Dynamic and Daily Partner-Specific Processes of Relationship Uncertainty and Enacted Relationship Talk

Relational Uncertainty and Relationship Talk

Relational uncertainty stems from three sources of ambiguity (Knobloch & Solomon, 1999) including self uncertainty (ambiguity about one’s involvement in the relationship), partner uncertainty (ambiguity about how one’s partner views the relationship), and relationship uncertainty (ambiguity regarding the status of the relationship itself; Solomon et al., 2016). Self, partner, and relationship uncertainty comprise the general construct of relational uncertainty (e.g., Bolkan et al., 2023; Goodboy et al., 2021; Solomon et al., 2016) which represents the global evaluation of a relationship in terms of the “degree of confidence people have in their perceptions of involvement” (Knobloch & Solomon, 1999, p. 264). Regarding its impact on communication, research points to the conclusion that uncertainty in relationships is associated negatively with relational talk—despite the potential for this type of communication to reduce relational uncertainty (Knobloch & Theiss, 2011a; Theiss & Solomon, 2006a, 2006b).

Research suggests that individuals who experience relational uncertainty may be reluctant to communicate openly about sensitive issues (e.g., Knobloch & Carpenter-Theune, 2004). For instance, people who experience relational uncertainty may have difficulty communicating with their partners because they perceive relational conversations as face threatening and embarrassing, or because they believe these conversations may make them feel vulnerable or may threaten their relationships (Knobloch & Theiss, 2011a; Solomon et al., 2016). Moreover, people may avoid specific topics in their relationships because they prefer hopeful ambiguity to certainty with respect to negative information (e.g., Knobloch & Carpenter-Theune, 2004). In some cases, relational uncertainty might be tolerated or even preferred if the information revealed during specific conversations is predicted to be undesirable or if it may reduce uncertainty in a manner that causes distress (Afifi & Burgoon, 1998).

Although research has revealed that uncertainty in relationships is associated with less relationship talk (Knobloch & Theiss, 2011a), researchers do not yet fully know how these variables associate (within-person) on a daily basis. Studying how individuals experience heightened relationship uncertainty at the daily level may be important to the study of RTT because people’s daily relational experiences impact their communication patterns. For example, Campbell et al. (2010) showed that higher variation in daily perceptions of relationship quality was linked to more destructive conflict discussions (e.g., being defensive or criticizing the partner), less positive emotion during conflict, and greater distress. Moreover, increases in daily variability in relationship quality were associated with higher self-reported relationship problems and a higher likelihood of perceiving disagreements as threatening to the relationship.

Considering the above, studying the within-person effect of relationship uncertainty on enacted relationship talk on a daily basis may shed light on RTT processes and help researchers obtain a more nuanced understanding of this association. Because relational uncertainty has been theorized to impact communicative engagement (Solomon et al., 2016) and has been associated with reductions in enacted relationship talk (Knobloch & Theiss, 2011a), we predicted that individuals’ daily experiences of heightened relationship uncertainty (for them, relative to their 30-day average month-long relationship uncertainty) on a specific day would negatively impact their enacted relationship talk that day (for them, relative to their 30-day average month-long enacted relationship talk). We used the following within-person hypothesis as a guide for our inquiry:

Along with within-person daily effects, we also examined between-person associations with (latent) 30-day averages. That is, we studied how individuals who experience more month-long average relationship uncertainty might enact less month-long average relationship talk compared to people who have lower average month-long relationship uncertainty. Similar to H1a (within-person effect), we predicted a between-person effect with month-long averages that differ between individuals, as articulated by hypothesis H1b:

Researchers who study relational dynamics often examine the consistency versus volatility of individuals’ experiences (Arriaga, 2001). Indeed, as Arriaga (2001) argued, “there is much to gain from assessing changes in key relationship variables, rather than assessing only the overall levels of such variables at a single point in time” (p. 756). Research has indicated that beyond their mean levels, relational constructs (e.g., satisfaction, commitment, ambivalence/uncertainty) have significant variability across daily experiences (Totenhagen et al., 2016). Similar to what Solomon et al. (2016) noted about RTT, this variability may reflect changes in the relational environment that occur during times of transition and may be captured at the daily level.

In daily life, person-specific innovations in romantic relationships occur as highs and lows that depart from typical (i.e., average) relational states, reflecting intraindividual variability around long-run averages. The variability in these innovations can be thought of as volatility (Jongerling et al., 2015). Though it has only recently been explored empirically (e.g., Goodboy et al., 2024), volatility aligns with relational experiences as they have been articulated by RTT. Specifically, Solomon et al. (2016) argued that relational turbulence occurs when individuals experience disarray in their relationships due to repeated exposure to relational events involving increased highs and lows that can have a negative effect on their judgments about the relationship. Thus, studying volatility over time, as the amplified highs and lows (or oscillations day-to-day) of relational experiences, is fundamental to RTT predictions because volatility captures individuals’ reactivity to daily relational events.

