|Home | About | Journals | Submit | Contact Us | Français|
Using a sample (N = 453) drawn from a representative sampling frame of couples who are married or living together and have a 3–7 year old child, this study investigates (a) the amount and specific areas of change desired by men and women, (b) the relation between relationship adjustment and desired change; and (c) the ways in which partners negotiate change. On the Areas of Change Questionnaire, women, compared with men, wanted greater increases in their partners’ emotional and companionate behaviors, instrumental support, and parenting involvement; men wanted greater increases in sex. Using the Actor-Partner Interdependence Model (Kenny, 1996, both men’s and women’s relationship adjustment predicted desired change (i.e., actor effects), over and above the effects of their partners’ adjustment (i.e., partner effects); partner effects were not significant. Each couple was also observed discussing the man’s and the woman’s top desired change area. Both men and women behaved more positively during the partner-initiated conversations than during their own-initiated conversations. Women, compared with men, were more negative in their own and in their partners’ conversations.
Modern relationships carry tremendous burdens. Partners must stoke romantic and sexual excitement and provide emotional support while fulfilling financial, household, and (in most cases) child nurturance responsibilities. In relationships that strive toward egalitarian decision-making, all of these areas must be negotiated in a climate in which social structures defining roles have been reduced or eliminated and expectations for a fulfilling relationship have risen (Counts, 2006). The shear scope of the financial and human resources and responsibilities to be allocated, re-allocated, and maintained makes conflict both inevitable and of intense interest to scientists and interventionists as a key determinant of relationship satisfaction and long-term health and viability.
Many theories have considered changes that partners want from each other and the conflicts that can ensue from those desires. For example, Social Exchange Theory (e.g., Thibaut & Kelley, 1959) posits that relationships operate quasi-economically; that is, because behavioral exchanges are dyadic, they produce rewards and/or costs of varying magnitudes. A key facet of this theory is that partners reference their dyadic outcomes to a comparison level (their expectations of the outcomes due to them in a relationship) and a comparison level for alternatives (their expectations of the outcomes they could receive in another relationship). Thus, desired change operates both on a specific level (e.g., cost/benefits related to housework distribution) and a relationship level (the overall perception of relationship costs/benefits and how that compares to alternatives). An extension of social exchange theory, Equity Theory (e.g., Walster, Walster, & Bercheid, 1978), emphasizes the fairness motive in dyadic exchanges, with inequitable relationships creating turmoil in both partners. According to Equity Theory, partners who believe they are overrewarded will feel guilty and those who believe they are underrewarded will feel angry; both are motivated to rectify the imbalance. The Ideal Standards Model (e.g., Fletcher & Simpson, 2000, p. 102) posits that evolution has favored humans who evaluate partners and relationships in three key, reproduction- and child-rearing-sensitive dimensions — “(a) warmth, commitment, and intimacy; (b) health, passion, and attractiveness; and (c) status and resources.” People who believe that their partners/relationships are falling short of the ideal levels in these areas are motivated to seek redress. Coughlin and Vangelisti (1999 Coughlin and Vangelisti (2000) have found support for both a personality-driven influence on change seeking (i.e., individual differences) and communication influence (i.e., couple communication style, comprising both partner’s individual styles in a dynamic interactive context). The thread that runs through these four models is that people track the rewards and costs in their relationships, that partners’ fates are intertwined in relationships, and that humans compare what they are getting to an internal model of what they believe they should be getting.
