Abstract
Background
Testing mechanisms of action (MoAs) hypothesized to drive behaviour change improves intervention efficacy and allows theoretical propositions to be evaluated, enabling evidence accumulation. However, clear correspondence between behavioural MoAs and associated measures is lacking, creating challenges for intervention evaluation.
Aims
To link well‐defined behavioural MoAs to multidimensional measures.
Method
Two researchers independently judged whether 44 measures (comprising 131 (sub)scales) in the Science of Behavior Change (SOBC) Measures Repository are suitable for measuring 270 MoAs from the Human Behaviour‐Change Project's MoA Ontology (2022). Links were categorized as ‘confirmed’ (aligned with a prior expert opinion study linking measures to 26 MoAs), ‘removed’ or ‘new’. Judgements were compared, discussed, reconciled iteratively and jointly reviewed for consistency.
Results
Six hundred and eighty‐six links between SOBC measures and MoAs were identified (397 ‘confirmed’, 289 ‘new’). Measures were found to tap into multiple MoAs, with 5.24 MoAs linked to each measure on average. These links demonstrated greater granularity than those identified in a previous expert opinion study because MoAs from the MoA Ontology were more specific than MoAs from this previous study. Commonly co‐occurring MoAs were identified (e.g., ‘self‐regulation process’ and ‘self‐regulation capability’) and MoAs potentially missing from the ontology were noted.
Conclusion
The refined measure‐MoA links provide more precise guidance for researchers when designing and/or selecting measures to assess the role of MoAs in theory‐based behavioural interventions. Future research should further explore measure‐MoA links by, for example, testing the discriminant content validity of ostensibly distinct measures that tap into the same or similar MoAs.
Keywords: behaviour, measure, measurement, mechanism, ontology
INTRODUCTION
Understanding the mechanisms by which behavioural interventions elicit desired change (e.g., decreased sedentary time, increased medication adherence) has the potential to increase their potency and scalability. As described by the Science of Behavior Change (SOBC) program (Sumner et al., 2018, 2019), the experimental medicine approach to behaviour change necessitates that a researcher (1) identifies a putative mechanism that underlies behaviour change, (2) measures that mechanism, (3) tests whether an intervention leads to a change in the measured mechanism and, finally, (4) tests whether change in the mechanism is associated with a change in behaviour. Although the value of this approach is widely recognized (Riddle, 2015), few randomized trials actually measure a hypothesized mechanism, and even fewer test whether mechanism change underlies behaviour change (Edmondson et al., 2018). Furthermore, there is not enough research that matches measures to their proposed mechanisms of action (MoAs) (see glossary of bold and italicized terms in Table 1), and many measurements and their subscales are complex and multi‐dimensional, meaning they do not comprehensively assess a specific MoA. Without a clear link between the measurements used and the MoAs that they tap into, there are significant challenges for behavioural scientists working to refine theories and develop robust, theory‐based interventions. To support the testing of interventions' MoAs, the SOBC has provided resources to behavioural scientists, including an online repository comprising putative measures of MoAs (https://measures.scienceofbehaviorchange.org/).
TABLE 1.
Glossary of terms.
| Label | Definition | References |
|---|---|---|
| Mechanism of action (MoA) | A process through which an intervention (e.g., a behaviour change intervention) influences an outcome behaviour | Schenk et al. (2024) |
| Repository | A centralized data storage in which files and resources can be kept | Gupta (2023) |
| Ontology | A classification framework that includes representations of entities, as well as unique labels, definitions and alphanumeric IDs for these entities and specifies the relationships between entities | Arp et al. (2015) |
| Entity | Anything that exists, including processes, objects and their objects. Entities are represented in ontologies, and these representations are referred to as ‘class’ | Arp et al. (2015) |
| Class | In ontologies, classes represent types or groupings of entities in the world. The terms ‘entity’ and ‘class’ are sometimes used interchangeably when referring to representations of entities in an ontology. Ontologies are hierarchically structure by linking a class to its parent class (see definition of parent class) | Arp et al. (2015) |
| Parent class | A class that has a hierarchical relationship to one or more subclasses, with the subclasses having all the properties of the parent and an additional more specific property | Arp et al. (2015) |
| Relationship | The way in which two things (e.g., classes) are linked | Arp et al. (2015) |
A complementary effort to the SOBC has been led by the Human Behaviour Change Project's researchers (HBCP; Michie et al., 2017; Michie et al., 2020; Wright et al., 2020), who focused on clearly operationalizing and organizing MoAs commonly targeted in these behavioural interventions. Clear operationalization of these MoAs is a critical component of conducting valid mechanistic hypothesis tests and a necessary precursor to implementing robust measurement (i.e., one must operationalize a construct before it is possible to validate and utilize a measure of that construct) (Derby et al., 2024). As a first step in addressing this gap, in a collaborative project between SOBC and the HBCP, we conducted an expert opinion study to identify links between self‐report measures in the online SOBC Measures Repository (Science Of Behavior Change (SOBC)) and 26 MoAs identified from behavioural theories and models (Bohlen et al., 2019; Carey et al., 2019; Connell et al., 2019; Johnston et al., 2018, 2021). This study identified 167 measure‐MoA links (Cornelius et al., 2023); however, the 26 MoAs included in this study comprised broad categories such as ‘Memory, Attention and Decision‐Making’ that did not capture more granular MoAs or less frequently investigated but potentially influential MoAs in behaviour change interventions. This limitation meant that the mapping cannot be used to precisely identify which specific MoAs each measurement taps into. For instance, the ‘Cognitive Reflection Test’ includes items related to self‐regulation and cognitive processes. To capture its cognitive aspects, the broad ‘Memory, Attention and Decision‐Making’ MoA was linked to this measurement. However, this link provides a general overview of the MoAs (e.g., cognitive process) associated with the ‘Cognitive Reflection Test’, which may limit its utility for users of the mapping.
