The Story of Us: Older and Younger Couples’ Language Use and Emotional Responses to Jointly Told Relationship Narratives

Abstract

Objectives

Social-emotional well-being is said to improve over adulthood, and studies of couples’ age differences have focused primarily on marital conflict. The way couples discuss their relationship story predicts marital quality among newlyweds and long-married couples alike, yet older and younger couples’ accounts have never been compared. The current study examined age differences in couples’ use of I/we-talk, emotion words, and immediacy (i.e., an urgent and unresolved style) during a relationship history discussion and their subsequent mood reactivity and appraisals.

Method

Married couples (N = 186 individuals within 93 couples, aged 22–77) recounted the story of their relationship then rated the discussion and their negative mood. Mediation models assessed the 3 linguistic features as parallel dyadic mediators linking couple age to negative mood responses and appraisals, controlling for global marital satisfaction, and baseline negative mood. Secondary analyses examined partners’ concordance in language use.

Results

Compared with younger couples, older couples used more positive than negative words and less immediacy which, in turn, was associated with husbands’ and wives’ less negative mood and more positive appraisals, only among husbands. Partners in older couples used more similar I/we-talk and emotional language, but these were unrelated to mood or appraisals.

Discussion

This study extends our understanding of how marital interactions differ by age in the understudied context of relationship history discussions, which may grow increasingly important for couples’ well-being with older age. Findings broadly align with social-emotional aging theories and uncover novel linguistic features relevant to the age-related emotional benefits of joint reminiscing.

As social networks shrink in middle and older age, close relationships grow increasingly important for health and well-being (Kiecolt-Glaser & Wilson, 2017). According to social-emotional aging theories, relationships also become more satisfying and less strained in older adulthood (Carstensen, 1995; Charles, 2010; Rook & Charles, 2017). In support of these theories, individuals’ self-reports of interpersonal conflict decline reliably across adulthood (Rook & Charles, 2017). In parallel, studies of couples’ age differences in marital interactions have focused primarily on behavior and reactions to marital conflict, with some notable exceptions (Smith, Berg et al., 2009; Wilson et al., 2019). However, if marital conflict decreases in frequency and intensity with older age as theories imply, interaction contexts beyond disagreement may grow more important for older couples’ health and well-being by comparison. Nevertheless, other marital interactions are understudied and more poorly understood.

Joint reminiscence may be such an activity that takes on greater centrality with age. Indeed, individuals engage in more frequent reminiscence and life review with older age, and this has direct ties to well-being (Bohlmeijer et al., 2007). Moreover, the language and behaviors couples use when they reminisce about their relationship, during what is called the oral history task, predict relationship function in older and younger couples alike. For example, in a sample of older couples, those who told their stories with more positive affect, less negative affect, greater communication skills, and higher levels of engagement rated their marriages as more satisfying (McCoy et al., 2017). Higher positive affect, lower negative affect, and greater engagement during the oral history task also foreshadowed higher marital satisfaction one year later. Studies with younger and middle-aged couples have offered similar evidence linking behavior during the oral history task to marital quality (Doohan et al., 2010; Holmberg et al., 2004). Furthermore, in a study of parents with young children, more negative behavior and use of fewer we pronouns, that is, we-talk, during the oral history discussion predicted a greater likelihood of divorce three years later (Buehlman et al., 1992). However, no studies to date have examined whether the characteristics of couples’ relationship narratives differ as a function of age and if these divergent patterns relate to differences in mood reactivity and post-discussion appraisals.

Building on this prior work, the current study aimed to compare couples’ jointly told relationship narratives and their associations with mood and discussion appraisals by age. We focused on couples’ average age because it provided the best proxy for the developmental framework that grounded our questions (Carstensen, 1995; Charles, 2010) in the context of the oral history task, a dyadic activity. We examined post-discussion negative mood as a primary outcome, rather than positive mood. Indeed, most positive affect scales, including ours in the current study, are biased toward high arousal (i.e., overt excitement; Crawford & Henry, 2004), and therefore unlikely to capture older adults’ emotional experience and many couples’ experiences in the oral history task.

