The interplay of partner influence and individual values predicts daily fluctuations in eating and physical activity

Abstract

Both social and individual factors play a role in shaping one’s diet and exercise habits. A total of 62 heterosexual couples reported on health behavior values (HBVs) and completed daily diaries assessing food intake and physical activity relative to their own normal behavior and the helpfulness of health-related influence from their partners. Repeated measures dyadic analysis showed that men in couples with high average HBV ate less than usual in response to positive partner influence. Also, in such couples, normal weight men engaged in more physical activity when positively influenced by their partners. However, normal weight men in couples with low average HBV engaged in less physical activity when influenced by their partners. Women who valued health less than their partners responded to partner influence by eating healthier. These results highlight the importance of considering both social and individual contributors to health behaviors.

Keywords

BMI eating exercise gender health values obesity overweight partner influence

Most people know that a healthy diet and exercise are good for them, so why do few adhere to a healthy lifestyle? Research indicates that social factors such as romantic partners’ attempts to influence one another (Homish & Leonard, 2008; Lewis & Butterfield, 2007; Lewis & Rook, 1999; Markey, Gomel, & Markey, 2008; Markey, Markey, & Gray, 2007; Umberson, 1992) and individual factors such as health values and health beliefs (Conner, Kirk, Cade, & Barrett, 2001; Jackson, Tucker, & Herman, 2007; Laffrey & Isenberg, 2003; Merrill, Friedrichs, & Larsen, 2002; Peltzer, 2000; Shepherd & Stockley, 1987; Watters & Satia, 2009) shape health behaviors. However, neither of these factors considered alone provides a satisfactory explanation for daily variation in health actions. We propose that to accurately capture the combined effect of social and individual influence on health behaviors, we need to investigate them in the context of close relationships. This means that we should consider not only the interaction between one partner’s health behavior values (HBVs) and the other partner’s influence but also take into account whether the partners’ HBVs are similar or different and whether overall couples’ HBVs are high or low.

Partner influence and health behaviors

Research shows that romantic partners can have an impact on health behaviors including eating and exercise (Markey et al., 2007). Relationship partners have been shown to exchange both positive and negative health habits over time (Homish & Leonard, 2008) and the positive correlation often found between spouses’ body mass indexes (BMIs) may be partially due to their shared eating and exercise habits (Jeffery & Rick, 2002). Although studies conclusively demonstrate that long-term partners affect each other’s health behaviors (e.g., Homish & Leonard, 2008; Lewis & Butterfield, 2007; Lewis & Rook, 1999; Umberson, 1992), they also suggest that those effects can take various forms. For example, partners may influence each other both through indirect means such as habit exchange (e.g., Homish & Leonard, 2008; Jeffery & Rick, 2002) and direct social control by one of the partners (e.g., Lewis & Rook, 1999; Umberson, 1992). Furthermore, the valence of partner influence may play a role in health outcomes. Lewis and Rook (1999) reported that positive social control tactics (e.g., rewarding behavioral change) were related to an increase in health-enhancing behaviors, whereas negative social control tactics (e.g., inducing guilt) were unrelated to health behavior change, but were associated with more psychological distress. Similarly, in a study by Tucker and Mueller (2000), participants whose partners modeled the desired health behavior or provided emotional support reported these influences as effective. In contrast, the use of strategies such as expression of negative effect or nagging was associated with less positive and more negative effect and lower self-esteem in the participants.

Importantly, although there is some objective agreement as to which tactics are effective or not, there is also considerable between-person variability (e.g., Tucker & Mueller, 2000) suggesting that an individual’s perception of his or her partner’s influence attempts may play a role. Specifically, the research on the valence of influence suggests that only influence attempts perceived in a positive light will have a desirable impact on health behaviors. To address this issue in the present study, we assessed people’s reports of how helpful they found their partners’ health-related influence to be, rather than the details of what the influence attempt entailed.

Health values and health behaviors

One broadly studied individual-level factor impacting health behaviors is health value. The diversity of definitions of health values found in the literature, however, calls for a brief clarification. The results of studies that define health value as valuing health in general are mixed. While some demonstrate a connection between health values and health actions such as maintaining a health-promoting lifestyle (Jackson et al., 2007), taking dietary supplements (Conner et al., 2001), being a runner (Walsh, 1985), and not being obese (Freedman & Rubinstein, 2010), others show that HBVs are not associated with physical activity (Laffrey & Isenberg, 2003) or dietary fat intake (Peltzer, 2000).

