A guide to the 93-technique classification system that transformed how we design, report, and evaluate behavior change interventions — and where it stands today.
Why Do We Need a Taxonomy of Behavior Change Techniques?
Imagine two research teams, one in London and one in Sydney, both designing interventions to improve medication adherence among diabetic patients. The London team reports using “motivational counseling,” while the Sydney team describes “behavioral support.” Are they doing the same thing? Without a shared vocabulary, no one can tell — and this ambiguity has haunted intervention science for decades.
The CONSORT guidelines for randomized trials call for precise reporting of intervention components, yet researchers and practitioners have historically described the active content of their interventions using inconsistent, overlapping, and often vague terminology. A technique labeled “self-monitoring” in one study might be called “daily diaries” in another; “behavioral counseling” in one protocol might refer to patient education, while in another it encompasses feedback, self-monitoring, and reinforcement combined. The result is a literature that resists replication, defies evidence synthesis, and frustrates implementation.
It was precisely this challenge that Susan Michie, Marie Johnston, Charles Abraham, and their international colleagues set out to address. Published in Annals of Behavioral Medicine in 2013, the Behavior Change Technique Taxonomy version 1 (BCTTv1) offered a systematic, consensually agreed classification of 93 distinct behavior change techniques (BCTs), organized into 16 hierarchical clusters. The paper has since become one of the most cited methodological contributions in behavioral science, providing a foundational tool for intervention specification across health domains worldwide.
What Is a Behavior Change Technique?
Before diving into the taxonomy itself, it is worth clarifying what the authors mean by a BCT. According to Michie and colleagues, a behavior change technique is an observable, replicable, and irreducible component of an intervention designed to alter or redirect causal processes that regulate behavior. In other words, a BCT is a proposed “active ingredient” — the smallest functional unit of an intervention that can be independently described and, in principle, independently tested.
BCTs are distinct from modes of delivery (face-to-face, digital, group-based) and from broader intervention labels (cognitive behavioral therapy, motivational interviewing). A single intervention may contain multiple BCTs, and the same BCT may be delivered through different modes. The taxonomy is concerned with what is delivered, not how it is delivered.
How Was the Taxonomy Developed?
The development process was methodologically rigorous, involving multiple iterative stages over several years:
Stage 1 — Prototype Development. The research team extracted BCTs from six previously published classification systems, yielding 124 candidate techniques. After removing composites and redundancies, 94 BCTs formed the initial prototype.
Stage 2–3 — Delphi Consensus Exercises. Fourteen international experts in behavior change participated in a two-round Delphi exercise. In each round, participants rated BCT labels and definitions for clarity, precision, distinctiveness, and confidence of use. BCTs failing to meet threshold criteria were revised, merged, divided, or removed.
Stage 4 — International Advisory Board Review. Sixteen members of a 30-person international advisory board from seven countries provided feedback through teleconferences, leading to further refinements. The board recommended empirical grouping of BCTs and a versioning system for the taxonomy.
Stage 5 and 7 — Reliability Testing. Study team members coded 85 published intervention descriptions across two rounds. Of the 26 BCTs occurring frequently enough for assessment, 23 achieved adjusted kappa scores of 0.60 or above — the threshold for acceptable inter-rater agreement.
Stage 6 — Hierarchical Clustering. Eighteen experts sorted BCTs into groups based on similarity of active ingredients. Hierarchical cluster analysis with bootstrap resampling (10,000 iterations) identified a 16-cluster solution as optimal, validated by Dunn’s index and figure of merit stability measures.
Stage 8 — Final Review. A lay reviewer ensured comprehensibility beyond behavioral science, and the team standardized all definitions to include active verbs specifying the action required for delivery.
The 93 BCTs Across 16 Groups
The taxonomy organizes 93 techniques into 16 empirically derived clusters. Below is the complete structure, with each cluster and its component BCTs:
1. Goals and Planning (9 BCTs)
This cluster addresses the cognitive and motivational foundations of intentional action. It includes goal setting for both behavior and outcomes, action planning (including implementation intentions), problem solving and coping planning, review of behavioral and outcome goals, commitment, behavioral contract, and identification of discrepancy between current behavior and goal standards. These techniques target the translation of motivation into structured, planful action.
2. Feedback and Monitoring (5 BCTs)
Monitoring and feedback are central to self-regulation. This group includes self-monitoring of behavior, self-monitoring of outcomes of behavior, monitoring of behavior or outcomes by others (with the person’s awareness), feedback on behavior, and biofeedback. These techniques provide the informational substrate upon which people can evaluate and adjust their actions.
