Physical activity–related social control and social support in older adults: Cognitive and emotional pathways to physical activity

Abstract

A survey of 217 older adults assessed physical activity–related positive and negative social control and emotional and informational support, using structural equation modeling to investigate mediational effects of emotional responses and behavioral intentions on physical activity. There were significant indirect effects of social control and social support on intentions as mediated by positive, but not negative, emotional responses, and significant indirect effects of emotional responses on physical activity as mediated by intentions. These findings help to identify the cognitive and emotional pathways by which social control and social support may promote or detract from physical activity in later life.

Keywords

emotions exercise health information intentions physical activity social control social support

Of all age groups, older adults are particularly prone to physical inactivity (Shaw et al., 2010) and the rate of obesity among older adults is increasing more rapidly than any other group (Ogden et al., 2014). Despite the high risk of inactivity among older adults, too little attention has been paid to social and psychological determinants of physical activity among this group. Altering older adults’ social environment is likely to be critical to improving their physical activity levels. There is general consensus that social network members can positively influence physical activity (e.g. Heaney and Israel, 2008; Mendonça et al., 2014), but we need to learn considerably more about which specific aspects of social interactions influence physical activity and how.

Two specific ways in which social network members may have a positive impact on older adults’ levels of physical activity are through the provision of physical activity–related social support (Franks et al., 2006) and through physical activity–related social control attempts (Cohen, 2004). Physical activity–related social support refers to attempts by network members to provide positive encouragement and feedback to individuals who are currently engaging in positive exercise behaviors (Sallis et al., 1987). Although definitions of social support can vary, ranging from perceived availability of support in times of need to actual received support, for the purposes of this study, we define social support as received support. Support may be encouraging and emotionally affirming when it is appraised positively (Newsom et al., 2005) or it may convey valuable knowledge (Rimal, 2001). It may not always be related to positive emotions or health consequences; however, as it sometimes results in negative consequences when it is too intrusive, critical, or demanding (i.e. “miscarried support”; Coyne et al., 1988). Social support constructs in more general contexts commonly include multiple functional dimensions such as emotional and informational support, but such multidimensional measures are not as commonly employed in research on health-related social support.

Another function that social network members engage in to promote positive health behaviors is health-related social control. In contrast to health-related social support, social control refers to attempts by network members to monitor and influence individuals who are not engaging in a healthy behavior, such as physical activity, either through positive tactics, such as gentle reminders and expressions of worry, or negative tactics, such as criticisms or restrictions of behavior (Lewis and Rook, 1999; Stephens et al., 2009). Social control may function to regulate an individual’s health by promoting healthy and preventing unhealthy lifestyle choices.

Although prior research has provided evidence that both social support and social control can be associated with healthier behavior (Anderson et al., 2006; Eyler et al., 1999), more work is needed on several fronts. Most research on how social support is related to physical activity has used global measures of social support or general measures of health-related social support (e.g. Franks et al., 2006; Kouvonen et al., 2011) rather than support that is provided specifically in connection with physical activity or exercise. When support is not measured with respect to a specific health behavior, the association between support and the behavior may be underestimated. The few studies that have focused specifically on physical activity–related social support have found that it is related to greater levels of physical activity (e.g. Anderson et al., 2006; Eyler et al., 1999), but few of these studies have used multidimensional conceptualizations of support (cf. Sallis et al., 1987). In contrast, most studies on health-related social control have usually assessed control strategies specific to a certain behavior, but most have not focused on physical activity (see Craddock et al., 2015 for a review). Much of the work conducted thus far has either focused on medical adherence for a specific chronic condition, such as diabetes or osteoarthritis, or on dietary behaviors as the outcome of interest (e.g. August and Sorkin, 2011; Helgeson et al., 2004; Stephens et al., 2010).

In addition to a need to better understand particular types of social interactions that may affect physical activity, it is important to better understand the psychological mechanisms by which social support and social control may have their effects on physical activity. Health behavior may be influenced by cognitive or emotional processes. In general, research investigating how social support may impact physical activity has tended to focus more on cognitive processes, such as behavioral intentions, and less on emotional processes (e.g. Ayotte et al., 2010). Support from network members may affect physical activity behavior because it alters attitudes and intentions (Ajzen et al., 2011; Rhodes et al., 2003). Less often considered is the role of the potential emotional responses to such support, yet emotional responses to support have the potential to impact physical activity directly or indirectly by affecting behavioral intentions. Positive emotional responses may increase the likelihood of engaging in physical activity (McAuley et al., 2007) and negative emotional responses may decrease the likelihood of engaging in physical activity (Dergance et al., 2005). Such emotional responses might affect behavior directly by broadening or narrowing one’s motivations to engage in new activities (Fredrickson, 2001) or more indirectly by affecting one’s behavioral intentions (Capella, 2007; Loewenstein et al., 2001). Loewenstein and colleagues argue that emotions serve as one input into decision making about behaviors, an indirect process that has received little attention in this realm. A third possibility is that cognitive and emotional processes are distinguishable, separate pathways to behavior, with some behavior determined by a more deliberative, intentional process and some behavior determined by a more immediate, affect-driven process (Smith and DeCoster, 2000).

