By Jason Hreha

Article

Behavior Change Models Compared: 16 Frameworks, Their Evidence and Practical Uses

By

Updated Published 32 min read

Behavior change models organize explanations of why people act and how an intervention might help them act differently. The label also gets used for tools that do other jobs: checklists, design processes and taxonomies. A 2015 scoping review by Rachel Davis and colleagues identified 82 theories potentially relevant to public health. Four accounted for 63% of the articles in that review—not 63% of all published behavioral research.

Illustration of a traveler facing several colored paths on a map beneath a drawing compass.

This comparison covers the main components, evidence, criticisms and practical uses of 16 models and frameworks. The aim is to help you choose what to investigate and how to design a useful test. It is not a league table of effect sizes: a correlation predicting behavior, a treatment effect and a framework's classification coverage measure different things.

If you're a practitioner trying to design an intervention, a student trying to understand the field, or just someone who's been told to "use a behavior change model" and doesn't know where to start, this is the guide. (For our complete guide to behavior change itself, see our behavior change guide.)

What Is a Behavior Change Model?

A behavior change model is a formal framework that identifies the key factors influencing human behavior and specifies how those factors interact to produce (or prevent) behavior change. Some models focus on individual cognition (what people think and believe). Others focus on motivation (why people want to change). Others focus on environment and context (the structures around people that make behavior easier or harder).

No single model captures everything. The field has been trying for 50 years, and the result is a fragmented landscape where different models emphasize different pieces of the puzzle. The smartest approach is to understand what each model does well, what it ignores, and when to reach for it.

Here's the honest comparison.

How I'm Comparing These Models

For each model, I cover:

What it is
the core components and how they relate
The evidence
specific meta-analyses with effect sizes, sample sizes, and named researchers
What it gets right
genuine strengths
What it gets wrong
named criticisms and evidence gaps
When to use it
the contexts where it actually helps

I've organized the 16 models into five categories based on what they emphasize:

Category Models Core Focus
Individual cognition TPB, HBM, SCT, PMT What people think and believe
Motivation and stages TTM, SDT, IMB Why people change (or don't)
Intervention design COM-B/BCW, Fogg B=MAP, BCT Taxonomy, PRIME How to design interventions
Environment and policy Nudge, EAST, MINDSPACE, Ecological How context shapes behavior
Integrative Behavioral State Model A selected behavior across eight person-and-context components

Part 1: Individual Cognition Models

These models emphasize beliefs, attitudes and expectations. That makes them useful sources of questions about what people think. It does not follow that changing a belief is sufficient to change the behavior, or that a model's predictive associations establish the effect of an intervention.

Theory of Planned Behavior (Ajzen, 1991)

The Theory of Planned Behavior is a widely studied model of intentions and behavior. Psychologist Icek Ajzen developed it in 1991 as an extension of his earlier Theory of Reasoned Action (1975, with Martin Fishbein). The core idea: behavior is driven by intentions, and intentions are shaped by three factors.

Core components:

Attitudes
Do I evaluate this behavior positively or negatively?
Subjective norms
Do the people around me think I should do this?
Perceived behavioral control (PBC)
Do I believe I can actually do this?

These three feed into intentions, which drive behavior.

The evidence:

Armitage and Conner's 2001 meta-analysis covered 185 independent studies. The TPB accounted for 27% of variance in behavior and 39% in intentions. The behavior figure was 31% for self-reports and 21% for objective or observed measures. These figures describe prediction within the reviewed data. They do not mean a TPB-based intervention changes behavior by 27%, or that TPB is the strongest model in every domain.

McEachan and colleagues' prospective review examined health behavior measured after the predictor assessment. Prediction varied with behavior, measurement and follow-up. That variation matters when choosing a model: evidence about one behavior and time interval cannot establish how well it will predict another.

Intentions and behavior should remain separate outcomes. A person can intend to exercise yet lack time, access or a workable plan. A survey can help identify that intention; it cannot show which practical change will make exercise happen. The next step is a comparison that tests the proposed change.

What it gets right: The TPB is genuinely useful for understanding what people intend to do. If you need to predict whether a population will adopt a new behavior, measuring attitudes, norms, and perceived control will get you reasonable predictions.

What it gets wrong: Intentions leave much of behavior unexplained. I would not use an intention survey as a complete intervention design process. But TPB does include perceived behavioral control and recognizes limits on actual control; it does not simply assume intentions always become action. Implementation intentions add a specific if-then plan, which can help connect an intention to a situation. Their effectiveness still needs evidence for the target behavior. The original TPB meta-analysis also found subjective norms relatively weak, partly reflecting how they were measured.

When to use it: Predicting behavioral intentions. Understanding what a population thinks about a target behavior. Not for designing interventions to change complex behavior.

Health Belief Model (Rosenstock, 1966)

The Health Belief Model grew out of work by a group of U.S. Public Health Service researchers in the 1950s, including Irwin Rosenstock, Godfrey Hochbaum, Stephen Kegeles and Howard Leventhal. They were investigating why people did not use preventive services such as tuberculosis screening. Rosenstock’s historical account describes that collaborative development; his widely cited 1966 paper was a later statement of the approach. Self-efficacy was added later.

Core components:

Perceived susceptibility
How likely am I to get this condition?
Perceived severity
How bad would it be?
Perceived benefits
Will the recommended action help?
Perceived barriers
What will it cost me (time, money, pain, effort)?
Cues to action
What triggers me to act? (symptoms, doctor's advice, news)
Self-efficacy
Can I actually do this? (added later)

The evidence:

Carpenter's 2010 longitudinal meta-analysis, covering 18 studies and 2,702 participants, found perceived benefits and barriers consistently stronger predictors than the threat-related constructs. Their predictive value varied with timing and the kind of health behavior. Carpenter explicitly advised against continued use of the simple direct-effects version of the model. That is a substantive criticism, not merely a request for more studies.

The design implication is to investigate the person's perceived barriers and expected benefits. A warning about a health risk may miss the reason someone cannot act. The model can help organize those beliefs, but describing a barrier does not remove it.