We examined three linguistic markers in the relationship narratives: first-person pronouns, emotional language, and verbal immediacy. These were treated as couple-level characteristics given that couples jointly constructed their relationship narratives in private, unstructured discussions. These features were of particular interest because they met three key criteria: (a) according to theory or prior work, they were likely to differ by age; (b) they shared empirical connections to well-being; and (c) they fit the context of the joint relationship narrative discussion. In prior work, older couples used more we-talk during a marital disagreement (Sillars et al., 1997), which aligns with aging theories that suggest older individuals increasingly prioritize social connection and employ proactive coping strategies (Carstensen, 1995; Charles, 2010). In turn, more we-talk and less frequent use of singular first-person pronouns (I, me, and my) have tracked with greater relationship closeness and quality, as well as higher emotional well-being in couples and individuals of all ages (Karan et al., 2019; Rentscher et al., 2013; Robbins et al., 2013; Sillars et al., 1997). Extending this work, we hypothesized that older couples’ relationship accounts would feature more we-talk relative to I-talk than their younger counterparts’ narratives; in turn, we predicted that more we- versus I-talk would relate to lower post-discussion negative mood and more positive discussion appraisals.

In line with aging theories and prior studies that suggest negativity declines and positivity may increase with age (Carstensen, 1995; Rook & Charles, 2017), we hypothesized that older couples’ relationship narratives would feature more positive emotional language relative to negative emotion words, and that this ratio would be larger than in younger couples’ narratives. Congruent with patterns in other oral history studies that more positive behavior predicts better emotional and marital outcomes (Buehlman et al., 1992; Doohan et al., 2010; McCoy et al., 2017), we expected that more positive language compared to negative would be associated with less negative mood and more positive appraisals. Third, we examined verbal immediacy, a well-characterized linguistic signature that captures a lack of resolution about past events (Cohn et al., 2004; Lee et al., 2011). Marked by excessive focus on the present and the self, as well as discrepancy language (e.g., should and would) and a lack of articles and long words, measures of verbal immediacy indicate that an experience is still actively being processed with some urgency and uncertainty. Higher levels of verbal immediacy about past events are reliably linked to unfavorable outcomes such as prolonged physiological reactivity and riskier health behaviors (Lee et al., 2011). Given older adults’ shift toward social connection and their strengths in emotion regulation (Carstensen, 1995; Charles, 2010), we predicted that older couples’ narratives would feature lower immediacy compared with younger couples’, and that this, in turn, would be linked to more positive appraisals and lower negative mood. To minimize the number of statistical tests, all three linguistic markers were tested as parallel mediators linking age to post-discussion negative mood and discussion appraisals.

Beyond older adults’ theorized focus on social connection and emotional fulfillment, aging couples have been shown to converge in their emotions, behaviors, and health (Kiecolt-Glaser & Wilson, 2017). Accordingly, secondary analyses tested whether age was associated with stronger linguistic concordance between the partners. Indeed, it is unknown whether age-related convergence extends to couples’ linguistic mimicry in shared conversations about their marital history. In turn, concordant levels of disclosure and language styles may promote both partners’ well-being (Hagedoorn et al., 2011; Karan, 2019). Thus, we also explored whether stronger linguistic concordance was associated with more positive discussion appraisals and less negative mood. Finally, supplemental analyses substituted relationship length for couple age in all models to determine whether a longer shared history would mirror the associations with developmental differences.

Method

Heterosexual couples (N = 186 individuals in 93 couples) married for at least 2 years were recruited from the local community for a larger study on marital stress and wound healing (Kiecolt-Glaser et al., 2005). We excluded couples if either spouse took blood pressure medication, smoked, or had immune- or endocrine-related health problems (e.g., cancer and diabetes). The Ohio State Research Review Committee approved the project; all participants gave written informed consent before participation. The parent study was powered to detect associations between marital behavior and wound healing. No relevant past work had used a similar study design or statistical methods proposed in the parent study. Thus, sample size calculations focused on assuring a sufficiently large sample for known associations involving key variables with N = 200 individuals as the primary basis. Using one-tailed tests and α = 0.05, N = 200 provided 100% power for the r = −0.39 correlation between perceived stress scores and interleukin (IL)-8 blister fluid levels observed in a prior study (Glaser et al., 1999); N = 36 provided 80% power. In lieu of post hoc power calculations for the present analysis, confidence intervals are provided for primary estimates (Dziak et al., 2020). The sample ranged 22–77 years old (M = 38.2) and had been married for 2–52 years (M = 11.9). More than half graduated college (60.7%), and most participants were white (90.9%). Most were in their first marriage (81.5%).