Another body of literature addresses valuing specific health-related behaviors (HBV; e.g., healthy diet and exercise) in staying healthy. The results of these studies generally show that those who value a specific health behavior, have a positive attitude toward this behavior, or believe that this behavior is important for staying healthy, are more likely to engage in this behavior (e.g., Merrill et al., 2002; Peltzer, 2000; Shepherd & Stockley, 1987; Watters & Satia, 2009). However, even specific HBVs do not always translate into health actions. In one study, 73% of all participants strongly agreed that it was important to maintain their health, 76% strongly believed that what they ate affected their health, and 69% strongly agreed that what they weigh affects their health; however, 48% of the participants were overweight or obese, suggesting a possible disconnect between HBVs and health actions (Freedman & Rubinstein, 2010). Similarly, Conner et al. (2001) showed that being concerned about health consequences of one’s diet was unrelated to dietary supplement use.

Although research evidence supports the link between HBVs and behaviors, the lack of consistency in results suggests the importance of looking at the interplay of HBVs and other factors that have documented relevance to health behavior. Based on existing findings, we can extrapolate that the connection between health values, health beliefs, and health behaviors may become stronger as beliefs become more specific (i.e., importance of a specific health behavior vs. importance of health in general) and as values become more focused on the individual (i.e., the value of health for the respondent vs. for general populations). Indeed, Laffrey and Isenberg (2003) showed that although physical activity was not related to the relative value of health as compared to other values, it was associated with perceived importance of exercise in one’s life. Similarly, Peltzer (2000) found no connection between exercising and avoiding dietary fat and general health values, but showed that beliefs about avoiding dietary fat and exercising were associated with corresponding behaviors. Based on these considerations, we conclude that to capture health values most predictive of specific health behaviors, we would need to assess values that are related to this specific behavior (vs. overall value of health) in a way that directly pertains to the individual (vs. the population in general).

The interactive effects of partner influence and health values on health behavior

We further suggest that investigating the links between HBVs, partner influence, and health behaviors in a dyadic context may help clarify the circumstances under which partner influence and individual HBVs may impact health behaviors. For example, we propose that if and how partner influence is acted on (e.g., following or dismissing the influence) will depend on whether partners assign similar or divergent values to health behaviors. Based on the idea that communal coping with health issues increases couples’ ability to deal with health problems (Lewis et al., 2006; Lyons, Mickelson, Sullivan, & Coyne, 1998; Rohrbaugh, Mehl, Shoham, Reilly, & Ewy, 2008), we hypothesize that the partners with similar values may be more likely to support each other in engaging in behaviors that are consistent with those values. However, when partners’ HBVs are different, adhering to health-beneficial activities might be harder and the outcomes may vary depending on which partner values the behavior more. We suggest that partners who value specific health behaviors are likely to accept influence from partners who share those values, but are less likely to be influenced by partners who are not as committed to those behaviors. In other words, people who value a health behavior more than their partners do will not alter this behavior based on partner influence. However, those who do not value health behaviors as much as their partners might be more willing to accept influence from their more health-conscious partner.

Hypotheses

Based on the rationale described above, the present study tests the following hypotheses:

H1. In couples with high-average HBVs, both partners will be more responsive to each other’s influence than partners in low-HBV couples. High-average values are indicative of both the partners valuing health behaviors at least somewhat, so partners in such couples are more likely to have a dyadic focus on engaging in those behaviors and being mutually supportive in adhering to them, making partner influence more likely to have an impact.

H2. People who value health behaviors more than their partners do will be less influenced by their partners than people who value health behaviors less than their partners. We reason that those committed to healthy behaviors are not going to change their habits based on influence from a less health-conscious partner. However, people who rate health behaviors as less important than their partners are more likely to be influenced by their partner’s stronger values and respond to partner influence by altering their health behaviors.

