This study investigated whether partner-related attachment characteristics differentially predict premature treatment termination as well as posttreatment and 1-year follow-up outcome in patients with social anxiety disorder treated with a manualized cognitive–behavioral therapy (CBT) or short-term psychodynamic therapy (PDT) in the SOPHO-NET (Social Phobia Psychotherapy Network) trial. Participants were 412 patients with social anxiety disorder (57% female) with a mean age of 35.4 years (SD = 12.1) who were randomized to either CBT or PDT. Partner-related attachment characteristics were measured using the revised Experiences in Close Relationships Questionnaire (ECR-R) at pretreatment. The Liebowitz Social Anxiety Scale was administered at pretreatment, posttreatment, and a 1-year follow-up. To address our research questions, linear regression models were applied. Furthermore, we compared CBT versus PDT patients within ECR-R quartiles. Treatment dropout did not differ between CBT and PDT and was not predicted by pretreatment attachment. In both treatment conditions, there was a trend for higher attachment anxiety to be associated with a more limited reduction in symptoms if controlling for pretreatment Liebowitz Social Anxiety Scale scores. Exploratory analyses showed that patients assigned to the highest quartile of the ECR-R-Avoidance distribution showed more benefit within the CBT condition posttreatment and at follow-up than the PDT condition. Our findings suggest that it may be useful to assess attachment characteristics in patients with social anxiety disorder before psychotherapeutic treatment. Patients characterized by very high pretreatment attachment avoidance (ECR-R-Avoidance >3.87) may specifically benefit more from CBT than from PDT. However, replication studies are needed that also should investigate nonlinear effects of pretreatment attachment.