Relational uncertainty and dyadic synchrony within the interaction of couples

Participants

The sample consisted of 97 heterosexual, mixed gender couples (N = 194 individuals; 97 cisgender women and 97 cisgender men). A total of 74 couples were married, 16 were cohabiting, and 7 were dating. Their average relationship length was 10.48 years (SD = 11.02 years, range = 3 months to 65 years); 68.8% had children. Participants ranged in age from 21 to 90 years old (M = 42.73, Mdn = 41.00, SD = 12.11). Participants identified as White (69.0%), Black (14.7%), Latinx (8.7%), Asian (5.4%), Native American/Pacific Islander (1.6%) and Biracial (.6%). Couples reported their annual household income as under $10,000 (5.4%), $10,000–$40,000 (24.5%), $41,000–$70,000 (19.8%), $71,000–$100,000 (22.8%), and over $100,000 (27.5%).

Measures

Observational assessment of interaction

Couples participated in a series of six standardized discussions that were videotaped and transcribed. The first discussion, which was designed to help couples get acclimated to the research laboratory and videotaping procedures, asked couples to plan their next vacation. The following two discussions invited couples to discuss recent conflicts in their relationship. Next, couples completed two conversations in which they were asked to describe what happens when each partner feels sad, down, or worried, and how they handle those situations together. We selected data for this study from the final discussion, in which couples were asked to talk about the three best things in their relationship. This conversation lasted for 5 minutes, and was designed to elicit expressions of self-disclosure, validation, and intimacy.

Preliminary analyses

Descriptive statistics are displayed in Table 1. The number of talk turns per participant within the 5-minute interaction ranged from 3–57 (men: M = 23.42, SD = 11.79, range = 3–57; women: M = 23.38, SD = 11.75, range = 3–57). The number of talk turns containing self-disclosure ranged from 0–32 (men: M = 6.30, SD = 5.23, range = 0–32; women: M = 5.65, SD = 4.15, range = 0–18), the number of talk turns containing validation ranged from 0–25 (men: M = 6.96, SD = 4.84, range = 0–23; women: M = 7.25, SD = 4.88, range = 0–25), and the number of talk turns containing hostility ranged from 0–22 (men: M = 3.24, SD = 4.66, range = 0–22; women: M = 4.02, SD = 4.36, range = 0–17).

OPEN IN VIEWER

Table 1.

Paired-sample t tests comparing men and women on the independent and dependent variables

	Women		Men		t (96)
	M	SD	M	SD
Relationship quality	83.036	23.508	90.123	20.035	−3.618***
Self uncertainty	2.862	1.294	2.552	1.083	2.507*
Partner uncertainty	2.955	1.481	2.758	1.113	1.406
Relationship uncertainty	3.059	1.271	2.760	1.083	2.648**
Total number of talk turns	23.381	11.748	23.423	11.792	−.575
Self-disclosure talk turns	5.650	4.146	6.299	5.234	−1.408
Validation talk turns	7.247	4.880	6.959	4.843	.638
Hostility talk turns	4.021	4.356	3.237	4.661	2.097*
Self-disclosure/hostility sequence	.948	1.318	.691	1.176	1.571
Validation/hostility sequence	1.010	1.237	.536	.842	3.467***

In a first preliminary analysis, we compared men and women on relationship quality; relational uncertainty; the number of talk turns containing self-disclosure, validation, or hostility; and the frequency of self-disclosure/hostility and validation/hostility sequences. Paired-sample t tests revealed that, compared to men, women exhibited lower levels of relationship quality, higher levels of self uncertainty and relationship uncertainty, more frequent talk turns containing hostility, and more frequent validation/hostility sequences. Second, we computed bivariate correlations among men, among women, and within couples (see Table 2). For both men and women, the three sources of relational uncertainty were positively associated with each other, the three sources of relational uncertainty were negatively associated with relationship quality, and the two types of interactional sequences were positively associated with each other. In addition, self uncertainty (for men) and relationship uncertainty (for both men and women) were positively correlated with talk turns containing hostility. Within couples, all variables shared strong positive correlations except the self-disclosure/hostility sequence. In a last preliminary analysis, we examined whether demographic covariates such as age, relationship length, and parental status were correlated with the three sources of relational uncertainty. No statistically significant correlations were apparent for men or for women.

