A useful explanation begins with the right question
The question “does treatment help?” differs from “what produced the improvement?” Randomizing a treatment does not automatically randomize the proposed mechanism. A person may report less threatening beliefs and fewer symptoms during the same weeks because either change influences the other or because both respond to a third process. Kazdin (2007) sets out the need for stronger designs and converging evidence when moving from a statistical mediator to a causal mechanism. For readers, the distinction can be expressed simply: a process associated with improvement is a candidate explanation, not yet a demonstrated reason that improvement occurred. This does not diminish clinical outcomes; it makes the explanation more precise.
Sources: Kazdin et al., 2007
Skills and the context in which they are used
Southward et al. (2024) organize putative mechanisms into treatment skills, processes shared across therapies, and processes that maintain psychological difficulties. Cognitive flexibility, behaviour change and acceptance may work alongside collaboration, expectations and confidence in using skills. Their interaction matters: a convincing rationale can make practice more likely, and successful practice can build confidence. The review supports studying how often and how well skills are used rather than assuming that simply naming a distortion produces change. These are hypotheses with varying empirical support, not a single established pathway shared by everyone. They also help explain why a well-designed exercise and a therapeutic relationship are not interchangeable components.
Sources: Southward et al., 2024
Learning provides a more specific account of exposure
The inhibitory-learning account of exposure proposes that new experiences can compete with prior threat predictions rather than erase a fear memory (Craske et al., 2014). It offers an explanation for why learning may depend on context and why fear can return. This is a clinical translation of experimental learning research, not a promise that exposure permanently removes anxiety. This explanation does not provide instructions for unsupervised high-intensity exposure. The useful scientific question is which learning targets improve meaningful functioning and durability, for which condition and under what support. A mechanistic account earns its value by guiding testable comparisons.
Sources: Craske et al., 2014
Digital mediation studies are progressing
Angerer et al. (2025) and Domhardt et al. (2025) analyze different domains within the same underlying set of 25 studies and 8110 participants. They should not be counted as independent replications. The former considers domains such as self-related and biophysiological variables; the latter focuses on cognitive and behavioural mediators. Their findings support candidate change processes, while methodological weaknesses, heterogeneity and timing constrain interpretation. Not every specific cognitive candidate showed a significant indirect effect. This is a useful corrective to the assumption that all benefit must pass through measured dysfunctional attitudes. It also shows why engagement with an exercise and mastery of its intended skill need separate measurement.
Sources: Angerer et al., 2025; Domhardt et al., 2025
Neuroscience adds observations rather than a shortcut to proof
König et al. (2025) reviewed 14 longitudinal task-fMRI studies in depression and described changes across emotion, reward and cortical processing, with inconsistent findings in some areas. The studies used different tasks and methods, limiting a single neural narrative. A subsequent naturalistic study of 59 patients and 60 healthy controls linked some hippocampal activity changes to CBT response (Borgers et al., 2026). Its design did not randomize patients to CBT versus another clinical treatment. Such a finding cannot by itself distinguish treatment-specific effects from other changes, validate an individual treatment-selection biomarker, or establish that symptom change was caused by that brain change. Both papers concern clinical treatment and imaging, not consumer-app use.
Sources: König et al., 2025; Borgers et al., 2026
What the mechanism evidence supports
The reviewed studies examine learning, appraisal, behaviour, and emotional processing during CBT. Terms such as “associated with” and “candidate mechanism” distinguish an observed relationship from an established causal pathway. Claims that CBT “rewires the brain,” that one brain region is the disorder, or that imaging proves a commercial app effective go beyond these study designs. This interpretation follows from the methods and limitations of the research; it is not a newly established biological finding.
Sources: Southward et al., 2024; Angerer et al., 2025; Domhardt et al., 2025; König et al., 2025; Borgers et al., 2026; Kazdin et al., 2007
What would make the next studies more informative
Useful mechanism studies would measure candidate processes before later outcomes, sample change repeatedly, distinguish within-person from between-person associations, and test specific intervention components experimentally. Digital products could support this work, but ordinary usage logs are not causal evidence. For example, users who complete more exercises may differ in motivation or initial severity before they complete them. A proposed company study should preregister the intended mediator and outcome, address missingness, and specify whether its question concerns learning, engagement or clinical benefit. Independent replication and testing in different populations would be needed before using such findings to personalize care or make biological marketing claims.
Sources: Angerer et al., 2025; Domhardt et al., 2025; Kazdin et al., 2007
References
Links open the original papers or official sources. Access labels describe the material assessed for this review.
Kazdin, A. E. (2007). Mediators and mechanisms of change in psychotherapy research. Annual Review of Clinical Psychology, 3, 1–27. 10.1146/annurev.clinpsy.3.022806.091432
Access assessed: PubMed abstract and bibliographic record
Southward, M. W., Kushner, M. L., Terrill, D. R., & Sauer-Zavala, S. (2024). A review of transdiagnostic mechanisms in cognitive behavior therapy. Psychiatric Clinics of North America, 47(2), 343–354. 10.1016/j.psc.2024.02.003
Access assessed: Published paper PDF on author institution website; selected conceptual and evidence sections inspected
Craske, M. G., Treanor, M., Conway, C. C., Zbozinek, T., & Vervliet, B. (2014). Maximizing exposure therapy: An inhibitory learning approach. Behaviour Research and Therapy, 58, 10–23. 10.1016/j.brat.2014.04.006
Access assessed: PubMed abstract and bibliographic record; primary published-paper excerpts
Angerer, F., Mennel, V., Grund, S., Mayer, A., Büscher, R., Sander, L. B., Cuijpers, P., Terhorst, Y., Baumeister, H., & Domhardt, M. (2025). Mechanisms of change in digital interventions for depression: A systematic review and meta-analysis of six mediator domains. Journal of Affective Disorders, 368, 615–632. 10.1016/j.jad.2024.09.055
Access assessed: Publisher structured abstract and author institutional bibliographic record; full-text audit unavailable
Domhardt, M., Mennel, V., Angerer, F., Grund, S., Mayer, A., Büscher, R., Sander, L. B., Cuijpers, P., Terhorst, Y., & Baumeister, H. (2025). Processes of change in digital interventions for depression: A meta-analytic review of cognitive and behavioral mediators. Behaviour Research and Therapy, 189, 104735. 10.1016/j.brat.2025.104735
Access assessed: Author institutional abstract and verified bibliographic record
König, P., Zwiky, E., Küttner, A., Uhlig, M., & Redlich, R. (2025). Brain functional effects of cognitive behavioral therapy for depression: A systematic review of task-based fMRI studies. Journal of Affective Disorders, 368, 872–887. 10.1016/j.jad.2024.09.084
Access assessed: PubMed structured abstract; publisher full text returned 403
Borgers, T., Zwiky, E., Klug, M., Enneking, V., Fisch, L., Klein, L., Neutz, L., Leehr, E. J., Opel, N., König, P., Schöniger, K., Küttner, A., Selle, J., Dannlowski, U., & Redlich, R. (2026). Changes in brain function during negative emotion processing following cognitive–behavioural therapy in depressive disorders. The British Journal of Psychiatry, 228(4), 324–331. 10.1192/bjp.2025.71
Access assessed: Publisher full text and PDF result excerpts; selected methods/results inspected
This evidence review is educational. It does not provide an individual diagnosis or treatment plan. Read the review methods, access limits, and clinical-review status.