An examination of the predictors of discord and divorce in the early stages of marriage

The work presented here represents a program of research designed to explore factors that predict marital discord and dissolution in the early stages of marriage. Chapter 1 reviews previous marital prediction studies and identifies consistent findings and common weaknesses in previous research. Chapter 2 offers a theoretical framework for examining change in relationships that is intended to place the prediction of marital outcomes in a larger conceptual context. Chapter 3 presents data from 85 couples in Munich, Germany. Results show that marital outcomes were predicted with 67% accuracy using a two-factor model in which hostility and neuroticism foreshadow divorce over the first 5 years of marriage whereas indices of negative communication predict dissatisfaction among the remaining marriages. Chapter 4 extends these analyses with data from 488 newlywed couples in Buffalo NY and 172 couples from Los Angeles CA. Using measures of verbal and physical aggression to predict divorce and measures of stress, trait-anger, and poor communication to predict marital dissatisfaction, 3-year marital outcomes were classified with 52% accuracy in the Buffalo sample. When race-specific (Caucasian or African American) predictive algorithms were developed, the level of accuracy rose to 57--58% in both groups. Similarly, when separate predictive algorithms were developed in couples with and without children at marriage, the level of accuracy rose to as much as 64%. Thus, race and premarital parenthood status seem to moderate marital prediction. Finally, the predictive algorithms developed in the Buffalo sample were applied prospectively to the couples from Los Angeles. The algorithms were able to predict with 60--66% accuracy in a new sample. These results extend the marital prediction literature by: (1) refining understanding of the factors involved in predicting marital outcomes during the first few years of marriage, (2) refining models of prediction to include moderating variables such as race and premarital parenthood status, (3) developing stable multivariate algorithms for prediction, and (4) providing the first evidence of prospective validation for predictive algorithms.