In contrast to research on social support, research on social control has rarely focused on cognitive responses and more often focused on emotional responses. Direct attempts to change another individual’s behavior may naturally be met with sadness, guilt, or resentment (Lewis and Rook, 1999). Lewis and Rook initially observed dual effects of social control, in which social control was found to be associated with negative emotional responses but healthier behavior. Other authors, however, have observed more congruent effects between emotional responses and behaviors, in which social control was found to be associated with negative emotional responses and less healthy behavior (e.g. Helgeson et al., 2004). Some of the discrepant findings are potentially clouded by the fact that a number of studies have not investigated the effects of both positive and negative social control strategies on both positive affect and negative affect. One proposition has been that a domain-specific mediational process, in which positive social control is associated with positive emotional reactions and healthier behavior and negative social control is associated negative emotional reactions and less healthy behavior, can account for the perplexing array of findings (Okun et al., 2007; Tucker et al., 2006). In a recent meta-analysis, Craddock et al. (2015) suggest that there is general support for such a domain-specific process, where on average studies in this area have found that positive social control was related to both positive affect and healthier behavior and negative social control was related to both negative affect and less healthy behavior. Their results, however, also indicate that there is some tendency for positive social control to be related to less negative affect and, likewise, some tendency for negative social control to be related to less positive affect in most studies, leaving some unanswered questions about the underlying processes involved.

One piece that has been missing from studies on social control thus far is the role that cognitive processes may play in explaining the connection between social control and health behaviors. Social control interactions may impact behavior by changing an individual’s mind about activity or increasing his or her intentions to be active, perhaps regardless of the impact on emotional responses. It is reasonable to expect, however, that behavioral intentions will be differentially impacted depending on whether social control strategies are positive or negative and whether these strategies are interpreted positively or negatively. Positive social control attempts seem to have a more consistent impact on health behaviors than negative social control attempts (Craddock et al., 2015), and it may be that changes in behavior are because responses to positive social control attempts are favorable, resulting in changed attitudes and intentions. Negative social control attempts are less often found to be related to health behavior, and this may be because negative emotional responses do not lead to changes in behavioral intentions. Some studies, in fact, have found that negative social control is associated with less healthy behavior or concealing unhealthy behaviors, known as “backfiring” (Tucker et al., 2006). A detrimental effect on behavior would suggest that negative emotional responses lead to intentions to defy the network member, perhaps because of reactance (Brehm, 1966). Although a better understanding of the impact of emotional responses on behavioral intentions would help to elucidate the behavior change process, to date there have been no direct investigations of the mediational pathway by which physical activity–related social control may impact behavioral intentions to engage in physical activity through emotional responses and whether these changes in behavioral intentions, in turn, impact physical activity behaviors.

This study seeks to improve our understanding of how social support and social control may influence physical activity among older adults in several respects. There is a need for improvement and expansion upon existing measures to better understand the independent effects of specific aspects of social support and social control on physical activity. In this study, we develop and use confirmatory factor analysis to test a multidimensional measure of physical activity–related social support and social control. More research is also needed to understand the key cognitive and emotional pathways through which these specific types of social interaction may impact physical activity. This study thus extends prior studies by investigating how both emotional responses and intentions may mediate the effects of specific social influence processes on physical activity.

Method

Sample

Study participants were 217 older adults enrolled in a university senior adult learning program. All participants were 65 years and older (M = 72.55). Participants had the following characteristics: 58.5 percent women, 6.7 percent minority (.5% African-American, 1.0% Hispanic, 1.9% Asian or Pacific Islander, and 2.9% other or mixed race/ethnicity), 7.2 percent high school degree or some college, 92.8 percent college degree or higher, and 47.9 percent married.

Procedure

Participants were contacted by email and invited to participate in a survey about “how individuals interact with family and friends about health, exercise, and nutrition.” Each respondent was paid US$10 for his or her participation. Of the 600 who were invited to participate, 217 (36.2%) completed the survey. The survey included questions about sociodemographics, social interactions, health, and physical activity, and took approximately 20 minutes to complete.

Measures

Sociodemographic

Basic sociodemographic information was collected, including age, gender (male = 0, female = 1), race/ethnicity (White/non-Hispanic = 0, minority = 1), marital status (unmarried = 0, married = 1), and education (1 = did not complete high school to 6 = doctoral degree), to be included as covariates in the analyses.

Health

Two aspects of health were measured to be included as covariates in the analyses. Respondents indicated (0 = no, 1 = yes) whether they had any of 11 health conditions (heart disease, congestive heart failure, high blood pressure, claudication, lung disease, visual impairment, diabetes, knee problems, arthritis, past or present injury that affects activities, depression). A sum of these conditions was used in the analyses. To assess whether the participant had a serious health condition that prevented physical activity, a binary variable (0 = no, 1 = yes) was constructed to represent whether or not participants reported one of three health conditions that have been known to limit physical activity (congestive heart failure, knee problems, injury).

Physical activity–related social interactions

In an effort to more comprehensively assess physical activity–related social interactions, we expanded upon existing measures by developing a series of questions about the frequency of interactions related to physical activity with network members across several domains. Some items were adapted from measures used in prior work (e.g. Lewis and Rook, 1999; Stephens et al., 2009), modified to specifically address support or control related to physical activity. Based on a review of the literature, we then supplemented with new items in order to obtain four items for each of the four domains (16 items total): emotional support, informational support, positive social control, and negative social control. Each question began with “During the past month, how often did someone you know …?” using a 7-point response scale (1 = never, 2 = less than once a month, 3 = once a month, 4 = 2–3 times a month, 5 = once a week, 6 = 2–3 times a week, and 7 = daily). Physical activity–related emotional support was assessed with questions such as “show they appreciated your efforts to do something active?” Physical activity–related informational support was assessed with questions such as “give you useful information about the benefits of physical activity?” Positive strategies of social control were assessed with questions such as “encourage you to be more physically active?” Negative strategies of social control were assessed with questions such as “criticize you or make you feel bad about not being active enough?” The full list of items is given in Appendix 1.

