The Big Five and enduring marriages

JOURNAL OF

RESEARCH IN PERSONALITY

Journal of Research in Personality 38 (2004) 481–504

www.elsevier.com/locate/jrp

The Big Five and enduring marriagesq M. Brent Donnellan,a,* Rand D. Conger,a and Chalandra M. Bryantb a

University of California, Davis, USA b Iowa State University, USA Available online 25 March 2004

Abstract The present investigation tested the relations between the Big Five dimensions of personality and the marital relationships of over 400 couples using both observational and questionnaire data. Four major findings emerged from these analyses. First, self-reports of neuroticism were positively correlated with negative interactions and negatively correlated with global evaluations of the marriage. Second, self-reports of agreeableness were negatively correlated with negative interactions and generally positively correlated with global evaluations of the marriage. Third, self-reports of openness were negatively correlated with observer reports of negative interactions. Fourth, self-reports of openness by wives were positively correlated with global reports of sexual satisfaction. These findings suggest that agreeableness and openness deserve increased attention as significant correlates of close relationships. Discussion emphasizes the importance of intrapersonal factors for understanding marital relationships. Ó 2004 Elsevier Inc. All rights reserved. Keywords: Big five; Personality; Marriage; Marital interaction; Close relationships

q

This research was supported by grants from the National Institute of Mental Health (Grants MH00567, MH19734, MH43270, MH48165, and MH51361), the National Institute on Drug Abuse (Grant DS05347), the Bureau of Maternal and Child Health (Grant MCJ-109572), the MacArthur Foundation Research Network on Successful Adolescent Development Among Youth in High-Risk Settings, and the Iowa Agriculture and Home Economics Experiment Station (Project No. 3320). Kali H. Trzesniewski and Dannelle Larsen-Rife provided helpful comments on an earlier draft. Xiaojia Ge and Richard W. Robins provided valuable background insight. Janet N. Melby graciously provided interobserver reliabilities. * Corresponding author. Present address: Department of Psychology, Michigan State University, East Lansing, MI 48823, USA. Fax: 1-517-432-2476. E-mail address: [email protected] (M.B. Donnellan). 0092-6566/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.jrp.2004.01.001

482

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

1. Introduction The present investigation explores whether or not personality characteristics influence observable marital interactions and global evaluations of the marriage. Our broad objective is to help integrate the intrapersonal perspective with the interpersonal perspective in the study of marital relationships. The intrapersonal perspective views personality dispositions as a major influence on marital functioning, whereas the interpersonal perspective views interactions between spouses as a major influence on marital functioning. Several scholars have noted that these perspectives are complementary and have advanced hypotheses linking personality traits to interpersonal interactions (e.g., Bradbury & Fincham, 1988; Caughlin, Huston, & Houts, 2000; Huston & Houts, 1998; Karney & Bradbury, 1995; Kelly & Conley, 1987). The common thread among these hypotheses is the proposition that personality traits influence how well couples interact with one another. For example, Karney and Bradbury (1995) propose that personality dispositions such as emotional instability or neuroticism create ‘‘enduring vulnerabilities’’ that affect how couples adapt to stressful experiences. This adaptation impacts overall relationship satisfaction. Similarly, Bradbury and Fincham (1988) argue that personality helps define aspects of the ‘‘distal context’’ that affect actual day-to-day marital interactions. Huston and Houts (1998) suggest that personality contributes to the ‘‘psychological infrastructure’’ of enduring relationships and are therefore key predictors of relationship success and/or dysfunction. These propositions all predict that marital interaction patterns are influenced by the personalities of the individuals in the relationship. Based on this theoretical perspective, we explicitly test the hypothesis that the links between the Big Five dimensions of personality (extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience) and global marital evaluations are mediated by negative marital interactions marked by high hostility and low warmth. Empirical support of this hypothesis will advance our understanding of the factors that foster dissatisfying marriages and help further integrate the study of marital interactions with one of the major organizing frameworks in personality psychology, the Big Five (e.g., Funder, 2001; John & Srivastava, 1999; but see Block, 1995, 2001). To our knowledge, few studies have attempted this sort of integration with multiple informant data.

2. A review of earlier research 2.1. Personality and relationship satisfaction In classic research, Terman and his colleagues (1938) used descriptors of neuroticism such as moody and irritable to describe unhappily married wives and husbands. A significant body of research linking neuroticism or negative emotionality to relationship satisfaction supports this insight (e.g., Bouchard, Lussier, & Sabourin, 1999; Botwin, Buss, & Shackelford, 1997; Caughlin et al., 2000; Eysenck & Wakefield, 1981; Karney & Bradbury, 1995, 1997; Kelly & Conley, 1987; Robins,

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

483

Caspi, & Moffitt, 2000, 2002; Watson, Hubbard, & Wiese, 2000). For instance, acquaintance-rated neuroticism was linked to marital satisfaction in a longitudinal study that spanned 50 years (Kelly & Conley, 1987) and Karney and Bradbury (1995) estimated that the impact of neuroticism on marital satisfaction using PearsonÕs r was ).19 for wives and ).13 for husbands in their meta-analysis. All told, the relation between neuroticism and relationship satisfaction has replicated across studies, methods, nations, and eras. Relatively less is known about how the other four factors relate to dimensions of marital relationships. This is because only a handful of studies have used all five factors to examine the relation between personality and measures of marital quality (e.g., Botwin et al., 1997; Bouchard et al., 1999; Watson et al., 2000). For example, Watson et al. (2000) found that agreeableness and conscientiousness were positively related to relationship satisfaction in dating couples while extraversion was positively related to satisfaction in married couples. In another study, spouse reports of general marital satisfaction were positively associated with partner reports of agreeableness and openness (Botwin et al., 1997). These same researchers demonstrated that wife reports of general marital satisfaction were associated with husband reports of conscientiousness. In summary, neuroticism has been linked to the quality of marital relationships in several studies and there are suggestions in the literature that many of the other Big Five factors are linked to close relationships. One goal of the present study is to examine the associations between all five personality factors and marital variables in a sample of enduring marriages using data from multiple informants. This design feature allows us to focus on cross-method associations (e.g., correlations between self-reports of personality and spouse reports of marital evaluations) to address the criticism that the associations between personality traits and martial variables are largely an artifact of shared method variance (e.g., Gottman, 1998). 2.2. Personality and marital interaction Over the last three decades, behavioral interaction research has accumulated an impressive body of evidence linking observed negative interactions between wives and husbands to marital dissatisfaction and divorce (for reviews see Gottman, 1994, 1998). This research suggests that unchecked and escalating hostile exchanges lacking in warmth are key predictors of marital dissatisfaction and dissolution. Therefore, one plausible way that personality dispositions influence global marital outcomes is by increasing or decreasing the frequency and/or intensity of these destructive interactions. Despite this conceptual understanding, few studies have examined this process directly with observational data. Caughlin et al. (2000) explored the relations between neuroticism and observed spousal negativity in a 13-year longitudinal investigation. Consistent with the above hypothesis, they concluded that much of the relation between neuroticism and marital satisfaction was attributable to negative communication patterns. Spouses higher in neuroticism were more likely to communicate their negative affect toward their partner, which in turn, had a deleterious impact on

484

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

relationship satisfaction. Few other studies have empirically linked Big Five traits other than neuroticism to observed marital interactions. Nonetheless, agreeableness may be an important personality predictor of marital functioning because there is evidence that this personality dimension is an important influence on interpersonal interactions (e.g., Graziano & Eisenberg, 1997; Graziano, Jensen-Campbell, & Hair, 1996; Tobin, Graziano, Vanman, & Tassinary, 2000). This literature suggests that more agreeable individuals are better able to regulate emotions during interpersonal interactions, which facilitates smoother interpersonal encounters. By extension, agreeable spouses may be better equipped to handle the conflicts that arise in marriages and this disposition may reduce the frequency and/or intensity of negative interactions. However, few studies have linked agreeableness to observable marital interactions or even to global evaluations of marriages. For example, Karney and Bradbury (1995) included only three studies linking agreeableness to marital satisfaction in their meta-analysis (the effect size using PearsonÕs r was .05 for wives and .03 for husbands). There may be links between the other Big Five traits and marital interactions, particularly conscientiousness and openness. Robins et al. (2000) speculated that individuals high in constraint (akin to conscientiousness) have the self-control to constructively manage the conflicts that inevitably arise in relationships. Individuals low in conscientiousness may inadvertently escalate negativity by responding impulsively or rashly to their partners. Moreover, conscientious spouses may simply evoke less spousal criticism, which in turn, may reduce the amount and/or intensity of negative marital interactions. Spouses high in openness may adopt a more intellectual approach to problem solving, may have a more flexible attitude towards change, and may be more willing to analyze their relationship. These behaviors and attitudes would likely facilitate the constructive management of conflict. It should be noted, however, that these links are speculative and need empirical support.

