A person values daily exercise, downloads a plan and sets an alarm, then skips the walk. They may genuinely want to do it and still struggle to start. Under what conditions does changing what we think help change what we do?

01

A favorable opinion is not yet a behavior

Wanting to go for a walk and actually doing it are different steps. That familiar gap is often treated as proof that intentions are useless. Mark Conner and Paul Norman ask a more precise question: what part of the path from an intervention to a changed behavior is carried by attitudes and intentions, and under which conditions? To answer that, they assess whether attitudes and intentions can be identified, changed and shown to help produce action.

An attitude is an evaluation of doing something: whether walking seems good or bad, enjoyable or unpleasant. An intention is a decision or self-instruction to act. The authors focus on behavioral intentions—'I will walk five times this week'—rather than broad aspirations such as 'I want to be fit.' These are different questions. I may regard exercise as worthwhile without deciding to do it; I may decide to do it while also anticipating that the experience will be unpleasant. Treating approval, commitment and action as synonyms conceals where a change effort failed.

02

What looks useful and what feels appealing

The review separates two components often hidden inside the word attitude. An instrumental evaluation asks whether a behavior is useful, healthy or valuable. An experiential evaluation asks whether doing it feels pleasant, interesting or satisfying. A person can judge a meal healthy while expecting to dislike eating it. Both evaluations matter, but they need not predict action equally well. In the prospective studies synthesized by Conner and Norman, experiential attitudes often have a stronger relationship with subsequent behavior than purely instrumental judgments. The distinction explains why a message about distant benefits may change agreement without changing the anticipated experience of doing the activity.

A further distinction separates an attitude's direction from its strength. Saying 'I like this' tells us the evaluation; it does not tell us whether that view will endure, resist counterarguments or guide a decision at the relevant moment. Stability, certainty, personal importance and ambivalence have been investigated as possible conditions of the attitude–behavior relationship. If an opinion changes between the survey and the opportunity to act, its earlier measurement cannot be expected to predict much. Evidence for some moderators is stronger in correlational work than in experimental tests, a limit the review keeps visible.

03

Prediction is not the same test as intervention

Suppose people with more favorable attitudes later exercise more. That correlation does not, by itself, show that persuading someone to like exercise will make them exercise. Past behavior, available time, health, money and other influences may shape both answers. The review therefore separates prediction from validation by intervention. Across six meta-analyses of prospective studies, the average attitude–behavior correlation was about .30, and the intention–behavior correlation about .42. Once past behavior was taken into account, the corresponding links with behavior change were smaller, about .14 and .22. These are averages across varied behaviors, not a forecast for one person.

The numbers support a meaningful but incomplete link. Experimental evidence addresses a different question: if an intervention actually changes an attitude or intention, does behavior also change? A meta-analysis discussed by the authors included 87 interventions that successfully changed attitudes and found a smaller but still detectable average effect on later behavior. Changes in intentions partly explained that link, yet did not explain all of it. This is stronger support for a causal role than correlation alone, while leaving room for other pathways and for differences among intervention designs.

04

Four questions for a change program

Conner and Norman organize the evidence through an experimental-medicine approach. First comes identification: choose a modifiable target plausibly linked to behavior. Second is validation: test whether changing that target affects action, rather than assuming that a correlation is sufficient. Third is engagement: determine whether the proposed message or activity actually changes the target. Fourth is intervention: test whether the resulting behavior change occurs through the proposed mechanism. Each stage can fail while another succeeds.

Imagine a campaign encouraging vaccination. It may correctly identify concern about side effects as one obstacle, but its message may leave that concern untouched. Or it may improve attitudes without overcoming the difficulty of getting an appointment. If uptake increases, researchers still need to test whether the shift came from changed attitudes, easier access or another route. This example is editorial, not one of the review's trials. It shows why an outcome alone does not establish how a program worked. The authors' operating-conditions framework adds a further question at every step: for whom, for which behavior and in which context does the link hold?

05

Why a strong intention may still remain on paper

The intention–behavior gap is real, but 'people lack willpower' is not an adequate explanation. An intention measured today may weaken before the opportunity to act. Competing goals may appear, the behavior may be habitual, or circumstances may leave little control. A plan to cook healthy meals faces a different test when a person's work schedule changes or fresh food is unavailable. The review examines the stability and strength of intentions, perceived control and the nature of the behavior as conditions that can alter the link to action.