More recently, the HBCP has undertaken a mammoth project to define and organize a greatly expanded set of MoAs hypothesized to underlie behaviour change into a sub‐ontology as part of its Behaviour Change Intervention Ontology (BCIO): The MoA Ontology (Michie et al., 2017, 2020; Schenk et al., 2024). This ontology's goal is to provide a clear and extensive classification system that labels, defines and organizes MoAs in behaviour change intervention research. To capture a wide range of MoAs and reflect various perspectives on MoAs, the MoA Ontology was developed by drawing on the constructs of 83 behavioural theories, applying the ontology to code MoAs in various published intervention reports and consulting international behavioural science experts (Schenk et al., 2024). This ontology was iteratively updated, with the published version including 284 classes (i.e., groupings or categories) of MoAs organized onto seven hierarchical levels. An example of the structure of the ontology's hierarchy can be seen in Figure 1. The ontology has since been further updated to more precisely capture the constructs of behavioural theories, now including 621 classes to capture MoAs (release date: 18 July 2025). The most up‐to‐date version of the ontology will always be available on the Human Behaviour Change Project's GitHub repository: https://github.com/HumanBehaviourChangeProject/ontologies/tree/master/MechanismOfAction.
FIGURE 1.
Snapshot of the Mechanism of Action (MoA) Ontology in its hierarchical format, with the arrows showing parent class–class relationships (e.g., self‐regulation of behaviour with its subclasses).
Ontologies are formally defined as classification systems that include classes (representations of entities, anything that exists in the universe, including objects, processes and attributes) with unique IDs, labels and definitions and relationships (Arp et al., 2015). These are often organized in a hierarchical format, with classes nested under a parent class, and each class definition should be unambiguous, clearly specifying a class's parent class and its distinguishing attributes from that parent class (Michie et al., 2019; Seppälä et al., 2017). For example, the definition of the class ‘motivation’ within the MoA Ontology first specifies that it is a ‘mental process’ (parent class) and then describes what distinguishes motivation from other subclasses belonging to the ‘mental process’ parent class. Ontologies are developed to be readable by both humans and computers, meaning they can be used to develop reasoning algorithms for applications such as data extraction and prediction about their classes and relationships (Hastings, 2017). Each class in an ontology has a unique alphanumeric ID (e.g., BCIO:006133), enabling them to be computer readable. They are also intended to be updated based on evidence and user feedback (He et al., 2018), so that they become increasingly useful and/or remain relevant resources for the research community over time.
The expert opinion study linking SOBC measurements to MoAs (Cornelius et al., 2023) was limited by the inclusion of only 26 MoAs, as well as the use of the term ‘related’ to determine measure–MoA links (i.e., it was not necessary that the measure actually assessed these MoAs to create a link). Furthermore, the SOBC measures were considered as a unit, with no recognition of potentially different MoAs across subscales. The present study was undertaken with the goal of improving the utility and specificity of this previous mapping work. Using the 167 measure–MoA links as a starting point (Cornelius et al., 2023), we aimed to confirm, remove and create new links between the measure subscales, with a focus on whether items in these scales actually intended to assess different classes of MoAs in the expanded MoA Ontology. We anticipated that the results of this study would provide valuable information regarding the content of these measures for assessing different MoAs, as well as data regarding potential overlap in assessment tools between definitionally non‐overlapping classes of MoAs. The study can, thereby, present a starting point to further investigate the measurements' content and validity and inform improvements to the MoA Ontology through its application. The ultimate goal of this work is to create more informed and efficacious interventions to improve health and well‐being.
MATERIALS AND METHODS
This study was reported according to consolidated criteria for reporting qualitative research (COREQ), where relevant (Tong et al., 2007). The study involved a deductive mapping of measurements guided by the MoA Ontology. Throughout the study, the judgements to refine the measure–MoA links were made by two researchers with doctoral degrees relating to psychology, one working as assistant professor (TC—White, cis‐gender female) and another as research fellow (PS—White, cis‐gender female). The first researcher had experience in applying and organizing measurement related to MoAs in repositories, while the second researcher had experience in developing and applying ontologies.
Materials
Measurements
The 167 measure–MoA links (Cornelius et al., 2023) between the 44 self‐report measures in the SOBC Measures Repository (https://measures.scienceofbehaviorchange.org) and the 26 MoAs identified by the HBCP (Carey et al., 2019; Connell et al., 2019; Johnston et al., 2018) served as the starting point for the present study. In contrast to the original study, based on expert feedback stating that some subscales within a given measure seemed to assess substantially different constructs, measure‐MoA links were separated out by subscale as applicable (see Supplementary Material S1). The starting assumption was made that if a measure was linked to one of the 26 MoAs, then so was its subscale. Thus, instead of 44 measures, there were 131 unique measure and subscale combinations, and the number of links increased from 167 to 529 (e.g., the 10‐Item Personality inventory was linked to 3 MoAs in the expert opinion study—considering these links separately across the five subscales resulted in 15 measure–MoA links). For simplicity, both measures without subscales and measures separated into subscales will be referred to as ‘measure’ in the Methods and Results sections.
MoAs
The classes of MoAs from the MoA Ontology part of the Behaviour Change Intervention Ontology were used in this study. As the ontology was not published at the start of the current work, an earlier version of the ontology (published on OSF in November 2022), with 270 classes, was used (see Version 5 here: https://osf.io/pkq4e). The ontology has been reported in a paper with 284 classes (see log of 12 added classes: https://osf.io/6gzuk). Since then, further updates to the ontology have resulted in 621 classes (https://github.com/HumanBehaviourChangeProject/ontologies/tree/master/MechanismOfAction; release date: 18 July 2025). However, such changes are expected, given the evolving nature of ontologies. The use of unique identifiers (IDs) for each ontology class ensures that the version applied in this study remains traceable and consistent with the latest version.
Procedure
The two researchers first independently judged whether each measure was linked to classes of MoAs from the MoA Ontology, recording their judgements on an Excel spreadsheet. This involved the researchers reading each measure's items and then separately coding which classes best captured these items. More than one class could be coded for each measure. To inform their coding, the researchers used iteratively developed guidelines (see Supplementary Material S2).
