How Do People Experience and Respond to Social Control From Their Partner? Three Daily Diary Studies

Scholz, Urte; Stadler, Gertraud; Berli, Corina; Lüscher, Janina; Knoll, Nina

doi:10.3389/fpsyg.2020.613546

ORIGINAL RESEARCH article

Front. Psychol., 13 January 2021

Sec. Personality and Social Psychology

Volume 11 - 2020 | https://doi.org/10.3389/fpsyg.2020.613546

This article is part of the Research Topic Individual versus Dyadic Processes: Health and Relationship Outcomes View all 18 articles

How Do People Experience and Respond to Social Control From Their Partner? Three Daily Diary Studies

$\r\nUrte Scholz,*$ Urte Scholz^1,2*

Gertraud Stadler³

Corina Berli¹

Janina Lüscher¹

Nina Knoll⁴

¹Applied Social and Health Psychology, Department of Psychology, University of Zurich, Zurich, Switzerland
²University Research Priority Program “Dynamics of Healthy Aging,” University of Zurich, Zurich, Switzerland
³Institute Gender in Medicine, Charité – Universitätsmedizin Berlin, Berlin, Germany
⁴Division of Health Psychology, Department of Education and Psychology, Freie Universität Berlin, Berlin, Germany

Positive and negative forms of social control are commonly used to regulate another person’s health-related behaviors, especially in couples. Social control efforts have been shown to result in desirable, but also undesirable effects on different outcomes. Little is known for which outcomes, when, and under which contextual conditions these different effects unfold in people’s everyday lives. Using the dual-effects model of health-related social control, we predicted that same-day and previous-day positive social control would result in desirable effects on target behavior, and same-day positive control on affect. Same-day and previous-day negative control was assumed to result in undesirable effects on reactant responses (i.e., doing the opposite of what the partner wanted and hiding the unhealthy behavior), and same-day negative control on affect. Further, we explored whether it makes a difference if one or both partners intend to change their health behavior. Three daily diary studies addressed these questions for smoking (Studies 1 and 2), and physical activity (Study 3). Receiving more positive control related to more desirable target behavior, and feeling better; more negative control was associated with more reactant responses and feeling worse. Social control unfolded its effects within 1 day, but hardly across days, indicating that control and its reactions to it are fast-acting processes in daily life. The pattern of results were the same for couples with one and both partners intending to change their behavior. Further, results replicated when using partner-reported provided control. Based on these results, social control cannot be unanimously recommended as a behavior change strategy in couples. Future studies should follow up on dyadic and temporal dynamics of social control in couples’ everyday lives in different contexts.

Introduction

Interest in the role of social relationships for health-related self-regulation has increased in social and health psychology in recent years, with a growing number of publications (e.g., Overall et al., 2009; Fitzsimons et al., 2015; Young et al., 2019). One interpersonal process involved in the social regulation of health behaviors is social control, defined as the influence on and regulation of another person’s behaviors (Lewis and Rook, 1999). Existing evidence on associations of social control with health behaviors and additional outcomes, such as affect or reactant responses, is mostly based on cross-sectional studies, while the few available longitudinal studies mainly speak to differences between persons (Craddock et al., 2015). It is thus largely unknown how and when, and for which outcomes social control effects unfold within individuals over time. Moreover, it is an open question whether effects of social control differ depending on context, i.e., whether one or both partners of a dyad intend to change their health behavior. Three daily diary studies presented in this article set out to address these questions.

Social Control: Definition, Theoretical Models, Empirical Evidence, and Research Gaps

Changing health behaviors is notoriously difficult, as research on self-regulation has shown (e.g., Sheeran, 2002). Although there is a long tradition of social psychological theories pointing to the importance of close others for self-regulation (van Lange and Rusbult, 2012), this issue has been largely neglected in health-related self-regulation research so far (Fitzsimons et al., 2015). In this report, we will highlight the role of health-related social control for two prominent health behaviors, as well as for affect, and reactant responses.