Despite intense public, scientific, prevention-oriented, and therapeutic interest in both desired changes and gender differences in relationship desires, surprisingly few studies have overtly studied specific areas of desired change. Previous investigations have found that women, compared with men, desire more change from their partners, in both community (Ball, Cowan, & Cowan, 1995; Margolin, Talovic, & Weinstein, 1983) and clinic (Doss, Simpson, & Christensen, 2004) samples. Further, differences in relationship adjustment may be qualified by an interaction with gender; Margolin et al. (1983) found that dissatisfied women, followed by dissatisfied men, desired the most change. Satisfied men, followed by satisfied women, desired the least amount of change from their partners. This study needs to be both replicated and extended using the more sophisticated analytic tools that did not exist in 1983; specifically, the Actor-Partner Interdependence Model (APIM; Kenny, 1996) allows investigators to simultaneously test (a) the impact of one’s own relationship adjustment on one’s own amount of desired change; (b) the impact of one’s own relationship adjustment on the partner’s amount of desired change; and (c) the impact of the partner’s relationship adjustment on one’s own amount of desired change. “Common sense” would predict that (a) one’s own relationship dissatisfaction would increase one’s desire for changes from the partner and (b) being involved with a dissatisfied partner would increase one’s desire for changes from the partner. Margolin et al.’s (1983) study offers tentative support for the supposition that one’s own adjustment drives one’s own desires for changes; the strongest support would come if that finding were replicated while controlling for the influence that partner’s dissatisfaction may have on one’s own desires for changes.
Questionnaires about desired changes are often used to set up observed conflicts between partners, an area of research that boasts several hundred studies (see Gottman & Notarius, 2000; Heyman, 2001 for reviews). In sum, these studies have found that distressed, compared with nondistressed, couples begin their conversations more hostilely and act more hostilely over the course of the observed conflict; are significantly more likely to reciprocate and escalate hostility and to carry on these hostility exchanges longer; display less positive behavior; have short and long-term ill effects of conflict on their bodily systems and health; and are more likely to meet requests for change with withdrawal/resistance (and vice versa) (Heyman, 2001).
Despite the extensive literature, two assumptions that are foundational for deriving meaning and inferring generalizability from observation of couples conflict have been underexplored. The first assumption is that women desire more change than men, leading some researchers (e.g., Coan & Gottman, 2007) to argue that one need only observe conflicts that women initiate. A related assumption — severely critiqued in Heyman’s (2001) psychometric review of couples observational research — is that studies that only observe women-initiated conflict or that do not control for the conflict initiator are generalizable to all couples conflict.
In the few studies that have investigated male- and female-initiated conversations, communication affect and behavior differed depending on whose topic was being discussed. One study found that women, compared with men, displayed more negative affect in both male- and female-initiated conversations and more positive affect only during female-initiated discussions (Johnson et al., 2005). Another series of studies investigated the demand/withdrawal pattern, which has been associated with poor relationship adjustment (Christensen & Heavey, 1990; Heavey, Layne, & Christensen, 1993). Previous research suggested that women, compared with men, were more typically the demanders during conflict discussions, whereas men were more likely withdrawers (Christensen & Heavey, 1990). This varied, however, depending on whose topic was discussed. During women’s conversations, female-demand/male-withdrawal was more likely, but during men’s topics, men and women did not differ on demand or withdrawal. Sagrestano, Christensen, and Heavey (1998) (using the same samples as Christensen & Heavey, 1990 and Heavey et al., 1993) reported that men’s and women’s (a) influence techniques and (b) attributions were influenced by whose topic was being discussed. These studies emphasize the importance of controlling who initiates discussions because who has a greater investment in the issue being discussed appears to influence the interaction.
This study focuses on both the changes that women and men desire and the ways in which they behave when trying to discuss possible changes, seeking to replicate and extend previous work using a fairly large sample of parents with children between ages 3–7 (N = 453) drawn from a representative sampling frame and thus reasonably generalizable. The earliest studies (Christensen & Heavey, 1990; Heavey et al., 1993; Sagrestano et al., 1998) used small samples (N = 31 and N=29) drawn from (a) targeted recruitment of families with a child with attention-deficit hyperactivity disorder, internalizing, or externalizing problems and (b) advertising and flyers or families with a nonproblematic child. The Johnson (2002) and Johnson et al. (2005) studies were of a sample of N = 172 couples drawn from a representative sampling frame of couples applying for marriage licenses.