Emotional responses

Following assessment of the frequency of each domain of physical activity–related social interactions, two questions were asked about how the individual felt in response to the interactions, with one question concerning positive emotional reactions and one question concerning negative emotional reactions. The topic of each social interaction question was reviewed and participants were asked “When someone … how pleased did you feel?” and “When someone … how upset did you feel?” Ratings were made on a 5-point rating scale (1 = not at all, 2 = just a little, 3 = moderately, 4 = very, 5 = extremely). The four questions about positive emotional reactions (one per domain) were averaged and the four questions about negative emotional reactions were averaged.

Cognitive responses

Behavioral intentions were measured with five questions, based generally on measures used in prior research (e.g. Prochaska and Velicer, 1997): “I am strongly committed to increasing my level of physical activity or maintaining an already active lifestyle,” “I do not plan to being physically active in the next 6 months,” “I intend to be physically active in the next 6 months,” “I intend to be physically active in the next month,” and “Over the past month, I have thought about particular physical activities that I might engage in or thought about ways to increase physical activities that I already engage in.”

Physical activity

Physical activity was measured with the Community Healthy Activities Model Program for Seniors (CHAMPS; Stewart et al., 1997). The CHAMPS assesses frequency of activities with 28 questions about exercise and recreational activities (e.g. dancing, aerobics, walking, swimming, light calisthenics). For each activity, the participant is asked whether he or she engaged in the activity in the past 4 weeks, how many times per week, and the duration (1 = less than 1 hour, 2 = 1–2.5 hours, 3 = 3–4.5 hours, 4 = 5–6.5 hours, 5 = 7–8.5 hours, and 6 = 9 or more hours). A subset of questions was used to compute a frequency of physical activity score weighted by the duration of the activity, scored according to that recommended by Stewart et al. (1997).

Overview of analyses

Analyses were conducted with Mplus 7.3 (Muthén and Muthén, 2012 [1998]) structural equation modeling software. Full maximum likelihood estimation for missing data with scaled chi-square and robust estimates for non-normal data was used (Yuan and Bentler, 2000). The maximum percentage of cases with univariate missing data was 6.5 percent and the maximum percentage of cases with bivariate missing data was 7.4 percent.

Because the physical activity–related social interactions measure was new, an initial step used confirmatory factor analysis to investigate the internal reliability and the hypothesized factor structure of the measure. A separate confirmatory factor model was used to assess the behavioral intentions measure. Final measures were then estimated as latent variables in three structural models in order to investigate the direct and indirect effects of social processes on intentions and on physical activity. The first model examined social control and social support factors together and included a set of covariates. The subsequent models investigated social control factors and social support factors separately, both controlling for the same set of covariates as in the first model. Indirect effects tests were based on maximum likelihood using confidence limits derived from 1000 bootstrap samples (Shrout and Bolger, 2002) without bias correction (Fritz et al., 2012).

Results

Confirmatory factor analyses

To examine the factor structure and internal reliability of the social interaction measures, we tested a four-factor confirmatory factor model using all four items for each of the four domains: physical activity–related emotional support, physical activity–related informational support, positive social control, and negative social control. This initial model did not reach commonly employed standards for acceptable fit (Hu and Bentler, 1999), χ² (98, N = 217) = 215.696, p = .0024, comparative fit index (CFI) =.932, standardized root mean square residual (SRMR) = .077. Standardized factor loadings from this model were all of acceptable magnitude, however, with values that ranged from .471 to .987. All standardized loadings are reported in the first column of Appendix 1 (Model 1). Inspection of modification indices suggested that the fit of the model could be improved substantially and that several items showed large modification indices if cross-factor loadings or correlated errors would be incorporated into the model.

A subsequent confirmatory factor model was tested eliminating several items (one item from each factor) that distinguished poorly among factors. The second model fit the data better although there was a large modification index for the correlated measurement residual between two items on the positive control factor, “Try to convince you to be more active out of concern for your health?” and “Encourage you to be more physically active?” The final model allowed these two measurement residuals to correlate, resulting in an excellent fit to the data with a substantially smaller chi-square value, χ² (47, N = 217) = 65.746, p = .0367, CFI = .986, SRMR = .040. Standardized loadings for this model (Model 2) are presented in the second column of Appendix 1. Inter-factor correlations for this final model can be found in Table 1. This final measurement model was used in the subsequent predictive analysis.

Table 1.

Correlations and descriptive statistics (N=217).

