Associations in behaviour change: why knowing is not doing

When someone says 'I support this', you are measuring an evaluation. 'I intend to do it' is a report about a future intention. Only when the person signs up, arrives, completes the procedure or repeats the activity do you observe a behavioural outcome. These outcomes are related but distinct. An employee may sincerely want to mentor and then discover that their next four shifts leave no room for it.

Ajzen's theory of planned behaviour links intentions to attitudes, perceived social norms and perceived behavioural control. It is useful for organising possible determinants, not for assuming that an intention fully determines action. Webb and Sheeran reviewed experimental evidence and found that interventions which changed intentions produced, on average, a smaller change in behaviour. Their reported effects describe the studies included in that review; they do not supply a conversion forecast for a particular Neurofactor project. [Ajzen, 1991; Webb & Sheeran, 2006]

A project should therefore begin with an observable outcome. 'Increase ownership' is too broad to evaluate directly. 'Offer one scheduled mentoring conversation to a new team member each month' specifies an act. Define who should do it, at what moment, under which circumstances, and how frequently. Otherwise a survey measuring positive attitudes can misleadingly become the main evidence that an intervention was successful.

The COM-B framework provides a practical way of checking what a behaviour requires. Its three interacting conditions are capability, opportunity and motivation. Capability includes relevant knowledge and skills. Opportunity encompasses physical and social conditions that enable or prompt an action. Motivation includes deliberate goals and assessments, as well as automatic responses and habits. The conditions interact: a difficult environment can change motivation, and repeatedly performing an action may build capability. [Michie, van Stralen & West, 2011]

For mentoring, the distinction matters. One experienced worker values the role but lacks the skills to guide a difficult conversation. A second could mentor effectively but receives no time allocation. A third has time and expertise yet associates the title with personal blame and reputational risk. A fourth may be ready to start but cannot access the sign-up form. These are not four marketing segments invented from a score; they are plausible explanations to investigate.

COM-B is a diagnostic framework, not an automatic root-cause detector. Researchers still need to examine real working conditions, consider alternatives and test whether reports match what happens on the job. Associations are especially relevant when exploring certain motivational processes, but the meaning attached to an environment can also influence whether it feels safe or workable. The framework helps prevent automatic processes from crowding out actual material constraints.

An association connects a cue, situation or concept with a meaning, expectation, memory or evaluation. The phrase 'become a mentor' might evoke expertise, recognition and contributing to the next generation. It might also evoke additional paperwork, exposure to mistakes or anxiety about difficult feedback. Different people can connect the same role to different experiences. Those links are potential subjects for investigation, not evidence that one group has a defective attitude.

Research on associative and propositional processes helps explain why an expressed belief may not tell the entire story of evaluation. People can endorse a goal while responding differently to cues in the moment. Strack and Deutsch's reflective-impulsive model describes interactions between more deliberative and more automatically activated processes. Such models should not be simplified into a literal two-brain contest or a claim that reflective thinking always loses. [Gawronski & Bodenhausen, 2006; Strack & Deutsch, 2004]

A properly designed implicit task may contribute information about relative, task-dependent associations. Its result is not a detector of a person's 'true desire', a diagnostic label or a guaranteed prediction of a subsequent choice. Procedure, sample, stimuli, context and psychometric properties all matter. Association research becomes useful when it helps generate falsifiable explanations alongside observed behaviour and the opportunities people actually have. [De Houwer et al., 2009]

Sometimes the barrier is attached to the very reason the behaviour matters. An employee may want to become a mentor because helping young people is meaningful. Precisely because quality matters, they may be worried about giving insufficient guidance. A resident may support more sustainable travel while being concerned about the impact of an unreliable connection on family or work. The positive motive does not mechanically cause the negative reaction; both may relate to something the person values.

That creates a different communication challenge. A campaign promising 'make a real impact' might increase interest and, for some people, increase the perceived stakes of getting it wrong. A study could test whether practical information about support, time commitments and shared responsibility changes this pattern. A message about pride may be useful, but perhaps it should be accompanied by a credible way to start safely and learn the role.

The distinction also prevents over-interpretation. Some people will have no anxiety at all; they simply have an impossible schedule. Others may disagree with the initiative for substantive and legitimate reasons. An association map should clarify differences between relevant groups and situations, not become a psychological story imposed on every refusal after the event.