3. The present investigation Drawing on the research and the theoretical ideas just reviewed, the present research addresses three major questions: (a) How well do the Big Five predict observable negative marital interaction patterns? (b) How well do the Big Five predict marital quality and sexual satisfaction using cross-informant data (i.e., how well do self-reports of personality predict spousal evaluations of the relationship)? (c) Do marital interactions mediate the relation between personality and global evaluations of the marriage? This investigation extends previous research in several ways. First, we test the relations between the Big Five traits and marital variables with a conservative approach using multiple informant data. This approach is superior to investigations that exclusively use self-reports because it rules out the possibility that shared method variance is responsible for the associations between personality and relationship variables. Moreover, relatively few studies have linked personality traits to observed interactions in close relationships (Cooper & Sheldon, 2002). Second, our initial

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

485

sample of 418 couples provides increased statistical power to detect relations between constructs. Third, we use four waves of prospective, longitudinal data so we can test for possible linkages between a broad set of intrapersonal individual differences and changes in marital outcomes over time. 3.1. Method 3.1.1. Sample The couples in this investigation were participants in a larger project designed to study the impact of economic stress on rural families and children (IYFP: Iowa Youth and Families Project). The study began in 1989 with a sample of 451 European American, lower-middle and middle class families living in north central Iowa and has been described in detail elsewhere (e.g., Conger, Rueter, & Elder, 1999). Analyses reported here are based on an initial sample of 418 wives and husbands with complete personality scores and marital outcomes from the 1990 wave of data collection (in which 424 total families participated or approximately 94% of the initial sample). We analyzed relationship information collected from the 418 participants during the 1990, 1991, 1992, and 1994 waves of data collection. Slight attrition occurred at subsequent waves (e.g., 96% of the families from 1990 participated in 1991) which caused missing data at one or more waves. Moreover, 25 of the 418 couples experienced divorce and/or separation by 1994 and therefore had missing data at one or more waves of assessment. Couples that divorced or separated did not differ from the other couples in terms of their personality (all tÕs <2.0, p ¼ ns). Although previous studies have linked personality to divorce (e.g., Jocklin, McGue, & Lykken, 1996), the small number of divorces in this sample greatly diminished our statistical power to detect these effects. In 1990, the average age of wives was 38.75 years (SD ¼ 4.08, Min. ¼ 30, Max. ¼ 54) and the average age of husbands was 40.86 years (SD ¼ 4.84, Min. ¼ 32, Max. ¼ 69). The couples had been married for an average of 18.93 years (SD ¼ 3.09, Min. ¼ 13, Max. ¼ 36) and this was the first marriage for the vast majority of the sample (approximately 94%). The modal category for highest level of education completed for wives was a high school diploma or a G.E.D. with a range from ninth grade to a M.A./M.S. degree. The modal category for husbands was the same as wives but ranged from eighth grade to a terminal or professional degree (e.g., Ph.D., M.D., D.D.S.). The median income for the families in 1989 was $39,969 and ranged from below the poverty line to quite affluent. The average household size was 4.91 (SD ¼ .95). 3.1.2. Procedures The families in the IYFP were recruited in 1989 from all 34 public and private schools with a seventh grade class in communities of 6500 or less in eight Iowa counties. The original design was to include families with a target seventh grade adolescent, one sibling within 4 years of age of the target, and both parents. After receiving a letter explaining the project, families were contacted by telephone and invited to

486

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

participate. A personal visit was made to families without telephones. About 78% of the eligible families agreed to take part in the study. Participating family members were paid approximately $10/h of participation. In each year of the study, families were visited twice in their homes. During the first visit, each family member completed a series of questionnaires that included the marital evaluations. In 1990, participants completed the self-report measure of the Big Five personality traits. The second visit occurred within a fortnight of the first visit and consisted of a series of four video-taped interactions. The first two interactions involved the whole family and the third task involved the siblings. For the present study, we used the fourth task, a 25 min marital interaction observation, when spouses discussed several topics including the history of their relationship, the current status of their relationship, areas of agreement and disagreement in their relationship, and their future plans together. In 1990, a total of 414 couples in our sample participated in the observational research task. The video-taped marital interactions were rated by trained observers who used the Iowa Family Interaction Rating Scales (Melby & Conger, 2001) to measure aspects of couple interaction that were of theoretical importance and interest for the broad goals of the IYFP. All observers received 200 h of training (20 h peer week for 10 weeks) and passed extensive reliability tests before coding taped interactions. Once coders were reliable, they attended at least two ‘‘maintenance’’ training sessions each week to ensure continued reliability. To assess interobserver reliability, approximately 25% of all video-tapes were independently rated by a second observer at each wave. These independent ratings were used to generate intraclass correlations (Suen & Ary, 1989). A complete description of all rating and task procedures along with scale definitions is available from the second author. 3.1.3. Measures Personality. The NEO-FFI (Costa & McCrae, 1989) was used to measure personality in 1990. The NEO-FFI is 60-item self-report questionnaire that assess each of the five factors of personality (extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience) with 12 items. Wives and husbands rated each statement using a 5-point scale (1, Strongly agree and 5, Strongly Disagree) and the scales were coded such that higher scores reflected greater extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience. Table 1 displays descriptive statistics and scale inter-correlations for the NEO-FFI. We found a few statistically significant associations between the personalities of husbands and wives (p < :05). Specifically, spousal reports of neuroticism and openness were positively correlated (r ¼ :16 for neuroticism and r ¼ :17 for openness). There were no other statistically significant correlations between the two spouses on the same personality traits. An examination of the cross-trait associations revealed that husband reports of agreeableness and conscientiousness were both negatively correlated with wife reports of neuroticism (r ¼
:11 for husband reports of agreeableness and r ¼
12 for husband reports of conscientiousness). Husband reports of neuroticism were also negatively related to wife reports of openness (r ¼
:12).

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

487

Table 1 Descriptive information for the NEO-FFI for wives and husbands Mean

SD

a

Scale inter-correlations E

A

C

N

O

Wife self-report Extraversion (E) Agreeableness (A) Conscientiousness (C) Neuroticism (N) Openness (O)

3.35 3.81 3.74 2.68 3.08

.45 .32 .44 .56 .43

.79 .69 .83 .87 .72

— .29 .34 ).43 .28

— .23 ).40 .09

— ).38 .10

— ).20

—

Husband self-report Extraversion (E) Agreeableness (A) Conscientiousness (C) Neuroticism (N) Openness (O)

3.21 3.57 3.64 2.48 2.95

.40 .33 .41 .51 .35

.75 .69 .82 .87 .60

— .27 .35 ).38 .19

— .27 ).33 .02

— ).48 .09

— ).18

—

*

p<.05.