Data Collection Procedure

At 7:00 a.m., couples arrived at the hospital research unit and provided baseline mood measures. After completing questionnaires, a blister wound procedure, and two social support discussions described in detail elsewhere (Kiecolt-Glaser et al., 2005), couples were given 30 min to complete an oral history task, wherein they jointly told the story of their relationship (Kiecolt-Glaser et al., 2005). A storyboard encouraged them to discuss the major milestones of the relationship—how they met, became interested in each other, as well as their wedding, major events since the wedding, how life was in the present day, and prospects for the future. Afterward, they privately rated the discussion and their mood. See Supplementary Material for a more detailed description of the oral history protocol.

Self-report Measures

Both partners reported their ages, which were averaged at the couple level (r = 0.95) and treated as a continuous variable. Participants rated their negative mood at the morning baseline period and immediately after the oral history exercise using the 10-item subscale of the Positive and Negative Affect Schedule (PANAS, Cronbach’s α = 0.83; Watson et al., 1988). Items ranged 1-Not at all to 5-Extremely, with higher scores indicating more negative mood (M_{Men’s Baseline} = 12.59, standard deviation [SD] = 3.50; M_{Women’s Baseline} = 13.37, SD = 4.31; M_{Men’s Post-discussion} = 11.90, SD = 3.75; M_{Women’s Post-discussion} = 11.84, SD = 3.38). Participants rated the quality of their oral history discussion on a 4-item, 9-point scale (Cronbach’s α = 0.94). Items assessed how satisfied the person was with the discussion’s outcome, the emotional tone, how well the partner had understood them, and how much the partner had supported them. Items were averaged such that higher scores indexed greater satisfaction with the discussion (M_Men = 7.79, SD = 1.60; M_Women = 7.67, SD = 1.47).

One of the most commonly cited marital quality scales, the 15-item marital adjustment test (MAT) was used because of its ability to discriminate between satisfied and dissatisfied couples (Cronbach’s α = 0.84; Locke & Wallace, 1959). Participants completed this measure at the time of screening, prior to the study visits. The items vary in their anchors and scoring, with larger scales given to those thought to carry more weight in characterizing a person’s overall marital satisfaction. Partners’ scores were strongly correlated (r = 0.70) and, thus, were averaged to parallel the other dyadic variables in the analysis. A sizable subset of couples (19.4%) met the cutoff for clinically significant marital distress (<100 on the MAT scale ranging 2–158).

Oral History Language Use

Transcribed oral history discussions were processed with Linguistic Inquiry and Word Count (Pennebaker, Austin, TX), text analysis software that produces percentages of words in psychologically relevant categories based on a standardized dictionary. Of interest for the current study were positive and negative emotion words, singular and plural first-person pronouns, and verbal immediacy. The ratio of negative-to-positive emotion word percentages reflected the emotional quality of the narrative, with larger values indexing a more negative emotional focus. The positivity effect suggests that older adults tend to recall more positive than negative content (Reed & Carstensen, 2012), which would result in a greater imbalance in negative compared to positive emotional information. The I/we ratio indexed the degree to which the narratives focused on the joint identity of the couple over their individuality. According to a few prior studies, the relative use of singular to plural first-person pronouns more strongly predicts health and well-being compared with we-talk alone (Rentscher et al., 2013; Robbins et al., 2013; Rohrbaugh et al., 2008; Rohrbaugh et al., 2012). Larger I/we values captured greater individual focus (i.e., larger percentage of singular pronouns compared with the percentage of plural pronouns). In addition to the conceptual impetus for constructing these two variables as ratios, the negative/positive and I/we ratios had the additional advantage of generating fewer hypothesis tests, limiting Type I error.