Present study

To test our hypotheses, we collected data from a community sample of heterosexual couples in committed romantic relationships. We obtained a fairly diverse sample in terms of age and relationship length. We believe that such a sample is less susceptible to recruitment biases than would be a convenience sample made up purely of college students. Participants first provided demographic data and a baseline measure of health behavior values, height, weight, and subjective ratings of their own weight. We reasoned that participants’ weight status (as assessed by BMI) may be affecting their health behaviors and their responsivity to partner influence; if so, their perceptions of their weight status may be as important as their actual BMI. We therefore explored whether BMI or perceived weight status moderated the relationship between HBVs, partner influence, and health behaviors. In addition, based on studies suggesting that people eat less healthily (de Castro, 1991; Rhodes, Cleveland, Murayi, & Moshfegh, 2007) and engage in more sedentary behavior (Lake, Townshend, Alvanides, Stamp, & Adamson, 2009) on weekends vs. weekdays, we accounted for days of the week in our analyses. Finally, in line with the literature pointing to gender differences in dietary habits (Beardsworth et al., 2002; Bonds-Raacke, 2006; Morse & Driskel, 2009; Prattala et al., 2007; Savoca & Miller, 2001; Wardle et al., 1994) and exercise (Darlow & Xu, 2011; Shifren, Bauserman, & Carter, 1993; Treiber et al., 1991), we explore the possibility of gender differences for both hypotheses.

One important methodological issue is that accurately reporting absolute amounts or types of food consumed is notoriously difficult, regardless of whether it is retrospective or in the context of a daily diary (e.g., Lichtman, Pisarska, Berman, & Pestone, 1992; Maurer et al., 2006). To avoid this problem, we focused upon variability in consumption and physical activity relative to the participant’s usual behavior, instead of assessing details of the food intake or the activities. Specifically, participants provided seven daily reports of how much they had eaten and how healthy it was, as well as the amount of physical activity they engaged in, all as compared to their own typical eating and activity. For example, one item we asked them was “Was the amount of food you ate today more/about the same/less than usual?” This focus on within-person variability removes the problem of establishing an absolute standard, does not require participants to accurately compare themselves to external guidelines, and minimizes response bias by focusing the participants upon their own variability rather than implied social comparisons. This method allowed us to test our hypotheses in the context of participants’ daily lives, as well as minimize memory effects and biases associated with delayed recall. In addition, by collecting ratings of HBVs from both participants, we were able to investigate whether couples’ combined overall levels of HBVs and the differences between their HBVs interacted with partner influence to predict daily variation in diet or activity.

Methods

Recruitment and participants

A community sample of committed heterosexual couples was recruited by offering academic credit to undergraduate students at the University of Arizona in exchange for distributing flyers to eligible couples in the community. Both individuals in participating couples had to be: (1) over the age of 18 years, (2) willing to participate in the study, and (3) in a romantic relationship for at least 6 weeks. On completion, participants were entered into a cash lottery in which they could win up to $70.

Participants initially included were 67 heterosexual couples; however, 5 couples did not complete any diary data and were excluded. The resulting sample of 62 couples (total N = 124) ranged in age from 18.8 to 65.5 years (M = 28.3, SD = 12.5); of which, 51% of the couples were living together, 30.8% were married, and 25% had children. Relationship duration ranged from 2 months to 44 years (M = 7.6 years, SD = 10.9 years). Of the total participants, 57% described themselves as European American, 1.6% as Asian American, 1.6% as African American, 0.8% as Native Hawaiian or Pacific Islander, 4.8% as Native American, and 32% as other (predominantly of Hispanic and mixed backgrounds; 2.2% of the data were missing). Of the sample, 28% described themselves as Hispanic. Education levels included high school completion or less (12%), some college (58%), a professional or an undergraduate degree (20%), and a graduate degree (10%).

Procedure

Participants received a flyer with instructions to log on to a secure Web site where they read a full disclosure invitation and registered for the study. The process of registration included informed consent. Online instructions asked the participants to complete all the baseline and daily measures individually and to abstain from discussing their answers with their partners. The first time the participants logged in, they completed an initial survey, including demographic variables and a measure of HBVs. They were then instructed to return to the Web site each day for seven consecutive days in the evening to complete daily diary measures. The participants could start logging the data at their convenience, but were asked to do so every day for 7 days once they had started. Reminder e-mails were sent to individuals every day of the study prompting them to work on the questionnaires separately from their partners. We checked the time stamps of the entries and no two partners’ surveys were filled out simultaneously, or immediately consecutively, suggesting independent completion by the participants. We later used these time stamps to determine the day of the week for each diary entry. We used data from all days when partners reported on their daily activities, yielding an average of 5.0 daily entries (SD = 1.9, range = 1–10; participants in three couples logged their data for 8, 9, and 10 days). Of all the participants, 23% completed seven or more diary days and completion rate was not associated with any of the key study variables. The dummy coded variable for completeness was weakly associated with educational level (r = .27, p = .003) and relationship duration (r = .21, p = .02). Controlling these variables did not change the results, so results are reported without controlling them. Additionally, we examined the association between relationship duration and key study variables and found no significant correlations. In total, there were 616 observations included in the daily analyses.