Generally, relationship science demonstrates that vacillations in relational experiences are detrimental to relational outcomes. For example, research has demonstrated that fluctuations in relational satisfaction were associated with relational dissolution, even after controlling for initial and mean satisfaction (Arriaga, 2001). As Arriaga (2001) noted, “even among individuals who, on average, were increasingly happy with the relationship, vacillation in their level of happiness increased the odds that their relationship would end” (p. 762). Likewise, fluctuations in relationship quality have been linked to higher average psychological distress and increased distress over time, lower reports of life satisfaction and decreased life satisfaction over time, and declining confidence in the relationship (Whitton et al., 2014). As it pertains to RTT, we might expect that volatility in daily experiences of relationship uncertainty has relational consequences as well. It could be argued that individuals who experience more inconsistency about their relationship uncertainty may be particularly unlikely to engage their partners and talk about their relationships on average. As such, we might expect that individuals who experience their relationships as intensely unpredictable are less likely to engage in relational conversations because the outcomes of these interactions may be exceptionally difficult to predict. Thus, if elevated daily relationship uncertainty is associated with less enacted relationship on a particular day (H1a), and if vacillations in relational experiences have a negative impact on relational outcomes, we might predict that individuals with more volatility (more amplified highs and lows compared to their average) in their day-to-day relationship uncertainty report lower mean levels of enacted relationship talk across a month. This longitudinal outcome was predicted in hypothesis two:

In addition to volatility, some experiences may be characterized by inertia as individuals have difficulty recovering from elevated ambiguity and their relationship uncertainty persists for longer periods of time. Inertia is an important person-specific parameter to model in intensive longitudinal processes because it reflects the person-specific daily carryover of relational experiences that may remain or linger over several days. Inertia is a novel but important element to add to the RTT framework because it reflects the delay in time it takes for elevated (or depressed) relational states to revert back to their person-specific long-run average (Goodboy et al., 2024).

Elevated relationship uncertainty that lasts for consecutive days may be detrimental to relational communication because it creates the potential for extended ambiguity about the meanings and consequences of relational conversations (Solomon et al., 2016). That said, if elevated relationship uncertainty persists over time for some people, it may increase the potential for them to distort their perceptions of their partners and their relationships which may make it more difficult to discuss relationship issues. Because individuals who experience lingering and persistent relationship uncertainty may be less willing or able to accurately interpret their partner’s actions, they may be more reluctant to have conversations about their relationships because they cannot predict how these will unfold. This possibility served as the foundation for our third hypothesis:

Attachment Dimensions

Researchers have presented evidence indicating that processes articulated by RTT are moderated by attachment dimensions (Goodboy et al., 2022). Specifically, attachment theory has been the subject of recent research examining how partners’ insecurities impact RTT processes (Goodboy et al., 2022, 2024). Attachment theory offers the potential to provide insight into individual differences in RTT processes because individuals’ relational histories guide their perceptions and expectations of their current romantic experiences (Collins, 1996).

Attachment is the manifestation of a behavioral system designed to promote relational bonds (and therefore security) when comfort and safety are required (Fraley & Shaver, 2000; Mikulincer & Shaver, 2016). Specifically, attachment theory is concerned with the process of human connection and the manner in which individuals seek support from attachment figures in times of distress. While particularly important in infancy, the attachment system is active throughout life and functions to establish a sense of security in the world when individuals encounter real or imagined threats (Mikulincer & Shaver, 2016; Shaver & Mikulincer, 2006). According to Fraley and Shaver (2000), “individual differences in adult attachment behavior are reflections of the expectations and beliefs people have formed about themselves and their close relationships on the basis of their attachment histories” (p. 135). More precisely, researchers note that individuals feel secure when their attachment figures have proven to be accessible and responsive to bids for help (Mikulincer & Shaver, 2016). If attachment figures are unpredictable in their availability, sensitivity, or responsiveness, people may experience anxiety and adapt their attachment behaviors to become hyperactive (i.e., intensified behaviors aimed at gaining attention; Mikulincer & Shaver, 2016). In contrast, if attachment figures punish or dismiss attachment seeking behaviors, people may experience avoidance and adapt their behaviors toward deactivation (i.e., attempting to deal with difficult circumstances on one’s own and/or inhibiting emotional experiences during times of distress; Mikulincer & Shaver, 2019).