Measure–MoA links from the expert opinion study (Cornelius et al., 2023) served as a starting point for these judgements. If a class from the ontology was coded for a measure and this class's definition captured the MoA definition from the expert opinion study, then the newly coded measure‐MoA link was considered to be ‘confirmed’. For instance, the measure ‘Future Orientation Scale’ was linked to the MoAs ‘Memory, Attention and Decision Processes’ in Cornelius et al. (2023). Because a related class from the ontology, ‘decision‐making’, mapped on to this measure in the present study, this was considered ‘confirmed’. An original measure–MoA link could be ‘confirmed’ by one or more classes (e.g., a link to the original MoA of self‐regulation could be confirmed by links to both ‘self‐regulation process’ and ‘self‐regulation capability’). Conversely, if the previously coded MoA was not captured by a class from the ontology, the link determined by expert opinion was considered ‘removed’. Classes that were coded and did not correspond to previously coded MoAs were ‘new’ links.
To draft the coding guidelines, the researchers first independently coded 15 measures (three scales with altogether 15 subscales) as a pilot and discussed and reconciled their disagreements. Given the diversity of the remaining measures (e.g., cognitive reasoning and environmental measures), they then independently coded 3–5 measurement scales at a time, discussed and reconciled disagreements and iteratively updated the guidelines. Once the coding process was complete, they jointly reviewed and discussed the final decisions for all measures to ensure that the guidelines had been applied consistently.
Data analysis strategy
We calculated percent agreement for all researcher judgments on measure‐MoA links made during the initial round of coding. We additionally calculated the number of codes that were changed during final review, after the iteratively developed codebook was complete and quality checks were done to ensure consistency.
Frequency of ‘confirmed’ and ‘removed’ links from the expert opinion study were computed; the frequency of MoA classes from the ontology for both ‘confirmed’ and ‘new’ links are presented in graphical format.
To highlight correspondence between the 26 original MoAs and classes of MoAs from the ontology, a mapping of the previously coded MoA to the ontology's classes for ‘confirmed’ links is presented. Summary statistics for the number of classes for each measure are also reported.
Although each class in the MoA Ontology is considered to be distinct, it is probable that these classes are not assessed separately by the measures (i.e., the measure is not a ‘pure’ measure of a given class). Thus, we identified commonly co‐occurring classes within a given measure (i.e., classes showed up in combination at least five times) for groups of two or more classes.
Finally, notes taken during the coding process were reviewed after reconciliation was complete to compile a list of classes potentially missing from the MoA Ontology. These suggested classes were submitted to the ontology developers for consideration in future iterations of the MoA Ontology.
RESULTS
The percent agreement between the two researchers' initial coding of classes of MoAs onto measures was 59.4% (see the researchers' coding record here: https://osf.io/pd9fc). Key disagreements related to the complexity of deciding which specific classes should be included to capture measurement items, particularly regarding (1) whether an item actually captures an MoA (e.g., whether a measure of self‐regulating emotion not only captures the self‐regulation process itself but the emotion being regulated) and (2) the granularity of the classes coded (e.g., ‘self‐efficacy belief for a behaviour’ vs. ‘self‐efficacy belief for a behaviour and its outcome’). Following reconciliations, 1122 judgements were recorded as links (678 judgements) or non‐links (444 judgements). During the final review, 71 judgements were updated when the researchers jointly reviewed their reconciled results (93.7% agreement between judgements in the initial reconciliation and review), and 55 additional links were recorded.
Of the 167 original links from the expert opinion study, 108 were considered to be confirmed (64.67%; i.e., the measure or one of its subscales corresponded to a class from the ontology that was captured by the original MoA), and 59 were removed (35.33%). These are summarized in Table 2.
TABLE 2.
Confirmed and removed links from the original expert opinion study.
| Measure | Original MoA | Link result |
|---|---|---|
| 10‐Item Personality Inventory (TIPI) | Emotion | ✓ |
| TIPI | Self‐Image | ✓ |
| TIPI | Social/Professional Role & Identity | ✗ |
| Behavioural Inhibition and Activation Systems (BIS‐BAS) Scales | Behavioural Regulation | ✓ |
| BIS/BAS | Goals | ✓ |
| BIS/BAS | Motivation | ✓ |
| BIS/BAS | Reinforcement | ✓ |
| BIS/BAS | Self‐Image | ✓ |
| Barratt Impulsiveness Scale (BIS) | Behavioural Regulation | ✓ |
| BIS | Beliefs About Capabilities | ✗ |
| BIS | Emotion | ✗ |
| BIS | Memory, Attention and Decision Processes | ✓ |
| BIS | Self‐Image | ✓ |
| Behaviour Rating Inventory of Executive Function (BRIEF)—Adults | Behavioural Regulation | ✓ |