The first major aim is to examine desired changes. First we hypothesized that women, compared with men, will desire more change from their partners (cf. Margolin et al., 1983). Second, because women engage in more instrumental household tasks than men do (e.g., housework and childcare tasks; Bittman, England, Sayer, Folbre, & Matheson, 2003), we hypothesized that women, compared with men, will desire more behavioral increases from their partners. Third, we will investigate the impact of gender and relationship adjustment on desired change. Margolin et al., (1983) found interactions between gender and adjustment on the amount of change desired. However, relationship adjustment in that study was summed between partners and then dichotomized; this is suboptimal both because of the limited variability of dichotomous variables and because variability from each partner is lost (Kenny, Kashy, & Cook, 2006). To address these limitations, the continuous association between men’s and women’s adjustment and desired change will be tested. We hypothesize that as women’s adjustment decreases, they desire more change from their partners than men do. In other words, we hypothesize that men and women who are more satisfied are more similar in the amount of change they desire than are men and women who are less satisfied. Fourth, as noted earlier, it is possible that one’s own adjustment and one’s partner’s adjustment predicts one’s desire for change; alternatively, one’s own adjustment may not predict desired change once the partner’s adjustment is taken into account. The APIM (Kenny, 1996) allows researchers to test such relations. We hypothesize that one’s own levels of adjustment will be negatively associated with one’s amount of desired change, even when controlling for partner’s adjustment. Finally, we will explore the specific areas of change to determine if men and women want change in that area, in what direction the desired changes are, whether men and women differ in their desired change in specific areas, and if gender differences in desired changes in specific areas remain after taking into account differences in total desired changes.
The second major aim is to investigate the discussion of change topics in an analogue conflict task (Heyman & Slep, 2004) in which couples are observed discussing key areas of desired change. In this sample derived from a representative sampling frame, we expect to replicate previous findings suggesting that whose topic is discussed affects observed communication behaviors (e.g., Christensen & Heavey, 1990; Johnson et al., 2005). We hypothesize that women will display more negative communication behaviors and affect during both male- and female-initiated conversations and that women, compared with men, will display more positive behaviors and affect only during their own conversations (Johnson et al., 2005)1.
This study included 453 couples (906 individuals) from Suffolk County, New York, recruited via random digit dialing, who were (a) married (and living together) or cohabiting2 for at least one year, (b) had a child between the ages of 3–7 years for whom either the respondent or the partner was the biological parent, and (c) able to speak English. Research assistants called 229,106 randomly generated telephone numbers and 12,009 respondents answered at least one question of the telephone screening survey. Respondents completed the telephone survey that assessed demographic characteristics as well as family functioning variables. Eligible and interested individuals (n= 2,212) were recontacted by staff and the study was explained in detail. Data collected from telephone respondents were compared with the 2000 U.S. Census data to evaluate representativeness; study participants were quite representative of the population (for complete details on sampling procedures and representativeness, see REDACTED FOR MASKED REVIEW). Comparisons of those who were eligible but chose not to participate to the participant demographic data revealed few statistically significant differences between participants and non-participants (and the few significant differences had small effect sizes).
Participants’ age was M = 37.16 (SD = 5.99, range 24–50) and M = 35.13 (SD = 5.18, range 21–47) for males and females, respectively. Participants were non-Hispanic white (80.6%), Latino or Hispanic (8.6%), African-American (6.2%), Asian (2.0%), Caribbean-American (1.4%), Native American (0.6%), Pacific Islander (0.1%), and Other (0.4%). The median family income for participants was $74,500 (range $4,700–$500,000).
The ACQ assesses 34 possible areas of change that one partner desires of the other (i.e., finances, housework, sexual relations). Participants indicated the degree of change on a 7-point Likert scale from −3 (much less) to +3 (much more), with 0 indicating no change desired. Absolute values for items were summed to create a variable indicating the total amount of change desired. Summary variables were also created for the amount of desired behavioral increases and decreases. The ACQ was added later in the larger study; therefore, scores are available for 145 men and 153 women. There were no significant (all t < 1.5, ns) demographic differences between those who completed the ACQ and those who did not. Cronbach’s α was .82 for men and .76 for women in this sample.