Third, we examined age differences in verbal immediacy, a linguistic signature marked by excessive focus on the present and on the self, reflecting a lack of resolution and of psychological distancing from past events (Cohn et al., 2004; Pennebaker & King, 1999). Developed in prior work using factor analysis and validated in subsequent studies (Cohn et al., 2004; Lee et al., 2011; Pennebaker & King, 1999), five categories (all percentages of the total words in the joint narrative) were z-scored and averaged together with equal weighting to represent immediacy: more present-tense verbs, more first-person singular pronouns, more words that cued discrepancy (e.g., should and would), fewer articles (i.e., the inverse of the article category, a, an, the, etc.), and fewer long words (i.e., the inverse of words with more than six letters).

Analytic Approach

The dyadic mediation macro MEDYAD (Coutts et al., 2019) used within SAS 9.4 (Cary, NC) evaluated associations between age and linguistic features of the relationship narrative, and in turn, whether the narrative features mediated links between age and negative mood reactivity and discussion satisfaction. The negative/positive ratio, I/we ratio, and immediacy served as parallel mediators, given that all three emerged from the same conversation. Because the protocol asked couples to jointly tell the story of their relationship, age and all three linguistic variables were examined at the couple level. The MEDYAD macro required the outcomes to be treated as distinguishable, at the individual level; thus, husbands’ and wives’ negative moods and discussion satisfaction were examined as separate outcomes in these dyadic models. Path A (age → narrative features) controlled couple marital satisfaction to account for the expectation that happier couples would tell stories with more positive than negative emotion, more collective than individual language, and less immediacy. This covariate also teased apart the role of age from the contribution of marital satisfaction (Smith, Berg et al., 2009). In addition to couple marital satisfaction, Path B (narrative features → outcomes) covaried baseline negative mood to enable conclusions about mood reactivity and account for the link between pre-task mood and discussion satisfaction. For indirect effects, standard errors and 95% confidence intervals were constructed with 5,000 bootstrapped samples. Effect sizes are described in Supplementary Material (pp. S2-S6, Figures S1-S4).

Secondary models explored the association between couples’ age and their linguistic concordance in heterogeneous-variance multilevel models using PROC MIXED in SAS 9.4. This was the best method for assessing the role of age in linguistic concordance because it allowed covariates in both fixed and random parts of the model (see also Wilson & Novak, 2021). The fixed effects of couple age and couple marital satisfaction served as covariates to account for systematic differences in their levels of negative/positive emotion word use, I/we use, and immediacy. Couple marital satisfaction also served as a random covariate to account for the anticipated association between higher marital satisfaction and stronger linguistic concordance between partners. To explore the mediating roles of linguistic concordance linking age to mood outcomes, we used the absolute value of couples’ difference scores as proxies for linguistic concordance in MEDYAD, with values closer to zero indexing greater concordance. Currently, there is no established method for carrying out parallel mediation in heterogeneous-variance models. We consider the difference-score method a preliminary step and encourage future research to extend heterogeneous-variance models to accommodate parallel mediation. Covariates in these secondary mediation models were the same as those in the primary mediation models.

Given the strong correlation between age and relationship length, age could not meaningfully be teased apart from marital duration in the analytic models. Instead, we treated couple age as a proxy for both broad developmental processes that may drive social-emotional changes and the couple’s shared history, and substituted relationship length for couple age.

Results

Descriptive Statistics and Preliminary Analyses

Consistent with prior marital studies (e.g., Carstensen et al., 1995), older couples in our sample had also been married longer than younger couples. The correlation between age and marital satisfaction was nonsignificant (Table 1). As expected, older age and greater marital satisfaction correlated with more positive than negative emotional content and lower immediacy in the relationship narrative (Table 1). Correlations with the I/we ratio trended in the anticipated direction but were nonsignificant. Finally, a stronger focus on the self-relative to the couple, that is, higher I/we scores, tracked with greater immediacy with moderate magnitude, which mirrors the fact that both indices included first-person singular pronouns in their calculations. Relationship narratives consisted of 1828-7157 total words (M = 4,739, SD = 1,076).