Health-related social control is defined as a set of specific interpersonal strategies meant to influence and regulate another person’s health behaviors (Lewis et al., 2004; Young et al., 2019). It is distinct from social support (cf. Rook, 1990) that aims at easing a challenging situation in the support recipient (Cohen, 2004). While this distinction is straightforward, partner regulation, a prominent concept from relationship science focusing on romantic relationships (Simpson et al., 2018; Baker and McNulty, 2020) is more closely related to social control. Partner regulation is defined as “an attempt to resolve relationship problems that arise” (p. 5, Baker and McNulty, 2020). Partner regulation and health-related social control thus address different targets. The only overlap between these two constructs occurs when it comes to influencing health behaviors that cause (potential) problems not only for one partner’s health, but also for the relationship. However, even in this area of overlap an important distinction between partner regulation and social control is that a prominent form of cooperative partner regulation is social support (Baker and McNulty, 2020) which again is distinct from social control. Owing to the differences in partner regulation and health-related social control the present work draws primarily on the health-related social control literature.

Several theoretical models on how health-related social control relates to health behavior and additional outcomes have been proposed (Okun et al., 2007), with the modified dual-effects model of social control receiving recent support by a meta-analysis (Craddock et al., 2015). The original dual-effects model postulates beneficial effects of social control on the target behavior, but at the same time costs in the form of reduced positive and increased negative affect in the social control recipient (Okun et al., 2007). The modified version of the dual-effects model distinguishes between positive and negative social control and has been tested in a number of studies (e.g., Lewis and Rook, 1999; Fekete et al., 2006; Scholz et al., 2013).

Positive control has been defined with a negotiation component: Positive control providers try to get recipients to agree with the change, using control strategies such as discussions about the behavior, complimenting recipients on change attempts or using reminders. Negative control in contrast is defined as lacking this negotiation component and as relying on using pressure and inducing guilt instead for making the target person change their health behavior (Lewis et al., 2004).

Empirical evidence indicates that positive control relates to the behaviors targeted at with a moderate-size effect, while negative control is unrelated to behavior (Craddock et al., 2015). Higher negative control was also moderately related to higher negative affect and lower positive affect. Positive control was unrelated to negative affect and moderately related to more positive affect (Craddock et al., 2015).

Recent research on the dual effects model of health-related social control has gone beyond the target behavior and affect as outcomes and additionally included reactant responses. The intertwined model of reactance postulates that reactance comprises negative affect (such as anger) and negative cognition components that are inseparable (Dillard and Shen, 2005; Rains, 2013). Reactance is triggered by a perceived threat to one’s autonomy (Brehm and Brehm, 1981); it is a common response to attempts of external persuasion and regulation (Dillard and Shen, 2005). Responses to social control, such as doing the opposite of what the control provider wanted the recipient to do (e.g., Tucker, 2002) and hiding undesirable behaviors (Tucker and Anders, 2001), can be considered consequences of reactance in that they represent different forms of direct restoration of one’s autonomy (Brehm and Brehm, 1981). An example for doing the opposite of what the control provider wanted is to smoke even more cigarettes, examples for hiding an undesirable behavior is to smoke in hiding or not telling the partner about skipping the exercise class. In health-related social control research, only a minority of studies has considered these additional, reactance-related outcomes of health-related social control (Craddock et al., 2015). Yet, investigating social control effects on these reactance-related responses is central for gaining a more comprehensive understanding of the consequences of social control. This study will address this issue by applying an extended version of the dual-effects model of social control: We relate both positive and negative social control to health behaviors and affect, and also to hiding the behavior and doing the opposite of the control provider’s intention as two reactance-related responses.

Between-Person and Within-Person Effects of Social Control

Most research on social control has tested hypotheses by focusing on differences between people: If Person A reports a higher level of negative control compared to Person B, Person A will on average also report higher levels of negative affect than Person B. The association between negative social control and negative affect within Person A or Person B over time, however, remains unknown (Hamaker, 2012), e.g., that at times of higher social control people will also experience more negative affect. In fact, the within-person association may be zero or even in the opposite direction as the between-person association (Hamaker, 2012). The strong focus on between-person associations is also in contrast to the causal claim most theories, including the dual-effects model of social control, make on associations within persons. The majority of studies on health-related social control followed cross-sectional designs and of the longitudinal ones few examined associations at the within-person level (Craddock et al., 2015). Those studies largely confirmed the hypothesized associations of the dual-effects model, but mostly while examining same-day relations (Khan et al., 2013). Consequently, an open question in health-related social control research is how effects of social control unfold over time within individuals. For answering this question, we will focus on micro-time dynamics from 1 day to another to gain a better understanding of the temporal dynamics of social control and its outcomes in people’s everyday lives (e.g., Scholz, 2019).

