Pronouns in Marital Interaction What Do "You" and "I" Say About Marital Health? Rachel A. Simmons, ' Peter C. Gordon," and Dianne L. Chambless University of Pennsylvania and University of North Carolina at Chapel Hill ABSTRACT-Recent studies in social psychology have found events. Marital researchers have included frequency of couples' that the frequency of certain words in people's speech and use of we in elaborate coding schemes of their oral-history writing is related to psychological aspects of their personal narratives. Although these measures predict longitudinal out- health. We investigated whether counts of "self" and comes, such as marital satisfaction and divorce (Buehlman, "other" pronouns used by 59 couples engaged in a prob- Gottman, & Katz, 1992; Gottman & Levenson, 1999), the in- lem-solving discussion were related to indices of marital dependent contribution of word counts has not been explored in health. One spouse in each couple had a diagnosis of ob- research on marriage. In the present study, we investigated sessive-compulsive disorder or panic disorder with ago- whether pronoun use during a focused, face-to-face conversa- raphobia; 50% of the patients and 40% of their spouses tion is related to the quality of a dyad's interaction and the reported marital dissatisfaction. Regardless of patients' overall health of their marriage. diagnostic status, spouses who used more second-person Recent research on language has yielded at least one finding pronouns were more negative during interactions, whereas that appears to have surprised social psychologists. Pennebaker those who used more first-person plural pronouns produced (2002) reported that his work originally focused on content more positive problem solutions, even when negative be- words, such as those conveying emotion, as sources of linguistic havior was statistically controlled. Moreover, use of first- insight into individual and social psychological processes. person singular pronouns was positively associated with However, his group discovered that more information was pres- marital satisfaction. These findings suggest that pronouns ent in what he termed junk words, such as prepositions, articles, used by spouses during conflict-resolution discussions and especially pronouns. Rather than regarding these words as provide insight into the quality of their interactions and junk, psycholinguists and sociolinguists have long regarded marriages. them as critical in shaping a shared world of meaning during lin- guistic communication. Pronouns provide an important means of establishing and maintaining the entities that are at the center of How people talk to one another clearly reflects the nature and a discourse (Gordon, Grosz, & Gilliom, 1993; Sanford & Garrod, quality of their relationships. Less obvious is which variables 1981). When those entities are humans, pronouns provide a are important to successful communication and relationship powerful way of indicating their centrality to the discussion, as health. Recent findings in social psychology suggest that sta- well as their social status (Brown & Levinson, 1987) tistical analysis of easily countable features of linguistic corpora Aware of pronouns' potential for capturing relationship atti- may provide some insight. Social psychologists, led by the work tudes, marital researchers have included pronouns in measures of Pennebaker (Campbell & Pennebaker, 2003; Cohn, Mehl, & of marital bond (Buehlman et al., 1992; Buehlman, Siler, Car- Pennebaker, 2004; Pennebaker, Mehl, & Niederhoffer, 2003), rere, & Gottman, in press). Buehlman's we-ness versus sepa- have shown that the relative frequency of different types of rateness dimension, which includes judges' ratings of spouses words in individuals' diaries and Web logs can predict their tendency to use we over he, she, and I during an oral-history physical health improvements and important characteristics of interview, predicted couples' concurrent interaction behavior their psychological adaptation, such as distancing from tragic and physiological reactivity, their marital satisfaction and likelihood of divorce at 3-year follow-up, and their interaction quality at 4-year follow-up (Buehlman et al., 1992; Gottman & Address correspondence to Rachel A. Simmons, University of Penn- Levenson, 1999). These findings suggest that we may capture sylvania, Department of Psychology, 3720 Walnut St., Philadelphia, important ways couples think about their marriages, but because PA 19104; e-mail: simmonsr@psych.upenn.edu. pronoun ratings are only one of several components of thewe-ness measure, these studies do not indicate whether pronoun resent markers of shared identity (e.g., we, us), "other" focus frequencies alone predict marital outcomes or interaction quality. (you), active self (1), and passive self (me) associated with marital Findings in social psychology have linked pronoun counts to health. Transcripts were analyzed using Linguistic Inquiry Word social processes. First-person plural pronouns (e.g., we, us, our) Count (LIWC; Pennebaker, Francis, & Booth, 2001). Pronoun appear to be markers of shared identity and affiliative motiva- proportions were correlated with measures of spouses' negativ- tion. Studies have found that people increase their use of first- ity, generation