Detecting emotional expression in face-to-face and online breast cancer support groups

Accurately detecting emotional expression in women with primary breast cancer participating in support groups may be important for therapists and researchers. In 2 small studies (N = 20 and N = 16), the authors examined whether video coding, human text coding, and automated text analysis provided consistent estimates of the level of emotional expression. In Study 1, the authors compared coding from videotapes and text transcripts of face-to-face groups. In Study 2, the authors examined transcripts of online synchronous groups. The authors found that human text coding significantly overestimated Positive Affect and underestimated Defensive/Hostile Affect compared with video coding. They found correlations were low for Positive Affect but moderate for negative affect between Linguistic Inquiry Word Count (LIWC) and video coding. The implications of utilizing text-only detection of emotion are discussed.