Social Control in Context

A further extension of previous research on social control is the inclusion of contextual factors to better understand under which conditions positive and negative social control unfold their beneficial or undesired effects (Tucker, 2002). So far, relationship quality has been studied most frequently as a contextual moderator of social control effects. Couples with higher relationship quality showed more beneficial and less undesirable outcomes of social control than less satisfied couples (Tucker, 2002; Knoll et al., 2012; Scholz et al., 2013). Another context factor that has not yet been addressed in research on effects of social control is the couple constellation with regard to the intended change in the health-related behavior, for instance, whether one or both partners intend to stop smoking or increase their physical activity. Couples with only one partner intending to change their behavior are more likely to rely on one-sided social control receipt and provision than couples with both partners intending to change jointly. This might be related to stronger threats to recipients’ autonomy, and consequently evoke stronger reactance-related responses, more negative affect, and less behavior change. In contrast, couples in which both members want to change their behavior, the joint endeavor might better legitimize the use of positive and negative control strategies, or enable reciprocation among partners (Gleason et al., 2003), thereby potentially buffering detrimental and intensifying beneficial effects of both positive and negative social control on behavior, affect, and reactance-related responses. The present research comprised different couple constellations with regard to individual or joint intended behavior change allowing to inspect different patterns of results.

Aims of the Present Research

Based on current evidence, it is largely unknown how, when and for which outcomes social control effects unfold within individuals over time. Moreover, it is an open question whether effects of social control differ depending on whether one or both partners of a dyad intend to change their behavior. This article addresses these questions using data from three daily diary studies, focusing on the within-person perspective.

In line with the modified dual-effects model (Butterfield and Lewis, 2002; Lewis et al., 2004; Scholz et al., 2013), we assumed positive associations between positive social control and the target behavior (i.e., less smoking or more physical activity, respectively). We hypothesized an effect of previous day (Hypothesis 1a, H1a) and same day (Hypothesis 1b, H1b) positive control on the target behavior. Positive control strategies are mostly prospective in nature (discussing, persuading, reminding) and may therefore likely unfold their effects from 1 day to the next.

Given the evidence from the dual-effects model, no hypothesis on associations between negative control and target behavior was formed. We refrain from generating explicit hypotheses for these kind of null effects, because proving that there is no or only a trivially small effect would require impossibly large samples (Cohen, 1990) that are beyond the scope of the intensive longitudinal dyadic studies reported here.

In terms of affect, we investigated whether or not people feel better or worse after receiving positive and negative social control from their romantic partners. Affective responses are likely to change quickly with changes in the situation, i.e., when control is present or not (cf. Gross, 2014). Therefore, we primarily assumed short-acting effects of social control on affect within a given day. We expected that people feel better in response to receiving positive control on the same day (Hypothesis 2; H2), and that they feel worse in response to receiving negative control on the same day (Hypothesis 3; H3).

Regarding associations between positive and negative control with reactant responses, we hypothesized that more negative social control received on the previous day (i.e., lagged effect; Hypothesis 4a, H4a) and on the same day (Hypothesis 4b, H4b) would be associated with more reactant responses, i.e., more doing the opposite of what the partner wanted and more hiding the behavior. Receiving high levels of negative social control is assumed to result in immediate and non-volatile reactance (anger and negative cognitions; Dillard and Shen, 2005). Attempting to restore one’s autonomy will take place the same day the reactance is experienced (i.e., same day effect). Further, the higher the level of negative control received the previous day, the higher the likelihood that this restoration of one’s autonomy might continue from the previous day to the next day (i.e., lagged effect). Positive social control strategies, such as discussions with the target person, still leave room for recipients to choose whether or not to adopt the target behavior. Consequently, the recipients of positive social control would experience no or only a weak threat to their autonomy. Therefore, no hypotheses are set for the associations between positive control and reactant responses.