OPEN IN VIEWER

Table 2.

Bivariate correlations among the independent and dependent variables for men, for women, and within couples

Variable	1	2	3	4	5	6	7	8	9
1. Relationship quality	.639***	−.689***	−.629***	−.733***	.142	.166	−.093	−.202*	−.242*
2. Self uncertainty	−.622***	.487***	.707***	.845***	−.099	−.156	.097	.240*	.298**
3. Partner uncertainty	−.643***	.752***	.461***	.743***	−.148	−.178	.127	.113	.146
4. Relationship uncertainty	−.666***	.826***	.798***	.565***	−.141	−.160	.206*	.266**	.324**
5. Self-disclosure talk turns	.083	−.158	−.179	−.135	.552***	.485***	.023	−.168	−.216*
6. Validation talk turns	.155	−.144	−.184	−.113	.449**	.580***	.059	−.218*	−.209*
7. Hostility talk turns	−.112	.216*	.194	.234*	−.140	−.136	.669**	.085	.064
8. Self-disclosure/hostility sequence	.060	−.059	−.115	−.103	.058	−.071	.173	.164	.527***
9. Validation/hostility sequence	.044	.127	.057	.102	.015	.000	.187	.537***	.205*

Hypothesis tests

We tested our hypotheses using multilevel modeling (Snijders & Bosker, 2012) to accommodate the dependency inherent in our dyadic data. All multilevel models were constructed such that individuals were nested within couples, and partners were distinguished by an actor’s sex (coded men = 1, women = −1). Following recommendations by Kenny et al. (2006), all multilevel models were estimated using restricted maximum likelihood, heterogeneous compound symmetry was employed as the covariance structure, and the predictors were grand-mean centered to facilitate interpretation of the intercepts. Each model included two random effects: variance in the intercepts and error variance. Standardized regression coefficients (βs) are provided as effect size estimates. All analyses were conducted using SPSS version 28.0.

Relational uncertainty predicting self-disclosure/hostility sequences (H1)

H1 suggested that, after covarying relationship quality, relational uncertainty is positively associated with self-disclosure/hostility sequences. To test this hypothesis, we constructed three multilevel models using self uncertainty, partner uncertainty, and relationship uncertainty as independent variables in separate models; we also employed sex and relationship quality as covariates. Thus, each model contained four Level 1 predictors: (a) sex, (b) relationship quality, (c) a source of relational uncertainty, and (d) an interaction term constructed as relational uncertainty x sex. See Table 3 for unstandardized parameter estimates from these models.

OPEN IN VIEWER

Table 3.

Results of multilevel models predicting the frequency of self-disclosure/hostility sequences

Effect	B	SE	95% CI	95% CI	t	p
			LL	UL
Self uncertainty
Intercept	.843	.095	.654	1.032	8.849	<.001***
Relationship quality	−.000	.005	−.011	.010	−.62	.951
Self uncertainty	.094	.099	−.102	.290	.947	.345
Sex	−.113	.082	−.276	.050	−1.376	.172
Self uncertainty * sex	.152	.071	.012	.292	2.152	.033*
Partner uncertainty
Intercept	.831	.097	.639	1.023	8.597	<.001***
Relationship quality	−.007	.005	−.018	.004	−1.304	.194
Partner uncertainty	−.079	.089	−.255	.097	−.883	.379
Sex	−.112	.081	−.272	.048	−1.384	.169
Partner uncertainty * sex	.118	.065	−.011	.247	1.805	.073

Effect	B	SE	95% CI	95% CI	t	p
			LL	UL
Relationship uncertainty
Intercept	.849	.096	.659	1.038	8.880	<.001***
Relationship quality	−.000	.006	−.012	.011	−.083	.934
Relationship uncertainty	.086	.105	−.120	.293	.823	.411
Sex	−.114	.081	−.275	.046	−1.415	.160
Relationship uncertainty * sex	.194	.070	.055	.332	2.770	.006**

Note. N = 97 couples.