The DAS is a 32-item measure of perceived relationship adjustment. Scores range from 0 to 151, with higher scores indicating greater perceived relationship quality; scores below 98 indicate clinical distress (Jacobson, Follette, & Revenstorf, 1984). Cronbach’s α was .93 for men and .94 for women.
The RMICS is a microanalytic coding system adapted from the Marital Interaction Coding System-IV (MICS-IV, Heyman, Weiss, & Eddy, 1995) with good reliability and validity (see Heyman, 2004 for a review of studies using the RMICS and a detailed description of code definitions and coding procedures). The basic coding unit is the speaker turn; if a speaker turns lasts longer than 30 seconds, it is coded in 30-second intervals. Coders assign only one of the eleven codes to each unit; if two or more codes are present during a speaker turn, a theoretically derived hierarchy (i.e., negative codes then positive codes then and neutral codes) indicates which code to retain.
Negative (psychological abuse, distress-maintaining attributions, hostility, and withdrawal) and positive (relationship-enhancing attributions, acceptance, self-disclosure, and humor) category frequencies were calculated and then divided by that individual’s total number of coded behaviors to create the proportion of the conversation that the individual was negative or positive3. Interrater agreement was good for each of the categories (Cohen’s kappa = .57 for positive codes, .60 for negative codes). Because of technical difficulties with audio recording during data collection, 1 out of 906 interactions was missing; multi-level modeling, however, uses robust estimation techniques and can adjust for missing data (Kenny et al., 2006).
Couples came to the laboratory for a total of six hours. Informed consent was obtained. Partners separately (a) completed the questionnaires and (b) were interviewed regarding areas in which change was desired (i.e., participants were asked to think of the things they had tried to get their partners to do, do differently, or change within the past year and identify those things that did not change as much as was desired). The staff member read a list of areas that are commonly problematic for couples and asked the participant to list any behaviors that fell into each category, to indicate how often that issue was discussed within the past year, and rate that desired change on an importance scale ranging from 1-not important to 6-very important. The list included changes desired in partner behavior in the areas of housework or yard work, children, extended family, relatives, and in-laws, money or finances, recreation, free time, or social activities, work, and how the partner treats the participant. Participants were also given an opportunity to provide issues of their own that were not covered by these categories. After both interviews were completed, the staff member used a randomized list to determine whose conversation would be first and selected topics of conversation based on the topics that received the highest rating of importance. If more than one topic had similar importance ratings (two or more issues with importance ratings of 6), participants were asked to select the one topic they considered to be the most important. Each partner was told to discuss the issue and “try to get somewhere with it” and to “deal with the issue the way you typically would at home;” only the participant bringing up the issue was told what topic was to be discussed. Partners were brought together in the video-recording studio and discussed the problem for 10 minutes. Afterward, partners were separated and completed questionnaires about the conversation and others unrelated to the current study. This procedure was repeated for the second conversation.
After the participants completed both conversations, the questionnaires and other procedures, they were paid, offered a list of community resources, and debriefed.
Multi-level modeling (MLM) within SPSS was used because partners’ adjustment scores were not independent (r = .63, p < .01). To test the first hypothesis (women desire more change than men do), gender was the MLM level one variable, couple membership was the level two variable, and total desired change was the dependent variable. (Table 1 shows descriptive statistics for the variables.) As hypothesized, women desired more change than men did, b = −1.66, t(152) = −2.90, p < .01 (d = .25)4.