As it pertains to RTT, Goodboy et al. (2022) noted that “attachment insecurity is theoretically and logically well-suited to explain individual differences” because “individuals with attachment anxiety and/or avoidance may be more susceptible to heightened experiences of relational uncertainty” (p. 320). To this point, Goodboy et al. found that the effect of relational uncertainty on relational turbulence through threat appraisal was stronger for anxious and avoidant individuals compared to more secure individuals. Relating to the aims of the current study, researchers have found that attachment is associated with daily perceptions of relationships to the extent that insecure attachment proclivities are associated with fluctuations in the way individuals appraise their romantic partners (Alfasi et al., 2010). Indeed, research indicates that insecure individuals are more reactive to daily spousal behavior (particularly negative behavior) compared to more secure individuals (Feeney, 2002). Considering that people with insecure attachment styles are more sensitive to daily relational events, it may be the case that, compared to individuals who are secure, anxious and avoidant individuals are also more sensitive to daily fluctuations in relationship uncertainty.

Readers may inquire about how individuals with anxious and avoidant tendencies might engage in enacted relationship talk when they experience uncertainty in their relationships. Research has revealed that anxious individuals tend to experience more worry about their relationships and are more likely to turn their attention toward rejection and desertion (Fraley & Shaver, 2000; Mikulincer & Shaver, 2016, 2019). As this perception of rejection manifests in relationship talk, researchers have found that, compared to secure individuals, people with anxious attachment styles report experiencing more escalation in daily conflicts, describe themselves as being more hurt by conflict, and predict that conflict would be more damaging to their relationships (Campbell et al., 2005). In addition, researchers have found that on days when conflict was higher than normal, anxious individuals were more likely than secure individuals to think their partners were less satisfied with, and less optimistic about, their relationships (Campbell et al., 2005). In the case of the current study, this could mean that when anxious individuals experience elevated relationship uncertainty on a particular day, they might be less likely to engage in enacted relationship talk because they may feel especially vulnerable with respect to these conversations. The following person-specific hypothesis was offered as a test of this prediction:

Research suggests that avoidant individuals adapt their behaviors toward deactivation by attempting to deal with difficult relationship circumstances on their own (Mikulincer & Shaver, 2016). Because avoidant individuals are likely to be self-reliant and respond to distress with interpersonal distancing and disengagement (Mikulincer & Shaver, 2016, 2019), we predicted that they would also be less likely to engage in enacted relationship talk on days when their relationship uncertainty was elevated. This idea served as the basis for our fifth hypothesis:

Researchers have demonstrated that individuals’ attachment styles are associated with fluctuations in relational experiences. For example, Cooper et al. (2018) found that women who scored higher in anxiety reported experiencing greater volatility in relationship quality. Similarly, Tammilehto et al. (2023) reported that attachment anxiety was associated with more variability in people’s reports of negative emotions. Considering that attachment anxiety appears to be related to the vacillations people experience within their relationships, we predicted that people who scored higher on attachment anxiety would report more volatility in relationship uncertainty and enacted relational talk over 30 consecutive days.

Individuals’ attachment styles might also be related to inertia pertaining to relationship uncertainty and enacted relationship talk because individuals who experience attachment avoidance may be less likely to experience distress in response to relationship-threatening scenarios (Collins, 1996) or to express their concerns (Mikulincer & Shaver, 2016). Thus, we might expect that their experiences of daily relationship uncertainty are more likely to carry over from previous days because they may distance themselves in response to threatening events (Mikulincer & Shaver, 2019) instead of working to repair abnormal relational states. On the contrary, because avoidant individuals may be less likely to engage in enacted relationship talk with their partners, we might expect that deviations from their typical states on a given day revert and recover more quickly.

Method

Results

To test our hypotheses, we used dynamic structural equation modeling (DSEM) with Bayesian estimation in Mplus 8.10 (Asparouhov et al., 2018). DSEM combines time series modeling, multilevel modeling, and structural equation modeling to examine dynamic processes of occasion-specific variables from consecutive time points with many repeated measures (Hamaker et al., 2023). McNeish and Hamaker (2020) explained the combination of these three modeling approaches in DSEM: time-series analysis handles the lagged effects of variables from previous time points to future time points (over many time points) for an individual (within-person, n = 1); multilevel modeling is used for the longitudinal analysis of all people in the sample with random effects (N = 202) allowing for interindividual (quantitative) differences in person-specific processes (parameters); structural equation modeling allows for path-analysis with multiple outcomes (see also Hamaker et al., 2023). DSEM also allows for multilevel location-scale modeling of time-invariant predictors of random intercepts (latent averages), slopes (individual within-person effects), and innovation variances (individual within-person variability). These flexible aspects of DSEM make it useful for studying daily life by specifying that people share the same dynamic model over time but have unique person-specific parameters that best characterize their day-to-day lives (unique averages or traits, unique magnitudes of lagged and daily effects, and unique residual variances or reactivity to daily life).