3.1.4. Negative marital interactions (high hostile/low warmth interactions) Observer reports. Trained observers rated wives and husbands on high hostile/low warmth behaviors expressed to one another during the marital observation task at each wave using a 9-point scale (1, the behavior is not at all characteristic of the individual and 9, the behavior is very characteristic of the individual). Seven rating scales were aggregated at each wave to form the observer rating of negative interactions: angry coercion, antisocial behavior, hostility, communication quality (reverse scored), listener responsiveness (reverse scored), prosocial behavior (reverse scored), and warmth (reverse scored). The average of the intraclass correlations at each wave is reported in parentheses (wife to husband/husband to wife). Angry coercion measured control attempts that included threats or coercion (.64/.59). Antisocial behavior measured the amount of defiant, resistant, and insensitive behavior (.64/.63). Hostility measured the amount of hostile and critical behavior (.78/.76). Communication quality measured the ability of the target spouse to clearly and appropriately express her/his feelings and points of view to her/his partner (.54/.51). Listener responsiveness measured the target spouseÕs verbal and non-verbal behaviors that indicated attentiveness to her/his partner (.54/.61). Prosocial behavior measured helpfulness and sensitivity (.44/.50). Warmth measured expressions of care, concern, and support (.59/.67). The observer ratings of negative marital interactions had adequate internal consistency for both wives and husbands at each wave (a range for wife as target ¼ .79–.84; a range for husband as target ¼ .77–.79). Negative interactions exhibited a large degree of rank-order consistency across time (for wives: e.g., r between 1990 and 1991 ¼ .55, r between 1991 and 1992 ¼ .55, and r between 1992 and 1994 ¼ .54; for husbands: the same coefficients were .48, .57, and .56, respectively). These correlations suggest that stability in negative interactions was an objective property of the relationship, evident to outside observers. Descriptive statistics for all marital variables are reported in Table 2.

488

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

Table 2 Descriptive statistics for relationship variables Wife

Husband

Correlation between wife and husband measures

Observer reports: negative interactions 1990 4.21 (1.11) 1991 4.62 (1.24) 1992 4.25 (1.32) 1994 4.73 (1.26)

4.17 4.60 4.26 4.61

(.96) (1.16) (1.12) (1.25)

.58 .62 .55 .63

Questionnaire reports: negative interactions Self-reports 1990 2.77 (.84) 1991 2.81 (.91) 1992 2.76 (.86) 1994 2.51 (.72)

2.81 2.78 2.77 2.52

(.74) (.78) (.74) (.67)

.50 .48 .47 .47

Partner reports 1990 1991 1992 1994

2.55 2.60 2.60 2.54

(.97) (1.10) (1.02) (.87)

.51 .54 .52 .48

3.61 3.52 3.54 3.36

(.66) (.70) (.68) (.66)

.62 .60 .59 .60

.00 .00 .00 .00

(.95) (.91) (.89) (.89)

.50 .52 .43 .48

2.70 2.70 2.70 2.66

(.96) (.98) (.99) (.88)

Global evaluations: sexual satisfaction 1990 3.64 (.75) 1991 3.56 (.76) 1992 3.55 (.76) 1994 3.36 (.72) Global evaluations: marital quality 1990 .00 (.96) 1991 .00 (.95) 1992 .00 (.92) 1994 .00 (.94)

Note. All correlations are significant at p < :05.

Questionnaire reports. At each wave, wives and husbands were asked how often their spouses engaged in eight behaviors during the last month using a 7-point scale (1, Always and 7, Never). Sample items include, ‘‘Shout or yell at you because she (he) was mad at you’’ and ‘‘Act loving and affectionate toward him (her).’’ These behaviors reflected the occurrence of interactions that were either high in hostility (e.g., getting angry) or low in warmth (e.g., infrequently acting loving and affectionate). This measure was coded so that higher scores reflected a greater frequency of high hostile/low warmth interactions. Partner reports by both wives and husbands had good internal consistency at each wave (a range for wife reports of husband behavior ¼ .89–.92; a range for husband reports of wife behavior ¼ .91–.92). Partner reports were correlated with observer reports (e.g., r ¼ :33 for wife target behavior in 1991 and r ¼ :41 for wife target behavior in 1992; r ¼ :35 for husband target behavior in 1991 and r ¼ :40 for husband target behavior in 1992). Wives and husbands answered the same set of questions regarding their own behaviors using the same 7-point scale. This measure was coded so that higher scores

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

489

reflected a greater frequency of high hostile/low warmth behaviors. Both wife and husband self-reports had good internal consistency at each wave (a range for wife reports ¼ .88–.90; a range for husband reports ¼ .85–.88). Self-reports and partner reports of the same target behaviors were significantly related indicating strong inter-reporter agreement (e.g., r ¼ :56 for wife target behavior in 1991 and r ¼ :54 for wife target behavior in 1992; r ¼ :52 for husband target behavior for both 1991 and 1992). Self-reports were also related to observer reports (e.g., r ¼ :36 for wife target behavior in 1991 and r ¼ :45 for wife target behavior in 1992; r ¼ :38 for husband target behavior for both 1991 and 1992). Questionnaire responses of negative interactions demonstrated a substantial amount of rank-order stability (for wife self-reports: e.g., r between 1990 and 1991 ¼ .77, r between 1991 and 1992 ¼ .82, and r between 1992 and 1994 ¼ .77; for husbands self-reports: the same coefficients were .78, .80, and .78, respectively). Partner reports exhibited a similar level of rank-order stability as self-reports (for wife reports of husband behavior: e.g., r between 1990 and 1991 ¼ .78, r between 1991 and 1992 ¼ .80, and r between 1992 and 1994 ¼ .74; for husband reports of wife behaviors: the same coefficients were .78, .83, and .79, respectively). Considering these retest correlations in conjunction with the retest correlations for observer reports suggests that the interaction patterns we measured were a relatively stable aspect of the marital relationships of our sample. 3.1.5. Global assessments of the marital relationship Marital quality. Marital quality was measured using 2 items that tapped the spousesÕ overall feelings of happiness and satisfaction with the marriage. Spouses were asked to respond to how happy they were with their marital relationship using a 6-point scale (0, extremely unhappy and 5, extremely happy) and to respond to how dissatisfied they were with their relationship using a 5-point scale (1, completely satisfied and 5, not at all satisfied). The question concerning dissatisfaction was reverse coded so that higher scores reflected greater satisfaction. Both questions were standardized and then averaged together to provide an index of each spouseÕs overall assessment of marital quality. This was defendable because the two items were strongly correlated at each wave (rs for wives ranged from .68 to .84; rs for husbands ranged from .57 to .78). Marital quality demonstrated rank-order consistency across time (for wives: e.g., r between 1990 and 1991 ¼ .68, r between 1991 and 1992 ¼ .65, and r between 1992 and 1994 ¼ .61; for husbands: the same coefficients were .70, .66, and .54, respectively). These retest correlations were consistent with previous research indicating stability in marital outcomes (e.g., Johnson, Amoloza, & Booth, 1992; Johnson & Booth, 1988). Marital sexual satisfaction. Sexual satisfaction was measured using an 8-item scale specifically developed for the IYFP. Spouses were asked how much they agreed or disagreed with a given statement about the quality of the sexual aspects of their marriage using a 5-point scale (1, strongly agree and 5, strongly disagree). This measure was scored so that higher scores reflected greater sexual satisfaction. Sample items include, ‘‘The sexual part of my marriage is just right for me’’ and ‘‘My husband (or wife) and I have a wonderful sex life.’’ This measure had good internal consis-