To test the second hypothesis (women will desire behavioral increases more than will men, even after controlling for relationship adjustment), gender and adjustment were the level one variables; couple membership was the level two variable. Desired change, the dependent variable, was calculated by summing items in which the respondent wanted a behavioral increase from the partner (i.e., ACQ item scores of +1 – +3). Women desired behavioral increases significantly more than men did, b = −1.60, t(151.6) = −3.12, p < .01 (d = .20) and lower adjustment was related to more desired behavioral increases rise, b = −.33, t(229.3) = −9.43, p < .001; the gender x adjustment interaction was not significant, b = −.03, t(185.3) = −0.94, ns. A parallel exploratory analysis was conducted for desired behavioral decreases. As before, as adjustment falls, desired behavioral decreases rise b = −.06, t(200.6) = −6.14, p < .001. There was no main effect for gender b = −.06, t(158.2) = −.35, ns nor a gender x adjustment interaction b = .00, t(195.9) = .10, ns.
To address the third hypothesis (women will want even more change [increases or decreases] as their adjustment declines than men will), gender and adjustment were entered as the level one variables; couple membership was the level two variable. The hypothesized gender x adjustment interaction was not significant, b = −.03, t(187.6) = −.78, p = .44.
Next, we tested the common sense beliefs that desire for changes from the partner is related to both (a) one’s own relationship dissatisfaction and (b) being involved with a dissatisfied partner. Before conducing the maximum likelihood structural equation modeling APIM analysis, we verified that the multivariate normality was not violated (Mardia’s coefficient = 2.02). As shown in Figure 1, only the first common sense belief was supported: when controlling for men’s relationship adjustment, women’s lower relationship adjustment was significantly associated with increased desire for change (r = −0.43 p < 0.001); similarly, when controlling for women’s adjustment, men’s adjustment was significantly associated with increased desire for change (r = −0.52, p < 0.001). Men’s adjustment was not significantly related to women’s desired change scores and vice versa.
Table 2 shows the amounts of change men and women desired (i.e., absolute value) in each specific area. To control for family-wise inflation of alpha, we applied a Bonferroni correction, such that a p < .00147 was necessary to be considered statistically significant. Both men and women wanted significant change on all items (i.e., t values that differ from 0 [no change]) except “hit me” and “engage in extramarital sexual relations.” Ten of the items showed significant gender differences in the amount of desired change — men, compared with women, wanting more change on women having meals ready on time and on two items about sex; women wanted more change on spending time with the children, household work (“accomplish his responsibilities promptly; “help with the housework when asked”), interpersonal connection (“spend time with me;” “give me attention when I need it”), expressing emotions clearly, help in planning free time, and working late. As a very conservative test, we tested if the gender differences at the item level remained after removing gender differences in overall change desired (by total desired change [minus that specific change item] on each desired change item and comparing the means). No item level gender differences remained.
We also re-ran the comparisons (see online supplemental Table 1), taking into account both the amount of change and the direction of change (i.e., behavioral increases or decreases). This was important to conduct because Table 2 showed whether change was desired, not the nature of the change (i.e., if the trend was for wanting more or less). Results indicated that for drinking, disciplining children, and having non-sexual relationships, both men and women have offsetting endorsements, with some people wanting their partners to discipline the children more, for example, but with other people wanting their partners to discipline the children less.
The second major purpose of this study was to replicate previous findings suggesting that whose topic is discussed affects communication between partners. Women, compared with men, were expected to evidence more negative communication behaviors during both male- and female-initiated conversations and women were expected to display more positive behaviors only during their own conversations. Two MLM analyses were conducted (one where the dependent variable was the proportion of RMICS positive behaviors and one where it was RMICS negative behaviors). The MLM level one variables were gender and who initiated the topic; the level two variable was couple membership. Women, compared with men, emitted more positive, b = −.004, t(205) = −2.17, p < .05 (d = .10), and more negative, b = −.02, t(53.6) = −7.10, p < .001 (d = .20), behaviors. Main effects for who initiated the topic were not significant (positive, b = .003, t = 1.27, p = .20; negative, b = −.001, t[451.7] = −0.55, ns), but there was a significant gender x who initiated the topic interaction for positive behaviors, b = −.02, t(833.07) = −11.17, p < .001 (see Figures 2 and and33)5. Post hoc comparisons (see Table 3) using the least significant differences test revealed that men emitted significantly more positive behavior than women did during female-initiated conversations; similarly, women emitted significantly more positive behavior than men did during male-initiated conversations, d = .27. Men were significantly more positive during female-initiated, compared with male-initiated, conversations (d = .30). Furthermore, men engaged in significantly fewer positive behaviors during female-initiated conversations than women engaged in during male-initiated conversations (d = .16) and men engaged in during male-initiated conversations (d = .47). Finally, women engaged in significantly fewer positive behaviors during female-initiated, compared with male-initiated, conversations (d = .43).
