Linguistic correlates of the Communication Patterns Questionnaire: The correspondence of language and quantitative measures of perceived communication patterns among emerging adult gay male couples

Abstract

Up to two-thirds of new HIV infections among sexual minority men are estimated to occur within main partnerships, with the greatest burden among younger partners ages 18 to 29. Various interpersonal processes, such as perceived communication patterns, influence HIV behavioral health among these couples. Although they represent correlates of health for gay couples, it is unclear how these abstract dynamics manifest in the context of everyday speech. The present study investigated linguistic predictors of perceived communication patterns among partnered gay men. Individual interviews from both members of 21 emerging adult couples were analyzed using Linguistic Inquiry Word Count. Couples’ total first-person singular (I-talk) and plural (We-talk) pronoun use were positively associated with total dyadic constructive communication scores. Negative emotion word use was negatively associated with constructive communication scores. Results suggest that language use provides important markers of interpersonal processes that predict HIV behavioral health among gay couples, suggesting the potential of language use to inform HIV prevention interventions for partnered gay men in emerging adult couples.

Gay, bisexual, and other men who have sex with men remain disproportionately affected by the HIV epidemic in the United States. In 2016, estimates indicate that 67% of all new HIV infections occurred in the context of condomless anal sex among sexual minority men (CDC, 2017). Of concern, research suggests that main partnerships are a critical context for HIV transmission, accounting for one-third (Goodreau et al., 2012) to two-thirds (Sullivan, Salazar, Buchbinder, & Sanchez, 2009) of new HIV infections among sexual minority men. Further, within these couples, younger partners are particularly are at risk. Specifically, young partnered men who fall within the developmental period known as emerging adulthood (i.e., those between 18 and 29 years of age; Arnett, 2000) are at increased risk for infection (Goodreau et al., 2012; Sullivan et al., 2009). For example, the CDC reports that, in 2016, young sexual minority men ages 15 through 24 accounted for 32% of all incident infections, which was 1.5 times the rate of older sexual minority men ages 30 to 44 and a staggering 21 times the rate of their heterosexual counterparts in the same age bracket (CDC, 2017). Further, it is estimated that as many as 79% of new infections among emerging adult sexual minority men ages 18 through 24 are likely to occur within main partnerships (Sullivan et al., 2009). As such, gay male couples that include emerging adult men represent an especially critical sub-context to consider for targeted HIV prevention efforts.

The high incidence of HIV infection among emerging adult gay male partners (and between main partners generally, regardless of age) has been attributed, in part, to high rates of sexual behaviors that may transmit HIV (e.g., anal sex) combined with low rates of HIV preventive behaviors among gay men in relationships (Sullivan et al., 2009). Research indicates that not only do these men report high rates of recent condomless anal sex with one another (e.g., Hoff, Chakravarty, Beougher, Neilands, & Darbes, 2012), but that gay men in main partnerships report substantial rates of various forms of non-monogamy (e.g., Grov, Starks, Rendina, & Parsons, 2014; Parsons, Starks, DuBois, & Golub, 2013), introducing additional HIV risk within the relationship. In addition to this sexual risk, gay men in relationships are more likely to perceive themselves at zero risk for HIV (Stephenson, White, Darbes, Hoff, & Sullivan, 2015) and less likely to have been tested for HIV recently (Mitchell & Petroll, 2012; Stephenson et al., 2015), additional factors, which when taken altogether, place gay men in relationships specifically at significant risk for HIV infection.

Various interpersonal processes appear to influence HIV behavioral health outcomes among gay couples, regardless of age. For instance, sexual communication has been linked to sexual HIV transmission risk behavior for partnered gay men (Starks, Tuck, Millar, & Parsons, 2016), and research suggests that even the perception of specific non-sexual communication patterns influence the risk of HIV infection among gay couples (Darbes, Chakravarty, Neilands, Beougher, & Hoff, 2014; Gomez et al., 2012). In particular, perceptions of two important communication styles have been examined with respect to sexual health among gay couples (e.g., Darbes et al., 2014): Constructive communication refers to a productive approach to contentious situations, characterized by mutual expression and teamwork, while avoidant communication describes an evasive, problematic approach to conflict, often marked by blaming and criticism (Christensen & Shenk, 1991).

