Critique of Research on Conflict Behaviors and Outcomes

Research in this general area has pro­duced many impressive results. Despite huge variation in how constructs like negativity have been measured (Heyman, 2001), there is ample evidence that negativity predicts dissatisfaction and divorce.

Nevertheless, there are concerns about the research in this area. Because the extant lit­erature is so large and varied, no particular concern applies to all studies, but each is prevalent. These concerns pertain to studies that predict divorce, research on the affect dimension of conflict engagement, and some general assumptions of the problem-solving paradigm.

Research on Predictors of Divorce

The research demonstrating that negativity predicts divorce has been fruitful, with some studies indicating the ability to predict divorce at a rate greater than 90% (e.g., Gottman, 1994). Not surprisingly, such findings have generated much attention, but there are rea­sons for this enthusiasm to remain somewhat tempered (Bradbury, Rogge, & Lawrence, 2001). First, the statistical analyses used to predict divorce are optimized for a particular sample. Unless the same prediction equa­tions are validated with another sample, it is impossible to know how many divorces can actually be predicted. Heyman and Smith Slep (2001), for example, used half of a nationally representative sample to compute a prediction equation that correctly labeled 90% of partici­pants as married or divorced.

Second, even if negativity predicts relational dissolution, this does not prove a causal rela­tionship (Glenn, 1998). Indeed, the existing evidence suggests that part of the association between negativity and divorce may be an arti­fact of including couples who are already mov­ing toward dissolution. Negativity appears to be a reliable predictor of divorces in the short term but not in the longer term. Huston, Caughlin, Houts, Smith, and George (2001) found that newlywed levels of negativity were significantly higher in couples who divorced in the first 2 years of marriage than for couples who divorced after at least 2 years. Similarly, in their 14-year longitudinal study, Gottman and Levenson (2000) found that negativity predicted divorces during the first half of the study but not in the second half. Given that relational dissatisfaction predicts heightened negative conflict resolution behaviors (Noller & Feeney, 1998), at least part of the associa­tion between negativity and relational dissolu­tion may reflect general distress rather than evidence that negativity causes divorce.

Research on the Affect

Dimension of Conflict Behaviors

A number of critiques have been made of research that focuses on the affect dimension of conflict behaviors. First, the well-documented finding that “negativity” is associated with relational dissatisfaction and relational dis­solution may seem obvious (Sillars & Canary, 2013). One counterargument to this point is that the general category of negativity sum­marizes a number of specific behaviors, some of which may be more important correlates of dissatisfaction and dissolution than are others (Gottman, 1994).

Another important point to be gleaned from the literature is the empirical evidence that negativity and positivity cannot be ade­quately described with a single dimension (Caughlin & Huston, 2006). If a single dimen­sion accounted for negativity and positivity, these behaviors would have a strong inverse correlation across all studies. However, the observed correlations are often quite small, and factor analyses support treating positive and negative behaviors separately (Huston & Vangelisti, 1991; D. A. Smith, Vivian, & O’Leary, 1990). Moreover, positive and nega­tive interaction behaviors often moderate each other’s association with relational satisfaction (Huston & Chorost, 1994; D. A. Smith et al.,

1990), a finding that would be unlikely if a single dimension adequately captured these constructs. Finally, Gottman and Levenson (2000) and Huston et al. (2001) found that the relational outcomes associated with positive and negative interaction behaviors are distinct (i.e., negativity and positivity do not demon­strate parallelism). Because of the empirical evidence that negativity and positivity are not part of a single dimension, it is usually not appropriate to collapse behaviors with posi­tive and negative affect into a single measure (e.g., Matthews et al., 1996).

Assumptions of the

Problem-Solving Paradigm

The prototypical study of conflict and relational outcomes involves observing a sin­gle episode of problem solving (Noller & Feeney, 1998).

1991). Nevertheless, the problem-solving par­adigm implies several questionable theoretical assumptions about relational conflict.

First, research in this area often assumes that conflict is inevitable. This assumption is often taken to mean that the amount of conflict and the difficulty of the issues do not matter; for example, Clements, Cordova, Markman, and Laurenceau (1997) argued, “The number and type of conflict areas... are less important than how couples handle these conflicts” (p. 342). The countless studies asking couples to discuss their most conten­tious issues implicitly assume that frequency and difficulty are unimportant. Also, using the behaviors produced in these episodes as indicative of “how couples handle” con­flict presumes that couples actually discuss their disagreements and that all couples have equally troubling yet solvable conflict issues. This implies a theoretical perspective of how conflict operates in romantic relationships that is, at best, suspect.

In contrast to the assumption that manag­ing conflict is more important than the dif­ficulty and frequency of conflicts, experienced relationship counselors attribute the majority of couples’ troubles to fundamental problems that often cause communication difficulties, rather than to problems managing conflict, per se (Vangelisti, 1994). Also, Sanford (2003) showed that the difficulty of married couples’ conflict issues is related positively to observa­tions of negativity and related inversely to relational satisfaction. Thus, the observed connections between negativity and dissatis­faction may be somewhat spurious and driven by the difficulty of the conflict.

Moreover, research outside laboratories suggests that the frequency of conflict is an important predictor of relational outcomes.

A second questionable assumption in the problem-solving paradigm is evident in the typical design of such studies. Most of this research, even investigations that assess other constructs longitudinally, measures conflict behaviors only once (Noller & Feeney, 1998). This is potentially problematic because behav­iors in a particular episode are likely to have a greater association with judgments of a similar time frame (e.g., satisfaction with the interaction) than with more molar judgments like overall relational satisfaction (Caughlin, 2010). Thus, the usual conflict observational paradigm assumes that the observed behaviors reflect what generally occurs in the relation­ship and that these behaviors are stable over time. This is a dubious assumption given that relational conflict does change over time (Caughlin & Huston, 1996).