3. Social Support (3 BCTs)
Social support is distinguished into three forms: general (unspecified), practical (tangible, instrumental), and emotional. Each represents a qualitatively different resource that others can provide to facilitate behavior change.
4. Shaping Knowledge (4 BCTs)
This cluster concerns techniques that build or restructure the knowledge base needed for behavior change: instruction on how to perform a behavior, information about antecedents (cues and triggers), reattribution (modifying causal attributions), and behavioral experiments (testing empirical hypotheses about behavior).
5. Natural Consequences (6 BCTs)
These techniques draw attention to the naturally occurring results of behavior: health consequences, social and environmental consequences, emotional consequences, salience of consequences, self-assessment of affective consequences, and anticipated regret. They are distinguished from “scheduled consequences” (cluster 10) in that they involve naturally occurring rather than externally arranged outcomes.
6. Comparison of Behavior (3 BCTs)
This group leverages social learning processes: demonstration of the behavior (modeling), social comparison, and information about others’ approval. Each technique draws on observational learning or normative influence to shape behavior.
7. Associations (8 BCTs)
Rooted in associative learning theory, this cluster includes prompts/cues, discriminative (learned) cues, fading, classical conditioning, exposure, satiation, escape learning, and time out. These techniques modify the stimulus environment or the learned associations between stimuli and responses.
8. Repetition and Substitution (7 BCTs)
Techniques in this group target behavioral skill acquisition and habit formation through practice: behavioral rehearsal/practice, behavior substitution, habit formation, habit reversal, overcorrection, graded tasks (progressive difficulty), and generalization of a target behavior.
9. Comparison of Outcomes (3 BCTs)
This cluster involves cognitive evaluation of potential outcomes: persuasive argument, pros and cons analysis, and comparative imagining of future outcomes. These techniques target decisional processes by presenting or eliciting comparative outcome information.
10. Reward and Threat (7 BCTs)
External motivational contingencies are captured here: material reward, social reward, non-specific reward, self-reward, incentive (advance commitment of reward), anticipation of future rewards or removal of punishment, and threat (informing of future punishment or withdrawal of reward).
11. Regulation (4 BCTs)
This cluster addresses the management of internal states that influence behavior: pharmacological support, regulation of negative emotions, conserving mental resources, and paradoxical instructions.
12. Antecedents (4 BCTs)
Environmental restructuring forms the core of this group: restructuring the physical environment, restructuring the social environment, avoidance/changing exposure to cues for the behavior, and distraction. These techniques modify the context in which behavior occurs.
13. Identity (5 BCTs)
Identity-based techniques target the self-concept in relation to behavior: identification of self as role model, reframing, cognitive dissonance, self-affirmation, and identity associated with changed behavior.
14. Scheduled Consequences (10 BCTs)
The largest cluster draws heavily from operant conditioning: punishment, response cost, extinction, discrimination training, shaping, overcorrection, counter-conditioning, differential reinforcement, thinning, and negative reinforcement. These techniques involve the systematic arrangement of consequences to modify behavior.
15. Self-Belief (4 BCTs)
Techniques targeting self-efficacy and confidence: verbal persuasion to boost self-efficacy, focus on past success, self-talk, and mental rehearsal of successful performance.
16. Covert Learning (3 BCTs)
The smallest cluster involves vicarious and imaginal learning processes: vicarious reinforcement, covert sensitization, and covert conditioning.
The Significance of BCTTv1
The contribution of BCTTv1 extends well beyond classification for its own sake. Its significance can be understood along several dimensions:
For systematic reviews, the taxonomy provides a reliable method for extracting and synthesizing information about intervention content. Through meta-regression analyses, researchers can identify which specific BCTs or BCT combinations are associated with effectiveness. This has been demonstrated across physical activity, healthy eating, smoking cessation, alcohol reduction, and sexually transmitted infection prevention.
For intervention design, the taxonomy serves as a comprehensive menu of options. Rather than relying on the limited set of techniques that any individual designer can bring to mind, developers can systematically consider the full range of available BCTs and select those most appropriate for their target behavior, population, and theory of change.
For intervention reporting, BCTTv1 provides the standardized vocabulary that CONSORT and MRC guidelines call for. Two researchers using the same taxonomy to describe the same intervention should produce the same specification, enabling the replication and implementation that cumulative science requires.
For competence frameworks, the taxonomy has enabled the specification of professional competences for delivering BCTs, forming the basis for national training programs — most notably the UK’s National Centre for Smoking Cessation and Training.