What it gets right: The focus on perceived barriers is genuinely useful. If you want to know why someone isn't doing a health behavior, their perceived barriers are a useful place to investigate.

What it gets wrong: Susceptibility and severity, the two "threat" components, are consistently weak predictors. This undermines the model's central premise that threat perception drives behavior. Its emphasis on perceived benefits and barriers is not a complete explanation of habitual action or how beliefs interact. Social influence is not wholly absent: Rosenstock explicitly described beliefs as influenced by group norms and pressures.

When to use it: Identifying perceived barriers to a specific health behavior. Designing health communications about new or unfamiliar threats. Not as a comprehensive intervention design framework.

Social Cognitive Theory (Bandura, 1986)

Social Cognitive Theory is Albert Bandura's comprehensive theory of human functioning, developed in his 1986 book "Social Foundations of Thought and Action." At its center is a concept Bandura introduced in 1977 that appears in many theories of behavior: self-efficacy.

Core components:

Self-efficacy
Belief in one's capability to perform a behavior (the star of the show)
Outcome expectations
What do I expect to happen if I do this? (physical, social, self-evaluative consequences)
Observational learning
Learning by watching others (modeling)
Reciprocal determinism
Person, behavior, and environment all influence each other bidirectionally
Self-regulation
Self-monitoring, goal-setting, self-reward

The evidence:

Stajkovic and Luthans's 1998 meta-analysis combined 114 studies with 21,616 participants and found a weighted average correlation of r = 0.38 between self-efficacy and work-related performance, with variation by task complexity and setting. That supports taking beliefs about capability seriously. It does not establish that self-efficacy is the strongest cause of change in every population, or that confidence training produces the same effect as the observed association.

Bandura describes four sources of self-efficacy: (1) mastery experience (actually succeeding at the behavior), (2) vicarious experience (watching someone similar succeed), (3) verbal persuasion (being told you can do it), and (4) physiological states (how you interpret arousal and stress).

What it gets right: Actual ability and confidence about using that ability are different. A person may know how to complete a task but doubt they can manage it under pressure. Conversely, confidence can exceed competence. Both questions deserve attention. This distinction is useful without claiming that one construct wins every comparison.

What it gets wrong: As a complete model, SCT is very broad and hard to operationalize. "Reciprocal determinism" sounds elegant but is essentially circular: everything affects everything. This makes it hard to test as a whole and hard to use as an intervention design guide. In practice, SCT is less a design framework and more a collection of useful constructs, especially self-efficacy, that other models borrow freely.

When to use it: Any time self-efficacy is the bottleneck. Training programs, skill-building interventions, coaching, mentoring. Use the self-efficacy construct even if you don't use the full model.

Protection Motivation Theory (Rogers, 1975)

Protection Motivation Theory was developed by Ronald Rogers in 1975 to explain how fear appeals work. He revised it in 1983 to add coping appraisal. The model is essentially a structured version of two questions: "How bad is the threat?" and "Can I do anything about it?"

Core components:

Threat appraisal:
Perceived severity
(how bad is it?)
Perceived vulnerability
(how likely am I to be affected?)
Coping appraisal:
Response efficacy
(will the recommended action work?)
Self-efficacy
(can I do it?)
Response costs
(what will it cost me?)

The evidence:

Floyd, Prentice-Dunn and Rogers's 2000 meta-analysis combined 65 studies involving approximately 30,000 participants across more than 20 health topics. It reported an overall standardized effect, d+ = 0.52, for protection-motivation variables. The synthesis combined measures concerning protective intentions and behavior. Its pooled result should not be read as the effect of one standardized intervention, or as evidence that making people more afraid is the best strategy.

What it gets right: PMT provides a clear, structured way to design health communications. If you're creating a message about a health threat, the model prompts questions about perceived threat and coping: does the person see the risk, trust the response and believe they can carry it out?

What it gets wrong: The model is limited to threat-based motivation. It can't explain positive behavior change that isn't fear-driven (exercise for enjoyment, healthy eating for taste). It also shares the rationality assumption with HBM and TPB. The evidence for actual behavior change (vs. intentions) is weaker.

When to use it: Health risk communication. Designing messages about specific health threats. Understanding why people do or don't protect themselves. Not for positive motivation or complex behavior change.

Part 2: Motivation and Stage Models

These models focus on why people change, not just what they think. They address the quality and dynamics of motivation itself.

Transtheoretical Model / Stages of Change (Prochaska & DiClemente, 1983)

The Transtheoretical Model is a prominent behavior change model with an extensive critical literature. James Prochaska and Carlo DiClemente developed it in 1983 by studying how smokers quit on their own, attempting to integrate across therapeutic orientations.

Core components:

Five stages of readiness:
  1. Precontemplation: Not thinking about changing
  2. Contemplation: Thinking about it but not ready
  3. Preparation: Planning to act soon
  4. Action: Currently making changes
  5. Maintenance: Sustaining change for 6+ months

Plus: 10 processes of change, decisional balance (pros vs. cons), self-efficacy, and temptation.

The evidence:

The TTM appeared in 33% of articles in the Davis scoping review. That establishes prominence within that sample, not effectiveness. Webb and colleagues' review of internet health interventions found different average effects among interventions drawing on different theories. Those were comparisons between studies, not randomized assignments of the same program to competing theories.

The criticism (and it's substantial):

Robert West published "Time for a change: putting the Transtheoretical Model to rest" in Addiction in 2005. His argument: the stages are arbitrary and unstable, the model ignores the biology of motivation (reward, habit, associative learning), and stage-matched interventions show no consistent advantage. West explicitly called for the model's abandonment.

Littell and Girvin's 2002 review questioned whether the stages were mutually exclusive and whether people moved sequentially through them. They also questioned the validity of the stage assessments. Those criticisms attack the formal categories, while leaving open their possible value as a way to discuss change.

Bridle and colleagues' systematic review found limited evidence for the effectiveness of stage-based interventions and important weaknesses in the available trials. That is a poor foundation for treating stage-matching as an established advantage over alternatives.

DiClemente responded in the same issue of Addiction, calling West's critique "a premature obituary." The debate continues, but the weight of evidence is not favorable to the TTM as an intervention framework.