Not every action involves a fresh deliberation. Through repetition, particular settings and moments can become cues for a familiar response. Someone who opens an old reporting spreadsheet after the weekly meeting may not be deciding, once again, that the new process is pointless. The familiar time, computer and routine are enough to trigger the established action. Wood and Neal describe how context-response associations influence behaviour in ways that interact with, but are not identical to, current goals. [Wood & Neal, 2007]

This is why another instruction email can fail to improve process compliance. The old shortcut remains prominent; the new form requires three additional screens. A better intervention may reposition the correct action within the existing workflow, remove obsolete links, simplify the first step or provide support at the exact moment a new routine should begin.

Habit formation also has no universal 21-day deadline. Lally and colleagues studied everyday repeated actions and observed substantial variation in the development of self-reported automaticity. Their work should not be turned into a fixed timetable for a different population or a complex professional procedure. Consistent repetition in an achievable context matters, and continued performance needs to be evaluated beyond the first successful attempt. [Lally et al., 2010]

Behaviour takes place among other people. Somebody may value mentoring but worry that colleagues will mock the initiative, that a manager will exploit their willingness or that mistakes will become publicly visible. What people believe their colleagues do, approve of and reward can matter as much as an official policy. A procedure may be safe on paper while participation still feels risky inside a team.

Research can therefore explore the meaning attached to terms such as 'ambassador', 'mentor', 'feedback' and 'change'. Does 'volunteer' signal autonomy or unpaid extra work? Does 'challenging' mean a chance to grow or a warning about failure? Are the answers different among team leaders, new hires and long-standing staff? The point is not to assign a stereotype to each role but to test whether these meanings help explain observed participation patterns.

Psychological safety should not be inferred from a single EEG measure, an implicit score or a few memorable quotes. Researchers should ask about specific episodes, observe decisions and pay attention to employees with less organisational power. A social-pressure campaign aimed at forcing visible compliance can aggravate the very obstacle it was designed to remove. Respectful behavioural design makes participation workable rather than turning reluctance into a personal failing.

Fictional example - not a Neurofactor client study or a measured result. A technical company launches a programme for experienced staff to mentor trainees. The initial message highlights the benefits for younger colleagues and the satisfaction of passing on expertise. Most participants in a short internal poll endorse the idea. Yet very few apply. The team provisionally describes the problem as a lack of motivation.

A behavioural diagnosis begins with the route to joining: filling out a form, gaining a manager's approval, finding capacity in the rota and holding the first conversation. Next, the organisation investigates what employees associate with 'placement mentor'. Pride and craftsmanship are plausible positive associations; unclear accountability, fear of mistakes and extra hours are plausible negatives. Researchers also examine scheduling data and the actual process, so that inability is not mistakenly labelled resistance.

Several interventions can then be compared. Perhaps prospective mentors benefit from a buddy system, formally protected time, clear limits of responsibility or a supported trial period. We cannot know which one will work from the fictional scenario. A pilot using a relevant comparison and actual application and retention outcomes would test whether the intervention improves behaviour. If stated willingness increases but enrolment does not, that distinction is a finding worth reporting rather than concealing.

Start with who needs to do what, where, when and how often. List competing explanations. For each one, specify evidence that would support it and evidence that could make it less credible. This matters because 'unconscious resistance' is too easy to claim after any disappointing result unless the concept is measured and competing causes are considered.

Potential barrier	What might be reported or observed	How to investigate	What not to conclude automatically
Skill deficit	'I do not know how to start'	Task observation, skills check, interviews	That the person does not care
Lack of opportunity	Willing person without time or access	Process mapping, rota and enrolment data	Hidden psychological resistance
Unfavourable association	Role linked with blame, embarrassment or lost control	Association tasks and contextual interviews	That every employee shares it
Established habit	Old procedure returns when a familiar cue appears	Repeated observation at the point of action	That another reminder is sufficient
Social expectation	Behaviour differs in front of colleagues	Team interviews and contextual comparisons	A fixed personality trait

Self-report alone is rarely a complete picture. People may remember a process differently from how it occurs. Observation, however, is not automatically causal either: two departments may differ in staffing, managers or workload. Combine complementary sources, document uncertainties and predefine the comparisons that will matter. [Michie et al., 2011; Ajzen, 1991]

Interviews can investigate perceived barriers and personal meaning. Behavioural observation shows what occurs at a critical point in a task. An appropriately validated implicit task may offer additional information about relative associations. EEG and eye tracking, where justified, can investigate electrical brain activity and visual attention under a specific protocol. None measures 'motivation' or 'resistance' directly and completely.