Among partnered gay men, perceptions of these two communication patterns are associated with HIV risk behavior and preventive attitudes. For instance, among HIV-seroconcordant couples, high levels of perceived constructive communication reduce the odds of condomless anal sex with outside sex partners (Darbes et al., 2014) and are negatively associated with substance use during sex within the relationship (Mitchell, 2015). Relatedly, Mitchell (2014b) observed among partnered gay men that, although perceived constructive communication was positively associated, perceived avoidant communication was negatively associated with willingness to engage in couples HIV testing and counseling (CHTC; also referred to as couples voluntary counseling & testing), an evidence-based behavioral couples’ HIV intervention (Burton, Darbes, & Operario, 2010; El-Bassel et al., 2011) recently adapted for gay male couples in the United States (Sullivan et al., 2014).

While important antecedents to sexual health for gay men in relationships, it is not clear how the perceptions of these communication styles manifest in everyday speech among partnered gay men. Many dyadic interventions, including those that employ traditional components of behavioral couples therapy, such as cognitive behavioral modules and spousal intervention, take the stance that dyadic functioning, which encompasses communication, is directly related to target health outcomes (Powers, Vedel, & Emmelkamp, 2008). In fact, enhancements in functioning are often seen as the mechanism by which such interventions may work (Davidovich, De Wit, & Stroebe, 2006). Therefore, these behavioral couples therapy interventions utilize dyadic speech as their behavioral target to influence communication.Intuitively, language use should correlate with actual communication patterns, but it is not clear how language use relates to perceptions of these communication patterns among gay male couples. As such, showing that measures of perceived quality of various communication styles are correlated with actual language among individual partnered gay men in couples at the level of the dyad may provide evidence of these measures’ validity and their potential utility as indicators of couple-level intervention effects. Specifically, an understanding of the linguistic profiles of perceived communication patterns among gay couples could enhance behavioral couples therapy interventions for HIV infection by providing concrete indicators of abstract relational processes that are directly associated with HIV behavioral health.

However, only recently have researchers investigated linguistic correlates of perceived communication styles among partnered individuals in couples navigating health issues at the level of the dyad and only among heterosexual couples. In that particular study (i.e., Rentscher, Rohrbaugh, Shoham, & Mehl, 2013), among couples with one member navigating congestive heart failure, substance abuse, or health-compromised smoking, dyadic trait perceived constructive communication was associated positively with the sum of partners’ first-person singular (i.e., I-talk) and negatively with total second-person (i.e., You-talk) pronoun use in a marital interaction task (in which members of a couple were prompted to discuss a disagreement related to health in attempt to resolve it; Rentscher et al., 2013). The role of language use, specifically dyadic pronoun use in relation to perceived constructive communication among couples in that particular study may be interpreted through the lens of Couples Interdependence Theory (Rusbult & Van Lenge, 2003). This theory holds that, in relationships, the goals of the individual partners, particularly those related to health, often shift (or are transformed) from the level of the partner to the level of the couple (i.e., become a mutual goal), a process that occurs through mutual accommodation (Lewis et al, 2006).

In this particular study, language used by couples during the interaction task may have been reflective of underlying goal transformation processes, as predicted by Couples Interdependence Theory (Rusbult & Van Lenge, 2003). Namely, it is possible that the increased prevalence of dyadic You-talk and I-talk simultaneously (Rentscher, Rohrbaugh, Shoham, & Mehl, 2013) may have been indicative of couples’ greater cognitive focus on the individuals therein (i.e., “you” and “I,” rather than “we”). However, in their association with perceived constructive communication as well, such language use might have indicative of the earliest stages of goal transformation to the level of the couple. Thus, through this lens, we might expect the simultaneous use of first-person pronoun use (both I-talk and We-talk) to be particularly important indicators of constructive communication patterns, perhaps reflecting the progression of goal transformation processes from the level of the individual partner to that of the couple.

Although no studies have examined associations between other language use, such as emotion-word content, and perceived communication patterns among partnered individuals, this language use also appears influential for adaptive relational functioning. For instance, in a study of families coping with cancer, negative emotion word use by patients was related to increased family conflict, while positive emotion word use on the part of the patients’ partners was associated with increased average family cohesion (Robbins, Mehl, Smith, & Weihs, 2013). Again, Couples Interdependence Theory (Rusbult & Van Lenge, 2003) might offer a straight-forward explanation for the association of the emotional valence of language with relational outcomes among couples in the context of health navigation: higher levels of negative word use within couples’ interactions might suggest increased distraction from the formation of constructive goals, while positive word use might be interpreted as building cohesion toward those goals.