490

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

tency for wives and husbands at each wave (a range for wife report ¼ .80–.89; a range for husband report ¼ .72–.87). Similar to the other relationship measures for this sample, wife and husband reports of marital sexual satisfaction were relatively stable across time (for wives: e.g., r between 1990 and 1991 ¼ .79, r between 1991 and 1992 ¼ .80, and r between 1992 and 1994 ¼ .77; for husbands: the same coefficients were .76, .81, and .74, respectively). 3.2. Results 3.2.1. Overview We first conducted our correlational and mediational analyses on the 1990 assessment to test our hypotheses about the associations between personality traits and marital variables. These results are reported and discussed below. We repeated these same analyses with the average of each construct across the four-waves of data (see Huston & Houts, 1998 for a similar analytic strategy) and the results were quite similar to those reported and discussed in this paper. In subsequent sections we report our attempts at longitudinal analyses using personality to predict changes in marital variables over time. Relationship measures for wives and husbands were correlated as seen in Table 2. For example, correlations between wife and husband reports of marital quality ranged from .43 to .52. These represent moderate to strong effect sizes and are quite reasonable because both partners are reporting about their evaluations of the same marriage. Accordingly, these variables should be treated cautiously in data analyses given their strong associations (e.g., Kashy & Snyder, 1995; Kenny, 1998). Given this interdependence between wife and husband reports of the marital variables, we also conducted analyses on the couple average of the marital variables (following KennyÕs, 1998 simple solution to the interdependence problem). We computed a couple average for questionnaire reports of negative interactions by taking the average of all four measures (see Tables 3 and 4). 3.2.2. Cross-sectional analyses 3.2.2.1. Does personality predict negative marital interactions and relationship evaluations?. Wife and husband reports of the Big Five in 1990 were correlated with relationship variables in 1990 as shown in Tables 3 and 4. We emphasize crossinformant associations (e.g., wife reports of personality and husband reports of his and her behavior) that were statistically significant (p < :05) because these associations are not as subject to concerns over shared method variance. We followed Cohen’s (1988) often cited rule of thumb for interpreting correlations: an r of .1 was considered small, an r of .3 was considered moderate, and an r of .5 or larger was considered large (see McCartney & Rosenthal, 2000). Negative interactions. Observed negative interactions by wives were associated with wife reports of agreeableness, neuroticism, and openness (r ¼
:13, r ¼ :17, and r ¼
:18, respectively) and husband reports of neuroticism (r ¼ :11). Observed negative interactions by husbands were associated with wife reports of openness

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

491

Table 3 Cross-sectional correlations between wife reports of personality and relationship variables Wife report of personality E

A

C

Observer reports: negative interactions Wife target .01 Husband target .07 Couple average .04

).13 ).06 ).11

).03 ).08 ).06

.17 .09 .15

).18 ).12 ).17

Questionnaire reports: negative Wife self-reports Report on wife by husband Husband self-reports Report on husband by wife Couple average

).31 ).17 ).14 ).16 ).23

).24 ).04 ).11 ).23 ).19

.37 .22 .21 .29 .33

).20 ).12 ).12 ).15 ).18

Global evaluations: sexual satisfaction Wife self-reports .15 Husband self-reports .03 Couple average .11

.17 .10 .15

.16 .02 .10

).24 .17 ).23

.21 ).17 .22

Global evaluations: marital quality Wife self-reports .11 Husband self-reports .07 Couple average .10

.11 .13 .14

.15 .07 .13

).23 ).20 ).25

.01 .04 .03

interactions ).22 ).09 ).04 ).10 ).14

N

O

Note. E, extraversion; A, agreeableness; C, conscientiousness; N, neuroticism; O, openness to experience. Coefficients in bold are significant cross-informant relations. * p < :05.

(r ¼
:12) and husband reports of agreeableness and openness (r ¼
:26 and r ¼
:13). Similarly, wife and husband reports of agreeableness, neuroticism, and openness were related to observed negative interactions at the couple level. Although these correlations were relatively small in magnitude, observational data provides a very rigorous test of the impact of broad personality traits on relationships. Moreover, self-reports of agreeableness, conscientiousness, and neuroticism were related to cross-informant questionnaire assessments of negative marital interactions (e.g., wife reports of conscientiousness predicted husband reports of his own negative interactions). Thus, it appears that agreeableness, conscientiousness, neuroticism, and openness are linked with behaviors related to the marriage and that shared method variance is not a viable alternative explanation for these relations. Global assessments of the relationship. Wife reports of agreeableness, neuroticism, and openness were associated with husband reports of sexual satisfaction (r ¼ :10, r ¼
:17, and r ¼ :17, respectively), whereas wife reports of agreeableness and neuroticism were correlated with husband reports of marital quality (r ¼ :13 and r ¼
:20). Husband reports of agreeableness, conscientiousness, and neuroticism were correlated with wife reports of sexual satisfaction (r ¼ :10, r ¼ :15, and r ¼
:15, respectively) whereas husband reports of conscientiousness and neuroticism were correlated with wife reports of marital quality (r ¼ :14 and r ¼
:13).

492

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

Table 4 Cross-sectional correlations between husband reports of personality and relationship variables Husband report of personality E

A

C

Observer reports: negative interactions Wife Target .03 Husband Target ).07 Couple Average ).02

).09 ).26 ).19

).08 ).03 ).06

.11 .08 .11

).08 ).13 ).11

Questionnaire reports: negative Wife self-reports Report on wife by husband Husband self-reports Report on husband by wife Couple average

).10 ).23 .29 ).14 ).22

).12 ).19 ).27 ).08 ).19

.11 .18 .29 .12 .21

).09 ).01 ).09 ).04 ).07

Global evaluations: sexual satisfaction Wife self-reports .04 Husband self-reports .10 Couple average .07

.10 .18 .16

.15 .21 .20

).15 ).20 ).19

.01 .06 .03

Global evaluations: marital quality Wife self-reports .05 Husband self-reports .18 Couple average .14

.10 .23 .19

.14 .19 .19

).13 ).23 ).21

.01 .02 .02

interactions ).02 ).09 ).19 ).03 ).09

N

O

Note. E, extraversion; A, agreeableness; C, conscientiousness; N, neuroticism; O, openness to experience. Coefficients in bold are significant cross-informant relations. * p<.05.

These cross-method results indicate that self-reports of personality predict spousal evaluations of the relationship. Similar results were evident when global evaluations of the relationship at the couple level were considered. 3.2.2.2. Do negative interactions mediate the relation between personality and global assessments of the marriage?. We correlated observed negative interactions in 1990 with relationship variables in 1990 to determine whether or not there was evidence for a link between our proposed mediator and relationship outcomes. It makes little sense to proceed with tests of mediation if there is no evidence for this link. Observer reports of the couple average of negative interactions were negatively correlated with the couple average of sexual satisfaction (r ¼
:32) and negatively correlated with the couple average of marital quality (r ¼
:39). Although we considered the couple level associations, the cross-partner associations were also statistically significant and were of a similar magnitude (e.g., the correlation between observer reports of wife negative interactions and husband reports of marital quality was ).29). These results indicate that observed negative interactions were associated with negative global evaluations of the relationship. Based on this evidence, we tested the hypothesis that the impact of broad personality traits on global evaluations of the relationship was mediated by observed negative interactions.

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

493

Overview of our approach to mediation. To address our research questions we followed the ‘‘causal steps’’ procedure identified by Baron and Kenny (1986: see also Holmbeck, 1997; MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002; Shrout & Bolger, 2002). We used the couple average of the marital variables to obviate concerns over the lack of independence between husband and wife reports. In the first step, a marital outcome was regressed on a given personality trait which provided an estimate of the direct effect of personality on global marital outcomes. The second step involved regressing negative interactions on personality, which provided evidence for a link between personality traits and the proposed mediator. At the third step, the marital outcome in question was regressed on personality traits and negative interactions. This step provided evidence linking negative interactions to marital outcomes and provided information concerning how controlling for negative interactions affected the relation between a given personality trait and a given marital outcome. We were interested in drawing one of two commonly drawn conclusions from these analyses, either that our data were consistent with complete mediation or partial mediation. Based on current conventions, a conclusion of complete mediation is typically drawn if there is no evidence that the coefficient for personality in Step 3 is different from zero (e.g., the confidence interval for the coefficient includes zero or the significance test is not statistically significant). A conclusion of partial mediation is typically drawn if there is simply evidence that the coefficient for personality in Step 3 is reduced from its value in Step 1. We did not expect to find consistent evidence for complete mediation because this situation is a rather atypical in psychological research (e.g., Baron & Kenny, 1986; Holmbeck, 1997; Shrout & Bolger, 2002). We expected to find evidence for partial mediation because we explicitly recognized that the ‘‘true’’ causal process linking broad personality traits to global evaluations of the relationship is likely to be complicated. It may also involve mechanisms not explicitly tested in this paper. Shrout and Bolger (2002) provide an extended explanation (pp. 432–434) of the various reasons why researchers often find evidence for partial mediation in their analyses. Indeed, we readily acknowledge that our process model linking personality traits to marital outcomes via observed martial interactions is literally false (e.g., MacCallum, 2003). At best, our model is a useful approximation of a complicated process. Given this situation, it would be unreasonable to expect to find widespread evidence of complete mediation. It is important to emphasize that the procedures and logic of mediational analyses are still evolving (e.g., Collins, Graham, & Flaherty, 1998; MacKinnon et al., 2002; Shrout & Bolger, 2002) and our results should be interpreted in light of a couple of these issues. First, MacKinnon et al. (2002) reviewed the various approaches to mediation and concluded that the Baron and Kenny (1986) approach has low Type I error rates but consequently it has relatively low statistical power. The technique we employed is thus very conservative. Although we acknowledged the increased risk of Type II errors with this method, we believed that taking such a conservative approach was important in light of the critical stance adopted by some researchers concerning the importance of personality traits for understanding marital functioning. However, the observed pattern of correlations was consistent with our underlying