The experimental evidence also suggests that effects of changing attitudes differ by behavior. In a meta-analysis reviewed by the authors, interventions to increase a behavior had larger average effects than those trying to reduce one. Interventions had smaller effects on frequent prevention behaviors, which are more likely to be embedded in routines, than on some less frequent actions. This does not make habits immune to change. It means an attitude-focused message may encounter a different barrier when behavior is repeated automatically in a stable setting. The review calls for testing these conditions more directly instead of turning them into universal excuses for an unsuccessful program.

06

Changing a thought requires more than delivering a message

A persuasive message has to be encountered, understood and connected to something the recipient cares about. The review brings together experiments and meta-analyses on changing attitudes and intentions, finding average effects but also substantial variation. Messages with stronger arguments tend to shift attitudes more when people engage with them. A large synthesis of motivationally matched messages found small to medium average effects on attitudes, intentions and observed behavior across health and other domains. Matching is promising; it is not proof that any personalized sentence will work.

Researchers have also catalogued behavior-change techniques, such as providing information or prompting a plan. Naming a technique does not tell us which psychological mechanism it activates. A program may be intended to change an intention but instead make an existing intention easier to remember at the moment of choice. The authors distinguish changing the value of a belief from activating or making it accessible. To learn which route matters, studies must isolate techniques, measure the proposed mental change and then observe behavior. The review finds that this full chain has been tested less systematically than broad claims about what interventions 'work.'

07

The middle step is the hardest to demonstrate

Suppose an intervention improves behavior. To say it worked through attitude change requires more than showing that attitudes and behavior both improved. The proposed mediator should change after the intervention and help account for the later action. The review describes trials where this chain was studied: an informational booklet about indoor tanning changed attitudes and later behavior, with mediation analyses supporting a role for attitudes. Another intervention about fruit and vegetable consumption traced a path through experiential attitudes and then intentions. These examples make the logic concrete, but they do not establish that every campaign works through the same route.

The authors stress that changes in intention account for only part of the effects in many syntheses. Other mechanisms, including routines or opportunities, may contribute. Some studies have measured attitudes but not tested the additional step through intentions; others have tested only a whole bundle of techniques. A mediator is therefore an empirical claim about a sequence, not a label to attach after a positive result. Better trials should specify the mechanism in advance, test it and measure conditions that may modify each link.

08

Context is a condition of the explanation

The relationship between attitudes, intentions and action does not have the same strength for every person or setting. The review examines perceived control, behavior frequency, message features and socioeconomic circumstances among possible moderators. A person may be persuaded that an action matters yet lack time, transport, money or institutional support to carry it out. Interventions centered entirely on individual belief can therefore miss structural constraints. Conner and Norman call for studies that test who benefits, under which conditions and by which mechanism instead of assuming one average effect applies everywhere.

The authors also warn that many proposed moderators have been observed repeatedly in correlational studies but remain thinly tested in experiments. Even fewer studies deliberately change a moderator to see whether the attitude–action link responds. Their conclusion is therefore balanced: attitudes and intentions are legitimate, modifiable targets with small to medium average effects, not powerless wishes and not guarantees. A useful behavior-change program must identify the obstacle, change the proposed target, observe what people actually do and check whether the explanation survives outside the easiest setting.

What to keep

  1. Intentions matter, but gaps remain.
  2. Experiments test causal change.
  3. Mechanisms should be measured, not assumed.
?

The nuance

What this story does not let us claim

Effects differ by person, context, behavior and how each construct is measured.

Go deeper: concepts and optional reading practice

Reversible cards

Test the concept

1 / 1

Topic assessment

Check whether you can explain the idea, not merely recognize it.

Every answer is grounded in the explanation above. Some questions introduce new cases to test whether you can apply and analyze the ideas, not merely recall them.

7 questions4 options per questionOne correct answer
01Which statement best captures the review’s central thesis?
02Which is the most appropriate operational definition of “Intention–behavior gap”?
03Which case correctly applies the central concept?
04If two studies obtain different results, which interpretation best preserves the logic of this guide?
05Which complementary conclusion belongs to the paper’s core argument?
06Which formulation best distinguishes “Intention–behavior gap” from a superficial label?
07Which limit should accompany any application of the paper to a new case?
Answer all seven questions.

Source & scope

Attitudes, Intentions, and Behavior Change. Mark Conner & Paul Norman. Annual Review of Psychology, vol. 77 (2026).

This editorial guide is an independent explanation. Consult the original publication for the studies, references and full context.

Open the original review ↗