Dynamics of Cognitions and Symptoms during Treatment: A Prospective Test of Beck's Cognitive Theory of Depression

Beck's cognitive theory posits that maladaptive beliefs causally drive depressive symptoms. Yet empirical support for the cognition-to-symptoms pathway is mixed, often due to sparse sampling, conflation of within- and between-person variance, and limited temporal modeling. We, therefore, aimed to address these limitations by testing reciprocal cognition-symptom dynamics in routine care using a more robust within-person longitudinal approach drawing on session-to-session assessments.

Adults in Norwegian primary care receiving CBT-oriented treatment (baseline N = 1,564; PHQ-9 ≥10) completed eight assessments across ≈16 weeks (PHQ-9; three-item negative self-beliefs measure). Primary analyses were observation-level dynamic-panel models (DPMs); residual-level models served as corroboration. We modeled non-linearity and time trends, tested equality-over-time and varying intervals, and reported model-implied long-run effects.

Symptoms and negative self-beliefs declined non-linearly. Time-lagged models supported reciprocity: lower negative self-beliefs predicted later symptom reductions, and lower symptoms predicted later belief change. In DPMs, a one-point improvement in negative self-beliefs forecast a long-run 0.46-point PHQ-9 reduction (95% CI 0.22-0.70); a one-point PHQ-9 improvement forecast a 0.16-point belief change (95% CI 0.09-0.22). Standardized long-run effects were 0.36 (cognition→symptoms; 95% CI 0.13-0.42) and 0.21 (symptoms→cognition; 95% CI 0.12-0.26); magnitudes did not differ (χ2(1) = 1.77, p = 0.18). Findings were robust across specifications.

This study provides credible longitudinal, within-person evidence in line with Beck's model: improvements in negative self-beliefs precede and accumulate into subsequent symptom relief, while symptoms also feed back on beliefs. As such, it fills an important empirical gap by demonstrating a robust cognition-to-symptoms pathway under rigorous temporal modeling.