494

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

process model: personality traits were linked to observed marital interactions and observed marital interactions were linked with marital outcomes. A second issue concerning our approach to mediation is also worth noting. Specifically, there is discussion concerning the importance of Step 1 for mediation (e.g., Collins et al., 1998; Shrout & Bolger, 2002). Holmbeck (1997, p. 602) pointed out that an indirect effect between personality and marital outcomes is still possible even if there is no evidence linking particular personality traits directly to marital outcomes. Shrout and Bolger (2002) recommend that Step 1 is ‘‘retained [only] in specific cases in which theory suggests that the associations are large or medium in size’’ (p. 430). We expected that effect sizes concerning the impact of broad personality traits on marriages would tend to be small to medium, both in terms of theory (e.g., the notion that personality helps form the ‘‘distal context’’ of the relationship: Bradbury & Fincham, 1988) and on previous results (e.g., Karney & Bradbury, 1995). Therefore, it was unclear that satisfying Step 1 was crucial or even reasonable for conducting our mediational analyses. Nevertheless, in the interest of taking a conservative approach, we followed Baron and KennyÕs (1986) original formulation and restricted our analyses to those cases where the correlational analysis provided evidence for a link between a particular personality trait and the couple average of the marital evaluations (e.g., wife reports of neuroticism and the couple average of marital quality). This decision also had the practical benefit of reducing the total number of mediational analyses we conducted. Accordingly, the criticism that our results capitalized on chance is less likely to be valid. We took a final step to ensure a very strict test of our mediational hypothesis. We used observer reports of martial interactions to obviate concerns surrounding inflated associations due to shared method variance. Considering these steps as a whole, we employed a relatively strict test of our mediational hypothesis. Our view was that if our hypothesis survives this rigorous of a test, then we would be in a better position to endorse the somewhat controversial notion that knowing something about each partnerÕs broad personality traits is useful for understanding their behavior in romantic relationships (see Gottman, 1998 and Robins et al., 2000 for a further discussion of this issue). Results of tests of mediation. Table 5 displays the results for wife reports of personality. Table 5 displays results from Step 3 of the regression analyses given that the results for Steps 1 and 2 are redundant with information already presented in Table 3 (i.e., links between personality and global evaluations and links between personality and negative interactions). We found support for the hypothesis that negative martial interactions partially mediate the association between the personality traits of wives and global evaluations of the marriage. All indirect effects were statistically significant at p < :05 based on Sobel tests (all values were greater than 2.03). Following Shrout and Bolger (2002) we computed the ratio of the indirect effect over the total effect for each mediational analysis. These ratios are displayed in Table 5 and ranged from .17 to .29. Table 6 displays the results for husband reports of personality. We found support for the hypothesis that the negative martial interactions partially mediate the associ-

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

495

Table 5 Tests of mediation for wife reports of personality Wife report of personality

Agreeableness

Couple average of the marital evaluation

Sexual satisfaction

Neuroticism Marital quality

Sexual satisfaction

Openness Marital quality

Sexual satisfaction

Step 3: predict DV from personality controlling for couple average of negative interactions b for personality trait .13 .11 ).19 ).20 .17 b for observed negative ).31 ).38 ).29 ).36 ).29 Interactions .12 .17 .14 .19 .13 R2 Assessment of the strength of mediation Indirect effect .03 Total effect .15 Ratio of the indirect effect .20 to the total effect Conclusion

Partial mediation

.04 .14 .29 Partial mediation

).04 ).23 .17

).05 ).25 .20

Partial mediation

Partial mediation

.05 .22 .23 Partial mediation

Note. All coefficients statistically significant at p < :05 unless otherwise noted.

Table 6 Tests of mediation for husband reports of personality Husband report of Personality

Agreeableness

Couple average of the marital evaluation

Sexual satisfaction

Neuroticism Marital quality

Sexual satisfaction

Marital quality

Step 3: predict DV from personality controlling for couple average of negative interactions b for personality trait .10 .13 ).16 ).17 b for observed negative interactions ).30 ).37 ).30 ).37 R2 .11 .17 .13 .18 Assessment of the strength of mediation Indirect effect Total effect Ratio of the indirect effect to the total effect Conclusion

.06 .16 .38 Partial mediation

.07 .19 .37 Partial mediation

).03 ).19 .16

).04 ).21 .19

Partial mediation

Partial mediation

Note. All coefficients statistically significant at p < :05 unless otherwise noted.

ation between the personality of husbands and global evaluations of the marriage. The indirect effects were all statistically significant at p < :05 (Sobel test values greater than 2.16) and the ratios of the indirect effects to the total effects ranged from .16 to .38. Ancillary analyses. We repeated our analyses using different endogenous variables to address concerns with shared method variance. For example, we used wife reports of personality to predict observer reports of her negative interactions and husband

496

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

reports of marital outcomes. The results for wife reports of personality were similar to those reported in Table 5 except that we found evidence for complete mediation in one instance. It appeared that the effect of wife reports of agreeableness on husband reports of sexual satisfaction was completely mediated by observer reports of her negative interactions (b for agreeableness controlling for negative interactions ¼ .08, p ¼ ns). We also used the same strategy for husband reports of personality and only needed to conduct the analysis for husband reports of agreeableness and wife reports of sexual satisfaction given the correlational findings. It appeared that this effect was fully mediated by observer reports of his negative interactions (b for agreeableness controlling for negative interactions ¼ .05, p ¼ ns). 3.2.3. Longitudinal analyses Overview. We explored the possibility that personality was related to changes in marital interactions and marital evaluations using two approaches. First, we conducted a set of growth models in the structural equation modeling (SEM) framework (Duncan, Duncan, Strycker, Li, & Alpert, 1999; Singer & Willett, 2003; Willet & Sayer, 1994). Growth modeling is one of the most highly recommended procedures for studying change (e.g., Rogosa, Brandt, & Zimowski, 1982; Singer & Willett, 2003; Willett, Singer, & Martin, 1998). Second, we conducted a more traditional regression-based approach to longitudinal data by regressing each of the 1994 marital variables on the respective 1990 marital variables and reports of personality. Growth curve modeling approach. We followed a two-step procedure for these analyses drawing on recommendations found in Singer and Willet (2003; see also Muthen & Curran, 1997 for similar advice). In Step 1, we estimated basic growth models for each of the marital variables. We introduced personality as a predictor of the growth curve parameters in Step 2 (e.g., the intercept and slope parameters in a linear change model) provided that the model in Step 1 met three criteria. First, the model in Step 1 had to produce sensible parameter estimates. Models that produced Heywood cases were naturally discarded. Lorenz, Wickrama, and Conger (2004) suggest that Heywood cases may indicate an absence of systematic change over time. Second, there needed to be statistically significant estimates of variances associated with ‘‘change’’ parameters (e.g., the slope parameter in a linear model). This condition is reasonable given that it makes little sense to try to examine predictors of change when there is little variability to predict. Given that we already demonstrated how personality was related to initial status in martial variables (i.e., the 1990 assessment) in Tables 3 and 4, we knew it would be redundant to focus on personality as a predictor of variability around the intercept. Finally, the growth models in Step 1 had to adequately fit the data (e.g., the covariances and means reproduced by the statistical model had to be ‘‘reasonably close’’ to their values in the original data set). To be sure, we relied on very liberal estimates of overall model fit such as RMSEA values around .10 and/or v2 /degree of freedom ratios of 3.0 or less (see Byrne, 2001). We specified the growth models so that the intercept represented values of the marital variables in 1990. In all analyses, we began with a linear model of change based on previous research (e.g., Karney & Bradbury, 1997; Kurdek, 1998).

