Assessing Longitudinal Change in Marriage: An Introduction to the Analysis of Growth Curves

Fundamental topics in understanding marriage involve issues of and change. For example, determining the causes of marital distress requires that marital quality be examined repeatedly in marriages that may be deteriorating or remaining satisfied over time. Investigating the effects of the transition to parenthood on marriage calls for describing the course of marriages across that transition. Evaluating interventions designed to alleviate or prevent-marital discord necessitates assessing the trajectory of marriages before and after treatment. Many other important marital variables, such as ways of dealing with conflict, depression, and the occurrence of stressful events, can be characterized similarly by their stability or change over time, suggesting that marriage, from any perspective, is an inherently temporal phenomenon. Progress in addressing questions about marriage therefore will depend on the quality of available techniques for assessing marriages overtime. The increasing use of longitudinal designs among marital researchers suggests growing consensus on the importance of examining marriages across (see Berscheid, 1994; Bradbury & Karney, 1993). However, recent review of 115 longitudinal studies of marriage (Karney & Bradbury, 1995) reveals surprising lack of consensus on how best to analyze longitudinal data in this domain. To date, at least 18 statistical techniques have been used to estimate longitudinal effects on marriage. Over 70% of the longitudinal studies of marriage use procedures that analyze only waves of data (e.g., zero-order correlations, t tests, analysis of variance) and, perhaps as result of the reliance on such techniques, more than half of these studies collect only waves of data. Multiple waves of data, when they have been collected, are typically analyzed as series of two-wave designs. A common approach is to use variance in measure taken at Time 1 to account for variance in later marital outcomes. To the extent that Time 1 factors account for variance in outcomes measured at later time, those factors are said to affect marriage over time. Regardless of how they are analyzed, data obtained in two-wave designs are limited in their ability to illuminate how marriages change. One limitation is that analyses based on waves of data may not reflect how marriages are actually experienced. Change in relationships is a constant (Furman, 1984, p. 29), and theorists have suggested that over time is the characteristic that distinguishes relationship from mere series of interactions (Duck & Sants, 1983, p. 39; see also Kelley et al., 1983). The focus on associations between initial measurements and final outcomes does not address the continuity of marriage and instead yields static view of marriage as series of discrete snapshots linked by incremental changes. By failing to describe or account for the patterns of that underlie those changes, common methods of analyzing longitudinal data may be overlooking defining feature of marriage. A separate methodological problem with these approaches is that two waves of data contain an extremely limited amount of information about the change of each individual (Rogosa, Brant, & Zimowski, 1982, p. 729). Estimates of longitudinal relationships based on only waves of data are therefore likely to be unreliable. These problems led Willett (1988) to contend that the inability of current data analysis techniques to describe change as continuous process has prevented empirical researchers in the social sciences from entertaining richer, broader spectrum of research questions, questions that deal with the nature of development (p. 347). Addressing such questions requires that marital researchers approach data collection and data analysis with the view that change is continuous process. With respect to data collection, this involves assessing marriages repeatedly over significant periods of with measures that retain their validity across time. …