We studied the evolution of therapeutic alliance and its relationship with attachment for 20 weeks, considering symptom improvement/worsening as a confounding variable.A sample of 30 adults completed the Working Alliance Inventory-Short Form eight times, the Experiences in Close Relationships Scale once, and the Clinical Outcomes in Routine Evaluation Outcome Measure at baseline and in week 20. We found a slightly growing linear evolution of therapeutic alliance. When associated with attachment, avoidance dimension did not reveal correlation. When associated with anxiety, it did show a stronger negative relationship at advanced phases of the therapy. No adjustment was necessary. Practical implications of these findings are discussed.