of positive problem solutions, and self-reported person plural pronouns after a large-scale trauma (Stone & marital satisfaction. Pennebaker, 2002) or after a victory of their home football team (Cialdini et al., 1999). Highly committed partners also use we METHOD pronouns more frequently than less committed partners when writing about their romantic relationships (Agnew, Rusbult, Van Participants Lange, & Langston, 1998). In contrast, first-person singular Archived transcripts of problem-solving interactions between pronouns (e.g., I, me, my) and second-person pronouns (e.g., married (n = 54) or cohabitating (n = 5) partners (hereafter you, your) have been interpreted as indicants of individuated referred to as spouses) provided our data. Participants were identity and of self-focus and other-focus of attention. Studies outpatients who had obsessive-compulsive disorder or panic have found that high self-monitoring (other-focused) individuals disorder with agoraphobia and were beginning treatment at use you more frequently than low self-monitoring individuals American University in Washington, DC, or Mclean Hospital in during unstructured peer interactions (Ickes, Reidhead, & Belmont, MA (see Chambless & Steketee, 1999, for details), Patterson, 1986) and that individuals high in trait anger use you along with their spouses. more frequently than individuals low in anger during mono- Participants averaged 37 years in age (SD = 9). Ninety logues (Weintraub, 1981). Similarly, numerous studies have percent of the sample was Caucasian and 7% African American. linked first-person pronouns to self-focus (for a review, see Fifty-seven (97%) of the couples were heterosexual. Socioeco- Pennebaker et al., 2003). nomic status of the couples spanned the range of Hollingshead's Marital theorists have argued that, with respect to marital inter- (1975) Four Factor Index of Social Position, from 16 to 66, with a action, / statements facilitate adaptive communication proc- median of 48 (e.g., minor professionals, small-business owners). esses, such as verbal immediacy and self-disclosure, whereas According to Dyadic Adjustment Scale (DAS; Spanier, 1976) excessive you statements signal blaming or distancing attitudes scores, 50% of the patients and 40% of the spouses were mar- (e.g., Hahlweg et al., 1984). However, in the only study of lan- itally distressed (scores less than 100); 57% of the couples in- guage use during marital interaction, Sillars, Shellen, McIntosh, cluded at least one spouse who reported distress. and Pomegranate (1997) found that happier couples used both fewer second-person pronouns and fewer first-person pronouns, compared with less happy couples. No studies to date have eva- Measures luated the relationship between spouses' pronoun use and their Interaction Coding actual behavior during interactions. The Kategoriensystem fur Partnerschaftliche Interaktion (KPI; Past studies have lumped together all first-person pronouns, Hahlweg & Conrad, 1983; Hahlweg et al., 1984) was used to including the nominative case / and the accusative case me. code dyadic interactions. One of four raters assigned each ut- However, factor analyses have yielded different factor loadings terance 1 of 10 verbal codes designed to measure behaviors for I and me (Campbell & Pennebaker, 2003). Inductive anal- considered by marital theorists and researchers to be func- yses also have found that me occurs during narrative episodes tionally important in communication. In addition, each utter- and vivid, clear speech associated with therapeutic change, where- ance was assigned a positive, negative, or neutral nonverbal as I is more common during vague, abstract, and ruminative code. Tapes of 40 couples were randomly selected for inde- speech associated with lack of improvement (Mergenthaler pendent coding to determine interrater reliability. Reliability as & Bucci, 1999). The roles of me and I in marital interaction calculated with intraclass correlation coefficients was satisfact have not been explored; however, within the context of a prob- tory to excellent (.58 to .89) for all codes except "negative lem-solving discussion, they are likely to be distinct. High levels problem solution" (.37), which was rarely used. A composite of me may reflect passive strivings or victimization narratives measure of negative interaction behavior was based on Hahlweg characteristic of poor-quality interactions and decreased satis- and Conrad's (1983) categories. Total negative interaction be- faction. In contrast, I may reflect positive self-disclosure and havior was computed as the average of the percentage of nega- perspective taking. tive nonverbal codes and the percentage of negative verbal In the present study, we explored whether pronouns used by codes (including criticism, disagreement, justification, and spouses during problem-solving interactions are correlated with negative-solution codes). The number of positive problem so- the quality and outcome of their interactions and their overall lutions (defined as constructive solutions and compromises) relationship satisfaction. We hypothesized that pronouns rep- generated by each couple was used to index interaction success.Rat Analysis Text analyses were carried out using the LIWC computer program (Pennebaker