There was also a significant gender by who initiated the topic interaction for negative behavior, b = .007, t(687.3) = 4.96, p < .001. Women, compared with men, were significantly more negative in both female-initiated (d = .30) and male-initiated conversations (d = .10). Women also emitted significantly more negative behaviors during female-initiated conversations than men did during male-initiated conversations (d = .21). Women were significantly more negative in male-initiated conversations than men were during female-initiated conversations (d = .20). Men’s negative behaviors did not differ significantly between male-initiated and female-initiated conversations. Finally, women’s greater negativity in their own, compared with their partner’s, conversations (p = .067; d = .11) fell just short of statistical significance.
The results of this study indicate that desired change within established heterosexual relationships is linked with gender and relationship adjustment. As hypothesized, women endorsed higher amounts of desired change than men did, replicating previous findings (Margolin et al., 1983). Also as hypothesized, women wanted behavioral increases more than men did. Specifically, women, compared with men, wanted significantly greater increases in their partners’ emotional and companionate behaviors (“spend time with me,” “give me attention when I need it,” “start interesting conversations with me,” “express his emotions clearly”), instrumental support (“accomplish his responsibilities promptly,” “help with the housework when asked,” “help in planning free time”), and parenting (“spend time with the children”); men, compared with women, wanted significantly greater increases in sex. In addition, women wanted men to work less more than men wanted that from women. Seen through the lens of the Ideal Standards Model, these results imply that there are bigger gaps for women than men in their partners falling short of their ideals in the “warmth, commitment, and intimacy” dimension (the emotional and companionate behaviors above) and the “status and resources” dimension (the instrumental support and parenting behavior desired increases coupled with desired reduction in men’s work hours. Although both men and women want more sex, women fall shorter than vice versa for men’s ideals for the “health, passion, and attractiveness” element of sexual frequency.
These findings suggest that women think that men are not investing enough of themselves (e.g. resources, time, attention, tangible support) in their relationships. Another possible explanation could be that women are more focused on relationships than are men (Gabriel & Gardner, 1999). This would suggest that relationship events may not be as salient to men, leading them to not remember possible changes they may want. If this is true, perhaps self-reports of desired change lead to higher rates of women wanting change as an artifact of gender differences in memories of dyadic events (Gabriel & Gardner, 1999, Study 4). If another method were used (e.g. daily diary, Stone, Kessler, & Haythornthwaite, 1991) perhaps the rates of desired change would be more equitable between partners because desired changes could be rated on a daily basis as problems arise and individuals would not need to track and summarize events over a long period of time. On the other hand, ephemeral desired changes may not be important to either daily nor distal relationship health, making summary self-reports such as the Areas of Change Questionnaire a more valid measure of the key elements of the construct.
The hypothesized interaction between gender and adjustment (women’s greater desire for change would be exacerbated at lower levels of adjustment) was not found. This result is inconsistent with that of Margolin et al. (1983). However, this cannot be attributed to different analytic strategies in the two studies — as a check, we replicated Margolin et al.’s distress/nondistress categorization strategy exactly and the interaction remained non-significant. The difference in findings could be due to the types of samples used in the two studies (our sample: parents of 3–7 year old children drawn from representative sampling frame; Margolin et al.’s sample: non-distressed couples recruited through advertising and distressed couples seeking therapy) or to the two decades that have elapsed between the studies.