Table 1.

Description of Study Predictors

Variable	1	2	3	4	5	6	M (SD)
1. Couple age	―						38.2 (12.4)
2. Years married	0.84*	―					11.9 (11.0)
3. Couple marital satisfaction	0.14	0.12	―				114.6 (21.0)
4. I/we ratio	−0.17	−0.09	−0.19‡	―			2.1 (0.9)
5. Negative/positive ratio	−0.26*	−0.15	−0.22*	0.20‡	―		0.3 (0.1)
6. Immediacy	−0.28*	−0.10	−0.28*	0.41*	−0.18‡	―	0.0 (0.6)

Notes: Ns = 89–93. *p < .05; ‡p < .08. SD = standard deviation.

Linguistic Features as Mediators of the Association Between Couple Age and Post-discussion Mood and Appraisals

Controlling for couple marital satisfaction, older couples’ narratives featured lower immediacy (Figure 1, B = −0.012, standard error [SE] = 0.005, p = .019, 95% confidence interval [CI; −0.022, −0.002]) and more positive compared to negative emotional language (B = −0.003, SE = 0.001, p = .026, 95% CI [−0.005, −0.0003]). Husbands reported lower post-discussion negative mood following narratives that focused more on the couple than the self (B = −0.956, SE = 0.452, p = .038, 95% CI [−1.855, −0.056]) and featured less immediacy (B = 1.768, SE = 0.685, p = .012, 95% CI [0.406, 3.130]). Wives’ negative mood reactivity did not track significantly with linguistic features, but associations fell in the expected direction with lower immediacy (B = 1.188, SE = 0.659, p = .075, 95% CI [−0.122, 2.499]). Tests of indirect effects revealed that lower immediacy in the relationship narratives served as a significant mediator linking older age and lower negative mood for both husbands (indirect estimate = −0.021, bootstrapped SE = 0.016, 95% CI [−0.060, −0.0001]) and wives (indirect estimate = −0.014, bootstrapped SE = 0.010, 95% CI [−0.038, −0.001]).

Figure 1.

Parallel mediation model tested using MEDYAD (Coutts et al., 2019). Solid black lines reflect statistically significant effects; dotted, gray lines denote paths that were tested and found to be nonsignificant. Couple marital satisfaction served as a covariate in Paths A and B; husbands’ and wives’ baseline negative mood also were controlled in Path B. *p < .05, ‡p = .075. Supplemental Table S1 provides all estimates and 95% confidence intervals.

As shown in Figure 2, lower immediacy related to husbands’ greater discussion satisfaction (B = −0.639, SE = 0.299, p = .036, 95% CI [−1.233, −0.044]), and served as a significant mediator linking older couple age to higher discussion satisfaction for husbands (indirect effect = 0.008, bootstrapped SE = 0.005, 95% CI [0.001, 0.020]), but not for wives.

Figure 2.

Parallel mediation model tested using MEDYAD. Solid black lines reflect statistically significant effects; dotted, gray lines denote paths that were tested and found to be nonsignificant. Couple marital satisfaction served as a covariate in Paths A and B; husbands’ and wives’ baseline negative mood also were controlled in Path B. *p < .05. Supplemental Table S2 provides all estimates and 95% confidence intervals.

Couple Age and Language Use Concordance During Oral History

According to random effects in heterogeneous-variance models, older couples used more similar I/we language during the oral history than did younger counterparts (Supplemental Table S3, estimate = −0.055, SE = 0.011, p < .0001; X(1) = 20.6, p < .0001). Likewise, older couples used more similar ratios of negative-to-positive emotion words compared with younger couples (estimate = −0.024, SE = 0.010, p = .021; X(1) = 4.8, p = .028). Older age was unrelated to greater couple concordance in immediacy (p > .250). In the absence of a validated method to estimate parallel mediation in dyadic heterogeneous-variance models, we explored the mediating roles of linguistic concordance using the absolute values of difference scores in MEDYAD. The associations between couple age and the difference scores (i.e., Path A) were not significant (Supplemental Tables S4-S5, ps > .250). In turn, the difference scores were unrelated to mood reactivity or discussion appraisal, with the exception of one association: women rated the oral history discussion as more satisfying when they and their husbands had more concordant immediacy scores (B = −0.997, SE = 0.422, p = .021, 95% CI [−1.836, 0.157]). No indirect effects were statistically significant.