For theory-mechanism linkage, BCTs provide the units of analysis needed to link intervention content to theories of behavior change. Preliminary work has mapped BCTs to theoretical domains and mechanisms of action, enabling theory-driven intervention design and evaluation.
Constructive Criticisms and Contextual Considerations
No methodological tool is immune to critique, and the authors themselves anticipated that BCTTv1 would require ongoing development. Several limitations deserve consideration — not to diminish the contribution, but to contextualize it within the evolving landscape of intervention science.
Expert Sample Composition
The Delphi exercise involved 14 experts, predominantly from Europe (eight from the UK) and overwhelmingly from psychology backgrounds. While this number is appropriate for Delphi methods, the relative homogeneity of the sample raises questions about whether the resulting taxonomy fully captures BCTs relevant to non-Western cultural contexts, community-level interventions, or disciplines outside psychology (e.g., nursing, social work, public health education, health management). The taxonomy’s definitions and examples may carry implicit assumptions that reflect a specific disciplinary and cultural lens.
Reliability Coverage
Of the 93 BCTs in the final taxonomy, only 26 occurred frequently enough in the coded intervention descriptions to permit reliability assessment. This means that 67 BCTs — over 70% of the taxonomy — were not formally tested for inter-rater reliability at the time of publication. Among the 26 assessed, three failed to reach the 0.60 kappa threshold. Subsequent reliability work by the same team (published in 2015) tested all 93 BCTs and reported good reliability for 80, which substantially strengthened the evidence base; however, the original paper’s reliability data remain partial.
Hierarchical Structure
The 16-cluster solution was statistically optimal, but the authors noted that 16 groups are too many for easy recall and that a simpler, higher-level structure would be desirable. Four of the 16 clusters had approximately unbiased (AU) p-values below 90%, suggesting that these groupings were less robustly supported by the data. The cluster labels, while empirically derived, sometimes overlap conceptually (e.g., “scheduled consequences” vs. “reward and threat” vs. “natural consequences”), which can create confusion for users unfamiliar with operant conditioning terminology.
Individual-Level Focus
BCTTv1 was developed primarily for interventions delivered to individuals whose behavior change is targeted. The taxonomy is less well suited for community-level, organizational, or policy-level interventions where the “active ingredients” may operate through structural, economic, or legislative mechanisms rather than through direct engagement with individuals. In health management contexts — hospital quality improvement, patient safety culture change, organizational green management — the relevant behavior change processes often involve systems-level dynamics that BCTTv1 was not designed to capture.
Absence of Theory-Mechanism Links
BCTTv1 is explicitly atheoretical: it classifies techniques without specifying the mechanisms through which they are hypothesized to work. While the authors noted this as a deliberate design choice (to avoid premature theoretical constraint), it means that users cannot directly infer from the taxonomy why a technique should work for a given behavior or population. The subsequent development of a Theory and Techniques Tool (2018–2021) addressed this gap by mapping BCTs to mechanisms of action, but this remains a separate resource.
Digital and Technology-Mediated Interventions
The taxonomy was developed in an era when most interventions were delivered face-to-face or via printed materials. The rapid proliferation of digital health interventions — smartphone apps, wearable devices, AI-driven adaptive interventions, chatbots — has introduced delivery modalities and technique combinations that were not envisioned in the original development process. Some researchers have found it challenging to code gamification elements, just-in-time adaptive interventions, or algorithm-driven personalization using BCTTv1 categories.
Dose and Sequencing
BCTTv1 captures the presence of techniques but does not address dose (how much of each technique is delivered), sequence (in what order techniques are delivered), tailoring (how techniques are adapted to individual characteristics), or fidelity (the degree to which techniques are delivered as intended). These dimensions are critical for understanding effectiveness but fall outside the taxonomy’s scope.
From Taxonomy to Ontology: The Ongoing Evolution
The story of BCTTv1 does not end in 2013. The authors intentionally named their work “version 1” to signal that development would continue. And it has — dramatically.
In 2023, a systematic collection of user feedback from six sources (the BCT website, user surveys, the Human Behaviour-Change Project, expert consultations, published research, and other classification systems) yielded 282 proposed changes to BCTTv1. This feedback highlighted needs for more precise definitions, additional BCTs, subdivision of existing BCTs, and a more flexible hierarchical structure.
The result is the Behaviour Change Technique Ontology (BCTO), published by Marques, Wright, Corker, Johnston, West, Hastings, Zhang, and Michie in 2024. The BCTO extends BCTTv1 from a flat hierarchical taxonomy into a formal ontology — a knowledge structure that defines not only entities (BCTs) but also the relationships between them.
