Affect pattern recognition: Using discrete hidden Markov models to discriminate distressed from nondistressed couples

Self-report affect sequences generated during a conversation between spouses were used to illustrate how Hidden Markov Model (HMM) methodology can classify couples according to marital quality. This pattern recognition technique allows an investigator to characterize processes that generate observable phenomena-in these data, expressed affect. I introduce the conceptual foundations and, briefly, the methodology of HMM and discuss its potential use in social and behavioral research. To illustrate the potential value of this method, I show how sequences of self-reported affect, derived in real-time during a laboratory interaction between 30 married partners, can successfully discriminate between distressed and nondistressed marital relationships.