Measures

Baseline measures

The participants’ self-reported weight ranged from 97 to 212 lbs for women (M = 134.6, SD = 24.3) and from 130 to 360 lbs for men (M = 192.7, SD = 43.5). Participants’ BMIs were calculated using a standard formula: weight(lbs)/height (in)²×703. Resulting BMIs ranged from 15.5 to 35.3 for women (M = 22.8, SD = 4.1) and from 19.3 to 47.4 for men (M = 26.9, SD = 5.2). Among women, 69.4% were normal weight, 9.7% were underweight, 16.1% were overweight, and 4.8% were obese. Among men, 35.5% were normal weight, 1.6% were underweight, 41.9% were overweight, and 21% were obese. Compared with the national averages, this sample has a lower prevalence of overweight and obesity, especially for women. Although literature suggests that self-reported BMIs tend to be slightly underestimated by both men and women, the correlations between subjectively and objectively measured BMIs are high (ranging between .95 and .98; e.g., Kovalchik, 2009; Mozumdar & Liguori, 2011). These correlations suggest that self-reported height and weight provide acceptable measures; however, results should be interpreted with these considerations in mind.

As suggested earlier, because participants’ perception of their own weight may have an impact on their health behaviors, we asked the participants to rate their own weight on a 7-point scale ranging from 1 = “very underweight” to 4 = “just right” to 7 = “very overweight.” For women, responses ranged from “very underweight” to “very overweight”; the average estimate was 4.6 (SD = 1.1), indicating the estimate between “just right” and “slightly overweight.” For men, responses ranged from “slightly underweight” to “very overweight.” The average estimate of men’s own weight was 4.8 (SD = 1.2), indicating that they rated their weight close to “slightly overweight” on average.

HBV was calculated as the mean score of two items in the baseline questionnaire: “How important is it to you to eat a healthy diet?” and “How important is it to you to engage in regular physical activity?” Both items were measured on a seven-point scale, which ranged from −3 = “not important at all,” to 3 = “very important.” Although these questions addressed health importance, as opposed to health value, we have found that in the existing literature, the constructs closest to the ones we are assessing are usually referred to as health values. To be consistent with this literature, we used primarily “health values” throughout the article, though we use health importance and health values interchangeably. The two items were strongly correlated (r = .59, p < .001). They were averaged to create the composite HBV score. HBV reported by men (M = 1.62, SE = .16) was not significantly different from that of women (M = 1.72, SE = .11).¹ Partners’ ratings of HBV were weakly positively correlated (r = .25, p = .006). To assess overall dyadic HBV and between-partner agreement on HBVs, we used partners’ averaged HBV scores and difference between HBV scores, respectively. The average HBV for couples was positive (M = 1.69, SD = .84, range = −.5 to 3) and partners were fairly similar in their HBV, again with a wide range (M difference = .10, SD = 1.31, range = −3 to 4.5).

Daily measures

The daily diaries included assessments of partner health-related influence, the relative healthiness and amount of food eaten during the day, and the relative amount of physical activity the participant is engaged in during the day. To avoid participant burden and increase compliance, all constructs were assessed with single face-valid items. For each item, participants responded with respect to the time period since they awoke that morning.

Partner health-related influence was assessed with a single item “Has your partner done or said anything that made it either easier or harder for you to behave in a healthy way, or in a way that makes you feel best?” The scale ranged from −2 = “He/she made it much harder” to 2 = “He/she made it much easier” and included 0 = “Neutral or not applicable” to capture responses of people whose partners made no such influence attempts. The average reported daily influence was .21 (SD = .69) for men and .10 (SD = .21) for women, indicating that most partner influence was perceived as neutral or helpful. Women’s influence was seen as more helpful by their male partners (b = .23, SE = .05) than vice versa (b = .11, SE = .04; F(1, 60) = 4.62, p = .04).

