The Dyadic Interplay Between Relationship Satisfaction, Perceived Positive and Negative Social Control, and a Reduction of Sedentary Behavior Time

Abstract

Background

Both the close relationship processes and health model and the dyadic health influence model posit that beliefs about the relationship (e.g., relationship satisfaction) and influence strategies (e.g., social control) serve as mediators of health behavior change. The evidence for such mediation is limited.

Purpose

This study investigated two competing hypotheses that arise from these models: (1) perceived use of positive and negative social control (attempts to influence the partner’s behaviors) predict sedentary behavior (SB) indirectly, via relationship satisfaction; or (2) relationship satisfaction predicts SB indirectly, via positive and negative social control.

Methods

Data from 320 dyads (target persons and their partners, aged 18–90 years), were analyzed using mediation models. SB time was measured with GT3X-BT accelerometers at Time 1 (T1; baseline) and Time 3 (T3; 8 months following baseline). Relationship satisfaction and social control were assessed at T1 and Time 2 (T2; 2 months following baseline).

Results

Higher T1 relationship satisfaction among target persons predicted target persons’ reporting of higher T2 negative control from partners, which in turn predicted lower T3 SB time among target persons. Lower T1 relationship satisfaction among partners predicted target persons’ reporting of higher T2 perceived negative control from partners, which predicted lower T3 SB time among target persons. On average, both members of the dyad reported moderate-to-high relationship satisfaction and low-to-moderate negative control.

Conclusions

In contrast to very low levels of negative control, its low-to-moderate levels may be related to beneficial behavioral effects (lower SB time) among target persons reporting moderate-to-high relationship satisfaction.

Sedentary behavior (SB) is defined as any waking activity characterized by an energy expenditure ≤ 1.5 METs while in a sitting, lying, or reclining posture [1]. This behavior is becoming prevalent across domains of human activity due to changes in workplace, the use of entertainment technologies, transportation, and communications [2]. Longer time spent in SB is associated with an increased risk of metabolic syndrome, type-2 diabetes, cardiovascular diseases, and lower physical quality of life [3, 4]. SB may be responsible for approximately 0.5 million deaths/year, representing 3.8% of all-cause mortality [4]. The World Health Organization (WHO) [5] recommends that adults limit time spent in SB and replace it with physical activity (PA) to help reduce detrimental effects of SB on health.

Social process variables are listed among key potential determinants of energy expenditure behaviors, including SB [6]. Social control is one type of social process variable that has the potential to influence SB, as hypothesized in the landmark publication by Lewis and Rook [7]. Social control is defined as any attempt by one partner to influence the other partner’s health or health behaviors [7, 8]. Positive social control refers to agents’ use of persuasion, rational logic, and positive reinforcement, while negative social control refers to expressions of negative emotions, or attempts to induce negative emotions in the target person to influence their behavior [9, 10]. Unlike social support, interactions involving social control need not be affirming or provide resources [9]. While positive and negative social control attempts are intended to elicit positive behavioral changes, improvements in health behavior may be accompanied by increases in distress, evoked by the ways control is delivered [7].

To date, dyadic studies investigating the associations between social control and health behaviors have reported mixed findings [9–11]. A meta-analysis found moderate beneficial effects of positive social control on health-promoting behaviors, but high levels of negative social control were associated with lower engagement in health-promoting behaviors (small effect sizes were observed) [8]. The findings were of high heterogeneity, focusing on within-individual associations, obtained mostly in cross-sectional studies, and SB was not investigated [8]. We identified only one dyadic study explaining links between social control and SB. Parental and child perceptions of the use of control-based strategies by parents were unrelated to self-reported child SB, assessed at an 8-month follow-up [12].

Relationship satisfaction is yet another relationship factor predicting health behaviors among people who are recommended to change their lifestyle and to become more active [13]. The associations between relationship satisfaction, social control, and health behaviors are explained by a framework for investigating dyadic relationship processes and health [14]. This framework [14] indicates that variables such as social support or control may predict relationship mediators (including relationship satisfaction), which in turn predict physiological states, affect, and health behavior. Importantly, the framework assumes that these variables may also be chained in a different order: relationship factors (including relationship satisfaction) may predict social process variables (provision and receipt of social support and control), which in turn explain health-related outcomes. In other words, relationship satisfaction, and social control may operate either as the predictors or as the mediators when explaining health behaviors [14]. The mediating role of social control is also in line with Hoffman et al.’s [15] framework, linking fluctuations in relationship satisfaction with mediator variables, including facilitators of effective self-regulation (e.g., social exchange processes), which in turn determine the achievement of behavioral goals.