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

497

We elected to modify the linear models if they had an overall poor fit to the data by correlating adjacent error terms in the basic linear model (Muthen & Curran, 1997) and/or by modeling a quadratic growth term. It should be noted that we were limited to relatively simple growth functions because we had only four data points (e.g., cubic models could not be evaluated). Separate analyses were conducted on the ‘‘happy’’ and ‘‘satisfaction’’ items of the marital quality measure because standardized scores are not suitable for growth modeling (Stoolmiller, 1995). Summary. Unfortunately, this approach produced disappointing results.1 It was difficult to get a good fitting univariate growth models in the first steps of our analyses and even when we relaxed our standards we found very inconsistent results. When personality traits significantly predicted parameters in the growth models, it was generally to the intercept parameters thereby duplicating the results displayed in Tables 3 and 4. There were only two cases where personality predicted the change parameters in growth models. Husband reports of extraversion predicted the slope and quadratic term in a growth model for sexual satisfaction that included a quadratic term (b predicting slope term ¼ .16 and b predicting the quadratic term ¼ ).18, ps < :05; Overall fit: v2 ¼ 10:73, df ¼ 2, p < :05; v2 =df ¼ 5:37; CFI ¼ 0.99; RMSEA ¼ 0.10, p close fit ¼ .06). Wife reports of neuroticism predicted the slope in a growth model for the questionnaire measures of negative interactions (b ¼ :
27, p < :05; Overall fit: v2 ¼ 61:15, df ¼ 7, p < :05; v2 =df ¼ 8:74; CFI ¼ 0.97; RMSEA ¼ 0.14, p close fit <.05). However, we were not very confident in these results given the relatively poor fit of the models and the seemingly paradoxical effects for extraversion. These paradoxical effects make a little more sense when we note that the slope and the quadratic terms were highly negatively correlated (r ¼
:95). Moreover, the results were essentially identical when we repeated our analyses using log transformed variables to address any concerns over non-normality in the marital variables. The one exception was that husband reports of openness predicted the slope in the basic linear model for marital happiness in a relatively poor fitting model (b ¼
:20, p < :05; Overall fit: v2 ¼ 56:54, df ¼ 7, p < :05; v2 =df ¼ 8:08; CFI ¼ 0.90; RMSEA ¼ 0.13, p close fit <.05). All told, the results for the growth models using SEM were not particularly illuminating. Regression approach. As a second approach, we regressed the couple average of each of the 1994 marital variable on personality traits and the respective martial outcome in 1990. We conducted separate models for husband and for wife reports of each personality trait predicting martial variables (observer reports of negative interactions, questionnaire reports of negative interactions, sexual satisfaction, and martial quality). There were three cases where personality traits predicted martial outcomes in 1994 controlling for the same outcome in 1990. Wife reports of agreeableness predicted observer reports of negative interactions in 1994 controlling for 1990 levels (b ¼
:10, p < :05) and husband reports of agreeableness predicted observer reports of negative interactions in 1994 controlling for 1990 levels (b ¼
:11, p < :05). These results suggest that agreeableness may predict relative

1

More extensive details are available from the first author.

498

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

changes in observed negative interactions. In addition, wife reports of agreeableness predicted the couple average in marital quality in 1994 controlling for levels of marital quality in 1990 (b ¼ :09, p < :05). However, the results for agreeableness were not robust across marital outcomes and were thus interpreted with sufficient caution.

4. Discussion We conducted a very methodologically conservative test of the associations between personality and relationships by using multiple informant data. Four major findings emerged from our data analyses. First, self-reports of neuroticism were positively correlated with negative interactions and negatively correlated with global evaluations of the marriage. Second, self-reports of agreeableness were negatively correlated with negative interactions and generally positively correlated with global evaluations of the marriage. Third, self-reports of openness were negatively correlated with observer reports of negative interactions. Fourth, self-reports of openness by wives were positively correlated with global reports of sexual satisfaction for both wives and husbands. In general, we found evidence linking self-reports of personality to observed marital interactions suggesting that negative marital interactions are partially rooted in the individual personalities of wives and husbands. Moreover, examining significant cross-informant correlations suggests that there is more to the links between personality and relationships than shared method variance (see also Watson et al., 2000). A specific goal of our investigation was to test a hypothesis derived from theoretical propositions (e.g., Bradbury & Fincham, 1988; Huston & Houts, 1998; Karney & Bradbury, 1995) that negative marital interactions mediate the relations between individual differences in personality and global evaluations of the marriage. We used a conservative strategy to conduct these analyses by using observer reports of marital interactions to help control for shared method variance. Results indicated that negative interactions partially mediate the relations between wife and husband reports of agreeableness and global evaluations of the relationship. Similarly, negative interactions partially mediate the relations between wife and husband reports of neuroticism and global evaluations of the relationship. We now turn to a few comments relative to each personality trait. Extraversion. We found little evidence linking self-reports of extraversion to observer and spousal reports of the relationship. We did, however, find a few significant associations between self-reports of extraversion and self-reports of marital variables. Unfortunately, these associations could be attributed to shared method variance. Moreover, the effect sizes we obtained were much smaller than the correlations reported by Watson et al. (2000). This was somewhat surprising because both investigations employed the NEO-FFI. Indeed, our results were more similar to Botwin et al. (1997) who reported rather negligible associations between extraversion and marital outcomes using a different measure. Tobin et al. (2000) noted that extraversion captures individual differences relating to social impact and this dimension of personality may not be such an important trait in shaping the qualities of marriages

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

499

and other intimate relationships. At any rate, given empirical inconsistencies and null results, future work is needed to generate a clearer picture of the role of extraversion in close relationships. Agreeableness. Agreeableness negatively predicts observable negative behavior to spouses during a marital observation task and these associations replicated across questionnaire assessments of negative behavior. It appears that agreeableness is an important intrapersonal influence on martial functioning and these findings contribute to the growing literature on the role of agreeableness in interpersonal relationships (e.g., Graziano et al., 1996; Tobin et al., 2000) and marriages (Botwin et al., 1997; Bouchard et al., 1999; Watson et al., 2000). Based on our results and others, we believe that agreeableness deserves increased attention as a significant predictor of close relationships. Indeed, agreeableness may be as important as neuroticism for understanding successful or distressed romantic relationships. Although more research is needed to explicate the processes linking agreeableness to marriages, it appears that agreeable people are less likely to engage in specific interactional behaviors that are deleterious for close relationships. We also found that agreeableness predicted relative declines in observed negative interactions from 1990 to 1994. Conscientiousness. We found that self-reports of conscientiousness were related to spousal reports of negative interactions and to global assessments of the relationship. In general, it appeared that husband reports of conscientiousness were a more consistent predictor of spousal global evaluations of the relationship, a finding consistent with evolutionary predictions (e.g., Botwin et al., 1997; Robins et al., 2000). However, future work is needed to explicitly test the mechanism that links conscientiousness to relationships. It is possible that conscientious individuals create fewer areas of disagreement because they are generally responsible, dependable, and hardworking. The link between conscientiousness and marital outcomes may also be mediated by processes such as a more equitable household division of labor or a reduction in problem behavior involvement such as drug/alcohol abuse. Future studies are needed to test these speculations. Neuroticism. Our results linking neuroticism and marital evaluations were consistent with previous research, both classic (e.g., Terman, 1938) and more recent (e.g., Bouchard et al., 1999; Botwin et al., 1997; Caughlin et al., 2000; Eysenck & Wakefield, 1981; Karney & Bradbury, 1995, 1997; Kelly & Conley, 1987; Robins et al., 2000, 2002; Watson et al., 2000). Neurotic wives and husbands are involved in less positively evaluated relationships and this effect holds for cross-informant correlations (e.g., wife reports of neuroticism and husband reports of martial outcomes). Importantly, we also found that neuroticism is linked with observer reports of negative interactions in the relationship. These findings linking neuroticism to observed marital interactions may help explain how neuroticism is linked to marital evaluations. Our findings also generally replicate the results of Caughlin et al. (2000) who linked neuroticism to communication negativity. It appears that one pathway whereby this broad personality dimension affects relationships is by its connection with day-to-day negative interactions involving both words and deeds. We believe that our results are also consistent with the theoretical orientation advanced by Caughlin et al. (2000) that neuroticism influ-