To model person-specific processes, we examined a two-level DSEM (McNeish & Hamaker, 2020) with random effects by estimating random intercepts, slopes, and residual (innovation) variances (see Figure 1). In DSEM, the measurement of relationship uncertainty and enacted relationship talk each day (t) are latent person-mean centered (Asparouhov & Muthén, 2019) for the decomposition of between-level (random intercept, or unique 30-day averages over time that differ between dating relationships) and within-level variance (person-specific deviations that differ from the unique 30-day average). Random (person-specific) slopes for our model included the daily effect of relationship uncertainty on enacted relationship talk today (lag0 within-person daily effect), the lag-1 autoregressive effect of yesterday’s relationship uncertainty on today’s relationship uncertainty throughout the month (within-person relationship uncertainty inertia), and the lag-1 autoregressive effect of yesterday’s enacted relationship talk on today’s relationship talk throughout the month (within-person talk inertia). These lag-1 autoregressive effects are a crucial specification in DSEM to account for a person’s unique time-adjacent effects (the autocorrelation of the previous and current time point) so that parameter estimates are unbiased (see Hamaker et al., 2023) as a person’s immediate past (i.e., yesterday) is controlled for when estimating the contemporaneous daily slopes in the current state (i.e., today). Beyond the statistical need to model lag-1 autoregression in DSEM, lagged effects provide person-specific estimates of inertia as some people persist longer in time as a portion of yesterday’s variable deviation will carryover today (i.e., they linger over days), but for others, yesterday’s variable has no impact on today (i.e., they recover or return to normal the next day).

OPEN IN VIEWER

Figure 1.

Dynamic structural equation model.

The random residual variances captured person-specific innovations or volatility of relationship uncertainty and talk over 30 days. Allowing each person to have their own residual variance is crucial in DSEM to avoid biasing parameter estimates that would result from assuming a constant residual variance (i.e., assuming that people are equally predictable and have the same variability in their daily upswings and downswings in relationship uncertainty and relationship talk). Instead, we allowed (and expected) each person to have unique innovations or daily reactions in their relationships (see Jongerling et al., 2015). Modeling person-specific residual variances allows some people to be more/less predictable in their swings (McNeish & Hamaker, 2020) and helped us determine why some people were consistent whereas others experienced volatile relationship uncertainty and relationship talk throughout the month. We allowed the random effects (random intercepts, slopes, and variances) to correlate at the between-person level (level 2; Hamaker et al., 2023).

We estimated our model with DSEM using 5,000 Markov Chain Monte Carlo (MCMC) iterations to produce a posterior distribution of parameters with 95% credibility intervals (CI). Bayesian credibility intervals are interpreted differently (and more intuitively) than confidence intervals in frequentist statistics; credibility intervals tell us there is a 95% probability that the parameter value is between the lower and upper limit (van de Schoot et al., 2014). We also calculated heterogeneity intervals based on the variances of slopes (±1.96√variance) which, assuming a normal distribution in the population, provide the model-implied range of expected slopes for individuals (Bolger et al., 2019). We ensured model convergence after 5,000 iterations with a potential scale reduction (i.e., a stopping criterion for two separate MCMC chains as estimates converged) of 1.05 (PSR = 1.012). Table 1 displays unstandardized means (fixed effects) and variances (random effects), 95% credibility intervals, and average within-level standardized effects.

OPEN IN VIEWER

Table 1.

Bayesian Parameter Estimates.

	Unstandardized	95% Credibility	Standardized	95% Credibility
	Posterior Median	Interval	Posterior Median	Interval
Daily Effect (Lag 0)
(DrU t → DrT t )
Mean	−.264	[−.358, −.169]*	−.128	[−.156, −.099]
Variance	.145	[.099, .212]*
Relationship Uncertainty Inertia (Lag-1)
(DrUt-1 → DrU t )
Mean	.332	[.281, .384]*	.332	[.298, .367]
Variance	.076	[.056, .101]*
Relationship Talk Inertia (Lag-1)
(DrTt-1 → DrT t )
Mean	.188	[.141, .233]*	.188	[.157, .219]
Variance	.060	[.041, .083]*
Relationship Uncertainty Volatility
Mean (log scale)	−1.990	[−2.258, −1.722]*	.822	[.794, .845]
Variance	3.616	[2.971, 4.499]*
Relationship Talk Volatility
Mean (log scale)	−.736	[−.919, −.552]*	.819	[.796, .841]
Variance	1.529	[1.240, 1.897]*

Note. Means are fixed effects and variances are random effects. DrU = daily relationship uncertainty, DrT = daily enacted relationship talk. For level 2 intercepts (30-day averages), the fixed and random effects were: relationship uncertainty (μ_DrU = 1.763; .561) and enacted relationship talk (μ_DrT = 3.790; 1.138). Standardized parameters are estimated per person and are within-level individually standardized estimates averaged across participants. Within-level R² averaged across individuals (clusters): relationship uncertainty = .178, enacted relationship talk = .181.