In each of the aforementioned studies, however, analyses of language were based in the context of a dyadic interaction, such as individual partners’ dialogue during the course of a conversation (i.e., both partners are present). However, as research suggests, among partnered individuals discussing their relationships (e.g., its dissolution; Blackburn, Brody, & LeFebvre, 2014), language use is context-dependent, and private accounts of events differ linguistically (e.g., different rates of pronouns; Blackburn et al., 2014) from those presented publicly, i.e. when they expected their responses to be read by others. Extending this finding to the context of partnered individuals’ in-person discussions of their relationships, we might conceptualize the public sphere as one in which an individual’s partner is present, whereas private speech might be characterized by an individual’s discussion without one’s partner present.

Analysis of private language use among partnered individuals discussing their relationships then might provide unique insight into communication processes not accessible when members of a couple present together. Further, practically speaking, and of relevance to the current study, most existing sexual healthcare services for sexual minority men, such as HIV testing and risk behavior reduction, occur at the individual-level. For instance, as CHTC has only recently begun to be implemented within the United States (Stephenson, Grabbe, Sidibe, McWilliams, & Sullivan, 2016), clinicians are more likely to encounter private speech among partnered gay men rather than that of members of the couple presenting for healthcare services together (i.e., publicly). Thus, in order to identify linguistic manifestations of dyadic processes that may inform the development of dyadic HIV interventions, such as perceived dyadic communication, it is important and perhaps currently more feasible to consider partners’ language use when they are speaking privately about the relationship.

To address these gaps in the literature, the current study reports the results of a linguistic analysis of gay male partners’ interviews about HIV behavioral health within their relationships, focusing specifically on couples that include at least one emerging adult male partner, an especially critical subpopulation whose high HIV prevalence (CDC, 2017) suggests consideration with respect to the development of novel HIV prevention efforts. The current study adds to the literature in a variety of ways. First, while health-related linguistic analyses have been conducted to study relational outcomes, no studies have examined these associations among gay couples. Second, among linguistic studies examining health outcomes with couples more broadly, only one (Rentscher et al., 2013) has focused on perceived communication patterns as the outcome associated with linguistic markers, and it was the first to use the couple as the unit of analysis.

The present study extends work from Rentscher and colleagues (2013) to examine the predictive power of linguistic markers in individual speech on perceptions of communication patterns among gay couples specifically. As perceived communication patterns have been associated with both HIV risk behavior and preventive attitudes at the level of the dyad among individual partnered gay male couples (Darbes et al., 2014; Mitchell, 2015; Mitchell, 2014b), these exploratory analyses aim to identify concrete, linguistic indicators of perceived communication patterns among partnered gay men, As such, in the investigation of the linguistic underpinnings of these communications patterns related to HIV behavioral health among gay male couples, this study initiates a line of research intended to identify distal indicators of these processes, which may inform the development of dyadic behavioral HIV prevention interventions, with particular focus on couples that include emerging adult gay men.

Methods

Recruitment and Enrollment

Twenty-one gay male couples (42 individuals) participated in the current study. These men were recruited after participation in a larger, online study on same-sex male relationships (Starks, Millar, & Parsons, 2015). Participants for the larger online couples’ study were recruited originally through a variety of field and online methods aimed to recruit gay and bisexual men across the U.S., with an emphasis on the New York City metropolitan area. Detailed recruitment procedures for the parent study are described elsewhere (Starks et al., 2015).

Eligibility for the current study required completion of the larger study by both members of the couple, a relationship length of at least three months, and recent anal sex together (at least once in the prior three months). Further, one member of the couple had to be between the ages of 18 and 29, both members needed to reside in the NYC metropolitan area, and both needed to be able to communicate in English. Couples expressing interest in the subsequent study were contacted, rescreened for eligibility over the phone, and scheduled for a baseline assessment at the research team’s community-based research center.