| BRIEF—Adults | Beliefs About Capabilities | ✓ |
| BRIEF—Adults | Memory, Attention and Decision Processes | ✓ |
| BRIEF—Adults | Self‐Image | ✗ |
| BRIEF—Adults | Skills | ✗ |
| Brief COPE | Behavioural Regulation | ✓ |
| Brief COPE | Beliefs About Capabilities | ✗ |
| Brief COPE | Emotion | ✓ |
| Brief COPE | Self‐Image | ✗ |
| Brief COPE | Skills | ✓ |
| Brief Risk‐Resilience Index for Screening (BRISC) | Behavioural Regulation | ✓ |
| BRISC | Beliefs About Capabilities | ✓ |
| BRISC | Emotion | ✓ |
| BRISC | Self‐Image | ✓ |
| BRISC | Skills | ✗ |
| Brief Self‐Control (BSC) | Behavioural Regulation | ✓ |
| BSC | Beliefs About Capabilities | ✓ |
| BSC | Self‐Image | ✓ |
| BSC | Skills | ✗ |
| Cognitive Reflection Test (CRT) | Behavioural Regulation | ✓ |
| CRT | Memory, Attention and Decision Processes | ✓ |
| CRT | Skills | ✗ |
| Consideration of Future Consequences (CFC) Scale | Behavioural Regulation | ✓ |
| CFC Scale | Beliefs About Consequences | ✗ |
| CFC Scale | Goals | ✓ |
| CFC Scale | Memory, Attention and Decision Processes | ✓ |
| CFC Scale | Motivation | ✓ |
| CFC Scale | Values | ✓ |
| Couple Coercion Scale | Behavioural Regulation | ✓ |
| Couple Coercion Scale | Emotion | ✓ |
| Couple Coercion Scale | Social Influences | ✓ |
| Daily Inventory of Stressful Events (DISE) | Emotion | ✓ |
| DISE | Environmental Context and Resources | ✓ |
| DISE | Perceived Susceptibility /Vulnerability | ✓ |
| Deferment of Gratification Scale | Behavioural Regulation | ✓ |
| Deferment of Gratification Scale | Memory, Attention and Decision Processes | ✗ |
| Deferment of Gratification Scale | Motivation | ✗ |
| Deferment of Gratification Scale | Reinforcement | ✗ |
| Dickman Functional and Dysfunctional Impulsivity Inventory (DII) | Behavioural Regulation | ✓ |
| DII | Beliefs About Capabilities | ✓ |
| DII | Memory, Attention and Decision Processes | ✓ |
| DII | Self‐Image | ✗ |
| Domain Specific Risk‐Taking Survey (DOSPERT)—Expected Benefits | Attitudes Towards Behaviour | ✗ |
| DOSPERT—Expected Benefits | Behavioural Regulation | ✗ |
| DOSPERT—Expected Benefits | Beliefs About Consequences | ✓ |
| DOSPERT—Expected Benefits | Memory, Attention and Decision Processes | ✗ |
| DOSPERT—Risk Perceptions | Attitudes Towards Behaviour | ✗ |
| DOSPERT—Risk Perceptions | Beliefs About Consequences | ✓ |
| DOSPERT—Risk Perceptions | Memory, Attention and Decision Processes | ✗ |
| DOSPERT—Risk Perceptions | Perceived Susceptibility /Vulnerability | ✓ |
| DOSPERT—Risk Taking | Attitudes Towards Behaviour | ✗ |
| DOSPERT—Risk Taking | Beliefs About Consequences | ✗ |
| DOSPERT—Risk Taking | Intentions | ✓ |
| DOSPERT—Risk Taking | Perceived Susceptibility /Vulnerability | ✗ |
| DOSPERT—Risk Taking | Self‐Image | ✗ |
| Ecological Momentary Assessment of Stressful Events | Emotion | ✓ |
| Ecological Momentary Assessment of Stressful Events | Environmental Context and Resources | ✓ |
| Emotion Regulation Questionnaire (ERQ) | Behavioural Regulation | ✓ |
| ERQ | Beliefs About Capabilities | ✗ |
| ERQ | Emotion | ✓ |
| ERQ | Memory, Attention and Decision Processes | ✗ |
| ERQ | Skills | ✗ |
| Emotion Regulation Strategies Scale | Behavioural Regulation | ✓ |
| Emotion Regulation Strategies Scale | Emotion | ✓ |
| Emotion Regulation Strategies Scale | Skills | ✗ |
| Five Facets of Mindfulness Questionnaire (FFMQ) | Behavioural Regulation | ✓ |
| FFMQ | Emotion | ✓ |
| FFMQ | Memory, Attention and Decision Processes | ✓ |
| FFMQ | Self‐Image | ✗ |
| FFMQ | Skills | ✗ |
| Future Orientation Scale | Behavioural Regulation | ✓ |
| Future Orientation Scale | Goals | ✗ |
| Future Orientation Scale | Memory, Attention and Decision Processes | ✓ |
| Future Orientation Scale | Skills | ✗ |
| Future Time Perspective Scale (FTP) | General Attitudes/Beliefs | ✓ |
| FTP | Optimism | ✗ |
| Generalized Self‐Efficacy Scale (GSE) | Behavioural Regulation | ✓ |
| GSE | Beliefs About Capabilities | ✓ |
| GSE | Optimism | ✗ |
| GSE | Self‐Image | ✗ |
| GSE | Skills | ✗ |
| Grit‐S | Behavioural Regulation | ✓ |
| Grit‐S | Beliefs About Capabilities | ✗ |
| Grit‐S | Goals | ✓ |
| Grit‐S | Motivation | ✗ |
| Grit‐S | Self‐Image | ✓ |
| I‐7: Impulsiveness and Venturesomeness Questionnaire | Behavioural Regulation | ✓ |
| I‐7: Impulsiveness and Venturesomeness Questionnaire | Emotion | ✓ |
| I‐7: Impulsiveness and Venturesomeness Questionnaire | Memory, Attention and Decision Processes | ✓ |
| I‐7: Impulsiveness and Venturesomeness Questionnaire | Self‐Image | ✓ |
| Kessler Psychological Distress Scale (K6+) | Emotion | ✓ |
| K6+ | Self‐Image | ✓ |
| Mindful Attention Awareness Scale (MAAS) | Behavioural Regulation | ✓ |
| MAAS | Memory, Attention and Decision Processes | ✓ |
| MAAS | Skills | ✗ |
| Multidimensional Assessment of Interoceptive Awareness (MAIA) | Behavioural Regulation | ✓ |
| MAIA | Emotion | ✓ |
| MAIA | Memory, Attention and Decision Processes | ✓ |
| MAIA | Self‐Image | ✗ |
| MAIA | Skills | ✗ |
| Multidimensional Personality Questionnaire (MPQ): Control vs. Impulsivity Scale | Behavioural Regulation | ✓ |
| MPQ: Control vs. Impulsivity Scale | Memory, Attention and Decision Processes | ✗ |
| MPQ: Control vs. Impulsivity Scale | Self‐Image | ✓ |