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
1. Age
2. Female	−.140*
3. Minority	−.060**	−.059
4. Married	−.247	−.141*	.102***
5. Education	.028	−.149*	.066	.237**
6. Health conditions	.153*	−.194**	.144	.075^a	−.004
7. Limitations	.136^a	−.289***	.061	.005	.017	.479***
8. Negative emotional responses	−.036	.068	−.030	−.070	−.074	−.031	−.100*
9. Positive emotional responses	−.129^a	.210**	−.087	.023	−.013	−.150*	−.205**	.035
10. Emotional support	−.107	.163*	−.077	−.023	.120^a	−.089	−.091	.136^a	.352***
11. Positive social control	.031	.022	−.117	.083	.162*	−.004	−.015	.306**	.253**	.601***
12. Negative social control	.038	−.024	−.009	.098***	−.050	.113	.055	.216**	−.215***	.047	.360***
13. Informational support	−.025	.101	.033	−.004	.092	−.094	−.121*	.206**	.341***	.582***	.592***	.154^a
14. Intention	−.151^a	.051	−.109	−.088	.146*	−.107	−.102	.048	.443***	.321***	.180**	−.172^a	.204*
15. Physical activity	−.140*	.017	−.061	.076	.144**	−.019	.061	−.070	.130^a	.307***	.006	−.008	.245**	.264***
Mean/%	72.63	58.6%	6.7%	47.9%	4.87	1.95	19.6%	.27	2.15	2.01	.37	.17	.92	4.35	59.41
SD	5.29	.49	.25	.33	.83	1.97	.40	.63	.91	1.28	.58	.59	.86	.77	35.90

SD: standard deviation.

Intentions, emotional support, positive social control, negative social control, and emotional support are latent variables.

p < .10, *p < .05, **p < .01, ***p < .001.

A separate confirmatory factor model was used to evaluate the fit and loadings for the behavioral intentions measure. This model, which specified that all five items loaded on a single factor, fit the data very well, χ² (5, N = 217) = 7.713, p = .1728, CFI = .979, SRMR = .021, and had acceptable standardized loadings, ranging from .509 to .904.

Descriptive analyses

Table 1 presents means, standard deviations, and intercorrelations for all variables used in the structural equation models. These results were generated from a structural equation model using the final latent variables for physical activity–related emotional support, physical activity–related informational support, positive control, negative control, and intentions.

The correlations in Table 1 show that physical activity had significant bivariate associations with being male, high levels of education, physical activity–related emotional support, physical activity–related informational support, and intentions, and a marginally significant association with positive emotional responses. Intentions had similar associations with each of these variables, but was also significantly associated with positive social control, and marginally associated (inversely) with negative social control.

Structural models

Mediational model for social control and social support

To investigate possible pathways by which social interactions are related to physical activity, we tested a mediational model with each of the social control and social support factors as predictors and positive and negative emotional responses and behavioral intentions as mediators. All covariates used in the predictive model (age, gender, race/ethnicity, marital status, education, health conditions, and physical limitations) were included in the model with paths to each of the outcomes (positive and negative emotional responses, intentions, and physical activity) but are not shown in the figure for simplicity. Results are summarized in Figure 1, which reports standardized coefficients and significance and omits paths from the covariates in the model. The model fit the data reasonably well, χ² (228) = 403.913, p < .001, CFI = .930, SRMR = .048, but the CFI did not quite reach the recommended cutoff of .95. The predictive model was fully saturated, however, and the measurement models fit well. We therefore concluded that any lack of fit was primarily a function of a complex model with a large number of covariates.

Figure 1.

Structural equation model of mediational pathways for the effects of social interactions on physical activity. Paths for covariates (age, gender, race, married, education, number of health conditions, and a binary variable for any of three limitations) predicting positive and negative emotional responses, behavioral intentions, and physical activity were included but are not shown. All path coefficients are standardized, ^ap < .10, *p < .05, **p < .01, ***p < .001.

There are several results to note from the figure. Negative social control strategies were significantly related to less positive emotional responses, β = −.490, SE = .085, β* = −.316, p < .001, and physical activity–related informational support was marginally associated with higher positive emotional responses, β = .215, SE = .114, β* = .202, p = .059. Although these were the only two significant independent effects on positive emotional responses, taken together, all of predictors accounted for approximately 28 percent of the variance in this variable, R² = .283, p < .001. None of the social interaction factors independently significantly predicted negative emotional responses although positive control strategies were marginally related to more negative emotional responses, β = .351, SE = .200, β* = .322, p = .078. Together, the predictors accounted for approximately 15 percent of the variance in negative emotional responses, however, R² = .145, p < .005. Substantial proportions of variance in behavioral intentions, R²= .293, p < .001, and in physical activity, R² = .273, p < .001, were accounted for by all of the predictors combined.

There were no significant independent direct effects on behavioral intentions, but physical activity–related emotional support was marginally positively associated with intentions, β = .106, SE = .059, β* = .176, p = .074. Behavioral intentions were significantly related to greater physical activity, β = .8.892, SE = 4.525, β* = .191, p < .05. Finally, direct effects of physical activity–related emotional support and physical activity–related informational support remained significant after accounting for intentions and all other variables in the model, β = 10.221, SE = 3.951, β* = .365, p < .05 and β = 11.565, SE = 5.474, β* = .276, p < .05, respectively.