What it gets right: The insight that change isn't all-or-nothing is genuinely useful. People are at different levels of readiness, and acknowledging this in clinical conversations helps build rapport.

What it gets wrong: The formal stages and the claim that matching an intervention to them improves outcomes need stronger support than their popularity suggests. The critiques above challenge stage boundaries, sequential progression and measurement. I would not let a stage label decide what help a person receives. Investigate the actual obstacle and the person's willingness to address it.

When to use it: As a clinical conversation tool to gauge readiness. Not as a formal intervention design framework. If someone insists on "stage-matching," know that the evidence doesn't support it.

Self-Determination Theory (Deci & Ryan, 1985)

Self-Determination Theory, developed by Edward Deci and Richard Ryan at the University of Rochester, asks a question the other models skip: not just "will this person change?" but "what kind of motivation is driving the change?" The answer matters enormously for whether behavior lasts.

Core components:

Three basic psychological needs:
Autonomy
Feeling that your behavior is self-chosen
Competence
Feeling effective and capable
Relatedness
Feeling connected to others
The motivation continuum (from least to most self-determined):

Amotivation → External regulation → Introjected regulation → Identified regulation → Integrated regulation → Intrinsic motivation

The critical distinction: autonomous motivation (identified, integrated, intrinsic) versus controlled motivation (external, introjected). The theory predicts different consequences from these forms of motivation. Whether an activity persists after external support ends remains an empirical question; the labels alone do not settle it.

The evidence:

Ng and colleagues' 2012 meta-analysis, covering 184 independent datasets, found relationships consistent with SDT among autonomy support, need satisfaction, motivation and health-related outcomes. These associations support studying the proposed processes. They are not the same as randomly manipulating each process to establish its causal contribution.

Ntoumanis and colleagues' review of 73 experimental studies of SDT-informed health interventions found modest, heterogeneous improvements in health behavior and some health outcomes. It offers intervention evidence as well as a reminder that an attractive theory does not imply large or uniform practical benefits.

What it gets right: The distinction between autonomous and controlled motivation asks an important question: does the person endorse this activity, or are they doing it only under pressure? For sustained activities, I want to know whether the activity offers a reason to return once the immediate reward or supervision ends. SDT gives that question a prominent place. It does not imply that every external reward undermines enjoyment or that every self-chosen activity lasts.

What it gets wrong: Positive motivation measures are not enough. An intervention still needs to show that behavior changes in a useful way and, when claimed, that health improves. The intervention review's modest and variable effects should restrain promises of easy, durable transformation. The framework also requires careful distinctions among motives that may overlap in a person's experience.

When to use it: Understanding why someone is (or isn't) motivated. Training healthcare providers, coaches, or managers in autonomy-supportive communication. Designing interventions where motivation quality matters more than motivation quantity. Long-term behavior maintenance.

Information-Motivation-Behavioral Skills Model (Fisher & Fisher, 1992)

The IMB model was developed by Jeffrey and William Fisher in 1992, originally to explain and improve HIV prevention behavior. It's the simplest of the major models: three constructs, clear causal pathways, and a direct translation to intervention design.

Core components:

Information
Knowledge about the behavior and the condition
Motivation
Personal attitudes and social norms about the behavior
Behavioral skills
Objective ability and perceived self-efficacy to perform the behavior

Information and motivation work primarily through behavioral skills to produce behavior change, though information and motivation can also directly affect behavior.

The evidence:

Fisher and colleagues tested an IMB-based intervention against a no-treatment comparison in a college population. They reported improvements in AIDS-preventive behaviors. That supports the specific package in that study; it does not isolate the contribution of each component or show that the model transfers unchanged to every domain.

What it gets right: Radical simplicity. Three constructs, three intervention components. If people don't know what to do, educate them. If they aren't motivated, address attitudes and norms. If they can't do it, build skills and self-efficacy. The direct mapping from assessment to intervention is elegant.

What it gets wrong: It's probably too simple. The model doesn't account for environmental constraints, habit, emotional regulation, or systemic barriers. Whether information is the limiting factor must be established for the particular task; an average association cannot diagnose an individual knowledge gap. The model was built for a specific problem (HIV prevention) and may not transfer well to complex, sustained behavior changes.

When to use it: Health education, medication adherence, sexual health, any domain where a clean assessment-to-intervention mapping is useful. Best for behaviors where information and skills really are the bottleneck.

Part 3: Intervention Design Frameworks

These aren't just models of behavior. They're tools for building interventions. They answer the practitioner's question: "I understand the problem. Now what do I actually do?"

COM-B + Behaviour Change Wheel (Michie, van Stralen & West, 2011)

The COM-B model and the Behaviour Change Wheel represent the most ambitious attempt to unify the field. Susan Michie, Maartje van Stralen, and Robert West developed them by systematically reviewing 19 existing frameworks and finding that none covered the full range of intervention types and policy options. Published in Implementation Science in 2011, the BCW is essentially a meta-framework built on top of all the others.

Core components:

COM-B (the behavioral diagnosis):
Capability
Physical (skills, stamina) + Psychological (knowledge, cognitive skills)
Opportunity
Physical (environment, resources, time) + Social (norms, social influence, cultural expectations)
Motivation
Reflective (conscious beliefs, plans, intentions) + Automatic (habits, emotions, impulses)

All six components interact to produce Behavior.

The Behaviour Change Wheel (the intervention toolkit):
  • 9 intervention functions: Education, Persuasion, Incentivisation, Coercion, Training, Restriction, Environmental restructuring, Modelling, Enablement
  • 7 policy categories: Communication/marketing, Guidelines, Fiscal measures, Regulation, Legislation, Environmental/social planning, Service provision

The evidence:

The original BCW paper evaluated how frameworks covered intervention and policy categories and whether its classification could be applied reliably. It explicitly called for further evaluation of whether using the approach improves intervention development and effectiveness. The Wheel does not inherit proof of effectiveness merely because it brings together 19 frameworks.

What it gets right: Comprehensiveness. COM-B forces you to consider capability, opportunity, AND motivation, not just one piece. The BCW then links each COM-B deficit to specific intervention types and policy options. It provides a structured process for moving from a behavioral question to intervention options.