Hypothesis 1a was confirmed with evidence of a within-person daily effect. For the typical (average) dating relationship, on days when individuals had more uncertainty about their relationship than they normally did (i.e., 1 unit above their person-specific month-long average), they enacted less relationship talk that day than they normally did, controlling for yesterday’s level of relationship talk (β = −.264 [95% CI = −.358, −.169]). The heterogeneity interval indicated that 95% of the population is expected to have within-person daily slopes ranging from between −1.010 to .482. Additionally, in the average dating relationship, there was inertia for daily relationship uncertainty (ϕ_DrU = .332 [95% CI = .281, .384]) with 95% of the population expected to range from −.208 to .872, and relationship talk (ϕ_DrT = .188 [95% CI = .141, .233]) with 95% of the population expected to range from −.292 to .668. Thus, we had evidence of an average within-person effect that was nonzero with heterogeneity in effects.

All between-level correlations of person-specific parameters are presented in Table 2. Correlations at the between-level (level 2) provided support for Hypothesis 1b, as well as Hypotheses 2 and 3. Specifically, in support of Hypothesis 1b, a level 2 correlation between random intercepts (individual means) showed that individuals who had more average relationship uncertainty over the course of 30 days participated in less average enacted relationship talk over 30 days (r = −.460 [−.575, −.330]). In support of our second hypothesis, a level 2 correlation indicated that individuals who had more volatility (within-person variability), or (log) innovation variance, in their relationship uncertainty engaged in less average enacted relationship talk over the month (r = −.218 [−.354, −.068]). In other words, individuals who had more amplified highs and lows in daily relationship uncertainty around their mean engaged in less (on average) enacted relationship talk throughout the month. Moreover, in support of our third hypothesis, results indicated that individuals who had more relationship uncertainty inertia over the course of 30 days also engaged in less average enacted relationship talk in the month (r = −.219 [−.393, −.025]). These results indicate that individuals who had difficulty recovering from perturbations in daily relationship uncertainty, or who stayed in an elevated state of relationship uncertainty, engaged in less enacted relationship talk over the course of a month (on average).

OPEN IN VIEWER

Table 2.

Level 2 Correlations Among Latent Variables (Parameters).

	DrU(b)	DrT(b)	Log DrU V	Log DrT V	DrU I	DrT I
DrT(b)	−.460*	—
	[−.575, −.330]
Log DrU V	.551*	−.218*	—
	[.437, .648]	[−.354, −.068]
Log DrT V	.174*	−.382*	.286*	—
	[.016, .320]	[−.506, −.240]	[.135, .418]
DrU I	.461*	−.219*	.260*	.071	—
	[.271, .621]	[−.393, −.025]	[.071, .428]	[−.127, .257]
DrT I	−.134	−.118	−.247*	.144	.127	—
	[−.353, .085]	[−.317, .101]	[−.429, −.061]	[−.035, .325]	[−.112, .383]
DrU t →DrT t	.413*	−.265*	.199	−.104	.326*	−.031
	[.186, .594]	[−.430, −.065]	[−.147, .468]	[−.288, .084]	[.071, .538]	[−.289, .222]

Note. Asterisks indicate significant correlations and brackets report 95% credibility intervals. DrU(b) = 30-day person-specific average of daily relationship uncertainty, DrT(b) = 30-day person-specific average of daily enacted relationship talk, DrU V = person-specific relationship uncertainty volatility (log scale), DrT V = person-specific enacted relationship talk volatility (log scale), DrU I = person-specific relationship uncertainty inertia, DrT I = person-specific enacted relationship talk inertia.

Attachment

Given the between-level differences in within-person parameters (random means, slopes, and variances reported in Table 1), we hypothesized that between-level (pretest) attachment anxiety and avoidance would predict person-specific longitudinal parameters of RTT (at level 2). Bolger and Laurenceau (2013) noted that it is theoretically important to explain variability in between-person differences with intensive longitudinal data and variance in random effects. DSEM accommodates Bolger and Laurenceau’s (2013) recommendation by allowing a between-person predictor of an individual’s person-specific means and residual variances—also known as multilevel location-scale modeling (McNeish & Hamaker, 2020). We used 5,000 iterations to estimate a multilevel location-scale model as depicted in Figure 2 and we allowed the level 2 latent variables to have correlated residuals. Unstandardized posterior medians and credibility intervals of the multilevel location-scale DSEM are reported in Table 3. Attachment dimensions (grand mean centered) predicted month-long person-specific averages, inertia, volatility, and daily effects.