*p < .05. **p < .01. ***p < .001.

Contrary to H1, after covarying relationship quality, self uncertainty did not predict self-disclosure/hostility sequences, β = .09, p = .35. However, an interaction emerged between self uncertainty and sex in predicting self-disclosure/hostility sequences, β = .15, p = .033. To probe this interaction, we constructed a two-intercept model to estimate separate coefficients for men and women (following recommendations by Kenny et al., 2006). Men’s self uncertainty predicted self-disclosure/hostility sequences at a level approaching significance, β = .24, p = .051, but this trend did not hold for women, β = −.06, p = .63.

No main effects emerged for partner uncertainty, β = −.08, p = .38, or relationship uncertainty, β = .08, p = .41, in the prediction of self-disclosure/hostility sequences. However, an interaction emerged between relationship uncertainty and sex in predicting self-disclosure/hostility sequences, β = .18, p = .006. To probe this interaction, we again constructed a two-intercept model to estimate separate coefficients for men and women. Although men’s relationship uncertainty predicted self-disclosure/hostility sequences, β = .27, p = .03, this association did not hold for women, β = −.10, p = .39.

Relational uncertainty predicting validation/hostility sequences (H2)

H2 predicted that, after covarying relationship quality, relational uncertainty is positively associated with validation/hostility sequences. To evaluate H2, we repeated the procedures we used to test H1 but substituted validation/hostility sequences as the dependent variable (see Table 4 for unstandardized parameter estimates from these models).

OPEN IN VIEWER

Table 4.

Results of multilevel models predicting the frequency of validation/hostility sequences

Effect	B	SE	95% CI	95% CI	t	p
			LL	UL
Self uncertainty
Intercept	.782	.080	.623	.941	9.748	<.001***
Relationship quality	.003	.004	−.006	.011	.617	.538
Self uncertainty	.200	.081	.040	.360	2.471	.014*
Sex	−.216	.069	−.352	−.079	−3.134	.002**
Self uncertainty * sex	.055	.058	−.059	.170	.958	.340
Partner uncertainty
Intercept	.777	.082	.614	.939	9.494	<.001***
Relationship quality	−.005	.004	−.013	.004	−1.092	.276
Partner uncertainty	.001	.073	−.143	.145	.013	.990
Sex	−.220	.070	−.359	−.082	−3.155	.002**
Partner uncertainty * sex	.034	.054	−.073	.140	.625	.533

Effect	B	SE	95% CI	95% CI	t	p
			LL	UL
Relationship uncertainty
Intercept	.785	.080	.625	.944	9.756	<.001***
Relationship quality	.004	.005	−.005	.013	.908	.365
Relationship uncertainty	.223	.085	.055	.392	2.616	.010**
Sex	−.219	.068	−.354	−.083	−3.203	.002**
Relationship uncertainty * sex	.077	.058	−.037	.191	1.335	.184

Note. N = 97 couples.

*p < .05. **p < .01. ***p < .001.

In support of H2, main effects for two of the three sources of relational uncertainty emerged: both self uncertainty, β = .23, p = .014, and relationship uncertainty, β = .25, p = .010, predicted validation/hostility sequences. In contrast, partner uncertainty was not associated with validation/hostility sequences, β = .01, p = .99, and no interactions between relational uncertainty and sex emerged.

