What the Evidence Says—and What It Does Not

Psychological flexibility at work has a stronger evidence base than a slogan but a weaker one than a settled prescription. This chapter focuses on how to read that evidence: what a large association between psychological inflexibility and wellbeing means, why workplace studies do not all answer the same question, and how study design and outcome selection limit generalization. The result is a practical standard for judging claims without either dismissing the process or promising that a change in mindset will repair a workplace.

The distinction matters because research headlines often compress several steps. “Flexibility is linked to wellbeing” can become “flexibility improves wellbeing,” which can become “employees who are struggling should become more flexible.” Each step adds a causal or practical conclusion that the underlying evidence may not support. A careful reading keeps data, interpretation, and recommendation separate.

A relationship is not a cause

The broadest quantitative evidence comes from a meta-analysis by Ong, Barthel, and Hofmann. It brought together 151 studies, 70,738 participants, at least 35 countries, 22 languages, 25 versions of the Acceptance and Action Questionnaire, and 43 wellbeing measures. Across 262 correlation analyses, the relationship between AAQ-measured psychological inflexibility and wellbeing was r = −.467, with a 95 percent confidence interval from −.488 to −.446.

A correlation summarizes how two measured variables vary together. Here, people reporting more psychological inflexibility also tended, on average, to report worse wellbeing. The negative sign describes the direction; the magnitude indicates a moderate relationship in the aggregated data. It does not say that inflexibility is a disease, that wellbeing falls by a fixed amount for every employee, or that changing an internal response will produce a predictable improvement.

The size of the synthesis should increase confidence that the relationship is worth taking seriously, but it should not erase the variation within it. Heterogeneity was very high: the reported I² was 95.11 percent. In practical terms, the studies differed substantially in participants, settings, measures, languages, wellbeing definitions, and methods. An average can describe a recurring pattern across that mixture while still being a poor forecast for any one worker or organization.

The authors estimate that the association corresponds to roughly 22 percent of the variance in wellbeing. That is not a personal risk percentage and cannot be converted into a treatment effect. Its most important implication may be the remainder: most variation in wellbeing was not accounted for by psychological inflexibility. Social support, environmental conditions, health, economic circumstances, and other psychological processes remain relevant. A model that explains part of wellbeing is useful; a model that claims to explain all of it is misleading.

The design also sets a firm limit. The meta-analysis used cross-sectional data, meaning the variables were measured at roughly the same point rather than tracked through a design that can establish temporal order. The finding cannot determine whether inflexibility contributes to poorer wellbeing, whether poor wellbeing makes flexible responding harder, or whether both are influenced by another factor. Longitudinal work may be consistent with a causal theory, but this meta-analysis itself establishes association, not direction or causation.

The moderators make the same point in a different way.