In accordance with recommendations for investigating mediation (MacKinnon, 2008), we computed indirect effects coefficients and tested for significance using a bootstrapping approach in order to investigate whether positive emotional responses mediated the effects of social interactions on behavioral intentions. In one set of tests, the independent indirect effects were estimated for each of the social interaction factors predicting behavioral intentions indirectly through positive and negative emotional responses. The indirect pathway between negative social control and behavioral intentions (β = −.154, β* = −.117) as mediated through positive affect responses was significant, with 95 percent bootstrap confidence intervals (CIs) that excluded zero (−.297, −.062). This effect was consistent with the notion that negative social control adversely impacts behavioral intentions to engage in physical activity by reducing positive emotional responses. Informational support had a marginally significant indirect effect on behavioral intentions as mediated by positive emotional responses (β = .067, β* = .075), because the 95 percent confidence limits included zero (−.007, .192) but the 90 percent confidence limits did not (.001, .158).

Another set of indirect effects tests investigated whether there was an independent indirect effect of each social interaction factor on physical activity as mediated through behavioral intentions. These indirect paths consisted of direct effects of each social interaction factor on behavioral intentions, independent of any effects mediated through affective responses, and the direct effect of behavioral intentions on physical activity. Such indirect effects might reflect mediation by intervening variables other than positive affect that might impact behavioral intentions but were not measured in this study. None of these indirect effects was significant.

In light of the reasonably high proportion of variance accounted for in both positive and negative emotional responses by all of the predictors taken together, it is possible that clearer indirect effects would emerge if social control and social support factors were not included in the same model. We considered the possibility that the hypotheses regarding the mediational role of emotional responses might still be appropriate even though the independent effects of control and support could not easily be distinguished from one another. The correlations among the factors were substantial in some cases, ranging from to .047 to .601 (Table 1), but these values do not approach the magnitude that would cause multicollinearity problems (Cohen et al., 2003). To investigate further, we tested two additional mediational models—a model with only emotional and informational support and a model with only positive and negative social control. All other variables and specifications of these models followed the model depicted in Figure 1, including the same covariates and mediator variables.

Mediational model for social control

The mediational model for social control had a similar fit to the previous model, χ² (120) = 240.933, p < .001, CFI = .930, SRMR = .047, which was near the recommended cutoff for the CFI and met the recommended cutoff for the SRMR. Figure 2 presents the standardized path coefficients. Both positive and negative social control were significantly related to positive emotional responses, β = .522, SE = .126, β* = .372, p < .001 and β = −.539, SE = .108, β* = −.349, p < .001, respectively, and positive social control was marginally associated with more negative emotional responses, β = .296, SE = .154, β* = .304, p = .054. As in the prior model, only positive emotional responses were related to behavioral intentions, β = .330, SE = .090, β* = .389, p < .001. Neither positive nor negative social control factors were directly related to intentions or physical activity in this model. The bootstrap analyses indicated a significant indirect effect of positive social control, β = .172 (CI: .049, .364), β* = .145, and negative social control, β = −.178 (CI: −.329, −.069), β* = −.136, on behavioral intentions as mediated by positive emotional responses. The corresponding indirect effects for negative emotional responses were not significant.

Figure 2.

Structural equation model of mediational pathways for the effects of social control on physical activity. Paths for covariates (age, gender, race, married, education, number of health conditions, and a binary variable for any of three limitations) predicting positive and negative emotional responses, behavioral intentions, and physical activity were included but are not shown. All path coefficients are standardized, ^ap < .10, *p < .05, **p < .01, ***p < .001.

Mediational model for social support

The fit of the mediational model for social support (Figure 3) was similar to the previous models, χ² (121) = 212.551, p < .001, CFI = .922, and SRMR = .043. Results indicated that emotional and informational support both significantly predicted positive emotional responses, β = .142, SE = .067, β* = .198, p < .05 and β = .221, SE = .097, β* = .296, p < .05, respectively, but not negative emotional responses. As in the model with just social control, positive emotional responses, β = .342, SE = .085, β* = .408, p < .001, but not negative emotional responses, were related to behavioral intentions. Emotional support did have a significant direct effect on behavioral intentions, β = .101, SE = .045, β* = .167, p < .001, and neither variable significantly predicted physical activity directly when behavioral intentions were considered. The results from the model with emotional and informational support indicated a significant indirect effect of emotional support on behavioral intentions as mediated by positive emotional responses, β = .048 (CI: .002, .114), β* = .079, and a significant indirect effect of informational support on behavioral intentions as mediated by positive emotional responses, β = .076 (CI: .008, .179), β* = .084. There were no significant indirect effects for negative emotional responses. Although there were significant direct effects from emotional support to behavioral intentions and behavioral intentions to physical activity, the indirect effect from emotional support to physical activity as mediated by behavioral intentions was not significant. These results suggest that there is evidence consistent with indirect pathways between social support and behavioral intentions as mediated by emotional responses when the effects of positive and negative social control are not held constant.

Figure 3.

Structural equation model of mediational pathways for the effects of social support on physical activity. Paths for covariates (age, gender, race, married, education, number of health conditions, and a binary variable for any of three limitations) predicting positive and negative emotional responses, behavioral intentions, and physical activity were included but are not shown. All path coefficients are standardized, ^ap < .10, *p < .05, **p < .01, ***p < .001.

Discussion

The purpose of this study was to build upon prior work that has investigated health-related social support and social control in the context of health behaviors. In particular, we sought to investigate the independent effects of several dimensions of these social interactions that are specific to physical activity. Furthermore, this study moves beyond much of the prior work that has examined only direct effects on physical activity by investigating some of the possible indirect pathways involving both cognitive and emotional processes by which social interactions may impact physical activity. Although cognitive processes, such as behavioral intentions, have received considerable attention in the social support literature and emotional processes have received considerable attention in the social control literature, there has been less attention devoted to the complex interplay between cognitive and emotional mechanisms in attempting to understand the independent effects of physical activity-specific social support and social control on physical activity.