Implications of the results

Ample scholarship suggests that intimacy develops through sequences of dyadic synchrony in which an individual self-discloses and a partner offers a validating response (e.g., Reis & Patrick, 1996; Reis & Shaver, 1988). Based on logic that people who are unsure about their relationship may have difficulty coordinating intimacy within interaction, we hypothesized that relational uncertainty corresponds with asynchronous sequences of self-disclosure met with hostility (H1). Our findings showed mixed support for this hypothesis: Men, but not women, who were experiencing relationship uncertainty (and to a lesser extent, self uncertainty) were more likely to engage in self-disclosure that was followed by hostility from their partner. Whereas prior work has documented a negative bivariate association between relational uncertainty and self-disclosure (Imai et al., 2023; Schrodt & Phillips, 2016; Theiss & Knobloch, 2013), our investigation is unique in evaluating relational uncertainty as a predictor of sequences of self-disclosure paired with an ensuing hostile response. On one hand, our data echo those of Imai and colleagues (2023), who identified a complex pattern of sex differences among relational uncertainty, self-disclosure, and relationship satisfaction. On the other hand, our findings will require replication before any conclusions can be drawn about differences in relational uncertainty between men and women in the context of interpersonal exchanges in which self-disclosure is met with hostility.

We also hypothesized that relational uncertainty corresponds with less dyadic synchrony via sequences of validation met with hostility (H2), and our results largely supported that prediction. Individuals experiencing both self uncertainty and relationship uncertainty were more likely to engage in asynchronous exchanges marked by validation followed by hostility. These findings both corroborate and extend prior work. People experiencing relational uncertainty report that they communicate with less warmth (Theiss & Knobloch, 2014) and more aggression (Delaney & Sharabi, 2020; Theiss & Knobloch, 2013), and they judge their partner to communicate less constructively (Knobloch, Miller, Bond, & Mannone, 2007) and more destructively (Knobloch, Miller, Bond, & Mannone, 2007; Theiss & Knobloch, 2013), and, to our knowledge, our study is the first to identify an association between relational uncertainty and sequences of validation/hostility. Most broadly, our findings for H1 and H2 suggest initial support for the logic that relational uncertainty may constrain the ability of couples to engage in synchronous, fluid, and coordinated interaction (Berger & Calabrese, 1975; Solomon et al., 2016).

Our results also provide an opportunity to compare the three sources of relational uncertainty. Unlike self uncertainty and relationship uncertainty, partner uncertainty was not associated with either type of asynchronous exchange. Both theory and research suggest that the three sources of relational uncertainty, despite substantial overlap, are unique constructs (Goodboy et al., 2021; Knobloch & Solomon, 1999; Solomon et al., 2016). Stated differently, the questions people have about their own involvement in the relationship (self uncertainty) and the status of the relationship (relationship uncertainty) are conceptually and empirically distinguishable from their questions about their partner’s participation in the relationship. A potential explanation may be that people who experience partner uncertainty may be more reluctant to self-disclose or express validation in the first place, thereby constraining the frequency of the sequence. However, no evidence emerged in our data to support that explanation, given the lack of statistically significant bivariate correlations between these variables (see Table 2). Another possibility is that individuals experiencing partner uncertainty may be motivated to match their spouse’s tone while they gather information about what their spouse is thinking and feeling (e.g., Knobloch & Satterlee, 2009). We await future research designed to investigate the divergence between self, partner, and relationship uncertainty in their associations with dyadic synchrony.

We considered the role of relationship quality as well. A major rival explanation for the link between relational uncertainty and dyadic synchrony is that couples experiencing relationship distress are both more unsure about their relationship (e.g., Goodboy et al., 2020) and more likely to communicate with hostility (see Balderrama-Durbin et al., 2020, for review). In fact, prior analyses from this same sample showed that poorer relationship quality corresponded with relational uncertainty (Knobloch & Knobloch-Fedders, 2010). We opted, then, for a particularly stringent test by covarying relationship quality when predicting sequences of self-disclosure/hostility and validation/hostility. By demonstrating an association between relational uncertainty and dyadic asynchrony above and beyond relationship quality (H1, H2), our findings not only rule out a potential confound, but also contribute to a growing body of evidence differentiating relational uncertainty and relationship quality (e.g., Knobloch et al., 2021; Knobloch & Theiss, 2017).

