The purpose of the present study was to examine the long-interval test–retest reliability of the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders (4th edn; DSM-IV) Axis II Personality Disorders (SCID-II) Japanese version. One hundred and twenty outpatients with anxiety disorders completed the self-report SCID-II personality questionnaire (SCID-II-PQ) and structured interviews, and then again 12 months later. In the SCID-II-PQ, 70.8% and 71.7% of the patients had a personality disorder (PD) at the first evaluation and second evaluation 12 months later, respectively, and Cohen's kappas ranged from 0.29 for paranoid PD to 0.83 for histrionic PD, and overall kappa was 0.56. In the SCID-II interviews, 47.5% and 41.7% of the patients fulfilled the criteria for PD at the first and the second evaluations, respectively. At least one PD was identified in 49 subjects (40.8%), of whom 65.3% had one PD, 30.6% had two PD, 2% had three PD, and 2% had four PD; the most frequently diagnosed PD were from cluster C (60.9%). The overall base rate of 12 PD was 7%, and overall kappa was 0.87. Cohen's kappas ranged from 0.86 for obsessive–compulsive PD to 0.93 for avoidant PD and schizoid PD, and were comparable with those in the previous interrater studies. The test–retest reliability of the SCID-II-PQ was moderately good, and after the SCID-II interview the test–retest reliability of the SCID-II appeared to be of almost perfect reliability. This first long-interval, large-sample, non-Western-language research on the test–retest reliability of the SCID-II for DSM-IV indicated its usefulness and excellent reliability.