Relationship Length as an Alternative to Couple Age

To determine the contribution of the couple’s shared history to hypothesized associations, relationship length was substituted for couple age in all models. Relationship length was unrelated to the three linguistic features (ps > .241), and thus was not indirectly linked to mood or appraisals through these features. Heterogeneous-variance models replicated stronger I/we concordance in longer-married couples (Estimate = −0.045, SE = 0.012, p < .0001; X(1) = 11.4, p = .001), but no significant differences emerged in emotional concordance (p = .197). Similar to age findings, relationship length was unrelated to immediacy concordance, nor was relationship length indirectly tied to mood reactivity or appraisals through a linguistic concordance.

Discussion

In a lab-based study of married couples ages 22–77, older couples recounted their relationship’s history with more positive than negative emotional language and lower immediacy (i.e., greater psychological distancing) compared to younger counterparts. In turn, this pattern of lower immediacy associated with older age related to less negative mood reactivity in husbands and wives, and more positive appraisals in husbands. Partners in older couples also matched each other’s language more closely, with more similar I/we ratios and emotional language; however, the degree of concordance was unrelated to mood responses. This study targeted relationship history discussions as a novel interaction context that is salient across the adult life span and may grow increasingly important with age, compared to other interactions. The findings identified linguistic features that may be related to emotional benefits of joint reminiscing for older couples.

Of the three linguistic features expected to be emotionally relevant and to differ with age, lower immediacy arose as the only factor linking older age to more favorable mood responses in husbands and wives, as well as more positive appraisals in husbands. From a clinical science perspective, speaking with high levels of immediacy about past events reflects unresolved problems. According to Lee et al. (2011), the heavy use of present-tense language, shorter words, and fewer articles point to an in-the-moment “overinvolvement with one’s own thoughts and feelings” (p. 287) that correlates with anxious attachment in close relationships as well as exaggerated physiological arousal. In a sample of people who separated from their partners within the previous year, talking about the separation with greater immediacy predicted larger systolic and diastolic blood pressure responses, particularly among anxiously attached individuals (Lee et al., 2011). This prior study included age as a covariate rather than a predictor of interest, but descriptive correlations between focal predictors and covariates revealed that older divorcees spoke about their separation with less immediacy, relative to younger counterparts. Extending this work, our data identified a similar adaptive pattern, that older couples told their relationship story with greater resolution compared with younger couples.

Notably, these associations emerged above and beyond couples’ marital satisfaction, except in wives’ discussion appraisals. Indeed, wives’ satisfaction with the oral history discussion only tracked with couples’ global satisfaction, a powerful lens through which marital behavior is viewed (Kiecolt-Glaser & Newton, 2001). If women are more sensitive to marital dynamics and thus serve as the emotional “barometer” of the relationship (Floyd & Markman, 1983), discussions may have tracked more closely with wives’ expectations, which are closely tied to global satisfaction (McNulty & Karney, 2002). Beyond this gender difference, it is important to note that studies designed to probe age differences in marital interactions have produced varying results in relation to global marital satisfaction. In a landmark study of 300 Utah couples, older adults reported higher global satisfaction, which in turn related to less negative mood reactivity to marital conflict (Smith, Uchino et al., 2009). However, a prominent Berkeley study recruited 156 middle-aged and older couples with separate subgroups of high and low marital satisfaction and found meaningful behavioral differences among the four groups (Carstensen et al., 1995). In this regard, the role of marital satisfaction in couples’ age differences partially depends on the correlation of age and marital satisfaction in the sample, which is influenced by sampling and recruitment strategies. The current study did not use a formal quota sampling approach to disaggregate age from marital satisfaction, yet the correlation between age and satisfaction was nonsignificant. This result implies that age differences in immediacy may arise from developmental patterns, regardless of satisfaction level.