Finally, we will address a gap in the literature, namely how contextual factors might affect associations between social control and its outcomes. Particularly, we will explore whether effects of social control differ depending on whether one or both partners of a dyad intend to change their behavior. This will be examined at an exploratory level as the couple constellations are part of the different studies. Study 1 tests hypotheses in a couple constellation with only one partner intending to change their behavior, whereas Study 2 and 3 examine social control in couples with both partners intending to change their behavior.

Hypotheses tests and exploratory analyses examined received positive and negative social control reported by target persons. Additionally, effects of partner-provided positive and negative control were examined to learn whether results hold across different perspectives and to exclude potential artifacts due to shared method variance.

Study 1: Smoker-Non-Smoker Couples

Participants were couples with one smoking partner undergoing a quit attempt. Study 1 focuses on 22 end-of-day diary entries (self-set quite date and 21 days after) of 70 smokers who relapsed after their self-set quite date (of a total of 100 quitting smokers). In line with prior research, relapse was defined as having smoked more than five cigarettes since the quit date at the project’s 1 month follow-up (West et al., 2005). The focus was on relapsers only, because during the time of assessment successful quitters smoked no cigarettes after the quit attempt, therefore they did not show within-person variability in smoking and could not contribute to a better understanding of within-person links between social control and smoking, reactance-related outcomes, and affective reactions.

Methods

Procedure and Sample

The data of Study 1 are from the larger project “Dyadic and Individual Regulation to End Chronic Tobacco Use (DIRECT)” funded by the Swiss National Science Foundation (100014_124516/1). Participants were adult smokers intending to quit during the study, and their non-smoking partners (for a full description of this study’s methods please see Ochsner et al., 2014). To be eligible, smokers had to smoke at least one cigarette daily, intend to quit during the study, be married or in a committed heterosexual relationship for at least 12 months, and be living with their non-smoking partner for at least 6 months. Smokers who attended a smoking cessation program, worked shifts of 24 h, were not fluent in German, or were pregnant were not eligible to participate. Couples received (100 Swiss Francs, about $109) for participating in the diary phase. Participants were provided with study smartphones and instructed to complete daily questionnaires within 1 h of going to bed for 22 consecutive days starting at the self-set quit date. Participants were treated in accordance with the ethical guidelines of the Helsinki Declaration (2000).

Participants had a mean age of 39.96 years (SD = 10.40), 22 (31.4%) were women. The majority (70%) reported 9 years of schooling and were currently employed (81.4%). Roughly half (55.7%) of the participants had children, with 35 (50%) living with the children in the same household.

Measures

Means and standard deviations across the diary phase, as well as the range of the scales, ICCs and number of participants providing data are displayed in Table 1. All item examples are translations from German.

TABLE 1

Table 1. Means, standard deviations, ranges across all diary days, and intraclass correlations (ICC) of main variables in Study 1, Study 2 for women and men separately, and Study 3.

Positive and negative social control were assessed with four items each by Butterfield and Lewis (2002) and adapted to the context of smoking (Ochsner et al., 2015) and daily assessments. A sample item for positive social control is “Today, my partner tried to influence my smoking behavior by trying to persuade me to reduce my smoking or to quit.” A sample item for negative control is “Today my partner tried to influence my smoking behavior by trying to make me feel guilty.” The response format ranged from 1 = never today to 4 = frequently today.

Partner provided positive and negative control was assessed with exactly the same items and response format, but from the perspective of the provider (e.g., Today, I tried to influence my partner’s smoking behavior by trying to make him/her feel guilty).

Target Behavior: Daily number of cigarettes smoked was assessed by two questions. “Did you smoke today (including only one puff)?” If participants indicated yes, they were asked to report how many cigarettes they had smoked (Heatherton et al., 1991); if they indicated no, daily number of cigarettes smoked was set to 0.

Affect after having received control was measured daily with the item: “How did you feel today after your partner tried to influence you this way?” Response options ranged on a seven-point scale from −3 “much worse” via 0 “unchanged” to + 3 “much better.” A further option for participants was to indicate that their partner did not try to influence their smoking behavior today, which was then coded as missing. This resulted in 49 relapsers providing data for this item across the 22 days.

