Coding systems have become popular methods of cataloging the verbal and nonverbal interaction occurring during marital and family therapy. One such system, Pinsof's (1981) Family Therapist Coding System (FTCS), was the first designed explicitly to identify and differentiate specific verbal behaviors of family therapists independent of their theoretical orientation. To test the system's interrater and intrarater reliability, data were coded from typed manuscripts of six audio-taped marital therapy sessions. Coders were two undergraduate students trained for about seven hours each. The code consisted of three categories (for verb, phrase, and speech clause); the codes are ranked hierarchically such that only one code is assigned to each of the nine scales within the categories. The results indicated low observer agreement for overall session reliability and for category reliability. The low reliability did not appear to be due to observer drift or actual therapy sessions, but to the expertise and experience of the coders. The primary contributor to low reliability appeared to be the individual codes. Most of the codes with lower percentage of agreement values appeared to be less clearly defined and more difficult to apply to the data. The FTCS does not appear to be a reliable or practical assessment tool for determining the effectiveness of the therapist's statements during ongoing marital therapy sessions. (LLL)