Although previous work has not examined immediacy through an aging lens, our finding aligns clearly with life course and social-emotional aging theories. First, older couples are likely to share a long history and, thus, may have spent a greater portion of their story recounting past events. However, shared history alone does not explain the findings, as relationship length did not replicate associations with lower immediacy in Supplementary Material. Indeed, for older couples, the major life stressors that pose the greatest risks for longstanding, intractable problems may also have passed. According to prior work, major life stressors peak in midlife, when couples are managing multiple roles at work, in their family, and in the community (Almeida & Horn, 2004). Greater immediacy among younger and middle-aged couples may reflect, in part, being in the midst of these challenging, unresolved stressors.

According to socioemotional selectivity theory (Carstensen, 1995), older adults’ motivations shift to prioritize emotional fulfillment over future-oriented, instrumental goals. With more life lived than yet to live, reminiscing about the past may grow increasingly central to emotional well-being. Indeed, older adults engage in reminiscence and life review with greater frequency compared with younger counterparts (Bohlmeijer et al., 2007). In tandem, autobiographical reasoning may also grow more sophisticated across the adult life span. From the personal narratives of adults ages 18–89, Pasupathi and Mansour (2006) found that older adults drew more connections between personal life events and their understanding of themselves compared with younger counterparts. Taken together, our finding that older couples told their relationship stories with lower immediacy—that is, less psychological urgency and greater resolution—and that this quality related to more positive appraisals and less negative emotional reactivity extends our understanding of couples’ aging-related dynamics in ways that align with theory.

Contrary to prediction, emotional language and the use of personal pronouns did not link associations of age with appraisals and mood. Older adults did use more positive emotion words relative to negative words compared with younger adults, which extends support for older adults’ positivity bias to the domain of couples’ unstructured retelling of their relationship history. In turn, however, the ratio of positive/negative emotion words did not uniquely relate to mood or appraisals. Nevertheless, prior work has not juxtaposed the positivity effect with other relevant factors, and its relative importance may depend on the nature of the task. Low immediacy may be particularly important in the context of co-constructing a story largely about the past, whereas emotional language may supersede immediacy in current experiences that evoke negative or positive emotion, such as an ongoing stressor or uplift.

In contrast to the positivity pattern, I/we language only related to husbands’ mood and was not associated with age. This gender difference mirrors the association found with husbands’ appraisals but not with wives’. Indeed, global satisfaction shared a particularly robust relationship with wives’ negative mood reactivity, suggesting that wives were especially attuned to the relationship’s overall state. Accordingly, wives responded more to global satisfaction, whereas husbands reacted to the specific features of their relationship narrative, with favorable mood responses following less individualistic, more communal styles. To the best of our knowledge, only one prior study identified a correlation between older age and more we-talk, which emerged during a marital disagreement discussion (Sillars et al., 1997). In contrast, our study examined I/we language alongside two other linguistic features that were weakly to moderately interrelated, which created a stricter test. Moreover, age differences in I/we-talk may emerge more clearly in discussions that highlight personal differences between partners, for example, marital conflict. In addition, the importance of personal pronoun use for well-being may depend on the study’s context. One study showed that we-talk sampled from individuals’ private thoughts tends to predict individual appraisals, whereas we-talk in couples’ conversations tracks more with their observed behavior (Wilson et al., in press). Accordingly, personal pronouns may have been more strongly associated with mood and appraisals if each partner had privately recounted the story of their relationship, rather than telling the story as a couple.

To capture the characteristics that emerged from couples’ joint relationship narrative, in our primary analyses, we examined the three linguistic features at the dyadic, whole-story level. In a secondary step, we explored whether partners in older couples mirrored each other’s language more closely, and if this concordance was important for mood and discussion appraisals. Indeed, partners converge in their emotions, health, and well-being with increasing age (Kiecolt-Glaser & Wilson, 2017). Extending this pattern, our results uncovered stronger concordance in I/we and negative/positive language among older couples. Longer marriages replicated associations with stronger I/we concordance, but not with emotional concordance, suggesting that a long shared history may track with converging language around interdependence, but not around emotionality. These dimensions of concordance were unrelated to mood or appraisals, but notably, we had to test the two sets of predictions with different models. To examine age as a predictor of linguistic concordance, we used heterogeneous-variance models, a powerful method that includes standard errors and residuals along with point estimates. However, there is currently no way to evaluate mediation in this kind of model, and thus, we had to examine linguistic concordance as mediators with different scores in the traditional dyadic multilevel model, using MEDYAD. Difference scores compound the error in both individuals’ measures, and thus may have lower reliability. Therefore, linguistic concordance may play a role in mood and discussion appraisals in a way that cannot be captured by the methods currently available—a limitation of the study.