The evidence-based dyadic health influence model (DHIM) [16] suggests complex indirect pathways through which beliefs about the relationship (such as relationship satisfaction) and social influence strategies (such as social control) may explain health behaviors of the target person. As proposed in the DHIM [16], the use of influence strategies (e.g., social control) by the partner may predict the target person’s relational beliefs (e.g., relationship satisfaction), which in turn are related to the target person’s health behaviors. For example, the partner’s use of social influence strategies may trigger relationship-relevant thoughts, such as the target person’s commitment to the relationship or beliefs about the importance of the relationship. Perceptions of high (or improved) relationship satisfaction and importance may prompt the target person to engage in a healthier behavior (e.g., reduce their SB time) because of the desire to obtain the affection of their partner and to maintain satisfactory relationship [16]. The review by Huelsnitz et al. [16] suggests that these hypothesized indirect associations have not been tested. Furthermore, the DHIM [16] proposes that the partner’s relational beliefs (e.g., anxiety about relationship, feeling dissatisfied) may prompt them to use influence strategies (including social control, such as guilt induction), which in turn may affect the target person’s health behavior. As indicated by Huelsnitz et al. [16] such indirect effects hypotheses have not been tested.

In line with the DHIM [16] a moderate, although not high, level of relationship satisfaction may prompt partners to use some social control strategies (positive or negative) in order to evoke a change in the target person’s behavior. In turn, the target person who is highly satisfied with the relationship may be sensitive to even small cues and likely to wait for their partner’s signaling a need for change (and thus perceive social control). Perceived social control may trigger target person’s willingness to act in line with the perceived influence strategy (e.g., negative social control), and engage in a healthy behavior to further satisfy the partner, and maintain the satisfactory relationship.

The majority of research on links between relationship satisfaction and health behaviors has been conducted in the context of romantic relationships [13, 17]. Behaviors such as SB occur across various settings and contexts, and may be undertaken without a romantic partner (e.g., at work, during leisure time [2]). Thus, research investigating people who intend to change their SB may also include types of dyads other than romantic, namely any types of dyads in which two individuals intend to reduce SB, or at least one person intends to become more active and the other person intends to support the target person during the behavior change process [18]. In any case, the type of the relationship should be controlled in dyadic research. Additionally, as indicated in health behavior change frameworks (e.g., implementation intentions) and dyadic research [18–21], intention is one of the key proximal determinants of health behavior, thus the strength of intention should be controlled in behavior change research.

In line with the DHIM and the framework for investigating dyadic relationship factors and health [14, 16], two competing mediation models were tested. First, we examined whether the target persons’ and partners’ perceived positive, and negative social control from the other person in the dyad (Time 1; T1) would predict their SB (measured at Time 3, T3; 8 months after T1) indirectly, with target persons’, and partners’ relationship satisfaction (Time 2, T2; 2 months after T1) mediating these associations. Second, we examined whether relationship satisfaction (T1; target persons and their partners) would predict SB (T3; target persons and partners) indirectly, with target persons’ and partners’ perceived positive, and negative social control from the other person in the dyad (T2) mediating these associations.

Method

Participants

At Time 1, participants were 640 adults forming N = 320 dyads (320 target persons and 320 partners). Time 3 measurement (8 months after T1) was completed by n = 288 target persons and n = 292 partners, indicating that the total longitudinal dropout was only 6.45%.