What it gets wrong: Complexity and circularity. Using the full BCW process properly requires training and time. The link from COM-B assessment to specific intervention functions involves professional judgment, not a mechanical algorithm. Some practitioners find it overwhelming. Critics note that the categories can overlap (is "social modeling" an intervention function or a motivation source?) and that the framework tells you what type of intervention to use but not the specific content. There's a deeper problem: COM-B is arguably tautological. Saying behavior requires capability, opportunity, and motivation is true by definition, but it doesn't specify which capabilities, which opportunities, or which motivational processes matter for any given behavior. It identifies categories but doesn't predict effect sizes or specify mechanisms.

When to use it: Complex, multi-level behavior change challenges. Intervention design at organizational or policy level. When you need to be systematically comprehensive. Pair it with the BCT Taxonomy for specific technique selection.

Fogg Behavior Model / B=MAP (BJ Fogg, 2009)

BJ Fogg, a Stanford communication researcher, introduced his behavior model in 2009 at a Persuasive Technology conference. He updated it in his 2019 book "Tiny Habits." The model has been enormously influential in Silicon Valley, product design, and the tech industry, while generating skepticism in academic behavioral science.

Core components:

B=MAP: behavior occurs, in Fogg’s model, when motivation, ability and a prompt converge. The notation is a mnemonic, not an additive equation.

Motivation
How much do you want to do it? (pleasure/pain, hope/fear, social acceptance/rejection)
Ability
How easy is it? (time, money, physical effort, mental effort, routine disruption)
Prompt
What triggers the behavior at the right moment?

In Fogg's account, these factors must converge at the moment of action. The action line represents combinations of motivation and ability at which a prompt can produce behavior; it is not a calibrated prediction equation. Tiny Habits applies this emphasis on making a starting action easier and connecting it with a recurring situation.

The evidence:

A 2025 scoping review found health interventions using the model and reported positive outcomes, alongside variation between studies and limited long-term follow-up. The authors are favorable toward the framework. Their review maps applications rather than showing that the Fogg model beats alternative frameworks in a controlled comparison.

What it gets right: The focus on reducing friction (ability) is genuinely useful. Most behavior change models overemphasize motivation and underemphasize how hard the behavior is to perform. Fogg emphasizes making the action easier instead of relying only on stronger motivation. The Tiny Habits approach (start absurdly small) has practical value for initiation.

What it gets wrong: Its strength is also its limit: focusing on a moment of action does not fully explain the organization of a complex, sustained activity. I would use it to examine a particular prompt or starting action, then investigate the social setting, resources, preferences and continuing effort separately. The 2025 review's limited follow-up is relevant here. Ability in the model can include practical constraints, so saying it ignores all context would overstate the criticism.

When to use it: Product design and UX. Simple, discrete behaviors. Getting someone started (first steps). Not for sustained, complex behavior change programs.

BCT Taxonomy v1 (Michie et al., 2013)

The BCT Taxonomy isn't a model or theory. It's a classification system: a standardized vocabulary for describing exactly what an intervention does. Susan Michie and colleagues developed it through a Delphi process with 14 international experts, drawing from 124 techniques across 6 existing classification systems.

Core components:

93 distinct behavior change techniques organized into 16 groupings, including:

Goals and planning
(goal setting, action planning, problem solving)
Feedback and monitoring
(self-monitoring, feedback on behavior/outcomes)
Social support
(practical, emotional, unspecified)
Shaping knowledge
(instruction, information about antecedents)
Natural consequences
(information about health, emotional, social consequences)
Comparison of behavior
(demonstration, social comparison)
Associations
(prompts/cues, associative learning)
Repetition and substitution
(behavioral practice, habit formation)
Reward and threat
(material incentive, social reward, self-reward)
Regulation
(pharmacological support, reducing negative emotions)
Self-belief
(verbal persuasion about capability)

The evidence:

Technique coding can reveal patterns across interventions. A 2009 healthy-eating and physical-activity review, using an earlier classification, associated self-monitoring plus another control-theory technique with larger average effects than other programs. It was a comparison between studies, not proof that adding a code produces the same benefit in a new program.

  • Describe content: record the technique actually delivered, not only the program name.
  • Compare programs: examine delivery, population and context as well as shared codes.
  • Test components: use a design that can separate a technique’s contribution from the rest of the package.

What it gets right: Precision. Broad labels such as "counseling" or "health education" can hide important differences in intervention content. BCTTv1 built on earlier classification systems to make descriptions more consistent. That helps researchers compare programs and test their ingredients; the codes alone cannot establish which ingredient caused an effect.

What it gets wrong: Ninety-three techniques is overwhelming. The taxonomy tells you what techniques exist but not which to use for your specific problem. It's a catalogue, not a decision tool. It requires training to code reliably. And it's atheoretical: it describes intervention content without specifying the causal mechanisms through which techniques produce change.

When to use it: Alongside COM-B/BCW for selecting specific techniques. When designing intervention protocols. When reporting intervention content in research. Not as a standalone design guide.

PRIME Theory (Robert West, 2006)

PRIME Theory is Robert West's attempt to build a general theory of motivation that accounts for what the classical models ignore: impulse, habit, and the fact that people frequently act against their own stated plans. West developed it in his 2006 book "Theory of Addiction" as an alternative to the TTM, which he had publicly critiqued.

Core components (hierarchical):

Plans
Conscious representations of future actions plus commitment
Responses
Starting, stopping, or modifying actions
Impulses/Inhibitory forces
Experienced as urges
Motives
Experienced as desires (wants and needs)
Evaluations
Evaluative beliefs about what is good or bad

The hierarchy matters: plans can generate motives, but impulses can override plans. This explains why someone can plan to quit smoking, genuinely believe smoking is harmful, and still light a cigarette when stressed. The lower-level system (impulses) overrides the higher-level system (plans).

The evidence:

West's own summary presents PRIME as an integration of conscious plans, evaluations, motives and impulses, developed in relation to addiction. That explains the framework's purpose. It is not a trial showing that a PRIME-based intervention outperforms alternatives; that claim requires evidence about the particular intervention.