et. al., 2001). Because we were interested in how spouses speak about themselves in relation to their partners, we created four variables to capture trade-offs between self and partner references. These were you-focus ( you, excluding generic you, as in the expression \"you know\"), we-focus (we, as, our), I-focus (I ), and me-focus (me). Each pronoun count was divided by the total number of rst and second-person pronouns. Dyadic Adjustment Scale The DAS (Spanier, 1976) is a self-report measure of marital satisfaction. The psychometric properties of this instrument are well established. Procedure Dyads were brought into a room with a video camera, given standardized instructions to consider the top issues facing their relationship, and asked to choose one to discuss for 10 min. Nearly a third (29%) of the couples selected problems related to the patient's anxiety disorder, whereas the other two thirds discussed more general sources of conict, such as communi- cation and division of labor. Couples were instructed to work toward a mutually satisfactory resolution and then left alone to discuss the selected problems. Videotapes were subsequently coded for interaction behavior and transcribed for text analysis. RESULTS Descriptive statistics are presented in Table 1. With the ex- ception of positive problem solutions, me-focus, and we-focus, measures were normally distributed. The positive-problem- solution distributions for patients and their spouses were positively skewed and not improved by any transformations; however, arcsine transformations (arcsine of the square root of p) improved the normality of the skewed pronoun distributions and TABLE 1 were used for all pronouns in the analyses. As predicted, I-focus and me-focus were not positively correlated for patients, rs = .1 (p = .4), or their spouses, rs = .08 (p = .6). Simple correlations between individuals' pronoun use, inter- action behavior, and marital satisfaction are presented in Table 2. Spearman rank correlations are reported for analyses involving positive problem solutions, because of its skewed distribution. Multiple regression analyses were also conducted to rule out in- teractions of diagnosis, gender, or conversation topic (anxiety related or not) with pronoun use. Predictors were centered to reduce collinearity with interaction terms (Aiken & West, 1991). Diagnosis, gender, and conversation topic did not interact with pronoun use and indices of marital health; thus, we interpret the simple correlations. With our sample size, power to detect a medium effect size (r) of .30 was only .61 for the interaction measures and .54 for marital satisfaction. Thus, alpha was re- tained at .05 despite the number of tests conducted. To reduce the likelihood of Type I error, we limit our interpretation to those ndings that were replicated across patients and their spouses. Marital Satisfaction and Interaction Quality As shown in Table 2, use of second-person pronouns was posi- tively correlated with negative interactions. Also, marital sat- isfaction was marginally positively correlated with I-focus during the interactions. Effect sizes were small to medium-large. Correlations between pronoun use and marital satisfaction did not achieve full statistical significance, perhaps because the power of those tests was reduced by missing data. Problem Solving We-focus was associated with more positive problem solutions. The effect size was medium for both patients and their spouses. To rule out the possibility that these relationships were due to the fact that positive problem solutions and we-focus were both correlated with negative interaction behavior (positive problem Means, Standard Deviations, and Ranges of Variables Included in the Analyses Patients Patients' spouses Variable Mean SD Minimum Maximum Mean SD Minimum Maximum Negative interaction behavior .32 .20 .02 .84 .28 .20 0 .87 Positive problem solutions 2 3 0 15 3 3 0 16 Marital satisfaction 102 22 47 141 101 22 52 135 Word count 765 381 140 1,896 799 341 211 1,582 You-focus .30 .12 .05 .54 .34 .15 .05 .78 Wefocus .12 .13 0 .78 .14 .12 0 .51 I-focus .45 .13 .08 .72 .40 .14 .13 .68 Mefocus .07 .04 0 .20 .06 .03 0 .13 Note. Negative interaction behavior is an average of the proportions of negative verbalbehavior and negative nonverbal behavior. Positive problem solutions is a raw count of constructive solutions and compromises suggested. Pronoun (focus) variables are the proportions of each word type over the total number of rst and secondperson pronouns. The number of patients and of spouses was 59 for all variables except for marital satisfaction, for which 54 patients and 47 patients' spouses provided data. TABLE 2 Correlations Between Individuals ' Pronoun Use and Their Interaction Behavior and Marital Satisfaction Negative Positive Pronoun interaction problem Marital variable behavior solutions satisfaction Patients Youfocus .41\" .16 .27* Wefocus .20 .32* .06 Ifocus .16 .06 25+ Mefocus .02 .02 .16 Patients' spouses You-focus .49\" .14 .22 Wefocus .33* .38\" .09 Ifocus .20 .12 .27+ Mefocus 27* .16 .13 Note. Pronoun ratios were transformed using the arcsine of the square root of p. Spearman rank correlations are presented for positive problem solutions. Marital satisfaction was measured with the Dyadic Adjustment Scale. +p