OPEN IN VIEWER

Table 3.

Bayesian Estimates and 95% Credibility Intervals for Multilevel Location-Scale Model.

	Posterior	95% Credibility
	Median	Interval
30-Day Individual Averages (Random Intercepts)
Anxiety → 30-day Relationship Uncertainty (DrU(b))	.131	[.045, .219]*
Avoidance → 30-day Relationship Uncertainty (DrU(b))	.354	[.243, .468]*
Anxiety → 30-day Enacted Relationship Talk (DrT(b))	−.116	[−.249, .017]
Avoidance → 30-day Enacted Relationship Talk (DrT(b))	−.324	[−.496, −.146]*
Inertia (Autoregressive Lag-1)
Anxiety → Relationship Uncertainty Inertia (DrU I)	.037	[−.005, .083]
Avoidance → Relationship Uncertainty Inertia (DrU I)	.070	[.012, .128]*
Anxiety → Relationship Talk Inertia (DrT I)	−.013	[−.055, .029]
Avoidance → Relationship Talk Inertia (DrT I)	.013	[−.042, .067]
Volatility (Log Innovation Variances)
Anxiety → Relationship Uncertainty Volatility (DrU V)	.377	[.144, .610]*
Avoidance → Relationship Uncertainty Volatility (DrU V)	.486	[.192, .768]*
Anxiety → Relationship Talk Volatility (DrT V)	.234	[.074, .392]*
Avoidance → Relationship Talk Volatility (DrT V)	−.024	[−.222, .175]
Daily Effect (Lag 0)
Anxiety → Daily Effect of Uncertainty on Talk (DrUt→DrTt)	−.079	[−.144, −.014]*
Avoidance → Daily Effect of Uncertainty on Talk (DrUt→DrTt)	.054	[−.030, .138]
Intercepts
Daily Relationship Uncertainty (DrU)	1.759	[1.663, 1.859]*
Daily Relationship Talk (DrT)	3.790	[3.638, 3.937]*
Daily Effect of Uncertainty on Talk (DrUt→DrTt)	−.255	[−.349, −.162]*
Relationship Uncertainty Inertia (DrU I)	.333	[.283, .381]*
Relationship Talk Inertia (DrT I)	.185	[.139, .233]*
Relationship Uncertainty Volatility (DrU V)	−1.992	[−2.247, −1.731]*
Relationship Talk Volatility (DrT V)	−.734	[−.913, −.561]*

Note. DrU(b) = 30-day person-specific average of daily relationship uncertainty, DrT(b) = 30-day person-specific average of daily enacted relationship talk, DrU V = person-specific relationship uncertainty volatility (log scale), DrT V = person-specific enacted relationship talk volatility (log scale), DrU I = person-specific relationship uncertainty inertia, DrT I = person-specific enacted relationship talk inertia. The correlation between attachment anxiety and avoidance is r = −.038 [−.172, .103]. Between-level R²: DrU(b) = .226, DrT(b) = .091, DrU I = .079, DrT I = .016, DrU V = .101, DrT V = .051, DrU_t → DrT_t = .084. Intercepts are expected scores for random effects when attachment anxiety and avoidance are at average values in the sample (both attachment dimensions are at a value of 0 as they are grand mean centered).

OPEN IN VIEWER

Figure 2.

Multilevel location-scale.

Attachment Anxiety

In support of Hypothesis 4, attachment anxiety predicted the person-specific daily effects of relationship uncertainty on enacted relationship talk (DrU_t→DrT_t; β = −.079 [95% CI = −.144, −.014]). In other words, individuals who were higher in anxiety had a stronger negative within-person daily effect of relationship uncertainty on enacted relationship talk. Hypotheses 6 and 7 were also supported. Attachment anxiety predicted 30-day person-specific relationship uncertainty volatility (DrU V; β = .377 [95% CI = .144, .610]) and relationship talk volatility (DrT V; β = .234 [95% CI = .074, .392]). That is, individuals who were higher in anxiety experienced greater variability in both their daily relationship uncertainty and enacted relationship talk over the month.

Attachment Avoidance

Hypothesis 5 was not supported. Attachment avoidance did not significantly predict the daily person-specific effect of relationship uncertainty on enacted relationship talk (DrU_t→DrT_t; β = .054 [95% CI = −.030, .138]). Hypothesis 8 was supported, but Hypothesis 9 was not. Attachment avoidance predicted person-specific relationship uncertainty inertia (DrU I; β = .070 [95% CI = .012, .128]), but not relationship talk inertia (DrT I; β = .013 [95% CI = −.042, .067]) indicating that partners who were higher in avoidance lingered longer in their relationship uncertainty over days before they returned to their typical state (i.e., their random intercept). These results demonstrate that, although avoidant partners had a stronger tendency to get “stuck” in an elevated state of relationship uncertainty over 30 days, their persistence in relationship talk remained unaffected.