Reactant responses. Doing the opposite of what the partner wanted was assessed with an item adapted from Tucker and Anders (2001): “Today I did exactly the opposite of what my partner wanted me to do with regard to my smoking.” Hiding smoking from the partner was assessed with an item adapted from Tucker and Anders (2001): “Today I hid my smoking from my partner.” For both items, response options ranged from 1 “today not at all” to 6 “today very frequently.” Again, a further option for participants was to indicate that their partner did not try to influence their smoking today, items were then coded as missing. Across 22 days, data for doing the opposite were provided by 49 participants; data for hiding smoking were provided by 56 participants.

Analytic Strategies

We used multilevel modeling to account for the hierarchical data structure, following recommendations by Bolger and Laurenceau (2013). Within-person (Level 1) predictors were person-mean centered. These centered variables provide information on the daily fluctuation around the person-specific mean, testing links between positive and negative social control and the different outcome variables within persons (Level 1). Between-person (Level 2) predictors, i.e., the average scores across the diary days of the respective variables were grand-mean centered at the sample mean. Due to the different number of participants for the outcomes affect, doing the opposite, and hiding in Study 1 and 2, the centering of the between-person variables was adjusted for the subsamples with available data.

To test Hypotheses 1–4, lagged analyses with a lag of 1 day were run. Models included between-person positive and negative control, within-person previous day, and same day positive and negative social control together with previous day’s outcome to predict the same day outcome. Testing effects of previous day positive and negative control on present day’s outcomes allow to establish temporal order. Further, including present day’s positive and negative control as predictors excludes the possibility that effects of previous day positive and negative control on present day’s outcome are artifacts due to strong associations of present day’s positive and negative control with present day’s outcome. Note that predictors are strongly correlated from one day to the next. Similarly, including previous day outcome in the analyses excludes the possibility that effects of previous day positive and negative control on present day outcomes are artificially inflated because of correlations between previous day positive and negative control and previous day outcomes. Due to lagged analyses, the second diary day served as the first outcome day. A linear time trend centered on Day 2 controlled for common time effects. One unit increase in time indicated 1 week representing all diary days (Day 1 = 0, Day 2 = 0.14, …, Day 7 = 1, Day 8 = 1.14, …).

Variables were standardized for better comparability of effects. While it is possible to standardize with the between-person standard deviation as was done here, it is also possible to standardize with the pooled within-person standard deviation or with the individualized standard deviation. For the present studies, standardizing with the between-person standard deviation was chosen for the following reasons: Standardizing with the between-person standard deviation allows comparing effect sizes (a) across the three studies for different couple constellations, (b) within our studies for between- and within-person effects, and (c) with other studies, e.g., from the meta-analyses on social control (Craddock et al., 2015). Although for between- person effects between-SD standardization and for within-person effects the within-SD standardization would have been ideal, using different standard deviations for standardizations would have prevented comparison of the different effect sizes of between- and within-person effects and with other studies. Thus, we standardized all effects with the between-person standard deviation by dividing all variables by their respective between-person standard deviation (SD). Of note, the between- and within-person SDs were largely comparable in size (see Table 1, and Supplementary Table P1 for provided control). For each SD increase in the predictor, the outcome changes in SDs as much as the regression weight indicates. This allows interpretation of effect sizes: b < 0.3 = small effect, 0.3 ≤ b < 0.5 = medium effect, b ≥ 0.5 = large effect (Cohen et al., 2003). As daily number of cigarettes smoked is a meaningful metric, this outcome was not standardized in Studies 1 or 2.

Daily number of cigarettes smoked (Studies 1 and 2) was a count variable, thus generalized linear mixed Poisson models with a logarithmic link function were used (Xie et al., 2013), resulting in rate ratios as the regression coefficients. Rate ratios (RR) indicate that a one-unit increase in the predictor results in percentage increase (distance above one) or percentage decrease (distance below one) in the criterion (Atkins et al., 2013). For the other three outcomes in Studies 1 and 2 (doing the opposite, hiding, and affect) linear mixed models were run.

For all analyses, a maximal random effects structure was specified (Barr et al., 2013). In case of non-convergence, the random effects structure was successively reduced until convergence was met. For parsimony, we reported results of the Level-1 random effects in the Supplementary Material S1 (and P2 for provided control). Intra-class correlations (ICC) for all measures were computed. The ICC is the amount of variance between second-level units, in our case persons, in relation to total variance (Bolger and Laurenceau, 2013). All analyses were conducted in SPSS 23, with a probability level of p = 0.05.