The baseline sample of target persons (64.4 % women) were 18–90 years old (M = 43.86, SD = 17.02). Their partners (64.1% women) were 18–84 years old (M = 42.32 years; SD = 16.55). The majority of target persons (61.6%) and partners (51.0%) were overweight or obese; 36.6% target persons and 47.1% partners had normal body weight. Regarding chronic diseases, 66.6% of target persons and 40.6% partners reported a diagnosis of type-2 diabetes or cardiovascular diseases (with or without comorbidities) or other chronic diseases (e.g., musculoskeletal disorders). Furthermore, 87.8% target persons declared that they did not meet PA recommendations [5, 22], and the remaining 12.12% reported that they received physician’s recommendations to improve their PA levels due to cardiovascular diseases/ type-2 diabetes. Among partners, 77.5% reported that they did not meet PA recommendations. Target persons and their partners reported that they intended to reduce their own SB levels at T1 (M_TP = 2.91, SD = 0.65; M_P = 2.89, SD = 0.65). Intentions of both persons in the dyads were similar in strength, paired t(1, 319) = 0.46, p = .694. The majority of dyads were in a romantic relationship (61.6%), whereas 38.4% of dyads were in other relationships, involving at least several face-to-face meetings every week (e.g., close friends, family members, workmates). All dyads were in a relationship for > 6 months.

About half of the participants (57.50% target persons and 56.80% partners) had completed higher education; 40.30 % of target persons and 41.90% of partners had a high school or a vocational diploma, or some post-secondary (non-tertiary) education; 2.20% of target persons and 1.30% of partners reported primary education. Half of the target persons (52.20%) and partners (49.40%) perceived their economic status as similar to the economic status of the average family in Poland, 42.20% target persons and 43.70% partners indicated that their economic status was above average; 5.60% target persons and 6.90% partners described their economic situation as worse than the economic status of the average family.

Procedures

This study reports secondary findings of a randomized controlled trial (pre-registered at ClinicalTrials.gov, #NCT03011385). The trial investigated the effects of PA planning interventions (7 planning/control procedures sessions delivered, between 1 week after T1, and 1 week after T2), combined with a healthy lifestyle education (addressing SB, PA, and healthy diet). The primary outcomes were PA and SB assessed over 8 months. To date, the published reports from this trial present the effects of the intervention on PA and SB, whereas social control, and relationship satisfaction were not analyzed [19, 20]. The findings indicated no effects of a planning intervention on SB time at T3 (8 months after T1), neither among target persons nor partners [20]. There was, however, a small effect of a collaborative planning intervention on a reduction of SB time at short-term (1 week after T1) among target persons. This short-term SB assessment was not accounted for in the present study.

Besides the planning interventions or the control condition procedures, all target persons and their partners took part in identical education sessions addressing SB. The education addressed SB definitions and patterns, SB health consequences, and ways to break SB bouts, and reduce overall SB time. No behavior change techniques addressing relationship satisfaction or social control were applied.

T1 self-report was followed by 6 days of accelerometer-based SB measurement, and by T2 self-report assessment, taking place at 2 months after T1. T3 was conducted at 8 months after T1 and included self-reports, followed by 6 days of accelerometer-based SB measurement. Data were collected individually (dyads completed questionnaires separately) during face-to-face meetings of one dyad with an experimenter.

The inclusion criteria were: (1) target persons and partners were ≥ 18 years old; (2) the dyad included a distinguishable target person (i.e., the individual who did not meet the recommended thresholds of PA [22] and/or was recommended by a specialist to reduce SB and increase their PA levels due to a chronic illness such as type-2 diabetes, cardiovascular diseases) and their partner; (3) target persons reported at least moderate intentions to initiate regular PA (4); the dyad was in a close relationship, defined as a romantic or other close relationship (family members, close friends, coworkers) involving several meetings each week; and (5) the relationship lasted > 6 months. Both target person and their partner could report strong intentions to reduce their SB levels or increase PA.

Data were collected between December 2016 and February 2020 in 24 urban locations and 7 rural locations in Poland. Participants were recruited via advertisements published in social media or on websites of non-governmental organizations; recruitment was also conducted during municipality-held health promotion events. Potential participants were informed about the study aims and procedures. After familiarizing themselves with the study goals, participants were screened for eligibility, and were asked to provide informed consent. Overall, 461 dyads were screened for eligibility; 141 either did not meet the inclusion criteria or decided not to take part in the study.

The study was approved by the Ethics Committee at the first author’s institution. All participants provided informed consent. There was no payment for participation; participants received a thank-you gift (value 5–10 EUR) after each measurement.

Measures

Means, standard deviations, and internal consistency coefficients are presented in Electronic Supplementary Material 1, Table S1.