What it gets right: It takes seriously the fact that human motivation is not purely rational. By including impulses and automatic processes in the formal model, PRIME Theory can explain phenomena (relapse, impulsive behavior, action against stated intentions) that the TPB and HBM cannot.

What it gets wrong: Less empirical testing than established models. Primarily applied in addiction and smoking cessation, with less evidence in other behavior change domains. The hierarchical structure, while conceptually elegant, can be difficult to translate into specific intervention design steps.

When to use it: Addiction and substance use. Understanding why people act against their own plans. When automatic/impulsive processes are clearly important.

Part 4: Environment and Policy Frameworks

These frameworks shift focus from what's happening inside the person to what's happening around them. They direct attention to changes in the setting: available options, practical barriers and the way a service works. Whether changing the context is the best response depends on the particular problem.

Nudge Theory / Choice Architecture (Thaler & Sunstein, 2008)

Economist Richard Thaler (Nobel Prize, 2017) and legal scholar Cass Sunstein published "Nudge" in 2008, introducing the idea of "libertarian paternalism": you can steer people toward better choices while preserving their freedom to choose. The tool is choice architecture, changing how options are presented rather than changing people's minds.

Core components:

Defaults
Pre-selected options (opt-in vs. opt-out)
Framing
How information is presented
Social norms
Highlighting what others do
Salience
Making key information prominent
Simplification
Reducing friction in choice processes

The evidence:

Dennis Hummel and Alexander Maedche reviewed 100 nudging publications with 317 effect sizes in 2019. Effectiveness varied substantially by nudge type and domain.

Mertens and colleagues' 2022 meta-analysis reported a positive average for choice-architecture interventions. It was subsequently corrected to remove observations from a retracted study and repair errors. The correction matters when identifying which data a reanalysis used.

But here's the catch. Maier and colleagues' reanalysis did not find convincing support for a positive overall effect after publication-bias adjustment. They also acknowledged heterogeneity, leaving room for specific interventions to work. In their reply, Mertens and colleagues acknowledged the bias problem and argued for better evidence about contexts. The evidence supports serious skepticism about a general nudge effect, not a universal numerical effect for every nudge.

Specific positive trials remain relevant. A large vaccination text-message experiment found an average increase in recorded influenza vaccination around an existing appointment. That result concerns a defined message opportunity, population and short-term endpoint. It does not validate every default or reminder, and it does not establish a lasting habit.

One commonly cited example deserves scrutiny: organ donation. An increase in recorded consent is not automatically an increase in organs transplanted. The latter also depends on identifying eligible donors, family decisions and the health system's capacity. A comparison about registration should be labeled as such. This outcome distinction is essential before using a default result to promise a health benefit.

What it gets right: The arrangement of a choice can matter. Making a usable option available, removing needless steps or setting an appropriate default can be worth testing. The result still belongs to that arrangement in that setting. Neither minimal cost nor a dramatic effect follows from calling something choice architecture.

What it gets wrong: The publication bias problem is real and serious. Beyond defaults, the evidence for other nudges is weaker than commonly claimed. Nudges also don't address root causes. They work by steering choices at the point of decision, which means they're best for one-time or infrequent decisions, not sustained behavior change. The "libertarian paternalism" framing has also drawn philosophical criticism about who gets to decide what's "better."

When to use it: Policy design at scale. Default setting for enrollment and consent. Simplifying complex decision environments. Not for sustained, complex behavior change.

EAST Framework (BIT, 2014)

EAST is the Behavioural Insights Team's practitioner-friendly distillation of behavioral science. Released in 2014, it was designed to be memorable, actionable, and usable by policymakers who aren't behavioral scientists.

Core components:

Easy
Reduce friction. Simplify. Use defaults. Pre-fill forms.
Attractive
Draw attention. Design rewards well. Use images and color.
Social
Show what others do. Use commitments. Leverage networks.
Timely
Prompt at the right moment. Consider present bias. Help people plan.

The evidence:

The EAST guide collects examples to help practitioners generate options. First published in 2014, it was updated in 2024. Evidence for one example should be assessed as evidence for that intervention, not as validation of all four principles together.

What it gets right: Simplicity and memorability. Four principles that anyone can apply. The emphasis on "Easy" first is well-supported by evidence. Reducing an unnecessary step is a sensible option to investigate when that step is preventing a wanted action.

What it gets wrong: EAST is atheoretical. It doesn't explain why these principles work, which limits its ability to generate novel predictions or guide complex intervention design. It's also limited to nudge-type interventions and can't address deep motivational or systemic challenges.

When to use it: Quick wins in service design. Government communications. When you need a simple framework for a non-specialist audience. As a complement to deeper models like COM-B, not as a replacement.

MINDSPACE Framework (Dolan et al., 2012)

MINDSPACE is a more granular version of the nudge approach, developed by Paul Dolan and colleagues for the UK Cabinet Office. The mnemonic captures nine behavioral influences that operate largely automatically.

Core components:

  • Messenger: We're influenced by who communicates information
  • Incentives: We respond more to losses than gains (loss aversion)
  • Norms: We do what others do
  • Defaults: We go with the pre-set option
  • Salience: We notice what's novel and relevant
  • Priming: Subconscious cues influence behavior
  • Affect: Emotions shape decisions
  • Commitments: We follow through on public promises
  • Ego: We protect our self-image

What it gets right: More specific than EAST. Its categories can prompt policy ideas. Each proposed application still needs its own evidence.

What it gets wrong: The Priming component is now on shaky ground. The psychological priming literature has been heavily criticized in the replication crisis, with many landmark studies failing to replicate. Including priming as one of nine "robust" effects looks outdated. The framework as a whole has not been tested as an integrated system. It's more a checklist than a model.

When to use it: Policy brainstorming. As a more detailed alternative to EAST. Treat priming claims with skepticism.

Ecological / Socio-Ecological Model (Bronfenbrenner, 1979; McLeroy et al., 1988)

The Socio-Ecological Model isn't a behavior change model in the same sense as the others. It's a framing that prevents a common error: assuming behavior change is entirely an individual problem.

