Purpose. The System for Observing Fitness Instruction Time (SOFIT) is a direct observation tool that was developed to assess class context and instructor behavior variables associated with physical activity levels during physical education classes. The purpose of this study was to modify the tool for use in adult group-exercise classes (SOFIT-X) and provide evidence of score validity and reliability. Methods. Three sequential studies were conducted to modify, validate, and assess the reliability of the SOFIT-X. Study 1 employed a nominal group technique using subject-matter experts to modify and validate the content of the SOFIT-X. Study 2 used an iterative cycle of pilot testing/measure revision using behavioral assessment experts to establish criterion validity across seven types of group-exercise classes. In Study 3, two trained observers collected SOFIT-X data on 15 live group-exercise classes to assess ecological validity and inter-rater reliability. Results. Seven subject-matter experts established the content validity of four coding categories, 22 sub codes and definitions, and 23 instructor behavior criteria to create a pilot version of the SOFIT-X. Three trained observers then established criterion-validated codes for two video segments of groupexercise classes (109 min; 31 min). Interobserver agreement among any two out of three observers was 97%, 100%, 99%, and 100% on posture, intensity, class context, and instructor behavior. Lastly, data collected on live group-exercise classes with the SOFIT-X indicated that posture, intensity level, class context, and instructor behavior could all be feasibly coded in real-time. Reliability data from nine of these classes revealed interobserver agreement of 91%, 76%, 92%, and 86% on posture, intensity, class context, and instructor behavior, respectively. Conclusions. Results support the SOFIT-X as a content- and criterion-valid measure that can be used to reliably code participant posture, class context, and instructor behavior in adult group-exercise classes. Further validation evidence is recommended for coding exercise intensity. Data from a small sample of classes also indicate the SOFIT-X has high face validity, and potential trends in posture, intensity, context, and instructor behavior across group-exercise format categories may exist.