After providing informed consent, both members of the couple completed a baseline survey including demographic, psychosocial, and health measures delivered via Qualtrics®, and participated in three separate interviews; members of the couple first met together for an initial joint interview, were separated and interviewed separately, and then were reunited for a final interview together. In this survey, participants self-reported perceived communication patterns via the Communication Patterns Questionnaire –Short Form (Christensen & Shenk, 1991), and critical incident measures (Leonard & Ross, 1997) were employed by assessors in each interview to capture detailed information regarding relationship dynamics, as well as HIV prevention and risk concerns within the relationship.

All data for the current analyses were collected at baseline assessments occurring between February and September of 2012. Assessments at the community-based research center were conducted in private rooms to ensure confidentiality, and all interviews were digitally recorded and subsequently outsourced to a professional transcription service to enhance fidelity and to ensure reliability. All procedures were approved by the Institutional Review Board of [MASKED FOR REVIEW].

Procedures

Linguistic measures were drawn from individual interviews. In these discussions, men reflected on their sexual agreements, sexual behavior within the relationship and, if applicable, outside partners. Moreover, men discussed the influence of HIV testing on condom use with main and outside partners, as well as the intersection of substance use and sexual behavior with these partners. Notably, only in these individual interviews did participants provide explicit discussions of HIV status disclosure and testing within the relationship. Specifically, individuals discussed how HIV status is broached in relationships (both anecdotally and more abstractly), as well the role of continual testing in the context of a committed relationship and its impact on HIV risk reduction strategies, such as condom use, with main and outside partners. Perceived communication patterns have been associated with sexual agreements (Gomez et al., 2012; Hoff, Beougher, Chakravarty, Darbes, & Neilands, 2010; Mitchell, 2014a), sexual risk behavior outside of the relationship (Darbes et al., 2014), and the use of substances during sex with main partners (Mitchell, 2015). As such, in these explicit discussions of HIV behavioral health, particularly their emphasis on HIV testing and status disclosure, individual interviews in particular may provide insight into perceived communication patterns to inform approaches to dyadic HIV prevention interventions for partnered gay men.

Text transcribed from individual partner interviews was analyzed with Linguistic Inquiry Word Count (LIWC; Pennebaker, Boyd, Jordan, & Blackburn, 2015), software that produces the percentage of total words falling into a variety of linguistic categories for a given block of text. Prior to the generation of these data, participant speech from individual interviews was isolated, checked for quality assurance, and prepared for analysis in line with recommendations (e.g., tagging of non-fluent phrases, such as “you know”).

Measures

Individual and Dyadic Characteristics

In the baseline survey completed prior to the study interviews, participants reported their age, race and ethnicity, HIV status, level of education, and annual income. For the purposes of these analyses, demographic characteristics were conceptualized at the couple-level. Concordance in age (both aged 18–29 or at least one member 30 years or older), race and ethnicity (both members identify as white, or at least one reported a non-white racial or ethnic identity), and HIV status (both HIV-negative/unknown or sero-discordant) were derived from this survey. Additionally, relationship length was derived from couples’ responses to this question during the initial joint qualitative interview and dichotomized at two years in length.

Couple-Level Linguistic Measures

In the current analyses, four measures were of primary interest: percentages of each partner’s first-person singular (I-talk; “I, me, my”) and plural (We-talk; “we, us, our”) pronoun speech, as well as positive (e.g., “commitment,” “share,” and “trust”) and negative (e.g., “confront,” “fight,” and “heartbreak”) emotion word use. Several studies have utilized this software to study relational dynamics among couples (Lin, Chen, & Li, 2016; Rentscher et al., 2013; Rohrbaugh, Mehl, Shoham, & Reilly, 2008). At the level of the couple, these measures were both added and subtracted to produce dyadic variables of total I-talk, We-talk, positive emotion talk, negative emotion talk (sums), as well as variables representing the discrepancy, or asymmetry (differences), of these four linguistic categories. The sum and difference in total word count were included as control variables, in order to account for the effect of discussion length on percentages of linguistic markers of interest.

Patterns of Communication

Two communication patterns (previously described) were measured with the Communication Patterns Questionnaire-Short Form (Christensen & Shenk, 1991). This 11-item instrument contains two subscales, which assess an individual’s perception of the use of constructive (3 items) and avoidant (8 items) communication styles in one’s main partnership. For each item, participants rated on a 9-point Likert scale the likelihood of a given verbal behavior during episodes of conflict with their partners. For each subscale, items were summed to produce a constructive and avoidant communication score for each participant in the dyad. Reliability for the constructive subscale was good (Cronbach’s α =.84) and adequate for the avoidant subscale (Cronbach’s α = .65).