Our findings provide several important new insights into how physical activity–related social control and social support impact physical activity in later life. Social control and social support are related to intentions to engage in physical activity through emotional responses, and, in particular, through positive emotional responses. There was no evidence of an effect of negative emotional responses on behavioral intentions or physical activity behavior. The greater importance of positive emotions may be because positive emotions motivate healthier physical activity behavior, whereas negative emotions neither promote nor discourage physical activity. Such an asymmetric effect is consistent with research showing that positive emotions promote openness to new activities, whereas negative emotional responses do not have the same motivationally broadening effect (Fredrickson, 2001) as well as research showing that positive messages are more effective in promoting physical activity than negative messages (Notthoff and Carstensen, 2014). Negative emotions have been found to be related to hiding unhealthy behaviors in other studies (Okun et al., 2007; Tucker et al., 2006), however; but we did not measure hiding behaviors in this study.

A further interesting finding was that positive emotional responses did not have any direct effects on physical activity once behavioral intentions were accounted for in the mediational model, a result that is consistent with a full mediational process in which positive emotions have their impact on physical activity only through more deliberate cognitive processes. It may be that positive emotional responses impact intentions, because they reduce perceived risk or enhance anticipated positive emotional consequences (Capella, 2007; Loewenstein et al., 2001). It was important to examine behavioral intentions, particularly in connection to social control strategies, as prior research on social control has not addressed this type of cognitive process. Our results add to the literature, because they suggest that emotional responses to social control have their effects on behavior only through behavioral intentions and not directly through emotional processes. Some insight also is gained from our study about the emotional processes by which physical activity–related social support may influence physical activity, a pathway that has been infrequently studied. The findings in this study suggest that both emotional and informational support affect behavioral intentions and, in turn, physical activity through positive emotional responses rather by directly impacting intentions.

Careful consideration of the mediational modeling results also provide a potential explanation about why negative social control attempts are often found to be unrelated to physical activity when their total effects are examined (Craddock et al., 2015), a finding we report in this study as well. The fact that negative social control was predictive of lower positive emotional responses and positive emotional responses are positively related to physical activity suggests a possible explanation for why positive social control tends to be associated with health behaviors and negative social control is not. When only total effects between social control strategies and health behavior are examined, countervailing effects of social control on emotions may cancel out one another, a phenomenon MacKinnon et al. (2007) refer to as “opposing” or “inconsistent” mediated effects. The opposing mediational effect of negative social control on physical activity might suggest that strategies that involve pressure or criticisms may reduce positive affect for some individuals, whereas the presence of positive social control strategies or support is associated with higher positive affect for other individuals. Because positive affect is associated with greater physical activity for these other individuals, the total effect of negative social control on physical activity is null. The significant negative indirect effect of negative social control on behavioral intentions through positive affect is congruent with such a process.

The analyses also further our understanding of the independent effects of physical activity–related social control and social support. Taken together, the multiple control and support factors we measured accounted for a substantial portion of the variance of emotional responses, but their independent effects on emotions were limited. Only when the effects of social control and social support factors were examined separately did their direct effects on emotional responses and their indirect effects on behavioral intentions through emotional responses become clear. Confirmatory factor analyses suggest that the support and control factors are distinct although modestly to markedly correlated, but it is simply that their independent associations on emotional responses are relatively indistinguishable. The weak independent effects may indicate that either direct social control attempts to influence behavior, such as suggestions or reminders, or more supportive actions, such as encouragement or information, may be sufficient to encourage behavior.

One result from our analyses was particularly intriguing because it may suggest an additional reason for the association between social control and health behavior. We found a negative direct effect of positive social control on physical activity after controlling for social support and mediating variables in the model. This negative independent relationship is counter to the total positive relationship we observed and that is commonly observed in the literature (Craddock et al., 2015). In reviewing studies on social control, Craddock and colleagues reported a moderate positive association between positive social control and health behaviors and no association between positive social control and backfiring (i.e. a worsening of behavior). The negative relationship between positive control and physical activity after controlling for support and mediating variables may be indicative of a process other than backfiring, however. The negative association may occur because social network members make positive control attempts in an effort to increase exercise among those who are inactive or have an insufficient level of activity. Such a process, which would be consistent with the general notion of mobilization of support (Eckenrode, 1983), suggests a causal direction opposite to that often hypothesized and requires further investigation using longitudinal data. The fact that positive social control had a weak independent relationship to behavioral intentions in the full model provides a clue as to the direction of the effect, because behaviors are more likely to be mediated by intentions if social interactions are the cause.

