We investigated the reliability of the Structured Interview for Schizotypy-Revised (SIS-R). The original interview (SIS) was developed by Kendler. We revised the SIS, primarily by standardizing the rating procedures. Operational definitions and explicit criteria for rating were given. We introduced a four-point scale and provided clear criteria for rating severity of symptoms and signs (frequency, duration, and level of conviction) to operationalize schizotypal features. We divided schizotypal signs of global affect and global organization of speech into three separate signs of affect and five separate signs of thinking and speech. The main goal of this study was the assessment of test-retest reliability of the SIS-R. A robust test-retest design using different interviewers at both times, with a mean interval of 19 days, was used. The sample consisted of 42 psychiatric patients, almost all with personality disorders. The strong linear-weighted kappa statistic was used to evaluate reliability. The first conclusion is that most schizotypal symptoms can be reliably assessed with the SIS-R. The second conclusion is that most schizotypal signs do not reach sufficient levels of reliability. After unreliable items are excluded, the shortened SIS-R is a reliable research instrument for measuring schizotypal features (as far as it concerns our mixed samples). It covers all three dimensions of schizotypy.