500

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

ences the enduring dynamics of the marriage. In addition, our results also parallel the findings of Bolger and Schilling (1991) linking neuroticism to interpersonal conflicts and general distress in daily lives. All told, the strategy of linking neuroticism to interpersonal processes appears to be a promising direction for future research. Openness to experience. Openness was negatively correlated with observed negative interactions. Moreover, the association between wife-reported openness and negative interactions and sexual satisfaction was one of the most interesting and perhaps unexpected results of our investigation. We should note, however, that Botwin et al. (1997) reported significant associations between openness and several marital outcomes. Unfortunately, these relations are a bit tricky to interpret because openness to experience is purported to be the most difficult of the five factors to understand (McCrae & Costa, 1997). We offer two speculations for our findings. First, greater openness may entail a more intellectual and flexible approach to problem solving which may facilitate a proactive and intellectual approach to the conflicts and disagreements that arise in relationships. This may explain the negative association between openness and negative interactions. Second, openness may be related to sexual satisfaction by a somewhat different process. One aspect of openness involves the motivation to acquire new experiences and the positive correlation between openness and scales measuring experience seeking supports this interpretation (McCrae & Costa, 1997). This motivation may extend to the exploration of sexuality by wives. Sexual exploration on the part of wives could translate to greater sexual satisfaction in the marriage. Although these interpretations are highly speculative and the basic finding requires replication, we believe that the intriguing relations between openness and close relationships warrants further examination. Our speculations also echo the call for more research concerning the influence of openness on family life by McCrae and Costa (1997).

5. Personality and changes in relationships We found little evidence linking personality to changes in marital outcomes over time using growth curve methodology. Few other studies using this technique have found that the personality traits captured by the Big Five predict changes in marital variables. For example, Karney and Bradbury (1997) used growth curve modeling to study the impact of neuroticism on marital satisfaction in a sample of newlywed couples followed over 4 years. They found that neuroticism predicted lower initial levels of marital satisfaction but did not predict changes in marital satisfaction. Apparently, neuroticism exerts a relatively constant effect on relationship satisfaction over several waves of assessments. Caughlin et al. (2000) reached similar conclusions using structural equation modeling. We have tentatively interpreted these findings as consistent with the enduring dynamics model of marriages (e.g., Huston & Houts, 1998). Personality traits shape the psychological infrastructure of the marriage from very early on in the relationship and this dynamic then persists as a relatively enduring aspect of the relationship. Indeed, our marital relationship measures demonstrated a high degree of both rank-

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

501

order and absolute stability which is consistent with this model. In fact, most measures of relationship satisfaction demonstrate relatively high levels of rank-order stability in continuing relationships (e.g., Johnson et al., 1992; Johnson & Booth, 1988). Based on these observations, Johnson et al. (1992) suggested that the stability of marital quality, as assessed by retest correlations, rivals that of personality traits. The implication is that a sample of enduring marriages may not be ideal for identifying personality correlates of changes in relationships. Perhaps investigations studying the courtship phase of relationships when the initial dynamics are being established will have better chances for finding robust evidence that personality predicts changes in relationships.

6. Limitations and conclusions A few caveats should be noted when considering the findings reported here. First, our measurement of personality was based on self-reports on one measure taken at one point in time. The ‘‘gold standard’’ design in this area would be to obtain personality data from multiple informants, with multiple instruments, and at multiple points in time. Because of this limitation, our analyses cannot address the interesting issue regarding the impact of relationship experiences on personality development (e.g., Robins et al., 2002). Future research on personality and marriage should strive for this gold standard. In our favor, we used a well-established and validated instrument to measure personality and we used multiple informants for the other variables in our analyses. In addition, current evidence indicates that personality is relatively stable in adulthood (e.g., Roberts & DelVecchio, 2000) so it is unlikely that we would observe major upheavals in personality from 1990 to 1994. A second limitation was that our sample was composed of rural European Americans in predominately first-time, long-term marriages. Unquestionably, our results should be replicated on diverse samples. However, Karney and Bradbury (1995) noted the need for studies that specifically focus on more homogenous samples in marital research to help focus and sharpen the conclusions that can be drawn from the literature. Consequently, our results demonstrate that personality is related to observable marital interactions and global evaluations in enduring marriages. Finally, concerns over the practical importance of the observed effect sizes in this investigation could be raised. The effect sizes we observed were generally small to medium in magnitude. This concern over small effect sizes should be tempered in light of several issues. First, our conservative approach may underestimate the impact of personality on marriages. Second, small effect sizes are to be expected in research on marital variables because marital outcomes and marital interactions are multiply influenced (Ahadi & Diener, 1989). Third, small effect sizes can have a major impact on outcomes over time (Abelson, 1985; Rosenthal & Rubin, 1979). Finally, we concur with Robins et al. (2000) who argue that the explanation of even a small amount of variance in a major life goal such as having a happy relationship is an important outcome of psychological inquiry. In that regard, we believe that our findings may have practical significance. Indeed, Meyer et al. (2001) recommended

502

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

rethinking the general interpretations of so-called small effect sizes in light of these kinds of concerns. In closing, we believe that the search for how personality influences marital relationships is both practically important and theoretically interesting. We found evidence that a broad range of intrapersonal factors are related to relationship interactions and global evaluations of enduring marriages. In particular, agreeableness and openness have significant links with relationship variables and have received much less attention in the literature relative to neuroticism. Moreover, these relations are ‘‘real’’ to the extent that shared method variance is not a viable alternative explanation for these associations. Accordingly, more work should focus on agreeableness and openness as important personality influences on close relationships. Future work is also needed to completely understand how a wide range of personality factors influence marriages. Although it is premature to draw strong conclusions, we believe that the study of the impact of personality on the day-today lives of couples is a good place to look for answers to questions concerning how personality impacts close relationships.

References Abelson, R. P. (1985). A variance explanation paradox: When a little is a lot. Psychological Bulletin, 97, 129–133. Ahadi, S., & Diener, E. (1989). Multiple determinants and effect size. Journal of Personality and Social Psychology, 56, 398–406. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator distinction in the social-psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Block, J. (1995). A contrarian view of the five-factor approach to personality description. Psychological Bulletin, 117, 187–215. Block, J. (2001). Millenial contrarianism: The five factor approach to personality description 5 years later. Journal of Research in Personality, 35, 98–107. Bolger, N., & Schilling, E. A. (1991). Personality and the problems of everyday life: The role of neuroticism in exposure and reactivity to daily stressors. Journal of Personality, 59, 355–386. Bouchard, G., Lussier, Y., & Sabourin, S. (1999). Personality and marital adjustment: Utility of the fivefactor model of personality. Journal of Marriage and the Family, 61, 651–660. Botwin, M. D., Buss, D. M., & Shackelford, T. K. (1997). Personality and mate preferences: Five factors in mate selection and marital satisfaction. Journal of Personality, 65, 107–136. Bradbury, T. N., & Fincham, F. D. (1988). Individual difference variables in close relationships: A contextual model of marriage as an integrative framework. Journal of Personality and Social Psychology, 54, 713–721. Byrne, B. M. (2001). Structural equation modeling with AMOS. Mahwah, NJ: Lawrence Erlbaum. Caughlin, J. P., Huston, T. L., & Houts, R. N. (2000). How does personality matter in marriage. An examination of trait anxiety, interpersonal negativity, and marital satisfaction. Journal of Personality and Social Psychology, 78, 326–336. Cohen, J. (1988). Statistical power for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum. Collins, L. M., Graham, J. W., & Flaherty, B. P. (1998). An alternative framework for defining mediation. Multivariate Behavioral Research, 33, 295–312. Conger, R. D., Rueter, M. A., & Elder, G. H. (1999). Couple resilience to economic pressure. Journal of Personality and Social Psychology, 76, 54–71.