Sedentary Behavior Time (T1 and T3)

SB time data were obtained with ActiGraph GT3X-BT accelerometers. Participants were instructed on how to use the devices and asked to report daily hours of wearing time during their waking hours for 6 days. Data obtained from each participant were used in the analyses only if devices had been worn for at least 8 hr per day, for a minimum of 3 days during the corresponding time period [23]. Data scoring methods were based on the Freedson VM3 [24] and the Freedson Adult [25] algorithms with the Actilife software [24]. Non-wear time was calculated using epoch-based algorithm based on Choi [26]; 10-sec epochs were used for a better distinction between SB and PA [27]. SB time was calculated as the average minutes of SB per every hour of device wearing time.

Perceived Positive and Negative Social Control (T1 and T2)

Seven items were used to assess if target persons and their partners perceived that the other person in the dyad used positive or negative social control to encourage SB reduction. Positive social control was assessed with 4 items based on Lewis and Butterfield [28] and Thorpe [29]: “How does your partner influence (motivate) you to limit the time you spend sitting? (1) repeatedly reminds you to take active breaks; (2) makes suggestions or drops hints; (3) uses humor; (4) uses praises and compliments.” Negative social control was assessed with 3 items based on Lewis and Butterfield [28] and Thorpe [29]: “How does your partner influence (motivate) you to limit the time you spend sitting? (5) being persistent; (6) trying to make you feel guilty; and (7) saying that you would change if you cared for him/her.” The responses, were provided on a 4-point scale ranging from 1 (totally disagree) to 4 (totally agree). Cronbach’s α coefficients ranged between.89 and.92 (see Electronic Supplementary Material 1, Table S1).

Relationship Satisfaction (T1 and T2)

To measure relationship satisfaction, a four-item version of the Couple Satisfaction Index (CSI-4) [30] was used. Participants were instructed to evaluate their relationship with the other person in the dyad using the following items: “Please indicate the degree of happiness, all things considered, of your relationship”, with answers ranging from 1 (very unhappy) to 4 (very happy);” “I have a warm and comfortable relationship with my partner,” with answers ranging from 1 (totally agree) to 4 (totally disagree); “How rewarding is your relationship with your partner?,” with answers ranging from 1 (not at all) to 4 (completely); “In general, how satisfied are you with your relationship?,” with answers ranging from 1 (not at all) to 4 (completely). Although CSI-4 [30] was developed in the context of the romantic relationship, our pilot study (n = 11) indicated that CSI-4 items were perceived as adequately describing satisfaction with the relationship in non-romantic dyads. Values of Cronbach’s α coefficient ranged between .87 and .93 for the total sample (see Electronic Supplementary Material 1, Table S1), between .84 and .94 in participants from dyads in a romantic relationship, and between .88, and .93 in participants from non-romantic dyads.

Control Variables

Sociodemographic covariates used in the sensitivity analysis were: (1) age; (2) gender; (3) education (elementary, vocational, high school, post-secondary, bachelor, master, other—please specify); (4) self-reported economic status, with responses varying from 1 (much above the average family in Poland) to 5 (much below the average family in Poland); (5) the type of relationship (romantic relationship = 1, vs. other, i.e., close family relationship, close friendship, work-related relationship = 0). T1 intention to reduce SB was assessed with 2 items [31], e.g., “I intend to sit for a maximum of 5 hr (in total) a day over the next week.” Responses ranged from 1 (definitely not) to 4 (definitely yes) (target persons: r =.23, p < .001, M = 2.91, SD = 0.65; partners: r =.10, p = .070, M = 2.89, SD = 0.65).

Data Analysis

The G*Power calculator (simulating a multiple regression model) was used to conduct a priori calculations of the sample size. Assuming small effect sizes f² = .05 (in line with previous dyadic longitudinal research [32, 33]), power of .80, Type I error rate of .05 and accounting for age and gender, the determined sample size was approximately 300 dyads.

Path analyses were performed using IBM AMOS versions 26, using maximum likelihood estimation. The two hypothesized models assumed that target persons and partners were distinguishable, and accounted for three measurement points, with the independent, mediator, and dependent variables assessed at separate time points, controlling for T1-level of the dependent variable. Several model-data fit indices were applied. A cutoff point of ≤ .08 for the root mean square error of approximation (RMSEA) was used [34]. A cutoff point ≥ .95 indicating good model-data fit, was applied for the comparative fit index (CFI) and the normed fit index (NFI) [34]. The indirect effects were evaluated with unstandardized effect coefficients, calculated with 10,000 bootstraps (95% CI). Missing data (including data missing due to drop-outs at T2 and T3) were accounted for by using the full information maximum likelihood procedure [34]. Little’s MCAR test indicated that the missing data patterns were systematic, Little’s χ²(N = 766) = 849.535, p = .019. Mardia’s coefficient of multivariate normality (values of 11.22 and 8.52) indicated moderate non-normality.