| NIH Self‐Efficacy Scale | Behavioural Regulation | ✓ |
| NIH Self‐Efficacy Scale | Beliefs About Capabilities | ✓ |
| NIH Self‐Efficacy Scale | Optimism | ✗ |
| NIH Self‐Efficacy Scale | Self‐Image | ✗ |
| NIH Self‐Efficacy Scale | Skills | ✗ |
| Parent Cognition Scale | Social Influences | ✓ |
| Parent–Child Coercion Scale | Social Influences | ✓ |
| Parent–Child Coercion Scale | Social/Professional Role and Identity | ✗ |
| Parent‐rated Stress (NIH Perceived Stress Scale) | Behavioural Regulation | ✓ |
| Parent‐rated Stress (NIH Perceived Stress Scale) | Emotion | ✓ |
| Pearlin Mastery Scale | Behavioural Regulation | ✗ |
| Pearlin Mastery Scale | Beliefs About Capabilities | ✓ |
| Pearlin Mastery Scale | Optimism | ✗ |
| Pearlin Mastery Scale | Self‐Image | ✗ |
| Positive and Negative Affect Scheduled (PANAS) | Emotion | ✓ |
| PANAS—Child | Emotion | ✓ |
| SIDES Affect Dysregulation Scale (Child‐Reported) | Behavioural Regulation | ✓ |
| SIDES Affect Dysregulation Scale (Child‐Reported) | Beliefs About Capabilities | ✗ |
| SIDES Affect Dysregulation Scale (Child‐Reported) | Emotion | ✓ |
| SIDES Affect Dysregulation Scale (Child‐Reported) | Self‐Image | ✗ |
| SIDES Affect Dysregulation Scale (Child‐Reported) | Skills | ✗ |
| Selection, Optimization, and Compensation (SOC) Questionnaire | Behavioural Regulation | ✓ |
| SOC Questionnaire | Beliefs About Consequences | ✗ |
| SOC Questionnaire | Goals | ✓ |
| SOC Questionnaire | Memory, Attention and Decision Processes | ✓ |
| SOC Questionnaire | Self‐Image | ✗ |
| Short Self‐Regulation Questionnaire | Behavioural Regulation | ✓ |
| Short Self‐Regulation Questionnaire | Beliefs About Capabilities | ✓ |
| Short Self‐Regulation Questionnaire | Goals | ✓ |
| Short Self‐Regulation Questionnaire | Self‐Image | ✓ |
| Short Self‐Regulation Questionnaire | Skills | ✓ |
| Theories of Willpower Scale | Behavioural Regulation | ✗ |
| Theories of Willpower Scale | Beliefs About Capabilities | ✓ |
| Theories of Willpower Scale | Beliefs About Consequences | ✓ |
| Theories of Willpower Scale | General Attitudes/Beliefs | ✓ |
| Three Factor Eating Questionnaire R‐18 (TFEQ‐r18) | Behavioural Cueing | ✓ |
| TFEQ‐r18 | Behavioural Regulation | ✓ |
| TFEQ‐r18 | Beliefs About Capabilities | ✓ |
| TFEQ‐r18 | Beliefs About Consequences | ✓ |
| TFEQ‐r18 | Emotion | ✓ |
| TFEQ‐r18 | Self‐Image | ✗ |
| UPPS‐P Impulsive Behaviour Scale | Behavioural Regulation | ✓ |
| UPPS‐P | Beliefs About Capabilities | ✓ |
| UPPS‐P | Memory, Attention and Decision Processes | ✓ |
| UPPS‐P | Self‐Image | ✓ |
| Zimbardo Time Perspective Inventory (ZTPI) | Behavioural Regulation | ✓ |
| ZTPI | Beliefs About Consequences | ✓ |
| ZTPI | Optimism | ✗ |
| ZTPI | Self‐Image | ✗ |
| Zuckerman Sensation Seeking Survey‐V | Behavioural Regulation | ✓ |
| Zuckerman Sensation Seeking Survey‐V | Self‐Image | ✗ |
Within the new ontology, and separated by subscale, there were a total of 397 links that confirmed expert opinion, 289 new links, and 309 removed links (https://osf.io/pd9fc). Figures 2 and 3 depict the frequency of the new MoAs when these were confirmed (Figure 2) or new (Figure 3).
FIGURE 2.
Frequency of new classes from the Mechanism of Action (MoA) Ontology that confirmed measure–MoA links from the earlier study.
FIGURE 3.
Frequency of new classes from the Mechanism of Action (MoA) Ontology that were part of new links.
There were 5.24 MoAs on average for the 131 measure/subscale combinations (SD = 3.78; Median = 5.0; Range 1.0, 28.0). The mapping of old MoAs to the MoAs from the ontology is in Supplementary Material S3. For example, for measurements relevant to the original MoA ‘Goals’, the new mapping included five classes from the MoA Ontology: ‘appraisal of goal importance’ (MFOEM:000072), ‘belief about consequences of goal attainment’ (BCIO:006021), ‘goal’ (BCIO:006049), ‘goal pursuit process’ (BCIO:006096) and ‘goal setting process’ (BCIO:006114).
To see whether certain MoAs in the ontology tended to be coded together (i.e., were often coded together within a single Measure/Subscale combination), we looked to see which MoAs were coded in groups of 2 through 10 at least 5 or more times. All common combinations are in Supplementary Material S4. No combinations of 7 or more MoAs were coded at least 5 times across each measure/subscale. The most common pairing, reoccurring for 36 measures, was ‘self‐regulation capability’ (BCIO:006005) and ‘self‐regulation process’ (BCIO:050268).
A total of 23 classes (e.g., self‐efficacy belief for mental process) were identified as potentially missing from the MoA Ontology and suggested labels and definitions for these classes (see the full list of classes here: https://osf.io/6gzuk). This list has been shared with the MoA Ontology development team for consideration. The need for greater specification of classes about self‐regulatory processes also became apparent throughout this process because self‐regulation processes involve two parts (i.e., process A to regulate process B). For example, one may use behaviour to regulate an emotional process or, vice versa, emotions may be used by an individual to regulate their behaviour. This issue has also been raised with the MoA Ontology developers.