The development process involved splitting 19 original BCTs into two or more sub-techniques, adding 27 new BCTs, and reorganizing groupings. Through iterative expert review and inter-rater reliability testing, the BCTO ultimately reached 281 BCTs organized into 20 higher-level groups across five hierarchical levels — a threefold expansion from the original 93. Inter-rater reliability testing showed good agreement (kappa values of 0.82 and 0.79 for researchers familiar and unfamiliar with the ontology, respectively).
The BCTO is part of a larger Behaviour Change Intervention Ontology (BCIO) that integrates BCTs with mechanisms of action, modes of delivery, population characteristics, and contextual factors. This broader framework, supported by the Theory and Techniques Tool, moves the field from classification toward causal modeling.
Implications for Future Research Design
For researchers planning new studies, BCTTv1 (and its successor, the BCTO) offers several practical implications:
Intervention specification. Any trial evaluating a behavior change intervention should specify its content using a recognized taxonomy. This is not merely a reporting recommendation; it is a prerequisite for replicability. Researchers should use the most current version of the classification system available and clearly state which version they used.
Systematic review methodology. Reviews synthesizing behavioral interventions should train coders using standardized BCT training resources (available at bct-taxonomy.com) and report inter-rater reliability for BCT coding. The shift from BCTTv1 to BCTO may require reviews to establish concordance between the two systems when synthesizing studies coded under different versions.
Theory-mechanism testing. The availability of BCT-to-mechanism-of-action mappings enables researchers to design studies that explicitly test hypothesized causal pathways. Mediation analyses examining whether a BCT produces its effect through the predicted mechanism represent a critical next step beyond simple effectiveness trials.
Digital health evaluation. As digital interventions become increasingly prevalent, researchers need to develop supplementary coding frameworks for technology-specific features (gamification, adaptive algorithms, passive sensing) that complement BCTTv1/BCTO categories.
Organizational and population-level interventions. There remains a significant gap in the classification of behavior change techniques operating at the organizational, community, and policy levels. Future research should extend the taxonomy to capture the active ingredients of quality improvement programs, patient safety culture interventions, and health system governance reforms.
Cross-cultural validation. The taxonomy’s applicability across non-Western cultural contexts has not been systematically tested. Collaborative international research programs are needed to evaluate whether BCT definitions, examples, and groupings are meaningful and useful in diverse cultural and linguistic settings.
Conclusion
BCTTv1 represents a landmark contribution to intervention science — one of those rare methodological papers that changes how an entire field thinks about its work. By providing a consensual, hierarchically structured vocabulary for the active ingredients of behavior change interventions, Michie and colleagues made it possible, for the first time, to reliably compare, synthesize, and replicate complex behavioral interventions across studies, populations, and settings.
The taxonomy is not perfect, and it was never intended to be final. Its limitations — the restricted expert sample, the partial reliability data, the individual-level focus, the absence of theory-mechanism links — are real but addressable. Many have already been addressed through subsequent work, culminating in the BCTO’s 281 techniques organized into a five-level ontological structure. The trajectory from BCTTv1 to BCTO mirrors the general maturation of behavioral science: from listing and classifying toward formal knowledge representation, causal modeling, and computational reasoning.
For health management researchers and practitioners, the message is clear: behavior change interventions are not black boxes. They can be opened, their components identified, their mechanisms hypothesized, and their effects predicted and tested. BCTTv1 gave us the first comprehensive toolkit for doing so. The ongoing evolution of that toolkit ensures that it will continue to serve the field for years to come.
Reference: Michie, S., Richardson, M., Johnston, M., Abraham, C., Francis, J., Hardeman, W., Eccles, M. P., Cane, J., & Wood, C. E. (2013). The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: Building an international consensus for the reporting of behavior change interventions. Annals of Behavioral Medicine, 46(1), 81–95. https://doi.org/10.1007/s12160-013-9486-6
Further Reading:
- Marques, M. M., Wright, A. J., Corker, E., Johnston, M., West, R., Hastings, J., Zhang, L., & Michie, S. (2024). The behaviour change technique ontology: Transforming the behaviour change technique taxonomy v1. Wellcome Open Research, 8, 308.
- Corker, E., Marques, M. M., Johnston, M., West, R., Hastings, J., Michie, S. (2023). Behaviour change techniques taxonomy v1: Feedback to inform the development of an ontology. Wellcome Open Research.
This post is part of the healthtopic.org academic blog series covering foundational frameworks in behavioral science and health management research.