Analytic Strategy for the Mediation Models

All models assumed that persons within dyads were distinguishable, with set roles as target persons, and partners. Although models were estimated in line with recommendations for actor-partner interdependence model with mediators (APIMeMs) [35], we refrain from using the terms “actor” and “partner” in describing the effects. The models were saturated in terms of the associations between the independent, mediator, and dependent variables, and their respective covariances (e.g., the residuals of independent variables, mediators, and outcome variables were assumed to covary) [35]. The SB indicators at T1, assessed in target persons and partners, were assumed to covary and predict T3 indicators of SB measured in both dyad members. Instead of using one model to test all mediation hypotheses, two hypothesized mediation models were calculated. This strategy allowed us to reduce the potential bias related to multi-collinearity and prevented a reduction of power of analysis related to a high number of parameters in the model (for a similar strategy see [33]).

Several indirect effects were tested: (1) those with the independent, mediator, and dependent variables measured in one person; (2) those with at least one variable in the chain of “the independent variable → the mediator → the dependent variable” measured in one person, and at least one variable in this chain measured in the other person. The total effects, total indirect effects, simple indirect effects, and direct effects were calculated, using the user-defined estimands function [35, 36]. To account for the dyadic interdependence, the independent variables’ indicators (T1) were assumed to correlate; SB indicators (T1) measured in the target persons, and partners were also assumed to correlate. Residuals of the mediators (T2) and SB (T3), measured in both persons in a dyad, were assumed to covary.

Sensitivity analyses were conducted in order to assess the robustness of the findings [37]. We examined whether the pattern of associations was similar in the hypothesized model and the model controlling for the type of relationship (romantic vs. other), target persons and partners’ age, gender, education, economic status (T1), and finally, the effects of the experimental group assignment (1 = PA planning intervention, 0 = the control group) on the mediator and dependent variables. Additionally, a two-group model assuming that direct and indirect effects are equal across two types of dyads (romantic vs. other relationship types) was compared with an unconstrained model [34]. The comparison allowed us to test if the observed direct and indirect effects were similar regardless of the type of the relationship. In case fit indices are good for the two compared models, the more parsimonious model (assuming equality of direct and indirect effects) should be accepted [34].

Results

Preliminary Analyses

Bivariate correlations among study variables, as well as means and standard deviations, are presented in Electronic Supplementary Material 1, Table S1.

Among target persons and partners, analyses for T1 data showed no differences between completers and drop-outs (see Electronic Supplementary Material 1).

There was no change in SB time from T1 to T3 among target persons, F(1, 319) = 3.33, p = .069, η² = .010 (the average SB time per hour time at T1: M_TP = 36.05, SD = 5.48; T3: M_TP = 35.56, SD = 5.53), or among partners, F(1, 319) = 0.75, p = .388, η² = .002.

On average, target persons and partners reported that they were satisfied with the relationship (T1 mean item response on a scale ranging from 1 to 4: M_TP = 3.50, SD = 0.56; M_P = 3.46, SD = 0.56). There were no significant differences in satisfaction between target persons and partners, either at T1, paired t(1, 319) = 1.48, p = .141, or at T2, paired t(1, 319) = ‐0.56, p = .580. Between T1 and T2 there was a small reduction in satisfaction among target persons, F(1, 319) = 10.81, p = .001, η² = .033, but partners reported stable relationship satisfaction across 2 months, F(1, 319) = 1.17, p = .281, η² =.004. At T1, in 90.3 % (n = 289) of dyads both target persons and partners indicated that they were satisfied or very satisfied with their relationship (mean item responses ≥ 2.6 on a scale ranging from 1 to 4).