Discussion

This study employed DSEM to test predictions consistent with RTT using a 30-day intensive longitudinal design. We tested theoretical hypotheses aligned with RTT processes including person-specific daily effects of relationship uncertainty on daily enacted relationship talk, day-to-day processes of RTT including inertia and volatility over time, and individual differences in attachment dimensions as these predicted within-person longitudinal parameters. By adopting a person-specific paradigm for testing RTT in daily life, where each person serves as their own control relative to themselves and individuals have their own unique longitudinal parameter estimates (Bolger & Laurenceau, 2013), this study permits us to offer a broad theoretical implication. Allowing each person to have their own unique estimates provided insights about the typical individual and how they changed over time (including those who have longitudinal processes that varied from the typical person) which helps us explain why some people have differences in their within-person processes. In this study we found that individuals with attachment insecurity (higher anxiety, avoidance) had more volatility (i.e., larger innovation variance) indicating that these individuals were less predictable than secure individuals (who had smaller volatility). Stated differently, because each person had their own person-specific effect size (within-person R²), the within-person predictions of RTT tested here appear to be more robust for secure individuals who were less reactive in daily life. Although this theoretical implication is tentative in the absence of additional empirical evidence, we might conclude that at least some of RTT’s predictions could produce larger person-specific effect sizes for secure partners who are more predictable on any given day compared to less secure individuals. Next, we discuss specific implications derived from our hypotheses.

In support of our first hypothesis (H1a), results indicated that on days when individuals experienced more uncertainty about their relationship than they normally did, they enacted less relationship talk that day than they typically did (holding constant their level of relationship talk from the day before). In support of H1b, we found that individuals who had higher levels of relationship uncertainty (averaged over a month) were also less likely to engage in enacted relationship talk (averaged over a month). These results are consistent with research revealing a negative association between relational uncertainty and relationship talk (Knobloch & Theiss, 2011a), and relationship uncertainty and communicative directness (Theiss & Solomon, 2006a, 2006b). As we argued in the introduction, individuals in our study may have preferred to disengage from relational communication on days when they perceived that these conversations would lead to negative outcomes. This may be the case because, although relationship uncertainty may seem unpalatable, participants in this study might have preferred to live with ambiguity rather than being confronted with distressing certainty (Knobloch & Solomon, 2005). Although scholars have previously demonstrated negative associations between relational uncertainty and communicative engagement through weekly longitudinal research designs (e.g., Knobloch & Theiss, 2011a), our study extends this line of research by revealing similar associations at the daily level.

Results from this study also revealed that volatility (person-specific log innovation variances) with respect to month-long relationship uncertainty was negatively associated with individuals’ average month-long enacted relationship talk (H2). These results indicate that individuals who experienced more vacillation in their daily relationship uncertainty were more likely to avoid engaging in enacted relationship talk. This outcome is in line with research demonstrating that fluctuations in relationship quality (Whitton et al., 2014) and relationship satisfaction (Arriaga, 2001) are linked to negative relational outcomes (e.g., relationship termination, psychological distress, etc.). These findings may indicate that individuals who experience inconsistency about their relationship uncertainty over time have more reactivity to daily events and may be less willing to talk openly about the status of the dyad due to the perceived negative consequences that could arise from these conversations. Perhaps individuals who experience their relationships as particularly unpredictable may be less likely to engage in relational conversations because the outcomes of these may be exceptionally difficult to forecast, and/or they suspect that the outcomes of such conversations would be more detrimental than living with elevated relationship uncertainty.

Individuals in this study who experienced persistent deviations from their long-run relationship uncertainty had a difficult time regulating it back to baseline after experiencing a perturbation. In a multilevel time series this is known as inertia, and our results supported the prediction (H3) that individuals’ relationship uncertainty inertia would be negatively associated with their average enacted relationship talk over a month. This result is similar to what Goodboy et al. (2024) reported with regard to daily anger and daily relational turbulence. In their study (which also used DSEM with daily measures), the authors found that when perturbed, individuals who remained angrier than normal (for a period of days) also experienced prolonged chaotic relational states. Our results represent a novel finding with respect to RTT insofar as they demonstrate that perturbations in relationship uncertainty that persist over time may have a unique impact on individuals’ experiences of, and behaviors related to, communication within their relationships. Specifically, results indicated that the harder it was for individuals to recover from uncertainty in their relationships, the more difficult it was for them to discuss the relationship with their significant others. As such, it appears that individuals who are unable to experience recuperation from the perspective of relationship uncertainty are particularly likely to suffer from its relational consequences.