The result that those lower in adjustment want more change from their partners is congruent with previous research (Margolin et al., 1983). Lower adjustment has been found to be associated with more negative cognitions and recall of negative partner behaviors (Bradbury & Fincham, 1987; Whisman & Delinsky, 2002); this may explain why those lower in adjustment desire more change. If adjustment is greater, then perhaps individuals are not as likely to notice specific things that they want their partners to change or they are more willing to let things go as they come up and not keep a running tally of things they want their partner to change (e.g., communal orientation, Clark, Oullette, Powell, & Milberg, 1987). On the other hand, high adjustment couples may just do more of the rewarding and fewer of the costly behaviors, thus coming closer to their partners’ relationship ideals and requiring fewer desired changes.
Although common sense would say that both one’s own and one’s partner’s adjustment would influence the amount of desired change, this turns out not to be true. As hypothesized, both men’s and women’s adjustment predicted amount of desired change (i.e., actor effects), over and above the effects of their partners’ adjustment (i.e., partner effects); however, partner effects were not significant. Although this could be due to common reporter variance (i.e., within-reporter associations are stronger than cross-reporter associations), it implies that one’s desire for change is related only to one’s one adjustment; if partner’s distress impacts one’s desire for change, it’s likely only through conflict that decreases one’s own adjustment.
Of particular clinical importance, in every domain in which men or women wanted a differential amount of increased or decreased partner behavior, both genders wanted significant change in the same direction (just not as much). For example, both men and women desire significantly more sex, but men want even more change than women do; both men and women desire increases in spending time together, but women want even more change than men do. Such “conflicts” about relative amounts are easier to resolve or at least accept (Jacobson & Christensen, 1998) than are conflicts with diametrically opposed desires.
Many studies have focused on one or two female-initiated conflict discussions, with the rationale being that these conversations are more generalizable because women desire more change than men do (e.g., Coan & Gottman, 2007). However, as with several previous studies (Christensen & Heavey, 1990; Heavey et al., 1993; Johnson et al., 2005), we found that a different pattern of results emerged when women brought up change topics than when men did. If behaviors occurring during female-initiated conversations were the only behaviors analyzed, one could have concluded that men were more positive than women were and women were more negative than men were. However, we found the hypothesized interaction between gender and who initiated the discussion. Specifically, women were more positive than men only during male-initiated conversations. Moreover, women were more positive in male-selected topics than men were during female-selected topics. Although women were more negative than men in both male- and female-initiated conversations, women were significantly less negative during male-initiated conversations than they were during their own conversations. This richer picture of couples communication would not have been found in studies using topics selected only by the female partner. Thus, women would have appeared generally less positive and more negative than they actually have the potential to be when the initiator of the discussion is changed. These results are in line with the conflict structure explanation of dyadic interactions that suggests that the way men and women behave during interactions can be attributed to the roles (e.g. problem initiator versus problem recipient) that they adopt during discussions (Peplau & Gordon, 1985).
The clinical implications of the observational findings suggest that clinicians should thoroughly assess problems that both partners have in the relationship because it is likely that the issues are multifaceted and communication skills can vary depending on who initiates the discussion. By observing both partners’ discussions about relationship problems (see Heyman, 2001), clinicians’ case conceptualizations and treatment plans will likely be impacted because therapists will have a more comprehensive understanding of how each partner acts and reacts depending on whose problem is discussed.
There are several strengths of the design and analytic strategy used in this study. First, the sample size was large for an observational study and participants were recruited via random digit dialing and were representative of the county’s population (REDACTED FOR MASKED REVIEW). Analytic strategies were used that control for interdependence of dyadic data so that effects could be isolated (i.e., Actor-Partner Interdependence Model using multi-level modeling and structural equation modeling). Finally, many of the results of the current study are concordant with previous research (e.g., women want more change than men do; Margolin, et al., 1983; there are behavioral differences in conversations depending on the interaction between conversation initiator and gender; Christensen & Heavey, 1990).