Target persons reported higher T1 perceived negative control than did their partners, albeit mean levels were low-to-moderate across participants (M_TP = 1.83, SD = 0.81; M_P = 1.65, SD = 0.73), paired t(1, 319) = 3.25, p = .001. Target persons reported higher T1 positive control than did their partners (M_TP = 2.28, SD = 0.84; M_P = 2.13, SD = 0.83), paired t(1, 319) = 2.80, p = .005. At T2, both persons in the dyad perceived similar low levels of negative control (p = .152), but target persons perceived higher positive control than did their partners (M_TP = 2.27, SD = 0.75; M_P = 2.17, SD = 0.75), paired t(1, 319) = 2.02, p = .044.

Findings for the “Control →Relationship Satisfaction →SB Time” Model

The hypothesized model, calculated for N = 320 dyads, had an acceptable fit, with χ²(14) = 25.39, p = .031, χ²/df = 1.814, NFI = .975, CFI = .988, RMSEA = .051 (90% CI [.015, .081]). The variables in the model explained 41.9% of variance in target persons’ SB (T3) and 51.5% of partners’ SB (T3). Associations between the independent variables (T1), mediators (T2), and the dependent variables (T3) are presented in Figure 1 and Table 1. For the values of covariance coefficients see Electronic Supplementary Material 1 (Table S2).

There were no significant simple indirect effects (Electronic Supplementary Material 1, Table S3). However, the analyses conducted for the hypothesized model yielded three direct effects on SB (T3). A higher level of target persons’ perceived positive control (T1) was related to more time spent on SB among target persons (T3), but also with less time spent on SB among their partners (T3). A higher level of target persons’ perceived negative control (T1) was related to less time spent on SB among target persons (T3). Relationship satisfaction indices (T2) were unrelated to SB (T3) of target persons or partners. High levels of target persons’ perceived positive control (T1) and low levels of target persons’ perceived negative control (T1) were associated with high levels of their own and their partners’ relationship satisfaction (T2). Partners’ perceived positive control (T1) was likewise positively associated with partners’ relationship satisfaction (T2).

The sensitivity analysis, accounting for gender, age, education, and economic status, (T1) of target persons and partners, the type of relationship (1 = romantic vs. 0 = other), and the effects of experimental group assignment on the mediator and dependent variables indicated patterns of effects similar to those obtained in the hypothesized model (Electronic Supplementary Material 1, Tables S4-S6). Thus, the robustness of the findings was confirmed.

The two-group model analysis, comparing dyads in romantic vs. non-romantic relationships, indicated that the model which assumed that all direct and indirect effects were equal across the two groups, had a good model-data fit (see Electronic Supplementary Material 1, Table S7). Thus, the more parsimonious model, assuming equality of the associations across two types of dyads, was accepted [34]. The two-group model yielded a similar pattern of associations to those found for the hypothesized one-group model (Electronic Supplementary Material 1, Table S8).

Findings for the “Relationship Satisfaction →Control →SB Time” Model

The hypothesized model, calculated for N = 320 dyads, had an acceptable fit, with χ²(14) = 30.34, p = .007, χ²/df = 2.167, NFI = .973, CFI = .985, RMSEA = .060 (90% CI [.031, .090]). The variables in the model explained 40.5% of variance of target persons’ SB (T3) and 50.5% of partners’ SB (T3). For associations between the independent variables (T1), mediators (T2), and the dependent variables (T3), see Figure 2 and Table 2. The values of covariance coefficients are presented in Electronic Supplementary Material 1 (Table S9).

The analysis of the hypothesized model showed two simple indirect effects (Table 2, see also Electronic Supplementary Material 1, Table S10). First, a higher level of relationship satisfaction among target persons (T1) was related to target persons perceiving higher levels of negative control (T2), which in turn predicted lower SB time among target persons (T3). The indirect effect coefficient was significant, b = ‐0.502, SE = 0.113, 95% CI [‐1.027, ‐0.142], p = .007. Second, partners’ reports of lower levels of relationship satisfaction (T1) predicted target persons’ reporting higher levels of perceived negative control (T2). In turn, higher levels of target persons’ perceived negative social control predicted lower levels of target persons’ SB (T3). The respective indirect effect coefficient was significant, b = ‐0.268, SE = 0.151, 95% CI [0.048, 0.668], p = .011.

