Background <p>Homework is an important component of cognitive therapy (CT), and has been associated with improved treatment outcome. However, patient-level differences, such as depression symptom severity, might account for the observed relationship between compliance and outcome—rather than a direct effect of compliance on outcome, per se. Therefore, the present study aimed to evaluate the relationship between homework compliance and next-session depression symptom severity, while accounting for patient-level differences.</p> Methods <p>Within the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial, 113 patients received CT after failing to remit from an initial course of citalopram treatment, with 73 receiving citalopram plus CT and 40 receiving CT only. We specified a multilevel growth model to analyze the relationship between homework compliance and next-session depression severity change. We also conducted dynamic panel structural equation modeling as a sensitivity analysis.</p> Results <p>Higher homework compliance significantly predicted lower next-session Quick Inventory of Depressive Symptomatology-Self-Rated (QIDS-SR) score. Treatment group was not significant.</p> Conclusions <p>The present study found that higher homework compliance in the preceding session was directly linked to depression severity improvement at the following session. These findings emphasize the importance for therapists to encourage their patients to complete their assigned homework.</p>