DISCUSSION
Building on the previous mapping of 44 self‐report measures from the SOBC repository onto 26 broad MoAs, the current study provided a more granular mapping of the 131 scales in these measures onto 270 clearly defined and distinct classes from an expanded MoA Ontology. This study also served to provide data regarding the distinctness of MoAs in practice (i.e., identify patterns of co‐occurrence) and to identify potentially missing classes from the MoA Ontology. There were a total of 529 links between measures (subscales) and MoAs. Of these, 397 were considered to confirm links from the expert opinion study (Cornelius et al., 2023), and 289 were new. Many classes were commonly coded together for a given measure (subscale), and 23 new classes were suggested for inclusion in future iterations of the MoA Ontology.
Study results provide more detailed guidance for researchers considering how to best measure an MoA hypothesized to underlie behaviour change and should encourage careful attention to correspondence between theory and methodology. MoAs should be clearly defined, distinct and measured both precisely and completely. Separating out measures by subscale and utilizing an expanded set of MoAs enabled greater precision. For example, although the two subscales of the Emotion Regulation Questionnaire, Cognitive Reappraisal (6 items; e.g., ‘When I want to feel a more positive emotion (such as joy or amusement), I change what I am thinking about’) and Expressive Suppression (4 items; e.g., ‘When I am feeling negative emotions, I make sure not to express them’), are related, the processes they measure are distinct (Gross & John, 2003). Furthermore, whereas the expert opinion study had only one broad MoA for ‘Emotion’, classes from the MoA Ontology were separated into more clear and distinct MoAs (e.g., ‘anger’ and ‘happiness’). The greater level of detail in the MoAs specified within the ontology serves as a precise and granular classification framework for describing MoAs in reports and for synthesizing and organizing evidence and measurements. Simpler frameworks, such as the Theoretical Domains Framework (TDF; Cane et al., 2012), have the advantage of being easier to apply, but the disadvantage of being high level and therefore lacking the detail of the MoA Ontology. Reasons for easier application include fewer categories, and definitions that capture a wider range of potential applications. Recent work has mapped the MoA Ontology's expanded classes to the 26 MoAs included in the Theories and Techniques Tool (TaTT) (Zhang et al., 2025) and the previous work to map measurements to MoAs (Cornelius et al., 2023). These 26 MoAs also include the 14 domains of the TDF, thereby providing a method for starting with a broad domain in the TDF (e.g., ‘Emotion’) and then identifying more detailed corresponding subclasses within the MoA Ontology.
Classes within the MoA Ontology are specified such that definitions do not overlap, but rather each class represents varying specificity within a hierarchical structure. Within this organizing framework, classes may commonly co‐occur (i.e., frequently be linked in common groupings to different measures), and this may be due to different reasons. First, it is possible that co‐occurrence indicates child classes within a shared parent class, such that a measure assesses multiple facets of an overarching construct. As an example, the Positive and Negative Affect Scales (PANAS; Watson et al., 1988) included the classes ‘feeling at ease’ (MFOEM:000107) and ‘feeling nervous’ (MFOEM:000124) both of which are children of the parent class ‘subjective affective feeling’ (MFOEM:000006). If all child classes are captured by a measure, this may also indicate that this measure captures the parent class completely. However, such claims should be made with caution because they assume that every possible class has been included within the ontology. Co‐occurrence can also indicate that certain classes may have non‐overlapping definitions but be practically indistinguishable in some measurement contexts. The most commonly co‐occurring pair, ‘self‐regulation process’ (BCIO:050268) and ‘self‐regulation capability’ (BCIO:006005), reflects this tautology (and is also reflected in the coding rules). Specifically, if someone is using a self‐regulation process, they must also be capable of using this process. Note that the converse is not true, as someone may be capable of self‐regulation but not implement these skills. This implies that some MoAs may not be separable using self‐report measures, and certainly merits greater attention to both improve measurement practices and parse what this means for existing theories.
After the coding process, potentially missing classes from the MoA Ontology were identified from meeting notes. The greatest difficulty was in coding MoAs that were implicit. For example, no class accurately described social influences via non‐conscious processes such as emotion contagion—as seen in their labels, both ‘inter‐personal behaviour’ (BCIO:036025) and ‘social influence behaviour’ (BCIO:006099) required a behaviour, defined as ‘A bodily process of a human that involves co‐ordinated contraction of striated muscles controlled by the brain’. It was also challenging to code complex self‐regulatory processes, in which process A regulates process B (e.g., reframing one's thoughts to regulate smoking behaviour vs. changing one's behaviours as a method of influencing thoughts). Greater specificity in entities related to self‐regulation could support more precise coding. A key advantage of the MoA Ontology is that it can be updated based on evidence and feedback (Arp et al., 2015; He et al., 2018), with new classes being added where needed, and with historical data allowing prior coding to be included in updated version due to backwards compatibility. For instance, since the mapping work reported here, the ontology has been updated to include ‘affective attitudes’ to more fully capture attitudes proposed within theories (Schenk et al., 2024). With the 23 classes proposed as a result of the mapping work reported here, this study supports expanding the applicability of the ontology to wider use cases.
LIMITATIONS AND FUTURE WORK
Results are limited by the inclusion of only the 44 self‐report measures included in the SOBC Measures Repository, which is a non‐comprehensive set of selected measures, so findings may not generalize. Furthermore, these measures were developed without reference to the MoA Ontology, so lack of correspondence can be expected. That said, this study was intended to be illustrative (rather than comprehensive), many of these measures are widely used (showing the breadth of the issue), and the insights gained from this effort can be used to inform the creation of new measures and refinement of existing ones to improve rigour and transparency and to inform theories of behaviour change. Insight into the unique challenges of cumulating the existing literature (e.g., through meta‐analysis) due to measure heterogeneity and overlapping constructs was also gained. Due to this complexity, other methodological approaches have advocated for developing new measures for specific studies, so long as constructs are clearly labelled and defined (Peters & Crutzen, 2022). Underspecified or ambiguous measure items (e.g., ‘I am constantly without money’) also contributed to difficulty in linking measures (subscales) to classes from the MoA Ontology and should raise concern regarding the clarity and quality of these measures. Some measure items referenced more stable influences on behaviour (e.g., structural environment, physical health), which would be better captured by the Setting or Population Ontologies part of the Behaviour Change Intervention Ontology (Michie et al., 2017; Norris et al., 2020). Finally, updates to the MoA Ontology are ongoing, and we used an earlier version of the ontology that can be referenced on the Open Science Framework (OSF; https://osf.io/pkq4e). However, the use of unique IDs for each class in the ontology also means that any changes (e.g., their labels) to these classes can be tracked.