Analyses yielded one additional direct effect, explaining partners’ SB (T4): a high level of relationship satisfaction among partners (T1) was associated with them spending more time on SB (T3). Direct effects of predictors on proposed mediators involved two positive associations of target persons’ relationship satisfaction (T1) with their own and their partners’ perceived positive control (T2). Higher levels of partners’ relationship satisfaction (T1) predicted lower perceived negative control among target persons.

The sensitivity analysis, controlling for sociodemographic variables (T1) of target persons and partners, the type of relationship (1 = romantic vs. 0 = other), and the effects of the experimental group assignment, indicated a pattern of effects similar to those obtained in the hypothesized model (Electronic Supplementary Material 1, Tables S11-S13). Thus, the robustness of the findings was confirmed. Two indirect effects obtained in the total sample were also significant in sensitivity analyses (Electronic Supplementary Material 1, Table S12).

The two-group model analysis, comparing romantic vs. non-romantic dyads, showed that the model assuming all direct, and indirect effects were equal across the groups yielded good model-data fit (Electronic Supplementary Material 1, Table S14). The more parsimonious model, assuming that the direct and indirect effects are equal across the two types of dyads, was accepted [34]. This model showed a pattern of associations similar to those obtained in the hypothesized one-group model (Electronic Supplementary Material 1, Table S15).

Discussion

This prospective study indicated an intriguing pattern of associations among relationship satisfaction, perceived positive and negative control, and accelerometer-assessed sedentary behaviors in dyads involving two adults who were family, friends, or in a romantic relationship. The study yields partial support for one of the formulated hypotheses based on two frameworks: the dyadic relationship factors and health [14] and the DHIM [16]. We found that higher relationship satisfaction of target persons and lower satisfaction of partners were linked with target persons’ reports of (relatively) higher use of negative social control by the partner, and, in turn, lower SB time among target persons. The “social control → relationship satisfaction → SB time” hypothesis was not confirmed.

Previous research indicated small unfavorable effects of negative social control on engagement in health-promoting behaviors [8]. Our findings show that these effects should be considered in the context of the relationship satisfaction in the dyad and the levels of perceived negative control. In particular, the indirect effects in our study have to be considered in the following context: (1) even those participants who were less satisfied reported moderate satisfaction with the relationship; (2) the “high levels” of the perceived use of negative control strategies meant that the participant reported perceiving an occasional use of negative control by the other person in the dyad. In line with the DHIM [16], it seems plausible that a moderately satisfied partner in such a dyad might use some negative control to influence the target person’s behavior, whereas the satisfied target person will perceive some negative social control and will comply with their partner’s wishes to secure the partner’s engagement with the relationship.

As suggested by Gleason [21], dyads in which both members intend to change their behavior may feel that their shared intentions legitimize the use of negative control and may benefit from this type of social control. In the present study people intending to participate in an intervention enhancing PA were enrolled (with at least target persons intending to increase their PA). Thus, as it might be expected, both participating dyad members reported at least moderate levels of intention to reduce their SB as well. The intention to change SB may have been a context in which perceived negative control facilitated target persons’ behavior change. We found that target persons’ perceived positive social control (T1) was related to lower SB time in partners 6 months later (T3). The perceptions of the use of positive control strategies indicate that a target person reported their partner reminding them of active breaks, making suggestions or dropping hints to reduce SB, or praising and complimenting a reduction of SB. Engaging in such control actions by the partners may require their awareness of time spent on SB by the other person in the dyad, possibly partners’ awareness of their own SB time, and engaging in modeling of SB reduction. Our findings are consistent with the results obtained in research on social support provision. Berli et al. [38] have found that among romantic partners, higher daily support provision to another person in a dyad was associated with higher own objective moderate-to-vigorous PA levels.

The indirect effects were found for SB of target persons only. This may be explained by the specific nature of the enrolled dyads. One dyad member was identified as the target person, either because they declared that their PA levels were below the recommended PA thresholds [24], or that they were recommended by a specialist to increase their PA levels due to a chronic illness, and reported at least moderate intentions to initiate regular PA. Thus, the SB time reduction was most likely to occur in target persons.