Data Analysis

Current analyses employed a “sums and differences” approach to dyadic analysis (Darnbes & Lewis, 2015; Judd, Kenny, & McClelland, 2001). Each linguistic predictor and both communication outcomes were calculated at the dyadic-level in two ways. Examination of the couple’s total linguistic contributions (e.g., sums of each partner’s We-talk percentages in a couple) allowed for a between-couple analysis of the predictive influence of language on the total and differences in both constructive and avoidant communication, while differences in partner’s linguistic proportions in each couple (e.g., differences in We-talk percentages for each couple) allowed for a within-couple investigation of linguistic influences on the sums and difference scores for each communication pattern. For each communication outcome (constructive and avoidant), in two multiple linear regressions, the sum and difference of communication scores were each regressed on both the sums and differences for each linguistic variable of interest, as well as dyadic demographic predictors.

In the current dataset, partners are exchangeable within the couple. In other words, there is not a variable that systematically differentiates partners within the couple. It is therefore necessary to systematically impose an ordering to partners within couples in order for analyses to proceed. This was done by first ranking the partners with respect to their scores on the outcome variable to facilitate interpretation of analyses. Partner 1 was therefore defined as the member of the dyad with the higher outcome score, and this ranking was done separately for each communication outcome. In cases of equivalent scores (one instance for each communication outcome; 4.8% of couples), numbering was arbitrary. This labeling approach preserved directionality in predictor difference scores. Couple-level sums and differences in a given communication style were then specified as the outcome variables in separate ordinary least-squares linear regression models. Both models included sums and differences of I-talk, We-talk, positive emotion talk, negative emotion talk, and total word count (included as a control variable).

Further, because discordance in key dyadic characteristics (i.e., demographic and relationship length) have been included as predictors to control for their influence on various outcomes in dyadic studies examining outcomes relevant to HIV prevention among gay couples (e.g., Gomez et al., 2012), discordance in age, race and ethnicity, HIV status, and relationship length were included as predictors to control for their influence on constructive and avoidant communication. All linguistic and dyadic characteristic predictors were entered simultaneously.

Results

Individual and Dyadic Characteristics

Table 1 shows demographic characteristics for the full sample. With an average age of 27.9 years (SD = 6.9), just over half of the men identified as white (52.4%), and the majority reported an annual income of less than $30,000 (61.9%). Less than half of participants were employed full-time (42.9%), and 9.5% reported an HIV-positive status. Table 2 contains means and correlations for key individual-level study variables. As can be seen in Table 3, for the majority of couples, at least one member identified as non-white (76.2%), both were between the ages of 18–29 (57.1%) and HIV-negative/unknown (81%), and relationship length was longer than two years (52.4%).

Table 1.

Individual Participant Characteristics (N = 42)

	n	%
Race
White	22	52.4
Black/African American	5	11.9
Latino	9	21.4
Other	6	14.3
Employment
Full-time	18	42.9
Part-time	12	28.6
Self-employed	8	19.0
Unemployed	4	9.5
Annual income
Less than $30,000	26	61.9
$30,000 - $49,000	8	19.0
$50,000 or more	8	19.0
HIV Status
Negative/Unknown	38	90.5
Positive	4	9.5

Table 2.

Correlations between Key Individual-Level Study Variables

	1	2	3	4	5	6	7	8	9	M(SD)
1. Age	–	–	–	–	–	–	–	–	–	27.4 (6.96)
2. Race (ref = white)	-.16	–	–	–	–	–	–	–	–	.48 (.51)
3. Income (ref = < $30,000)	.05	−.36*	–	–	–	–	–	–	–	.38 (.49)
4. HIV status (ref = HIV-negative)	.24	.02	.08	–	–	–	–	–	–	.10 (.30)
5. I-talk	.24	−.10	.23	.35*	–	–	–	–	–	7.52 (1.33)
6. We-talk	−.11	.14	.00	−.16	−.42**	–	–	–	–	2.28 (1.27)
7. Positive emotion talk	−.09	−.07	.00	−.29	−.21	.17	–	–	–	3.09 (.91)
8. Negative emotion talk	.09	−.08	.17	.01	.26	−.12	.04	–	–	1.41 (.46)
9. CPQ-Constructive	−.15	.12	.18	.02	.05	.20	.16	−.43*	–	21.43 (4.74)
10. CPQ-Avoidance	−.03	−.13	.15	−.21	−.02	.26	−.01	−.02	−.18	32.86 (10.31)

^*p <.05;

^**p < .01

Table 3.