In interpreting the findings of this study, limitations and implications for future research should be noted. The results of this study are based on cross-sectional data and therefore questions remain about causal directionality that can only be addressed with longitudinal research. Our study used an extensive self-reported leisure-time physical activity measure, which incorporated frequency and duration, but additional work is needed that uses objective assessments of physical activity (e.g. using accelerometers) to obtain more sensitive assessments of energy expenditure (Strath et al., 2003). Also, our measure of emotional responses to social support and social control differs from measures commonly employed in other studies on social control. The measure developed by Lewis and Rook (1999), which has been widely used to gauge emotional responses to social control, consists of questions about specific emotions in response to social control attempts (e.g. “anger,” “resentment,” “appreciative,” “optimistic”), whereas our measure more generally asked whether the respondent was “pleased” or “upset” in response to social support and social control interactions. Our more general approach was used primarily to accommodate the multiple-item and multidimensional assessment of social interactions involving both support and control. This more general approach is not far removed from the approach of Lewis and Rook (1999), but, to the extent that our findings deviate from that of other studies, it could be a function of the different measure of emotional responses. Our results overall, however, seem to be generally consistent with findings elsewhere in the literature, in that positive social control strategies were significantly related to positive emotional responses and negative social control strategies were significantly related to negative affect (Craddock et al., 2015), in addition to the substantial associations with other variables found in our study that suggest reasonable validity for this measure. Other limitations involve the representativeness of the sample. Participants in this study were highly educated, and there was a higher proportion of non-Hispanic Whites than in the population. Despite these limitations, the findings add to the literature by delineating specific pathways that are involved in promoting or detracting from physical activity in later life, a period in which physical inactivity is especially detrimental to health.

Footnotes

Appendix

Appendix 1.

Standardized factor loadings for four-factor confirmatory factor analyses of the social influence processes measures.

Factor	Model 1	Model 2
Physical activity–related emotional support
Act supportive of something you did that was physically active?	.759
Compliment you for doing something physical active?	.846	.819
Show they appreciated your efforts to do something active?	.913	.939
Sympathize with your efforts to be physically active?	.769	.767
Physical activity–related positive social control
Suggest doing something physically active together?	.471
Try to convince you to be more active out of concern for your health?^a	.758	.572
Encourage you to be more physically active?^a	.881	.734
Stress the importance of having an active lifestyle?	.682	.784
Physical activity–related negative social control
Criticize you or make you feel bad about not being active enough?	.967	.968
Try to make you feel guilty about how much physical activity you do?	.709
Question or express doubts about your lack of physical activity?	.987	.985
Express disappointment that you are not physically active enough?	.967	.968
Physical activity–related informational support
Give you useful information about the benefits of physical activity?	.744	.759
Mention an experience related to physical activity that was helpful to you?	.813	.800
Suggest a useful source of information, such as an article, a brochure, or website, about the health benefits of physical activity?	.768	.764
Recommend a person, place, or activity, such as a trainer, exercise group, physical therapist, or gym that would make it easier to be physically active?	.533

CFI: comparative fit index; SRMR: standardized root mean square residual.

Model fit information: Model 1—χ² (98, N = 217) = 215.696, p = .0024, CFI = .932, SRMR = .077; Model 2—χ² (47) = 65.746, p = .0367, CFI = .986, SRMR = .040.

Measurement residuals were allowed to correlate for these two items in Model 2.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was partially supported by a grant to the Center for Social and Demographic Analysis (CSDA) from the National Institute of Child Health and Human Development (R24-HD044943, Shaw), by a grant from the National Institutes of Health (1R21HD080828, Strath), and by internal funds provided by Portland State University (Newsom).

References

Ajzen

Joyce

Sheikh

. (2011) Knowledge and the prediction of behavior: The role of information accuracy in the theory of planned behavior. Basic and Applied Social Psychology 33(2): 101–117.

Anderson

Wojcik

Winett

. (2006) Social-cognitive determinants of physical activity: The influence of social support, self-efficacy, outcome expectations, and self-regulation among participants in a church-based health promotion study. Health Psychology 25(4): 510–520.

August

Sorkin

(2011) Support and influence in the context of diabetes management: Do racial/ethnic differences exist? Journal of Health Psychology 16: 711–721.

Ayotte

Margrett

Hicks-Patrick

(2010) Physical activity in middle-aged and young-old adults the roles of self-efficacy, barriers, outcome expectancies, self-regulatory behaviors and social support. Journal of Health Psychology 15(2): 173–185.

Brehm

(1966) A Theory of Psychological Reactance. New York: Academic Press.

Capella

(2007) The role of discrete emotions in the theory of reasoned action and its successors: Quitting smoking in young adults. In: Ajzen

Albarracín

Hornik

(eds) Prediction and Change of Health Behavior: Applying the Reasoned Action Approach. Mahwah, NJ: Lawrence Erlbaum Associates, pp. 43–52.

Cohen

West

. (2003) Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. New York: Lawrence Erlbaum Associates.

Cohen

(2004) Relationships and health. American Psychologist 59: 676–684.

Coyne

Wortman

Lehman

(1988) The other side of support: Emotional over involvement, and miscarried helping. In: Gottlieb

(ed.) Social Support: Formats, Processes, and Effects. Newbury Park, CA: SAGE, pp. 305–330.

10.

Craddock

vanDellen

Novak

. (2015) Influence in relationships: A meta-analysis on health-related social control. Basic and Applied Social Psychology 37: 118–130.

11.

Dergance

Mouton

Lichtenstein

. (2005) Potential mediators of ethnic differences in physical activity in older Mexican Americans and European Americans: Results from the San Antonio longitudinal study of aging. Journal of the American Geriatrics Society 53: 1240–1247.

12.

Eckenrode

(1983) The mobilization of social supports: Some individual constraints. American Journal of Community Psychology 11(5): 509–528.

13.