Strengths, limitations, and future directions

Our study makes several contributions to scholarship on relational uncertainty and couple interaction. By theorizing about sequences of self-disclosure/hostility and validation/hostility within interaction, we direct attention to the dynamic interplay between partners to supplement prior research on the behavior of individuals (e.g., Solomon et al., 2016). Moreover, by harnessing observational methods, our investigation moves beyond self-report to directly examine sequences of behavior (e.g., Theiss, 2018), and appears to be the first to corroborate a link between people’s relational uncertainty and their sequential exchanges within conversations designed to facilitate expressions of intimacy. In addition, our clinical sample of couples experiencing considerable relationship distress complements the high-functioning samples typical of research investigating the link between relational uncertainty and communication (e.g., King & Theiss, 2016; Knobloch, Miller, Bond, & Mannone, 2007).

Limitations of our study exist as well. First, because we recorded couples’ interactions in a laboratory setting, their exchanges may not fully represent their naturally-occurring behavior. Web-based observational methods designed to assess the interactions of couples in their own homes show promise for bolstering ecological validity (Perry et al., 2021). Second, participants reported relatively low levels of relational uncertainty (see Table 1) despite the relationship distress they were experiencing. An alternative measure of relational uncertainty, developed after we began data collection for this study, appears to be superior to the original scale in maximizing variance (Solomon & Brisini, 2017); we encourage researchers to employ the new measure in future research to combat possible floor effects. Third, our approach to coding involved evaluating each speaking turn for the presence or absence of self-disclosure, validation, and hostility, regardless of the frequency or intensity of those behaviors. Future studies evaluating the degree of self-disclosure, validation, and hostility are necessary given data linking relational uncertainty to both the frequency and intensity of language (King & Theiss, 2016; Knobloch, 2006). Fourth, our cross-sectional design prevented us from testing reverse temporal pathways. We positioned relational uncertainty as a predictor of dyadic synchrony based on extant theorizing, but people’s communication behavior also has reciprocal effects on their relational uncertainty (Solomon et al., 2016). For example, when individuals engage in self-disclosure or express validation, they are likely to expect positive support or affirmation in return (Collins & Feeney, 2000; Holmes & Rempel, 1989), so receiving a hostile response may violate their expectations in ways that spark relational uncertainty (e.g., Ramirez & Wang, 2008). Finally, although participants were diverse with respect to age, race/ethnicity, relationship length, and household income, we did not collect information on disability, and our sample was limited to mixed-gender couples. Investigating dyadic synchrony among participants with disabilities, as well as among same-gender couples, is important given the wealth of sociopolitical factors that give rise to relational uncertainty within marginalized groups (Monk & Ogolsky, 2019).

Our study illuminates two avenues for future research. First, observational studies of dyadic interaction are essential for advancing scholarship across a variety of domains, including relational uncertainty (King & Theiss, 2016; Theiss, 2018), interpersonal communication (Caughlin & Basinger, 2014), the assessment of couple functioning (Balderrama-Durbin et al., 2020), and couple therapy (Friedlander et al., 2019). We encourage scholars to build on this project by leveraging the SASB circumplex model for coding interaction (Benjamin & Cushing, 2000). Grounded in a rich tradition of conceptual and empirical support (Benjamin, 2018), SASB is a sophisticated measurement tool designed to assess moment-by-moment interactional behavior. It provides an accessible way of probing turn-by-turn exchanges that are essential for sequential analysis (Knobloch-Fedders et al., 2014), and measures behavior with the specificity necessary for clinical assessment and treatment planning (Benjamin, 2006, 2018).

Second, we call for additional research triangulating the perceptions of insiders versus outsiders. Couple members do not always agree with third-party coders in their appraisals of interaction, and partners experiencing relational uncertainty may be particularly prone to biased appraisals of each other’s behavior (Knobloch, Miller, Bond, & Mannone, 2007). More work is needed to examine how couples judge their exchanges in real time compared to the cues visible to independent observers such as family members, friends, and practitioners (Friedlander et al., 2019). As an added benefit, such studies would inform guidelines for clinicians to help romantic partners manage relational uncertainty more effectively.