When mindfulness is not the mechanism: what actually drives change in OCD
- Across three distinct treatments for obsessive-compulsive disorder – CBT with exposure and response prevention, inference-based CBT, and mindfulness-based stress reduction – only reductions in inferential confusion uniquely predicted symptom improvement when all candidate mechanisms were modelled together.
- Dispositional mindfulness did not change significantly over the course of treatment, including in the mindfulness-based stress reduction arm, and did not account for unique variance in outcome once inferential confusion was entered.
- The mediating role of inferential confusion was modality-independent: it held in the cognitive, exposure-based, and mindfulness conditions alike, suggesting a shared cognitive pathway rather than a technique-specific effect.
- The association survived control for concurrent anxiety, depression, and psychotropic medication; adjusting for medication strengthened rather than weakened the contribution of inferential confusion (partial eta-squared near 0.29).
Mechanism-of-change research asks a question that outcome trials cannot: not whether a therapy works, but through what process. This secondary analysis of a Canadian multicentre randomized trial (N = 111 adults with primary OCD, McGill University and Université de Montréal) took three treatments built on different theories of pathology and tested their proposed active ingredients side by side. The design is unusually clean because it pits a cognitive mechanism (inferential confusion, the tendency to treat an imagined possibility as if it were real), a belief-based mechanism (inflated responsibility, overestimation of threat, perfectionism), and a process-based mechanism (mindfulness) against one another in the same statistical model.
The headline result is a dissociation. At the bivariate level, change in all three constructs tracked symptom reduction – the familiar picture in which everything correlates with everything. But once the mechanisms competed in a mixed-effects model, only inferential confusion retained an independent relationship with improvement. Obsessive beliefs and mindfulness fell away as unique predictors. Most striking for contemplative practice, dispositional mindfulness did not even shift reliably over treatment, including among patients who received eight weeks of mindfulness-based stress reduction.
This matters because it separates a plausible mechanism from a load-bearing one. Mindfulness can produce symptom relief without operating through measured trait mindfulness; the benefit of an MBSR course for OCD may travel along a more general cognitive route – loosening the reasoning that fuses imagination with reality – rather than through cultivated present-moment awareness as such. The finding aligns with a broader unease in the field that our self-report mindfulness scales may not capture the construct that therapy actually moves.
A transdiagnostic reasoning target
Inferential confusion is not a symptom; it is a style of reasoning. If a single cognitive process underlies improvement whether the patient is doing exposure, cognitive restructuring, or meditation, then that process becomes a candidate transdiagnostic target. Clinicians can listen for the moment a patient slides from "this could be contaminated" to behaving as though it certainly is, and can make that inferential leap, rather than the feared content, the focus of work.
The limits of a single contrast
The cleanliness of the dissociation also carries a caution. A model in which one mechanism wins can reflect measurement quality as much as causal priority: inferential confusion may simply be measured more sensitively than mindfulness in an OCD sample. The result invites better mindfulness instrumentation before concluding that present-moment awareness is therapeutically inert here.
Mindfulness-based treatment helped patients with OCD, yet measured mindfulness did not change and did not explain the gains; the work was done by a shift in how imagined possibilities were treated as real.
Secondary analysis of a completed trial, modest sample (N = 111), and reliance on self-report mindfulness measures that may under-detect change. Mechanisms were assessed at the symptom-level over time rather than via fine-grained session sampling, so temporal precedence between mechanism change and symptom change cannot be firmly established. Findings are specific to OCD and may not generalize to disorders where mindfulness has stronger measured engagement.