Sensitivity analyses were conducted for all models. We adjusted for nicotine dependence at baseline (Fagerström test of nicotine dependence; Heatherton et al., 1991), age, and gender in Study 1 and nicotine dependence and age in Study 2 (where we ran separate analyses for men and women). In Study 3, we adjusted for gender, age, BMI, intervention group vs. control group, intervention phase vs. follow-up phase, and weekday vs. weekend, and weartime of accelerometers in the analyses of physical activity. For parsimony, we report the unadjusted analyses in this article. Tables depicting sensitivity analyses are reported in the Supplementary Material S2 (and Supplementary Table P2 for provided control).

Results Study 1

Results of Study 1 are displayed in Table 2. For random effects see Supplementary Tables S1-1. Results of partner-reported provided social control, largely reflecting the results of received social control, are displayed in Supplementary Table P1-1. Sensitivity analyses showed the same patterns of results, see Supplementary Table S2-1 for received and Supplementary Table P2-1 for provided control.

TABLE 2

Table 2. Study 1: within- and between-person effects of negative and positive social control on daily number of cigarettes smoked, affect, doing the opposite, and hiding smoking after the quit date for relapsing smokers.

Does Positive Social Control Predict Daily Number of Cigarettes Smoked (H1a and H1b)?

In contrast to H1a assuming a previous-day effect, but in support of H1b assuming a same-day effect, positive control on the same day, but not the previous day was significantly related to smoking: On days with higher than usual same-day positive control smokers reported 5% less cigarettes smoked (see Table 2). As expected from the theoretical assumptions of the extended dual-effects model, same day and previous day negative control were unrelated to daily number of cigarettes smoked.

Does Positive and Negative Social Control Predict Affect (H2 and H3)?

In support of H2, same day positive control was associated with feeling better. Additionally, a positive effect of previous-day positive control on feeling better emerged. In line with H3, same-day, but not previous-day negative control was related to smokers feeling worse.

Does Negative Social Control Predict More Reactant Responses (H4a and H4b)?

Supporting H4b, but contrasting H4a, only same day, but not previous day negative control was related to doing the opposite, indicating that higher than usual levels of negative control were related to more doing the opposite within the same day. Similarly, only same day but not previous day negative control significantly predicted hiding: More than usual same day negative control related to more hiding. Further, same day positive control was significantly related to less hiding, but –as expected from the extended dual-effects model- unrelated to doing the opposite.

Brief Discussion of Study 1 Results

Overall, results mainly confirmed our hypotheses on same-day effects (see Table 3 and Supplementary Table P6 for provided control for a color-coded overview of the results across all studies with regard to confirmation/disconfirmation of Hypotheses 1–4). Positive control predicted fewer cigarettes smoked (H1b) and less hiding. Negative control predicted more doing the opposite and hiding (H4b). However, these effects emerged only for the same day, but not the previous day. This speaks in favor of social control being a fast process linked with rather immediate outcomes in people’s everyday lives. For the reactant responses this might be explained by relationship maintenance issues: displaying maladaptive behaviors across several days may prove negative for the relationship (Burke and Segrin, 2017). Thus, smokers might try to avoid prolonged negative behavioral reactions for the sake of their relational well-being.

TABLE 3

Table 3. Overview of results for study 1, study 2 (for men and women), and study 3 for received positive and negative social control.

Hypotheses 2 and 3 for affect were largely confirmed and highlight the applicability of the assumptions on affective correlates in the extended dual-effects model to smokers’ everyday lives. The more consistent link of affective reactions to receiving positive and negative control compared to the behavioral outcomes might also be explained by better measurement precision: Affect was assessed as a direct affective reaction to receiving control from the partner. All behavioral outcomes were assessed in a less targeted manner, leaving room for other influences.

The results of Study 1 emerged in a couple constellation with only one partner intending to change their behavior. Studies 2 and 3 addressed the same hypotheses in the context of both partners intending to change their behavior.

Study 2: Dual-Smoker Couples

The second study had the same research design as Study 1, but participants were dual-smoker couples intending to quit jointly on a self-set quit date. This allows for a comparison between effects of social control in a different couple context. Due to this planned comparison, data for female and male smokers were analyzed separately to ensure the same structure of results as in Study 1.