Core components (five levels):

  1. Individual: Knowledge, attitudes, beliefs, skills
  2. Interpersonal: Family, friends, social networks
  3. Organizational: Workplace policies, institutional rules
  4. Community: Relationships between organizations, community norms
  5. Policy: Laws, regulations at local to national level

The evidence:

The CDC uses a four-level social-ecological model for violence prevention: individual, relationship, community and societal influences. The five-level formulation outlined above separates organizational and policy questions differently. These are related organizing approaches, not one standardized model with a single effect size.

What it gets right: It systematically directs attention beyond the individual, including institutions and policy. If someone isn't exercising, the ecological model asks: Is the problem individual (knowledge, motivation)? Interpersonal (no workout partner)? Organizational (no gym at work, no time off)? Community (unsafe neighborhoods, no parks)? Policy (no physical education requirements, car-dependent infrastructure)?

What it gets wrong: It's too broad to guide specific interventions. It tells you to "think about all levels" but doesn't tell you what to do at any of them. It needs to be combined with a more specific model (like COM-B) to be actionable.

When to use it: Framing complex behavior change challenges. Ensuring you don't fall into the "individual blame" trap. Program planning and evaluation. Always use alongside a more specific model.

Part 5: Integrative Frameworks

The models above share a common problem: each captures some drivers of behavior while ignoring others. The following framework attempts to fix that by integrating the pieces into a single comprehensive system.

The Behavioral State Model (Hreha, developing framework)

The Behavioral State Model is my eight-component framework for investigating why a selected behavior is or is not occurring. It separates six Personal Components—Personality, Perception, Emotions, Abilities, Social Status/Situation, and Motivations—from two Context Components: the Social Environment and Physical Environment.

The extra resolution can be useful when a broad label hides the practical difference. Distrust and lack of desire may both land inside a broad motivation category, but they call for different evidence and different tests. The model asks a team to record its evidence, identify candidate constraints, test a targeted change, and revise the diagnosis.

Evidence boundary: The component constructs draw on established areas of psychology, but BSM has not been independently validated as a complete model or measure. A component rating is a structured judgment, not a behavioral probability. The lowest rating is a candidate constraint, not proof of a cause.

When to use it: After you have selected a specific target behavior and need a detailed set of diagnostic questions. Use Behavior Matching first when the larger question is which behavior to choose.

The Master Comparison Table

Model Year Type Core Focus Key Evidence Biggest Limitation
Theory of Planned Behavior 1991 Cognitive Attitudes → Intentions → Behavior Predictive associations; measurement and follow-up matter Prediction does not establish intervention effects
Health Belief Model 1966 Cognitive Threat perception → Health behavior Longitudinal evidence stronger for benefits and barriers Simple direct-effects version criticized
Social Cognitive Theory 1986 Cognitive Self-efficacy → Behavior Self-efficacy associated with work performance Association is not a universal intervention ranking
Protection Motivation Theory 1975 Cognitive Threat + Coping → Protection Evidence on protective intentions and behaviors Pooled outcomes do not specify one intervention
Transtheoretical Model 1983 Stage Stages of readiness → Change Stage boundaries and stage-matching disputed Limited evidence for the formal stage-based approach
Self-Determination Theory 1985 Motivational Autonomous motivation → Lasting change Modest, heterogeneous health-intervention effects Motivation measures are not health outcomes
IMB Model 1992 Motivational Info + Motivation + Skills → Behavior Specific intervention trials, initially in HIV prevention Package evidence does not isolate each component
COM-B / BCW 2011 Design Capability + Opportunity + Motivation Original classification and reliability study Outcome superiority still requires evaluation
Fogg B=MAP 2009 Design Motivation + Ability + Prompt Health applications mapped in a scoping review Limited long-term and comparative evidence
BCT Taxonomy 2013 Design 93 techniques in 16 groups Shared definitions and coding reliability Classification does not select effective techniques
PRIME Theory 2006 Design Plans → Motives → Impulses → Responses Integrative theoretical account of motivation Intervention-specific efficacy must be established
Nudge / Choice Architecture 2008 Policy Change context, not minds Specific effects; disputed bias-adjusted pooled evidence No universal effect for a heterogeneous category
EAST 2014 Policy Easy, Attractive, Social, Timely Practitioner guide, updated 2024 Examples do not validate the mnemonic as a whole
MINDSPACE 2012 Policy 9 automatic behavioral influences Checklist of candidate influences Each proposed application needs evidence
Ecological Model 1988 Framing 5 levels: individual → policy Organizes influences across levels Needs specific actions and evaluation
Behavioral State Model Developing Integrative diagnostic Eight personal and contextual components Component constructs have relevant research literatures Complete framework and scoring method are not independently validated

Which Model Should You Use?

The honest answer: it depends on what you're trying to do.

If you're designing a health communication about a specific threat: Use Protection Motivation Theory. It organizes questions about perceived threat, response effectiveness and capability. Establish the actual concern before choosing the message.

If you need to diagnose why a behavior isn't happening: Start with COM-B. It forces you to consider capability, opportunity, AND motivation. Most other models only address one or two of these.

If you're designing a sustained behavior change intervention: COM-B can organize the investigation, the BCW can suggest intervention functions, and the BCT Taxonomy can describe the chosen techniques. SDT adds questions about whether the person endorses the activity. These tools can structure the work; none ensures autonomous motivation or lasting change.

If you're a coach, therapist, or healthcare provider: Use SDT's autonomy-supportive approach for how you communicate. Use TTM stages as a conversational gauge of readiness (but don't rely on formal stage-matching). Focus on building self-efficacy (from SCT).

If you're designing a product or app: Fogg's B=MAP is a reasonable starting point for feature-level design (reduce friction, add prompts). But for the overall behavior change strategy, use COM-B or the ecological model to make sure you're not missing the bigger picture.

If you're making policy: Start with the ecological model to ensure you're addressing multiple levels. Use EAST or MINDSPACE for specific policy design. Use nudge/choice architecture for default and enrollment decisions. Use COM-B/BCW if you're designing a comprehensive policy package.

