Abstract: Supportive Expressive Therapy (SET) improves psychosocial outcomes in breast cancer patients, yet it has not reliably been differentiated from community support group models based on emotional content. We compared SET and The Wellness Community (TWC) support groups using observational coding of videotapes and human- and computer-based coding of text transcripts. Participants were 45 women with primary breast cancer (87% Caucasian, mean age=57 yrs), randomized to two TWC (N=20) and two SET (N=21) support groups that met for 16 weekly sessions. Four group therapy sessions, were coded for emotional content by trained coders using Specific Affect Coding (Gottman & Krokott, 1989) modified for breast cancer patients (Giese-Davis et al., 2006), and Linguistic Inquiry and Word Count (LIWC; Pennebaker & Francis, 1997). Comparisons were made based on proportions of primary negative affect, positive affect, defensiveness/hostility, constrained anger, and neutral affect between therapy modes and assessment modalities using nonparametric statistics. Results indicated that SET groups expressed more primary negative affect and defensiveness/hostility, compared to TWC groups. Less defensiveness/hostility was detected in text transcripts than in videotapes suggesting important emotional content may be lost in the absence of behavioral cues. SPAFF and LIWC primary negative affect scores in transcripts were significantly positively associated, supporting limited convergent validity. This study established that SET and TWC differed on emotional content in theoretically-consistent ways and demonstrated limitations of text-based analysis in detecting emotional content in interactive group therapy contexts.