In addition to the drawback in testing linguistic concordance as a predictor, the sample consisted of heterosexual, mostly White, educated couples, another limitation. Follow-up studies should evaluate these questions in a more diverse sample representing sexual minorities and people of color, across the socioeconomic spectrum. Socioeconomic hardship strains relationships, and many couples of color and sexual minority couples experience unique stressors related to discrimination and stigma. Reflecting on their shared journey may provide a source of strength and resilience, particularly among older couples who have faced these challenges. Given the strong correlations between partners’ ages and marriage length, their unique contributions could not be isolated. To tease apart the contributions of partners’ individual ages, future studies will need to recruit partners whose ages are less redundant, and use more individualized tasks. To disambiguate the role of age from shared relationship history, studies will need to recruit larger numbers of older couples in newer relationships and middle-aged couples in long-lasting marriages.

Practical Significance

It is valuable to consider the findings in practical terms. To illustrate, for every 100 positive words a younger couple (i.e., age 25) shared in their relationship narrative, they also uttered 33.6 negative words (Supplementary Material, Supplementary Figure S1). In contrast, an older couple (i.e., age 75) expressed only 20.6 negative words for every 100 positive words. As shown in Supplementary Figure S3, for a husband with average values on all covariates, an average level of immediacy translated to a negative mood score of ~12 (Supplementary Figure S3). According to Crawford and Henry (2004), this corresponded to the ~18th percentile in a large, life span sample. A score 1 SD above the mean on immediacy translated to a negative mood score of ~13, or the 38th percentile. Likewise, the ~1-point mood difference between older and younger couples attributable to associations with immediacy would translate to a 6–10-point percentile difference on the PANAS scale, a meaningful contrast according to Crawford and Henry’s norms.

If couples engage in similar dynamics in daily life, these differences shown in the laboratory may build over long periods of time to affect well-being and relationship function. Indeed, researchers have long studied reactivity to lab-based marital interactions to understand the mechanisms that link marriage to health and well-being (Heyman, 2001). Emotional responses to lab-based marital conflict reflect important differences in marital satisfaction, for example (Whisman et al., 2002). In a cutting-edge study, heart rate reactivity to marital conflict at home exceeded reactivity to similar discussions in the laboratory (Baucom et al., 2018), indicating that the responses in the lab may underestimate the effect of marital interactions in daily life. Nevertheless, future work must empirically assess whether the current findings extend to couples’ daily lives—how frequently couples reminisce together and negotiate their joint relationship story, as well as whether this practice affects daily and long-term well-being.

Conclusions

In summary, the current study uncovered that older couples jointly told their relationship history with less immediacy and more positivity than negativity, compared to younger couples. In turn, lower immediacy was related to husbands’ and wives’ subsequent negative mood, as well as husbands’ discussion appraisals. In this way, the study identified novel linguistic features relevant to appraisals and emotional benefits of joint reminiscing. These findings broadly align with social-emotional aging theories and extend our understanding of couples’ age differences in the context of joint reminiscence, an understudied type of interaction that may grow increasingly important for couples’ well-being with age.

Funding

This work was supported by the National Institutes of Health (UL1RR025755, P50 DE13749, P01 AG16321 to J. K. Kiecolt-Glaser; R01 AG057032 to J. K. Kiecolt-Glaser; K05 CA172296 to J. K. Kiecolt-Glaser; R00 AG056667 to S. J. Wilson; and L30 AG06025 to S. J. Wilson).

Conflict of Interest

None declared.