The judgements made during coding were based on two researchers' knowledge and subjective interpretations of the measurement items and the ontology classes. Both researchers have led work on measure classification and ontology development, including the measures and MoAs examined in the current study. Despite this and the application of a standardized framework, personal biases may have unintentionally impacted study results and generalizability. This study is an extension to prior work using similar methodologies (e.g., mapping measures to a smaller set of MoAs, but via contribution from a larger group of experts; Cornelius et al., 2023). To illustrate this programmatic approach, the current study includes a transparent comparison in the two mappings produced between MoAs and their potential measurements. The mapping generated in the current study should be viewed as a detailed starting point for linking measures to relevant MoAs. Quantitative studies are underway that map items to mechanisms with larger sample sizes, for example, using discriminant content validity. These heterogeneous methodologies allow for a nuanced and comprehensive exploration of the complex issues surrounding measurement of MoAs. This work should continue to be expanded across a broad range of measures, with potential to improve rigour, transparency, theory development and intervention efficacy and potentially even provide more formal guidance on measure selection in behavioural process research.
CONCLUSION
The present study builds on previous research by providing more precise guidance for behavioural scientists when selecting measures of MoAs for use in research studies. Results also highlight the critical importance of operationalizing behavioural MoAs to ensure rigorous and valid hypothesis testing. For example, the finding that different measures often tapped into multiple MoAs, as well as the common co‐occurrence of MoAs, suggests challenges in measuring certain MoAs in isolation. Future efforts should continue to tackle the development and refinement of measures that are valid for assessing behavioural MoAs. Insights were also generated regarding missing classes, which can be used as feedback to inform future iterations of the MoA Ontology.
AUTHOR CONTRIBUTIONS
Paulina M. Schenk: Conceptualization; data curation; formal analysis; investigation; methodology; project administration; validation; writing – original draft; writing – review and editing; visualization. Susan Michie: Supervision; writing – review and editing; funding acquisition. Marie Johnston: Funding acquisition; writing – review and editing; supervision. Talea Cornelius: Funding acquisition; conceptualization; data curation; formal analysis; investigation; methodology; project administration; software; supervision; validation; writing – original draft; writing – review and editing; visualization.
CONFLICT OF INTEREST STATEMENT
The authors declare that they have no conflicts of interest.
Supporting information
Data S1:
ACKNOWLEDGEMENTS
This work was supported by the National Institutes of Health Science of Behavior Change Program through an award administered by the National Institute on Aging (U24AG052175), by a Wellcome Trust collaborative award as a part of the Human Behaviour‐Change Project: Building the science of behaviour change for complex intervention development (201,524/Z/16/Z) and by a National Institutes of Health award ‘Accelerating Behavioral and Social Sciences Through Ontology Development and Application: NIH Research Network Project (APRICOT)’ [1U01CA291884‐01].
Schenk, P. M. , Michie, S. , Johnston, M. , & Cornelius, T. (2025). Measuring processes of change in behavioural interventions: Insights gained from linking mechanisms of action to associated measures. British Journal of Health Psychology, 30, e70015. 10.1111/bjhp.70015
Contributor Information
Paulina M. Schenk, Email: paulina.schenk.13@ucl.ac.uk.
Talea Cornelius, Email: tmc2184@cumc.columbia.edu.
DATA AVAILABILITY STATEMENT
There is no registration or pre‐registration plan available for the study or analysis. All data, code and materials are available at https://osf.io/mt6vw/.
Data set: https://osf.io/qt5ay/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
R code: https://osf.io/ebazx/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
Coding rules: https://osf.io/xdgy9/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
Measurement items and judgements on their mapping to the MoA Ontology for each researcher: https://osf.io/pd9fc/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
Suggested changes to the MoA Ontology: https://osf.io/6gzuk/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
REFERENCES
- Arp, R. , Smith, B. , & Spear, A. D. (2015). Building ontologies with basic formal ontology. Mit Press. [Google Scholar]
- Bohlen, L. C. , Carey, R. , de Bruin, M. , Rothman, A. , Johnston, M. , Kelly, M. P. , & Michie, S. (2019). Links between behaviour change techniques and mechanisms of action: An expert consensus study. Annals of Behavioral Medicine, 53(8), 708–720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cane, J. , O'Connor, D. , & Michie, S. (2012). Validation of the theoretical domains framework for use in behaviour change and implementation research. Implementation Science, 7, 1–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Carey, R. N. , Connell, L. E. , Johnston, M. , Rothman, A. J. , De Bruin, M. , Kelly, M. P. , & Michie, S. (2019). Behavior change techniques and their mechanisms of action: A synthesis of links described in published intervention literature. Annals of Behavioral Medicine, 53(8), 693–707. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Connell, L. E. , Carey, R. N. , De Bruin, M. , Rothman, A. J. , Johnston, M. , Kelly, M. P. , & Michie, S. (2019). Links between behavior change techniques and mechanisms of action: An expert consensus study. Annals of Behavioral Medicine, 53(8), 708–720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cornelius, T. , Derby, L. , Connell Bohlen, L. , Birk, J. L. , Rothman, A. J. , Johnston, M. , & Michie, S. (2023). Linking measures to mechanisms of action: An expert opinion study. British Journal of Health Psychology, 28(1), 98–115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Derby, L. , Bohlen, L. C. , Michie, S. , Johnston, M. , Birk, J. L. , Rothman, A. J. , & Cornelius, T. (2024). Linking measures to mechanisms of action in behavior change: A qualitative analysis of expert views. Social Science & Medicine, 352, 117023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Edmondson, D. , Falzon, L. , Sundquist, K. J. , Julian, J. , Meli, L. , Sumner, J. A. , & Kronish, I. M. (2018). A systematic review of the inclusion of mechanisms of action in NIH‐funded intervention trials to improve medication adherence. Behaviour Research and Therapy, 101, 12–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gross, J. J. , & John, O. P. (2003). Individual differences in two emotion regulation processes: Implications for affect, relationships, and well‐being. Journal of Personality and Social Psychology, 85(2), 348–362. [DOI] [PubMed] [Google Scholar]