Dyadic Characteristics (N = 21)

	n	%
Racial Concordance
One member non-white	16	76.2
Both members white	5	23.8
Age Concordance
Both members 18–29	12	57.1
Only one member 18–29	9	42.9
HIV Status Concordance
Both HIV-Negative/Unknown	17	81
One member HIV-Positive	4	19
Relationship Length
More than two years	11	52.4
Two years or less	10	47.6

Constructive Communication

Table 4 shows results of a linear regression predicting the sums of constructive communication. As can be seen, sums of I-talk (β = 0.68, p < 0.05) and We-talk (β = 0.44, p < 0.05) were positively related with the sum of constructive communication. In contrast, the sum of negative emotion talk (β = −0.71, p < 0.01) was negatively associated with the sum of constructive communication. Regression coefficients for differences in I-talk and We-talk, positive emotion talk, and negative emotion talk, as well as all demographic predictors were non-significant. Because linguistic difference scores are not predictive of communication sums (but sums are), results suggest that patterns of linguistic-communication associations are the same at the individual-level as the couple-level. The final model was significant (F (14,6) = 4.28, p < .05) and accounted for 91% of the variance in constructive communication sums. The model predicting constructive communication difference scores was not significant (R² = .87; F (14,6) = 2.83, p = .10).

Table 4.

OLS Linear Regression of Couple-Level Characteristics and Linguistic Predictors on Constructive Communication Sums

	B	95% CI	β
Age Discordance	−2.04	(−10.66, 6.58)	−.13
Racial Discordance	−7.40	(−16.84, 2.03)	−.40
HIV Discordance	−5.89	(−19.59, 7.82)	−.30
Relationship Length	−3.43	(−9.83, 2.96)	−.22
We-talk
Sum	1.71*	(.19, 3.23)	.44
Difference	1.91	(−.65, 4.47)	.37
I-talk
Sum	3.02*	(.24, 5.80)	.68
Difference	.97	(−1.02, 2.95)	.28
Positive Emotion Talk
Sum	.16	(−2.28, 2.60)	.03
Difference	−1.37	(−5.86, 3.13)	−.20
Negative Emotion Talk
Sum	−7.61**	(−11.50, −3.73)	−.71
Difference	−2.19	(−10.82, 6.43)	−.15
Total Word Count
Sum	−.02	(−.01, .00)	−.39
Difference	.00	(.00, .00)	.04

^*p <.05;

^**p < .01

Avoidant Communication

Models of linguistic and demographic variables did not predict avoidant communication scores. The models predicting sum (R² = .71; F (14,6) = 1.04, p = .52) and difference (R² = .73; F (14,6) = 1.15, p = .46) scores on avoidant communication were not significant.

Discussion

These results illustrate a correspondence between the content of speech observed in individual interviews with partnered gay men in emerging adult couples and responses to measures assessing the perceived style of communication within their relationships. First-person pronoun use in individual speech (i.e., I-talk and We-talk) was positively associated with perceived constructive communication; meanwhile, negative emotion word use was negatively associated with perceived constructive communication. These findings build on previous work supporting the application of linguistic approaches to couples navigating various health-related issues (Rentscher et al., 2013; Rohrbaugh et al., 2008; Rohrbaugh et al., 2012). Specifically, in addition to replicating the use of linguistic approaches to health for use in dyadic analyses (Rentscher et al., 2013), these exploratory analyses expand the scope of linguistic approaches to couples’ health to include gay men in main partnerships (focusing specifically on those couples including emerging adult men) by examining the correspondence between language use in individual speech and ratings of perceived dyadic communication, which have been associated with HIV risk behavior and preventive attitudes (Darbes et al., 2014; Mitchell, 2014b; Mitchell, 2015).

