physical opportunity gap). For others, cycling is simply not what people like them do (a social opportunity gap). Some do not want to (reflective motivation), and most are quietly on car autopilot (automatic motivation). The rigorous version of this diagnosis is the method we laid out in article L5-14, and it is worth doing properly, because the single most common failure in applied work is skipping it and reaching for the intervention you already know how to run. Suppose the diagnosis comes back clear: for this group the binding constraint is physical opportunity, there is nowhere safe to ride.

‍

2. Then reach for the right lower-level tool for that region. This is where the map pays off. A physical-opportunity barrier calls for the environment, protected lanes, secure parking, a shower at the office (the choice-architecture and friction work of L5-13 and L5-16), and no amount of "cycling is healthy" messaging will substitute. Had the diagnosis instead pointed at social opportunity, the tool would be norms and social proof, showing how many colleagues already ride (L4-03, L5-18). At reflective motivation, you would be in Theory-of-Planned-Behaviour and messaging territory (L6-01, L5-17). At automatic motivation, habit and cue work rather than persuasion (L1-04, and Fogg's prompts in L6-02). At capability, confidence sessions and traffic-skills training. The map's whole purpose is to stop you spending the budget on a motivation campaign when the real problem was that the road felt too dangerous to ride.

‍

3. Use it as a shared language. Real cycling programmes are run by several teams at once, and COM-B lets each describe its bit in the same words instead of launching three disconnected "cycling initiatives." The transport planners own physical opportunity (the lanes and parking), the communications team owns social opportunity (making it look normal), and HR owns capability (the confidence sessions); because all three are named on one map, they can see they are covering different regions rather than duplicating a single one. In a large organisation that common vocabulary is quietly one of the framework's biggest practical wins.

‍

4. Treat it as an anti-tunnel-vision checklist. Every discipline has a reflex: the communications agency reaches for a campaign, the planner reaches for infrastructure, the HR lead reaches for training. Running down capability, opportunity, and motivation in turn is a cheap guard against each of them backing its favourite lever regardless of what the would-be cyclists actually needed. It is the same lesson the rest of this layer keeps teaching: the Theory of Planned Behaviour warns you to aim at the factor that is actually missing, Fogg warns you to check the prompt before the motivation, and COM-B simply makes that discipline systematic across every kind of barrier at once.

‍

For companies, the same logic runs from a car park to a checkout: consult the map before the creative brief, diagnose the region, then commission the intervention that fits it, and use the shared vocabulary so UX, marketing, and operations solve one problem rather than three. For campaigns, it turns "why won't people turn out" into a diagnosable question, is the barrier capability (they do not know when or how), opportunity (voting is genuinely hard, or nobody around them is doing it), or motivation (they do not care or do not feel it is for them), with a different tool behind each. For government, this is the framework's home ground, adopted by national health and transport bodies (the cycling case above is squarely their territory) precisely because it disciplines the design process, but two cautions apply hard here: the coercion and restriction functions carry political and cultural weight that a map cannot adjudicate for you, and a comprehensive framework's blessing is not evidence that a specific intervention will work, so the map still has to be followed by real evaluation.

‍

The reason COM-B anchors this layer is that it is the closest thing behavioural science has to a single picture of everything that moves behaviour. That makes it invaluable as a map but unreliable as an oracle, and knowing the difference is the whole skill. A field full of rival frameworks finally has one that everyone can read. Just remember that reading the map is not the same as walking the ground.

‍