Besides hypothesized mediation effects, it is possible that relationship satisfaction may act as a moderator of effects of social control on health behavior, as suggested in the contextual model [39]. Among dyads with low relationship quality, both positive and negative social control may lead to unfavorable changes in health behaviors [39]. Previous studies conducted among romantic dyads indicated that people with high relationship quality report more beneficial behavioral outcomes of social control than those in less satisfied dyads [17]. Future research investigating the role of relationship satisfaction should involve dyads with a high variation of relationship quality and test the competing mediation and moderation models. Participants enrolled in our study reported high levels of relationship satisfaction, therefore the moderator hypothesis could not be tested as the alternative model.

As this is one of the first studies testing the indirect (mediating) associations between social control, relationship satisfaction, and behavior change in dyads [16], implications for practice may be premature. Further research assessing the two competing hypotheses in dyads is needed, for example to confirm if moderate levels of relationship satisfaction in partners may be linked to the perception of some negative control by the satisfied target persons, and, consequently, with a positive change in other health behaviors.

The present study has several limitations. The majority of participants were people with higher education and medium or higher economic status, which limits any generalizations. The findings cannot be generalized to individuals with weak intentions to exercise, or to dyads that are dissatisfied with their relationship. In contrast to the majority of previous studies [8], we did not focus exclusively on romantic relationships but included other dyads who were family members or close friends as well. Triaxial hip-worn accelerometers were used to capture SB, whereas more preferable devices would involve instruments allowing for a better differentiation between sitting, standing, and light-intensity PA. Although sensitivity analyses indicated that the associations obtained in the hypothesized models were similar after controlling for assignment to the experimental condition, further indirect effects of the intervention on the mediators/dependent variables are possible. Another limitation refers to a lack of testing for complex underlying social exchange or self-regulation processes, that may explain health behavior change.

Conclusions

Among dyads participating in an intervention to increase PA, both higher levels of satisfaction with the relationship among target persons and (relatively) lower levels of satisfaction among partners were related to (relatively) higher negative control perceived by the target persons. In turn, the (relatively) higher levels of negative control were related to better behavioral outcomes in target persons. Findings held for dyads in romantic and other close relationships, with family or friends. Overall, participants reported moderate-to-high relationship satisfaction and low-to-moderate perceived negative control.

Funding

This study was supported by grant no. 2014/15/B/HS6/00923 from National Science Center, Poland, awarded to Aleksandra Luszczynska

Compliance with Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards

Conflict of interest Authors Maria Siwa, Zofia Szczuka, Anna Banik, Ewa Kulis, Monika Boberska, Dominika Wietrzykowska, Nina Knol, Anita DeLongis, Bärbel Knäuper, and Aleksandra Luszczynska declare that they have no conflict of interest.

Research involving human participants and/or animals All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional research committees and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent Informed consent was obtained from all participants included in the study.

Transparency Statement

1. Study registration: This study reports secondary findings of a randomized controlled trial pre-registered at ClinicalTrials.gov (#NCT03011385).

2. Analytic plan: The analysis plan was not formally pre-registered.

3. Data availability. De-identified data from this study are not available in a public archive. For research purposes, de-identified data from this study can be obtained from the corresponding author.

4. Analytic code availability. Analytic code used to conduct the analyses presented in this study are not available in a public archive. Analytic codes used to conduct the analyses can be obtained from the corresponding author.

5. Materials availability. Materials used to conduct the original randomized control trial (including the protocol and inttervention/education procedures) are publicly available at the Open Science Framework platform (https://osf.io/68gp2/). The remaining materials can be obtained from the corresponding author.

CRediT Author Statement

Maria Siwa: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Visualization, Writing—original draft, Writing—review and editing.

Zofia Szczuka: Data curation, Investigation, Project administration, Resources, Visualization, Writing—review and editing.

Anna Banik: Data curation, Investigation, Project administration, Resources, Writing—review and editing.

Ewa Kulis: Data curation, Investigation, Project administration, Resources, Writing—review and editing.

Monika Boberska: Data curation, Investigation, Project administration, Resources, Writing—review and editing.

Dominika Wietrzykowska: Data curation, Investigation, Project administration, Resources, Writing—review and editing.

Nina Knoll: Conceptualization, Formal analysis, Methodology, Writing—review and editing.

Anita DeLongis: Conceptualization, Writing—review and editing.

Bärbel Knäuper: Conceptualization, Writing—review and editing.

Aleksandra Luszczynska: Conceptualization, Funding acquisition, Methodology, Software, Supervision, Validation, Writing - original draft, Writing—review and editing.

References