If you're trying to predict behavior: Choose constructs and measures relevant to a specified action, population and interval. TPB provides one testable starting point. Evaluate predictions against appropriate alternatives rather than borrowing a headline variance figure from another domain.

If you need to choose among possible behaviors: Use Behavior Matching to compare candidates before intervention design. After selecting one, use the Behavioral State Model to investigate its eight Personal and Context Components. Neither method turns an initial rating into proof; the selected behavior and proposed constraint still need testing.

If confidence appears to be a problem: Distinguish perceived capability from actual capability. Check whether the person needs practice, resources, clearer expectations or a different action. Evidence that self-efficacy predicts performance does not identify a universally highest-return intervention.

Example: Choosing a Model for a Team Handoff

Imagine a small team that keeps losing track of decisions between shifts. Someone proposes a daily reminder to complete a handoff form. Before building the reminder, ask what the next shift actually needs: the current status, the next action, and who owns it.

First, compare possible behaviors. The outgoing person could fill in a separate form, add three lines to the task the team already uses, or give a short verbal handoff. A verbal update requires both people to be available. A separate form creates another place to check. An update on the existing task may fit better if everyone can access it. Ask the people doing the work and try the feasible options. This is the behavior-selection question behind Behavior Market Fit.

Then investigate the obstacle. Suppose the team chooses the three-line update but people still skip it. COM-B gives you three areas to investigate: Can people write a useful update (capability)? Do they have access and time before the shift ends (opportunity)? Do they see a reason to do it, or does another task take priority (motivation)? Watch a handoff and ask what happened. An empty field alone cannot tell you which explanation is right.

Use prompt design for a prompt problem. If people want to leave an update, can do it, and simply miss the moment, a reminder beside the task's “End shift” action is a reasonable idea to test. That is a specific application of Fogg's motivation, ability, and prompt model. If the incoming person cannot open the task, fix access first.

Check the handoff, not just the checkbox. Record whether the update was completed and whether the next person could identify the next action without asking for clarification. More completed fields with no improvement in that second measure would be a reason to rethink the format. This is a hypothetical design example, not a reported study or a claim that either model guarantees success.

Document the intervention, then test the handoff

Once the team has a plausible design, the BCT Taxonomy can describe its content. Agreeing to leave an update, specifying when to do it and adding a reminder are different ingredients. Recording them makes it possible to explain what changed, rather than calling the whole package “better communication.”

Compare the revised process with the existing one over a defined period. If teams share work, account for that in the assignment and analysis. Measure whether the incoming worker can act without clarification, the time both shifts spend, and problems caused by missing or misleading updates. The research-methods guide helps choose an appropriate comparison.

The frameworks now have distinct jobs. Behavior Market Fit prompts selection of a suitable action; COM-B structures the investigation; Fogg focuses attention on a moment of action; the taxonomy describes the content. None supplies the result of the test. This worked selection is the purpose of this article; the framework glossary provides the shorter reference.

The Uncomfortable Truth About Behavior Change Models

After reviewing the evidence across all 16 models, three patterns emerge that the field doesn't talk about enough.

First, prediction is not explanation of everything a person does. A model can account for some variation in a measured behavior without identifying every cause of it. Armitage and Conner's results also varied with how behavior was measured. The unexplained proportion is not a direct inventory of unknown causes, and one review cannot establish which model is best across all domains.

Evidence can also look different across research settings. DellaVigna and Linos found average effects of 1.4 percentage points in trials from two US nudge units, compared with 8.7 points in a separate academic sample. These were different sets of trials, not the same intervention shrinking after leaving a laboratory. The finding challenges optimistic forecasts based on selected academic results while leaving the individual interventions to be assessed on their own evidence.

Second, naming a theory is not the same as using it well. Webb and colleagues associated more extensive theory use with larger effects in internet health interventions, but the authors called for experiments to establish cause and effect. A strong design should state which construct the intervention targets, how the content targets it and what result would challenge the explanation. Putting a model's name in the introduction does none of that.

Third, the most practically useful insights cut across models. Self-efficacy appears in SCT, PMT, TTM, IMB, and COM-B. Reducing barriers appears in HBM, Fogg, EAST, and choice architecture. Environmental restructuring appears in COM-B, ecological models, and nudge theory. The most powerful constructs aren't owned by any single model.

The field's fragmentation into 82+ competing theories may actually be the problem. COM-B offers one attempt at integration; whether integration improves results still needs to be tested. Find the core constructs that reliably predict and change behavior across contexts, regardless of which theoretical tent they live in.

But picking the target behavior comes before polishing the intervention. Behavior Matching compares possible actions against the goal, people, and setting. The Behavioral State Model then helps examine the selected action in more detail. That sequence can expose a poor fit before a team spends months optimizing the wrong request.

For a deeper dive into what actually works for lasting behavior change, see our complete guide to behavior change.

Frequently Asked Questions

What is the most effective behavior change model?

No single model is established as most effective across all purposes. A theory can help predict behavior; a design framework can organize investigation; a taxonomy can describe intervention content. Choose the job first. Then inspect evidence about the relevant behavior and comparison. COM-B is a useful organizing framework, but its breadth is not proof of superiority.

How many behavior change models are there?

Davis and colleagues identified 82 theories potentially relevant to public health in their 2015 scoping review. Four accounted for 63% of that review's articles. The authors explicitly state that the search was designed to identify theories, not exhaustively measure all their research use. The count is a useful map, not a final census of behavioral science.

What is the difference between COM-B and the Behaviour Change Wheel?

COM-B sits at the center of the Behaviour Change Wheel and organizes questions about capability, opportunity and motivation. The Wheel adds nine intervention functions and seven policy categories. COM-B helps frame an investigation; it does not identify the cause of a problem automatically. Selecting an intervention still requires evidence and judgment.

Does it take 66 days to form a habit?

The 66-day figure is the median modeled time to reach 95% of an estimated automaticity plateau in the 39 selected cases in Lally and colleagues' study. The range was 18–254 days, while observation lasted 84 days. Selection depended on data and curve criteria, so the group is not a simple validated success count. See the full Lally study analysis for the measurement and exercise-subgroup qualifications.