In particular, in the present sample, total first-person pronoun use among partners was associated with total perceived constructive communication among these men. In other words, greater use of both I-talk and We-talk on the part of both partners was associated with greater total perceived constructive communication. As previously introduced, these results are consistent with, and could be interpreted through the lens of, Couples Interdependence Theory (Rubult & Van Lange, 2003). Again, within intimate relationships, individual goals and interests (i.e., I-talk) may be “transformed” and adapted to suit those of the couple (i.e., We-talk). Further, these dyadic accommodation processes are important for the adoption of health preventive and risk reduction behaviors among couples (Lewis et al., 2006) and among gay male couples navigating HIV-related behavioral health specifically (Lewis, Gladstone, Schmal, & Darbes, 2006).

In the current sample, when individual members of the couples discuss their relationships and the role of HIV risk behavior reduction and testing behaviors therein, it is possible that the personal (I-talk) and dyadic (We-talk) orientations embodied in pronoun use are reflective of a constructive, accommodative process to health within the relationship. The negative correlation (r = −.42, p <.001) between individuals’ I-talk and We-talk suggests that it is not those men who use both I-talk and We-talk in discussions of their relationship who communicate more effectively with their partners. Rather, those who discuss their relationship experiences in either dyadic (We-talk) or personal terms (I-talk) perceive more constructive communication with their partners and are perhaps more adept communicators. This linguistic interpretation is consistent with previous research on accommodation with respect to health goals in the context of couples interdependence theory (Lewis et al., 2006a; Lewis et al., 2006b). Namely, individuals capable of discussing their relationships in either personal or joint perspectives are particularly capable of communicating constructively with their partners, which is necessary in order to accommodate another’s health goals into one’s own. As such, it may be that first-person pronoun use (whether singular or plural) may serve as one linguistic mechanism by which accommodation inherent in constructive communication arises between members of gay couples and facilitates favorable HIV behavioral health, such as reductions in HIV risk behaviors and more favorable attitudes towards HIV prevention services.

Relatedly, partners’ total negative emotion word use was negatively associated with total perceived constructive communication. That is to say, couples in which men used a high proportion of negative emotion words in discussions of HIV behavioral health within the relationship scored lower on total constructive communication. Although no studies have examined associations of emotion word speech and perceived communication among couples, these findings are in line with research in which individual negative emotion word use has been associated with poor relational outcomes, such as family conflict (Robbins et al., 2013). Insofar as negative emotion word use may be associated with conflict more generally, it follows then that such language use might be associated with less constructive communication processes that are reflective of conflict within the relationship. As such, in the present study, greater use of negative words among couples may be indicative of less constructive communication dynamics among couples.

In interpreting these associations, it is important to consider that only the sums of first-person pronoun and negative word use were associated with total perceived constructive communication, while within-couple differences in these variables were not significant. This pattern suggests the importance of linguistic symmetry, which may underlie greater total constructive communication. Research suggests that linguistic style matching (i.e., similarity in function word use, such as pronoun and emotion word use within dyads; as pioneered with LIWC in Niederhoffer & Pennebaker, 2002) predicts productive relational outcomes, such as relationship stability (Ireland et al., 2011) as well as interpersonal processes relevant to communication, such as group cooperation (Gonzalez, Hancock, & Pennebaker, 2009). This interpretation makes sense in light of recent research in which asymmetrical language use (i.e., in the present study, differences in percentages of linguistic categories) within couples was shown to be negatively associated with optimal relational outcomes, including perceived constructive communication (Rentscher et al., 2013). In the present study, concordant (i.e., symmetrical) use of first-person pronoun and negative emotion words among partnered gay men may reflect greater or diminished dyadic constructive communication, respectively.

Results from the current study have clinical implications for partnered gay men. As perceived constructive communication has been linked to a variety of HIV risk behaviors and preventive attitudes (Darbes et al., 2014; Mitchell, 2015; Mitchell, 2014b), language use associated with this perceived communication could inform couple-level approaches to HIV prevention and sexual health for partnered gay men. According to Pennebaker (2015), language use reflects rather than drives ongoing psychological processes. In fact, previous studies examining language and health among couples hazard that these linguistic measures should serve as proxies for dyadic processes themselves that account for clinical change among couples rather than the intervention target itself (e.g., Davidovich et al 2006; Rohrbaugh et al., 2012). As such, given the current results, clinicians and healthcare providers could be trained to identify certain language patterns in their interactions with gays couples in the context of HIV prevention interventions. Changes over time in first-person pronoun and negative emotion word use likely are indicative of changes in constructive communication, a dyadic process hypothesized as one mechanism of change in such dyadic interventions (Davidovich et al., 2006).