From a person-specific perspective, it is important to explain why partners have different experiences in their daily lives and month-long outcomes. Accordingly, our findings contribute to the growing body of literature linking individual differences in attachment anxiety and avoidance to RTT processes. Although not predicted with specific hypotheses, our results indicated that anxious individuals reported elevated average month-long relationship uncertainty. Moreover, we found that avoidant individuals reported higher month-long averages of relationship uncertainty and lower month-long averages of enacted relationship talk. These results are in line with studies showing that individuals with insecure attachment styles are more likely to experience uncertainty in their relationships (e.g., Goodboy et al., 2022), and that people who are avoidant are more likely to engage in distancing and disengagement when experiencing threatening events (Mikulincer & Shaver, 2019).

Research has shown that people with attachment insecurity are more likely to be engaged in destructive communication (e.g., escalating conflict, attacking, criticizing), give less support (i.e., forgiveness and caregiving), and are involved in fewer constructive interactions (behaviors linked to successful conflict resolution such as cooperation and compromise; Li & Chan, 2012). In line with these outcomes, results from our study demonstrated that individuals who were higher in attachment anxiety experienced a stronger negative within-person association between daily relationship uncertainty and daily relationship talk (H4). These results make sense in light of research that shows anxiously attached people are more likely to report escalation in their conflict, being hurt by conflict, and to believe that conflict would be more damaging to the relationship compared to securely attached individuals (Campbell et al., 2005). That said, people who are high in attachment anxiety may be particularly likely to abstain from relationship talk if they wish to avoid these negative experiences. Moreover, anxious individuals may be less likely to engage in enacted relationship talk when they feel uncertain about the relationship because they are more likely to “worry and ruminate about being rejected or abandoned by their partners” (Campbell et al., 2005, p. 511).

We also found that attachment anxiety predicted volatility in relationship uncertainty (H6) and enacted relationship talk (H7) throughout the month. These findings corroborate the position of researchers who argue that people who experience attachment anxiety are more “sensitive to transient events” (Alfasi et al., 2010, p. 607). As Campbell et al. (2005) argued, because anxious people are more concerned about their relationships, they may be more sensitive to, and place more meaning on, daily relational incidents. In fact, Campbell et al. (2005) reported that “highly anxious individuals rely more heavily on daily perceptions of relationship events to assess the current and future quality of their relationships” (p. 526). Our findings bolster this conclusion regarding relationship uncertainty and add to the literature by demonstrating that people who experience attachment anxiety may also be more behaviorally reactive to daily events with respect to relational communication.

Our hypotheses about attachment avoidance were not all supported. Attachment avoidance did not significantly moderate the daily effect of relationship uncertainty on enacted relationship talk (H5). Attachment avoidance predicted relationship uncertainty inertia (H8), but it did not predict enacted relationship talk inertia (H9). Pertaining to hypothesis eight, our findings revealed that avoidantly attached individuals were more likely to persist in their experiences of relationship uncertainty. This result may be explained by the notion that avoidantly attached individuals tend to be less bothered by potential relational threats compared to people who are securely attached (Collins, 1996). As such, our results may reflect a lack of exigency to reduce relationship ambiguity because avoidant individuals may be less bothered by this perception in their relationships compared to individuals with different attachment proclivities.

In summary, our findings add to the body of research on RTT by demonstrating that individuals’ daily experiences of relationship uncertainty are important considerations. This includes individuals’ daily experiences of relationship uncertainty, the volatility of experiences regarding relationship uncertainty, and the difficulty that individuals have with respect to recovering from relationship uncertainty after experiencing it on a particular day. Taken together, results from this study indicate that, along with general experiences of relationship uncertainty, the perturbations and persistence that individuals experience in terms of daily relationship uncertainty have important implications for communicative engagement as outlined by RTT. Moreover, this study adds to the growing body of literature supporting the integration of differences in attachment anxiety and avoidance within the framework of RTT. As noted, individuals who are insecurely attached may be particularly reactive to daily events and may therefore respond to fluctuations in their experiences of relationship uncertainty with differing levels of communicative engagement compared to securely attached individuals.

Conclusion

Taken together, results from this study reinforce the theoretical argument that partner communication (or lack thereof) is due to individuals’ reactions to relationship uncertainty that occurs on a daily basis. Specifically, we found that individuals’ daily relational experiences provide critical information with respect to RTT predictions and we showed that attachment proclivities are fundamental to these relationship-specific predictions. As researchers continue to study RTT, they may consider how daily interactions influence individuals’ experiences within their relationships while keeping in mind attachment insecurities that are predictive of how relationship uncertainty and communication processes unfold day-to-day over the course of a month.