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

503

Cooper, M. L., & Sheldon, M. S. (2002). Seventy years of research on personality and close relationships: Substantive and methodological trends over time. Journal of Personality, 70, 783–812. Costa, P. T., & McCrae, R. R. (1989). The NEO PI/FFI: Manual supplement. Odessa, FL: Psychological Assessment Resources. Duncan, T. E., Duncan, S. C., Strycker, L. A., Li, F., & Alpert, A. (1999). An introduction to latent variable growth curve modeling: Concepts, issues, and applications. Mahwah, NJ: Lawrence Erlbaum. Eysenck, H. J., & Wakefield, J. A. (1981). Psychological factors as predictors of marital satisfaction. Advances in Behavior Research and Therapy, 3, 151–192. Funder, D. C. (2001). Personality. Annual Review of Psychology, 52, 197–221. Gottman, J. M. (1994). What predicts divorce? The relationship between marital processes and marital outcomes. Hillsdale, NJ: Erlbaum. Gottman, J. M. (1998). Psychology and the study of the marital processes. Annual Review of Psychology, 49, 169–197. Graziano, W. G., & Eisenberg, N. (1997). Agreeableness: A dimension of personality. In R. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of Personality Psychology (pp. 795–824). San Diego: Academic Press. Graziano, W. G., Jensen-Campbell, L. A., & Hair, E. C. (1996). Perceiving interpersonal conflict and reacting to it: The case for agreeableness. Journal of Personality and Social Psychology, 70, 820–835. Holmbeck, G. N. (1997). Toward terminological, conceptual, and statistical clarity in the study of mediators and moderators: Examples from the child-clinical and pediatric psychology literatures. Journal of Consulting and Clinical Psychology, 65, 599–610. Huston, T. L., & Houts, R. M. (1998). The psychological infrastructure of courtship and marriage: The role of personality and compatibility in romantic relationships. In T. N. Bradbury (Ed.), The developmental course of marital dysfunction (pp. 114–151). New York: Cambridge University Press. Jocklin, V., McGue, M., & Lykken, D. T. (1996). Personality and divorce: A genetic analysis. Journal of Personality and Social Psychology, 71, 288–299. John, O. P., & Srivastava, S. (1999). The Big Five trait taxonomy: History, measurement, and theoretical perspectives. In L. A. Pervin & O. P. John (Eds.), Handbook of personality: Theory and research (2nd ed., pp. 102–138). New York: Guilford Press. Johnson, D. R., Amoloza, T., & Booth, A. (1992). Stability and developmental change in marital quality. Journal of Marriage and the Family, 54, 582–594. Johnson, D. R., & Booth, A. (1988). Marital quality: A product of the dyadic environment or individual factors? Social Forces, 76, 883–904. Karney, B., & Bradbury, T. N. (1995). The longitudinal course of marital quality and stability: A review of theory, methods, and research. Psychological Bulletin, 118, 3–34. Karney, B., & Bradbury, T. N. (1997). Neuroticism, marital interaction, and the trajectory of marital satisfaction. Journal of Personality and Social Psychology, 72, 1075–1092. Kashy, D. A., & Snyder, D. K. (1995). Measurement and data analytic issues in couples research. Psychological Assessment, 7, 338–348. Kelly, E. L., & Conley, J. J. (1987). Personality and compatibility: A prospective analysis of marital stability and marital satisfaction. Journal of Personality and Social Psychology, 52, 27–40. Kenny, D. A. (1998). Couples, gender, and time: Comments on method. In T. N. Bradbury (Ed.), The developmental course of marital dysfunction (pp. 410–422). New York: Cambridge University Press. Kurdek, L. A. (1998). Developmental changes in marital satisfaction: A 6-year prospective longitudinal study of newlywed couples. In T. N. Bradbury (Ed.), The developmental course of marital dysfunction (pp. 180–204). New York: Cambridge University Press. Lorenz, F., Wickrama, K. A. S., & Conger, R. D. (2004). Modeling continuity and change in family relations with panel data. In R. D. Conger, F. O. Lorenz, & K. A. S. Wickrama (Eds.), Continuity and change in family relations: Theory, methods, and empirical findings (pp. 15–62). Mahwah, NJ: Lawrence Erlbaum. MacCallum, R. C. (2003). Working with imperfect models. Multivariate Behavioral Research, 38, 113–140. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7, 83–104.

504

M.B. Donnellan et al. / Journal of Research in Personality 38 (2004) 481–504

McCartney, K., & Rosenthal, R. (2000). Effect size, practical importance, and social policy for children. Child Development, 71, 173–180. McCrae, R. R., & Costa, P. T. (1997). Conceptions and correlates of openness to experience. In R. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of personality psychology (pp. 825–847). San Diego: Academic Press. Melby, J. N., & Conger, R. D. (2001). The Iowa family interaction rating scales: Instrument summary. In P. K. Kerig & K. M. Lindahl (Eds.), Family observational coding systems: Resources for systematic research (pp. 33–57). Mahwah, NJ: Erlbaum. Meyer, G. J., Finn, S. E., Eyde, L. D., Kay, G. G., Moreland, K. L., Dies, R. R., Eismane, E. J., Kubiszyn, T. W., & Reed, G. M. (2001). Psychological testing and psychological assessment: A review of the evidence and issues. American Psychologist, 56, 128–165. Muthen, B. M., & Curran, P. J. (1997). General longitudinal modeling of individual differences in experimental designs: A latent variable framework for analysis and power estimation. Psychological Methods, 2, 371–402. Robins, R. W., Caspi, A., & Moffitt, T. E. (2000). Two personalities, one relationship: Both partnersÕ personality traits shape the quality of their relationship. Journal of Personality and Social Psychology, 79, 251–259. Robins, R. W., Caspi, A., & Moffitt, T. E. (2002). ItÕs not just who youÕre with, itÕs who you are: Personality and relationship experiences across multiple relationships. Journal of Personality, 70, 925– 964. Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency of personality traits from childhood to old age: A quantitative review of longitudinal studies. Psychological Bulletin, 126, 3–25. Rogosa, D., Brandt, D., & Zimowski, M. (1982). A growth curve approach to the measurement of change. Psychological Bulletin, 92, 726–748. Rosenthal, R., & Rubin, D. B. (1979). A note on percent variance explained as a measure of the importance of effects. Journal of Applied Social Psychology, 9, 395–396. Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: New procedures and recommendations. Psychological Methods, 7, 422–445. Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change and event occurrence. New York: Oxford University Press. Stoolmiller, M. (1995). Using latent growth curve models to study developmental processes. In J. M. Gottman (Ed.), The analysis of change (pp. 103–138). Mahawah, NJ: Lawrence Erlbaum. Suen, H. K., & Ary, D. (1989). Analyzing quantitative behavior observation data. Hillsdale, NJ: Erlbaum. Terman, L. M. (1938). Psychological factors in marital happiness. New York: McGraw-Hill. Tobin, R. M., Graziano, W. G., Vanman, E. J., & Tassinary, L. G. (2000). Personality, emotional experience, and efforts to control emotions. Journal of Personality and Social Psychology, 79, 656–669. Watson, D., Hubbard, B., & Wiese, D. (2000). General traits of personality and affectivity as predictors of satisfaction in intimate relationships: Evidence from self- and partner-ratings. Journal of Personality, 68, 413–449. Willet, J. B., & Sayer, A. G. (1994). Using covariance structure analysis to detect correlates and predictors of individual change over time. Psychological Bulletin, 116, 363–381. Willett, J. B., Singer, J. D., & Martin, N. C. (1998). The design and analysis of longitudinal studies of development and psychopathology in context: Statistical models and methodological recommendations. Development and Psychopathology, 10, 395–426.