Eyler

Brownson

Donatelle

. (1999) Physical activity social support and middle-and older-aged minority women: Results from a US survey. Social Science & Medicine 49(6): 781–789.

14.

Franks

Stephens

MAP

Rook

. (2006) Spouses’ provision of health-related social control and support to patients participating in cardiac rehabilitation. Journal of Family Psychology 20: 311–318.

15.

Fredrickson

(2001) The role of positive emotions in positive psychology: The broaden-and-build theory of positive emotions. American Psychologist 56(3): 218–226.

16.

Fritz

Taylor

MacKinnon

(2012) Explanation of two anomalous results in statistical mediation analysis. Multivariate Behavioral Research 47(1): 61–87.

17.

Heaney

Israel

(2008) Social networks and social support. In: Glanz

Rimer

Viswanath

(eds) Health Behavior and Health Education: Theory, Research and Practice. San Francisco, CA: Jossey-Bass, pp. 189–210.

18.

Helgeson

Novak

Lepore

. (2004) Spouse social control efforts: Relations to health behavior and well-being among men with prostate cancer. Journal of social and Personal Relationships 21(1): 53–68.

19.

Bentler

(1999) Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A multidisciplinary Journal 6(1): 1–55.

20.

Kouvonen

De Vogli

Stafford

. (2011) Social support and the likelihood of maintaining and improving levels of physical activity: The Whitehall II study. The European Journal of Public Health 22: 514–518.

21.

Lewis

Rook

(1999) Social control in personal relationships: Impact on health behaviors and psychological distress. Health Psychology 18: 63–71.

22.

Loewenstein

Weber

Hsee

. (2001) Risk as feelings. Psychological Bulletin 127(2): 267–286.

23.

McAuley

Morris

Motl

. (2007) Long-term follow-up of physical activity behavior in older adults. Health Psychology 26: 375–380.

24.

MacKinnon

(2008) Introduction to Statistical Mediation Analysis. New York: Routledge.

25.

MacKinnon

Fairchild

Fritz

(2007) Mediation analysis. Annual Review of Psychology 58: 593–614.

26.

Mendonça

Cheng

Mélo

. (2014) Physical activity and social support in adolescents: A systematic review. Health Education Research 29: 822–839.

27.

Muthén

(2012 [1998]) Mplus User’s Guide: Statistical Analysis with Latent Variables (7th edn). Los Angeles, CA: Muthén & Muthén.

28.

Newsom

Rook

Nishishiba

. (2005) Understanding the relative importance of positive and negative social exchanges: Examining specific domains and appraisals. Journals of Gerontology: Psychological Sciences, 60B, P304–P312.

29.

Notthoff

Carstensen

(2014) Positive messaging promotes walking in older adults. Psychology and Aging 29(2): 329–341.

30.

Ogden

Carroll

Kit

. (2014) Prevalence of childhood and adult obesity in the United States, 2011–2012. Journal of the American Medical Association 311(8): 806–814.

31.

Okun

Huff

August

. (2007) Testing hypotheses distilled from four models of the effects of health-related social control. Basic and Applied Social Psychology 29(2): 185–193.

32.

Prochaska

Velicer

(1997) The transtheoretical model of health behavior change. American Journal of Health Promotion 12(1): 38–48.

33.

Rhodes

Courneya

Jones

(2003) Translating exercise intentions into behavior: Personality and social cognitive correlates. Journal of Health Psychology 8: 447–458.

34.

Rimal

(2001) Longitudinal influences of knowledge and self-efficacy on exercise behavior: Tests of a mutual reinforcement model. Journal of Health Psychology 6(1): 31–46.

35.

Sallis

Grossman

Pinski

. (1987) The development of scales to measure social support for diet and exercise behaviors. Preventive Medicine 16: 825–836.

36.

Shaw

Liang

Krause

. (2010) Age differences and social stratification in the long-term trajectories of leisure-time physical activity. Journal of Gerontology Series B: Psychological and Social Sciences 65(6): 756–766.

37.

Shrout

Bolger

(2002) Mediation in experimental and nonexperimental studies: New procedures and recommendations. Psychological Methods 7(4): 422–445.

38.

Smith

DeCoster

(2000) Dual-process models in social and cognitive psychology: Conceptual integration and links to underlying memory systems. Personality and Social Psychology Review 4(2): 108–131.

39.

Stephens

Fekete

Franks

. (2009) Spouses’ use of pressure and persuasion to promote osteoarthritis patients’ medical adherence after orthopedic surgery. Health Psychology 28(1): 48–55.

40.

Stephens

MAP

Rook

Franks

. (2010) Spouses use of social control to improve diabetic patients’ dietary adherence. Families, Systems & Health 28: 199–208.

41.

Stewart

Mills

Sepsis

. (1997) Evaluation of CHAMPS, a physical activity promotion program for older adults. Annals of Behavioral Medicine 19(4): 353–361.

42.

Strath

Bassett

Jr Ham

. (2003) Assessment of physical activity by telephone interview versus objective monitoring. Medicine and Science in Sports and Exercise 35(12): 2112–2118.

43.

Tucker

Orlando

Elliott

. (2006) Affective and behavioral responses to health-related social control. Health Psychology 25(6): 715.

44.

Yuan

K-H

Bentler

(2000) Three likelihood-based methods for mean and covariance structure analysis with nonnormal missing data. Sociological Methodology 30: 165–200.