There are several limitations to the current study. First, because participants were cohabiting or married heterosexual couples with children between ages three and seven, the results of this study cannot necessarily be generalized to other kinds of couples (e.g., those without children or with only older children; same-sex couples). Additionally, the sample for this study was largely White. Predominantly ethnic minority samples may demonstrate different relationship processes than the current sample. As noted earlier, results about desired change from questionnaires may differ from those using methods such as daily diaries.
In conclusion, in relationships with at least one young child, both men and women desire change, but in several key areas — instrumental support, emotional and companionate behaviors and parenting — women want more behavioral increases than men do; men want more increases in sexual interaction than women do. In the popular press or in therapeutic shorthand, significant gender differences often are described (and thus become reified) dichotomously: “Women want instrumental support, emotional and companionate behaviors and parenting; men want sex.” This would be a misinterpretation of the findings. Overall, men and women want changes in the same directions on a host of behaviors. There are gender differences in the degree of change wanted on some behaviors. Preventionists and therapists can use these nomethic findings to generally guide interventions and to look for similar patterns with individual couples. Couples in conflict may have an easier time negotiating solutions if they perceive that they desire similar changes, with a sticking point being the matter of degree. Finally, clinicians and researchers who watch couple conflicts to infer where communication difficulties lie will be well served to remember that who desires the change makes a difference in men’s and women’s behavior when they discuss it. When observing couples’ communication for clinical or research purposes, both men and women should be asked to initiate a change conversation for maximum validity and generalizability.
Preparation of this article was supported by National Institute of Mental Health Grant R01MH57985. We would like to thank Tessa West for her statistical consultation on multi-level modeling. We also wish to acknowledge the monumental efforts of all those involved in this project, including Cheryl Van Dyke, Susan O’Leary, Bonnie Rainey, Camilo Ortiz, Michael Lorber, Debbie Leung, Evelyn Flaherty, Patti Fritz, Jeff Snarr, Mat Williams, and Heather Foran. We would also like to thank the hundreds of undergraduate research assistants who coded the videos, conducted the random digit dialing survey, and managed the data. Finally, we would like to thank the hundreds of families who volunteered to expose some of their most vulnerable issues for the long-term benefit of helping others.
1One can be both more positive and more negative than one’s partner because one’s partner can display more neutral behaviors.
2Couples had to be married and living together (n = 428, 94.5%) or cohabiting (n = 25, 5.5%) for at least one year to qualify.
3Dysphoric affect was not incorporated in the negative cluster because it is theoretically distinct from other negative behaviors (i.e., self vs. other focused, Biglan, Lewin, Hops, & Patterson, 1990). Because constructive problem discussion is categorized as a neutral code (and this was not of interest in the current study), it was not included in the positive or negative clusters.
4Effect sizes (Cohen’s d) were calculated for results using two methods. Largely, effect sizes were calculated by inputting the means and standard deviations for each of the two comparison groups into an effect size calculator retrieved from a website created and maintained by Lee Becker (http://web.uccs.edu/lbecker/Psy590/escalc3.htm). To calculate Cohen’s d for paired t-tests, the mean difference between the pair was divided by the standard deviation of the differences (Kotrlik & Williams, 2003).
5Conversations were counterbalanced to ensure that order effects would not be present. Still, data were analyzed using multi-level modeling to determine if the order of conversations affected positive and negative conversation behaviors. There were no significant results for positive behaviors. In terms of the proportion of negative behaviors, people were more negative during the second conversation than the first conversation, b = 0.005, t(452.22) = 2.14, p < .05. However, this did not interact with gender or who initiated the conversation.
Publisher's Disclaimer: The following manuscript is the final accepted manuscript. It has not been subjected to the final copyediting, fact-checking, and proofreading required for formal publication. It is not the definitive, publisher-authenticated version. The American Psychological Association and its Council of Editors disclaim any responsibility or liabilities for errors or omissions of this manuscript version, any version derived from this manuscript by NIH, or other third parties. The published version is available at www.apa.org/journals/fam.