Relatedly, and of particular importance, another implication of the current results is its relevance at both the dyadic- and individual-level. Because linguistic difference scores are not predictive of perceived communication sums (but sums are), results suggest that patterns of linguistic-communication associations are the same at the individual-level as the couple-level. Because of this property, the aforementioned clinical strategy could apply to interactions with individual partnered gay men as well. As CHTC has not been widely adopted within healthcare agencies across the United States (Stephenson et al., 2016), many partnered gay men are likely to present individually for sexual health services, such as HIV testing. In these contexts, particularly when subjects of main partnerships and HIV testing are broached, partnered gay men who demonstrate low levels of first-person pronouns (e.g., I-talk or We-talk) and high levels of negative emotion word use may indicate individuals in relationship in which there is little constructive communication who would be good candidates for interventions, such as CHTC, to promote HIV behavioral health among gay couples.

Despite clinical potential, the present study must be viewed in light of several limitations. First, because effect sizes were moderate to large even in non-significant models (i.e., those predicting dyadic avoidant communication scores), null findings should be viewed cautiously. These findings represent a preliminary examination of the correspondence between individual language patterns and perceptions of communication styles. Future studies, particularly of avoidant communication, are warranted. Second, in the current study, linguistic variables were derived from individual interviews, with private language use conceptualized as consistent with the individualized current nature of practitioner-client interactions in healthcare settings. While such an approach represents a novel method to analyze interview data, as most studies examining language use have been within the context of a dyadic interaction, future research should also investigate the potential influence of the partner on language use (i.e., publicly; when members of the couples are interacting together).

Another important limitation of the current study is the conceptualization of communication patterns. In the present study, only perceived constructive and avoidant communication were examined. These two communication subtypes, while often the focus of important research investigating relational processes related to HIV behavioral health among gay couples (e.g., Darbes et al., 2014; Mitchell, 2014b; Mitchell, 2015), arguably are one-sided. In particular, constructive and avoidant communication patterns are considered patterns of positive communication (Christensen & Shenk, 1991), representing positive engagement and benign disengagement, respectively. However, perceived communication style encompasses both positive and negative approaches to interaction, the latter of which are not measured with the CPQ. Namely, absent in the current study are negative engagement and negative disengagement. As such, future research into the linguistic underpinnings of perceived communication patterns should supplement the CPQ with other measures of perceived negative communication style (i.e., negative engagement and negative avoidance) in order to inform more fully the development of HIV prevention interventions for partnered gay men.

Although the current study sought to examine the linguistic underpinnings of perceived communication among gay couples as distal indicators of HIV behavioral health among couples, future studies should examine direct associations of language use and actual HIV preventive risk and behavior. These behavioral outcomes may include a variety of behaviors associated with perceived communication patterns, such as sexual risk behavior (Darbes et al., 2014), substance use during sex (Mitchell, 2015), as well as willingness to participate in CHTC (Mitchell, 2014b). Further, while the present analyses provide insight into the linguistic underpinnings of perceived communication associated with traditional HIV behavioral health outcomes among couples, it must be noted that the majority of these data were collected prior to the roll-out of pre-exposure prophylaxis (PrEP) in the summer of 2012 as an HIV prevention strategy (USFDA, 2012). Thus, in addition to the examination of direct relationships between dyadic language and behavioral outcomes, future research should replicate the current examination of language use and its association with perceived communication patterns in conversations surrounding pre-exposure prophylaxis (PrEP) as a prevention method among gay male couples.

Despite these limitations, the present study adds to a growing body of literature with respect to linguistic approaches to dyadic health navigation. In particular, this study is the first to investigate the correspondence between individual language use and perceptions of dyadic communication among gay couples and, after Rentscher and colleagues (2013), represents only the second study to examine the dyad as the unit of analysis among studies examining health navigation for couples. Results indicate that there is a correspondence between the language utilized by individual partnered men when discussing HIV-related issues and perceptions of specific types of communication that have been linked to HIV-related risk and prevention behavior. As partnered gay men, particularly those ages 18 to 29, represent a group particularly vulnerable to HIV infection, this study points to the importance of language use among gay couples in relation to HIV prevention, laying the foundation for future research to examine the influence of language use on key relational processes, such as perceived constructive communication, which predict HIV-related health among gay couples.