Also, in a study using both diaries and self­reports, Caughlin and Huston (1996) found that marital stability over a 13-year period was related to changes in the experiences of conflict over the first 2 years of marriage. According to daily diary reports of conflicts, dyads who divorced and those who stayed married were similar in the first 2 years in terms of the number of conflict episodes. Divorced and married couples did not differ significantly in conflict frequency, and the frequency declined over the first 2 years of marriage, regardless of whether a couple even­tually divorced or not. However, retrospective reports taken first when the couples were new­lyweds and again after 1 and 2 years indicated that couples who eventually divorced reported significant increases in conflict over time, whereas couples who stayed married reported stable levels of conflict.

Our conceptual framework (see Figure 6.1) implies one possible explanation for the discrepancy between the two assessments of conflict. The two different measures ask people to report on different time frames, with retrospective reports reflecting a 2-month period and the diaries a single day. The differ­ent time frames might cue individuals to count conflicts differently. Consider, for example, the notion of serial arguing (e.g., Malis & Roloff, 2006; Roloff & Johnson, 2002; Roloff & Reznik, 2008), which recognizes that con­flict about a particular issue can extend past a single episode. For a diary measure of con­flict on a given day, participants may report only the overt conflict episodes, which means they would exclude ongoing serial arguments that were not explicitly discussed that day. However, when asked about the amount of conflict over the past 2 months, spouses would likely include ongoing serial arguments, even if they are rarely discussed on a given day. Having many (or particularly frustrating) serial arguments could explain why dyads who end up divorcing reported fewer daily episodes of conflict over time while also reporting that the amount of conflict in their relationship was increasing. Even if a couple does not discuss an issue often, the knowledge that the issue is ongoing and difficult to resolve may be just as important as overt conflict episodes (Lloyd, 1990; Roloff & Johnson, 2002). These findings show the importance of considering more than a single episode: Patterns of conflict can develop and unfold over periods of time much longer than the typical observational study (Christensen & Heavey, 1993; Roloff & Johnson, 2002).

A third assumption of the problem­solving paradigm is that the sample of con­flict behaviors obtained during a laboratory encounter is valid and representative. Much has been written about the external validity of typical observational studies (Heyman, 2001). The general conclusion typically is that laboratory behaviors are not as nega­tive as conflict behaviors at home, but there is some validity to observational methods because they can predict dissatisfaction and divorce (Heyman, 2001).

Nevertheless, observational studies are often not ideal for addressing certain impor­tant questions about relational conflict (Noller & Feeney, 2004). Consider, for example, the difficulties of studying conflict avoidance in a laboratory setting. Individuals are less likely to withdraw if they are directed by a researcher to discuss an issue for a preset amount of time (Kluwer, Heesink, & Van de Vliert, 1997; Roberts, 2000). The setting also constrains the types of avoidance behaviors partners engage in (e.g., they are unlikely to leave the room, make a phone call, or turn on an electronic device). Moreover, much conflict avoidance occurs before an issue is even introduced, mak­ing it difficult to observe such avoidance.

Even if behaviors can be validly assessed in an observational episode, there are questions about whether a single assessment is adequate (Fincham, 2004). Particular kinds of critical conflicts are unlikely to be captured in a labo­ratory. A couple’s “first big fight,” for instance, can have a large impact on the development of a relationship, leading to dissolution of some dyads and heightened interdependence in others (Siegert & Stamp, 1994). Unless scholars assume that there are no behavioral differences between the most crucial conflict episodes and ones observed in research, laboratory studies are likely to miss important aspects of conflict.

Also, research using diary methods has shown that relational partners’ moods influ­ence how negatively they treat each other (Schulz, Cowan, Cowan, & Brennan, 2004). Having a negatively stressful day at work, for example, is associated with more mari­tal conflict (Bolger, DeLongis, Kessler, & Wethington, 1989), greater expressions of anger from wives, and more withdrawn behavior from husbands ( Schulz et al., 2004). Such variations probably add a source of unreliability to assessments that rely on a single encounter. Given this potential for unreliability, analyses based on these mea­sures may actually underestimate the asso­ciation of conflict behaviors with relational outcomes.

In short, the assumptions of the prob­lem-solving paradigm may obscure important aspects of relational conflict. The typical labo­ratory designs make the untenable assump­tions that the frequency and difficulty of conflict is irrelevant; they focus primarily on observable conflict behaviors in a particular episode, which may be less important than perceptions of conflict, such as the belief that a serial argument is not resolvable; and they usually presume that the behaviors in a single session are representative of conflict behaviors in a relationship generally, a dubi­ous assumption given the situational demands inherent in observational studies. Although more research using the problem-solving para­digm is certainly warranted, it is important not to treat the results from this paradigm as representative of the overall impact of conflict on romantic relationships. Common observational methods capture some of what is important about relational conflict, but not everything. Indeed, the concerns about this paradigm suggest that previous studies may have underestimated the complete impact of relational conflict.

Individuals' Influences

on Relational Conflict

Conflict scholars investigating how individu­als shape their conflicts have sought to explain how conflicts emerge and develop and have investigated how individuals’ cognitions influ­ence the course and outcomes of relational conflict. Again, the literature in this general area is enormous, even rivaling the research on associations between conflict behaviors and relational outcomes. Our focus here is on reviewing several common explanations for conflict and on briefly discussing the impor­tance of interpretations of relational conflict.