- Gupta, V. (2023). What is repository? (Definition, Tutorial, How to Clone). Built In. https://builtin.com/software‐engineering‐perspectives/repository [Google Scholar]
- Hastings, J. (2017). Primer on ontologies. In Dessimoz C. & Škunca N. (Eds.), The Gene Ontology Handbook. Springer New York, 3–13. [Google Scholar]
- He, Y. , Xiang, Z. , Zheng, J. , Lin, Y. , Overton, J. A. , & Ong, E. (2018). The eXtensible ontology development (XOD) principles and tool implementation to support ontology interoperability. Journal of Biomedical Semantics, 9, 1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Johnston, M. , Carey, R. N. , Connell Bohlen, L. , Johnston, D. W. , Rothman, A. , de Bruin, M. , & Michie, S. (2018). Linking behavior change techniques and mechanisms of action: Triangulation of findings from literature synthesis and expert consensus. Annals of Behavioral Medicine, 53(8), 708–720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Johnston, M. , Carey, R. N. , Connell Bohlen, L. E. , Johnston, D. W. , Rothman, A. J. , De Bruin, M. , Kelly, M. P. , Groarke, H. , & Michie, S. (2021). Development of an online tool for linking behavior change techniques and mechanisms of action based on triangulation of findings from literature synthesis and expert consensus. Translational Behavioral Medicine, 11(5), 1049–1065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Michie, S. , Thomas, J. , Johnston, M. , Aonghusa, P. M. , Shawe‐Taylor, J. , Kelly, M. P. , Deleris, L. A. , Finnerty, A. N. , Marques, M. M. , & Norris, E. (2017). The human behaviour‐change project: Harnessing the power of artificial intelligence and machine learning for evidence synthesis and interpretation. Implementation Science, 12(1), 1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Michie, S. , West, R. , Finnerty, A. N. , Norris, E. , Wright, A. J. , Marques, M. M. , Johnston, M. , Kelly, M. P. , Thomas, J. , & Hastings, J. (2020). Representation of behaviour change interventions and their evaluation: Development of the upper level of the behaviour change intervention ontology. Wellcome Open Research, 5, 123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Michie, S. , West, R. , & Hastings, J. (2019). Creating ontological definitions for use in science. Qeios. [Google Scholar]
- Norris, E. , Marques, M. M. , Finnerty, A. N. , Wright, A. J. , West, R. , Hastings, J. , Williams, P. , Carey, R. N. , Kelly, M. P. , & Johnston, M. (2020). Development of an intervention setting ontology for behaviour change: Specifying where interventions take place. Wellcome Open Research, 5, 124. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Peters, G.‐J. , & Crutzen, R. (2022). Knowing what we're talking about: facilitating decentralized, unequivocal publication of and reference to psychological construct definitions and instructions.
- Riddle, M. (2015). Science of behavior change working group news from the NIH: Using an experimental medicine approach to facilitate translational research. Translational Behavioral Medicine, 5(4), 486–488. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Schenk, P. M. , Wright, A. J. , West, R. , Hastings, J. , Lorencatto, F. , Moore, C. , Hayes, E. , Schneider, V. , & Michie, S. (2024). An ontology of mechanisms of action in behaviour change interventions. Wellcome Open Research, 8, 337. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Seppälä, S. , Ruttenberg, A. , & Smith, B. (2017). Guidelines for writing definitions in ontologies. Ciência da Informação, 46(1), 73–88. [Google Scholar]
- Sumner, J. A. , Birk, J. L. , Cornelius, T. , Derby, L. , Edmondson, D. , & Davidson, K. W. (2019). The NIH science of behavior change mechanism‐focused approach to behavior change research. Psychosomatic Medicine, 81(4), A178. [Google Scholar]
- Sumner, J. A. , Carey, R. N. , Michie, S. , Johnston, M. , Edmondson, D. , & Davidson, K. W. (2018). Using rigorous methods to advance behaviour change science. Nature Human Behaviour, 2(11), 797–799. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Tong, A. , Sainsbury, P. , & Craig, J. (2007). Consolidated criteria for reporting qualitative research (COREQ): A 32‐item checklist for interviews and focus groups. International Journal for Quality in Health Care, 19(6), 349–357. [DOI] [PubMed] [Google Scholar]
- Watson, D. , Clark, L. A. , & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54(6), 1063–1070. [DOI] [PubMed] [Google Scholar]
- Wright, A. J. , Norris, E. , Finnerty, A. N. , Marques, M. M. , Johnston, M. , Kelly, M. P. , Hastings, J. , West, R. , & Michie, S. (2020). Ontologies relevant to behaviour change interventions: A method for their development. Wellcome Open Research, 5, 126. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Zhang, L. , Schenk, P. M. , Santilli, M. , Wright, A. J. , Marques, M. M. , Johnston, M. , & Michie, S. (2025). Linking behaviour change techniques to mechanisms of action: Using the theory and techniques tool alongside the behaviour change intervention ontology: [version 1; peer review: 1 approved, 1 not approved]. Wellcome Open Research, 10, 192. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1:
Data Availability Statement
There is no registration or pre‐registration plan available for the study or analysis. All data, code and materials are available at https://osf.io/mt6vw/.
Data set: https://osf.io/qt5ay/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
R code: https://osf.io/ebazx/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
Coding rules: https://osf.io/xdgy9/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
Measurement items and judgements on their mapping to the MoA Ontology for each researcher: https://osf.io/pd9fc/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).
Suggested changes to the MoA Ontology: https://osf.io/6gzuk/?view_only=9e48ae0353e04de99df0859e16e6f214 (link view only for masked review).