What is the best behavior change model for exercise?

Choose an activity and a feasible way to perform it before choosing a framework. COM-B can organize questions about skill, access and motivation; SDT can help examine whether the person endorses the activity. Neither label guarantees adherence. Lally's exercise subgroup had a median modeled estimate of 91 days among 13 selected participants, extending beyond the study's 84-day observation period. That does not establish the best framework for exercise.

Are behavior change models evidence-based?

The kind of evidence matters. Predictive associations, trials of a particular program and tests of classification reliability answer different questions. Several models have substantial research histories; that does not validate every application. The profiles above distinguish these evidence types and link to the supporting reviews. For a compact reference to what each tool does, see Behavior Change Frameworks Compared.

References

Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179-211.

Armitage, C. J., & Conner, M. (2001). Efficacy of the Theory of Planned Behaviour: A meta-analytic review. British Journal of Social Psychology, 40, 471-499.

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Prentice-Hall.

Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2), 191-215.

Bridle, C., Riemsma, R. P., Pattenden, J., Sowden, A. J., Mather, L., Watt, I. S., & Walker, A. (2005). Systematic review of the effectiveness of health behavior interventions based on the transtheoretical model. Psychology & Health, 20(3), 283-301.

Carpenter, C. J. (2010). A meta-analysis of the effectiveness of health belief model variables in predicting behavior. Health Communication, 25(8), 661-669.

Davis, R., Campbell, R., Hildon, Z., Hobbs, L., & Michie, S. (2015). Theories of behaviour and behaviour change across the social and behavioural sciences: A scoping review. Health Psychology Review, 9(3), 323-344.

Deci, E. L., Koestner, R., & Ryan, R. M. (1999). A meta-analytic review of experiments examining the effects of extrinsic rewards on intrinsic motivation. Psychological Bulletin, 125(6), 627-668.

Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human behavior. Plenum.

DellaVigna, S., & Linos, E. (2022). RCTs to scale: Comprehensive evidence from two nudge units. Econometrica, 90(1), 81-116.

Dolan, P., Hallsworth, M., Halpern, D., King, D., Metcalfe, R., & Vlaev, I. (2012). Influencing behaviour: The mindspace way. Journal of Economic Psychology, 33(1), 264-277.

Fisher, J. D., & Fisher, W. A. (1992). Changing AIDS-risk behavior. Psychological Bulletin, 111(3), 455-474.

Floyd, D. L., Prentice-Dunn, S., & Rogers, R. W. (2000). A meta-analysis of research on protection motivation theory. Journal of Applied Social Psychology, 30(2), 407-429.

Gollwitzer, P. M., & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 69-119.

Hreha, J. (2026). The Behavioral State Model: Eight Components of Behavior. The Behavioral Scientist. https://www.thebehavioralscientist.com/articles/the-behavioral-state-model

Hummel, D., & Maedche, A. (2019). How effective is nudging? A quantitative review on the effect sizes and limits of empirical nudging studies. Journal of Behavioral and Experimental Economics, 80, 47-58.

Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010). How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology, 40(6), 998-1009.

Littell, J. H., & Girvin, H. (2002). Stages of change: A critique. Behavior Modification, 26(2), 223-273.

McEachan, R. R. C., Conner, M., Taylor, N. J., & Lawton, R. J. (2011). Prospective prediction of health-related behaviours with the Theory of Planned Behaviour: A meta-analysis. Health Psychology Review, 5(2), 97-144.

McLeroy, K. R., Bibeau, D., Steckler, A., & Glanz, K. (1988). An ecological perspective on health promotion programs. Health Education Quarterly, 15(4), 351-377.

Mertens, S., Herberz, M., Hahnel, U. J. J., & Brosch, T. (2022). The effectiveness of nudging: A meta-analysis of choice architecture interventions across behavioral domains. Proceedings of the National Academy of Sciences, 119(1), e2107346118.

Michie, S., Richardson, M., Johnston, M., Abraham, C., Francis, J., Hardeman, W., ... & Wood, C. E. (2013). The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques. Annals of Behavioral Medicine, 46(1), 81-95.

Michie, S., van Stralen, M. M., & West, R. (2011). The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implementation Science, 6, 42.

Ng, J. Y., Ntoumanis, N., Thogersen-Ntoumani, C., Deci, E. L., Ryan, R. M., Duda, J. L., & Williams, G. C. (2012). Self-determination theory applied to health contexts: A meta-analysis. Perspectives on Psychological Science, 7(4), 325-340.

Ntoumanis, N., Ng, J. Y., Prestwich, A., Quested, E., Hancox, J. E., Thogersen-Ntoumani, C., ... & Williams, G. C. (2021). A meta-analysis of self-determination theory-informed intervention studies in the health domain. Health Psychology Review, 15(2), 214-244.

Prochaska, J. O., & DiClemente, C. C. (1983). Stages and processes of self-change of smoking: Toward an integrative model of change. Journal of Consulting and Clinical Psychology, 51(3), 390-395.

Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change. The Journal of Psychology, 91(1), 93-114.

Rosenstock, I. M. (1966). Why people use health services. Milbank Memorial Fund Quarterly, 44(3), 94-127.

Sheeran, P. (2002). Intention-behaviour relations: A conceptual and empirical review. European Review of Social Psychology, 12(1), 1-36.

Stajkovic, A. D., & Luthans, F. (1998). Self-efficacy and work-related performance: A meta-analysis. Psychological Bulletin, 124(2), 240-261.

Thaler, R. H., & Sunstein, C. R. (2008). Nudge: Improving decisions about health, wealth, and happiness. Yale University Press.

Webb, T. L., Joseph, J., Yardley, L., & Michie, S. (2010). Using the internet to promote health behavior change. Journal of Medical Internet Research, 12(1), e4.

West, R. (2005). Time for a change: Putting the Transtheoretical (Stages of Change) Model to rest. Addiction, 100(8), 1036-1039.

West, R. (2006). Theory of addiction. Wiley-Blackwell.

Wood, W., & Runger, D. (2016). Psychology of habit. Annual Review of Psychology, 67, 289-314.