Method	Potentially useful for	Does not establish on its own
Interviews and open questions	Expectations, explanations and perceived risk	The frequency of actual behaviour
Association or implicit tasks	Relative evaluations under specified task conditions	Hidden true intent or an individual's future action
Observation and process data	Execution, errors, delays and drop-off	The cause of the observed pattern
Eye tracking / EEG	Visual attention / electrical brain activity in a defined task	Emotion, conviction or behaviour change directly
Comparative or randomised field trial	Changes in predefined behavioural outcomes	Universal effects outside the tested context

Choose methods because of what you need to learn, not their technological appeal. A broken form should be fixed, not sent for EEG analysis. If a role continues to trigger unexpected hesitation despite sufficient opportunity, carefully designed association research may provide useful hypotheses. [De Houwer et al., 2009; Michie et al., 2013]

Without diagnosis, organisations tend to choose the intervention they are most comfortable producing: a poster, webinar or reminder. But a skills problem may need guided practice. An opportunity problem may need protected time or simpler access. A social concern may need a real shift in how support and responsibility work. A habit problem may need changes to contextual cues and assistance during early repetition.

Michie and colleagues' Behaviour Change Wheel links a behavioural diagnosis to classes of intervention. Their subsequent Behaviour Change Technique Taxonomy provides a vocabulary for specifying active components of an intervention so that it can be reported, compared and replicated. Instruction, practice, feedback, environmental restructuring and action planning are not equivalent tools. 'We will nudge people' is not a sufficient specification of what will change. [Michie et al., 2011; Michie et al., 2013]

Choice architecture research suggests that the way options are arranged can influence decisions in some contexts. Effects vary by technique and domain and findings are affected by study design and publication bias. No default, reminder or framing device is a guaranteed behavioural solution. The ethical standard matters too: preserve autonomy and avoid concealing important trade-offs, particularly when people may be vulnerable or decisions have substantial consequences. [Mertens et al., 2022]

An intention often lacks a time, location or first step. Implementation intentions address this by specifying an 'if-then' plan in advance. For example: 'If I review the next month's rota with my manager, then I will ask which hours are protected for mentoring.' Or: 'If the new form appears in my work queue, then I will follow the agreed checklist before proceeding.' The technique links a recognisable opportunity with an intended response.

Gollwitzer and Sheeran's meta-analysis reports that implementation intentions can support goal attainment. It does not mean that a plan can compensate for a missing budget, an unsafe workplace or an impossible workload. This approach is most relevant when a viable goal exists and the main difficulty involves noticing the opportunity or initiating the action. [Gollwitzer & Sheeran, 2006]

Specify how the plan will be evaluated. Was the cue encountered? Did the planned response follow? How often did the opportunity arise? More complex behaviours may still require practice and feedback, even after the first successful action. The right question is whether the technique improves the intended behaviour under the conditions of this setting.

An intervention is not validated simply because participants like the concept. Relevant outcomes may include applications submitted, correct procedure use, repeated participation, safety performance or retention of an improved practice. Define a primary behavioural outcome and observation period before the pilot. Also look for unintended effects: stress, poorer quality, exclusion or damage to trust.

Where feasible, compare a pilot group with a suitable control or introduce the change in phases. Record baseline performance, exposure to the intervention and changes occurring simultaneously. If one department also receives new leadership and a revised rota, the evaluation must consider those factors. Behaviour improving after a campaign is not, on its own, proof that the campaign caused the improvement.

Small samples require particular care. Report how many people had an opportunity to act, how many did so, and whether the behaviour continued. Include negative or null findings. They may show that the original problem was misdiagnosed or that the intervention did not reach the point of action. A meaningful evaluation is designed to learn, not merely to produce a success story. [Webb & Sheeran, 2006; Michie et al., 2013]

Knowing is not doing. Between a favourable evaluation and an observable action may lie insufficient capability, lack of opportunity, conflicting associations, social expectations and established habits. These are not interchangeable barriers, and they do not all require more persuasion. A sound approach starts with the target behaviour, investigates its context, tests plausible explanations, designs a proportionate intervention and measures behavioural results.

For Neurofactor, the most productive question is not 'how do we persuade people harder?' It is 'what makes the action attractive, possible, safe and practicable here - and what prevents it?'. That question supports better research design, better decisions and more credible claims.

The series will develop these themes in separate articles: why people fail to act on sensible intentions (NF-AS-BLOG-G-001), when to choose communication, process or environmental change (NF-AS-BLOG-G-013) and how to evaluate an intervention in a real-world setting (NF-AS-BLOG-G-069). Link to them only when they are published. Existing knowledge-base resources cover behavioural diagnosis, measuring associations, conflicting associations and types of resistance.