The dissolution of marriages has serious consequences. The basic facts on divorce are reviewed in this chapter. The basic task of my research effort is to predict and understand the longitudinal course of marriages. In my laboratory, I now can tell from a brief interview with a couple about the history of their marriage, from a few questionnaires they fill out, and from three brief 15-minute samples of videotape what the eventual fate of a sample of marriages is likely to be. This chapter introduces my basic paradigm for studying marriages. This paradigm consists of eight components. From my videotapes, 1 code (a) problem solving; (b) affect; (c) power; (d) cross-situational responses (pervasiveness of discord and rebound); (e) the following sequences: start up, continuance, positive reciprocity. During the interaction, I also measure a couple's (f) physiological responses. From questionnaires, I assess the construct of (g) distance and isolation. From coding the tapes of an interview, I assess (h) how the couple views its past history. In this chapter, I discuss why I think that observation of behavior is the cornerstone of this research effort. I also note that it is not at all obvious what to observe, and that early researchers were somewhat befuddled by the huge array of possibilities. However, the early data on marriages led researchers to recognize the importance of problem solving, affect, and power. Most recently, the study of affect has become more precise, and we now are reliably measuring specific affects and patterns of affect expression. To illustrate these, two examples, contempt and defensiveness, are described.