To assess eating behavior relative to each participant’s own normal behavior, participants reported each day whether they had eaten more, less, or about the same amount as they usually do using the following scale: −1 = “In the past day, I have eaten less than usual,” 0 = “about the same amount as usual,” and 1 = “more than usual.” The participants also reported whether the food they had eaten was −1 = “less healthy than what I normally eat,” 0 = “of about the same healthiness as what I normally eat,” or 1 = “more healthy than what I normally eat.” The relative amount of physical activity was assessed on a scale similar to the one above: −1 = “I did less physical activity than I usually do,” 0 = “I did about the same amount of physical activity as I usually do,” and 1 = “I did more physical activity than I usually do.”

As expected, the mode of all the three outcome variables was 0, showing that most of the time people reported eating the same amount and quality of food as usual, and getting a usual amount of physical activity. The participants reported eating about the same amount as usual 58% of the time, eating more than usual 16% of the time, and eating less 26% of the time. They reported eating as healthily as usual 67% of the time, less healthily than usual 22% of the time, and more healthily than usual 11% of the time. Finally, for exercise, the percentages were 56% for the usual amount, 29% for less than usual, and 15% for more than usual. The intraclass correlations (ICCs) for the item assessing amount of eating was .06, suggesting that 6% of the variance was due to between-person differences. Similarly, the ICC for eating quality was .12 indicating that 12% of variance was due to between-person differences and for exercise amount, ICC was .04. These small ICCs show that only a small amount of variance could be explained by between-person differences.

We included separate reports of relative amount of food eaten, relative healthiness of food, and relative amount of exercise to assess whether different health behaviors are susceptible to partner influence among men and women. The correlation between eating amount and eating quality was low and negative (r = −.15, p = .01 for men and r = −.21, p < .001 for women); the correlation between eating amount and exercise was small for men (r = .12, p = .04) and not significant for women (r = −.06, p = n.s.), and between eating healthy and exercise was small for women (r = .19, p = .001) and not significant for men (r = .08, p = n.s.). The relatively small correlations and the different correlational patterns for men and women suggested that eating quality, eating amount, and exercise amount were relatively independent from each other and so we treated them separately in subsequent analyses.

Statistical analyses

Data collected from partners over a period of days is subject to multiple sources of interdependence, including autocorrelation within persons as well as between-partner correlations of average scores, slopes, and daily fluctuations (Kenny, Kashy, & Cook, 2006; Laurenceau & Bolger, 2005). This is likely to result in nonindependent residuals; therefore, standard regression techniques are inappropriate (Kenny et al., 2006). We thus employed a longitudinal dyadic model which is detailed in Laurenceau and Bolger (2005). This model includes separate fixed and random intercepts for men and women. It also estimates between-partner correlations of intercepts, slopes, and residuals, while simultaneously modeling autocorrelation over days within persons. We checked for fixed and random effects of time for all three outcome variables (eating amount, quality, and activity). None of them were significant, suggesting that monitoring these behaviors did not alter them over the course of the study. Based on this, we did not include time in our models. Having accounted for the various sources of interdependence, hypothesis testing can be conducted by specifying appropriate fixed effects models in the same way as is done in standard multiple regressions. We used SAS Proc Mixed to estimate all models. The estimation procedure used by Proc Mixed handles missing data by including all observations that contain all variables in the model (Singer & Willett, 2003). The estimation procedure then adjusts degrees of freedom and standard errors to take into account the actual number of observations included in each analysis. We used person-centered daily predictor (partner influence) and sample-centered dyad level predictors (average HBV and the difference of HBV within the couple; Enders & Tofighi, 2007). Model assumptions, including normality of dependent variables and residuals, were assessed using standard procedures prior to conducting hypothesis tests.

We used an average–difference model for our analyses. This model is a derivative of the actor–partner interaction model (APIM) and we chose it for several reasons. First, it perfectly fits our hypothesis that both shared values and the difference between partners’ values matter. Kenny et al. (2006) suggest that interactions should be modeled in terms that make the most theoretical sense, as opposed to being automatically modeled as the product. Since we are examining the dyadic-level effects of shared versus divergent partner HBV, it is important to look at the difference scores, which are captured with the average–difference model but not the APIM. Second, preliminary analysis showed that actor effects for both participants in the couples were not statistically different and therefore can be well-represented by the average score. Furthermore, by including both the average and the difference scores into our analysis, we address concerns about the validity of difference scores used in isolation (Kenny et al., 2006). Finally, average–difference models contain fewer terms than the APIM, and since our hypotheses are mainly about interactions, the APIM becomes unwieldy.

We tested our hypotheses with three parallel models: one with eating amount as the outcome, the second with eating quality, and the third with physical activity as the outcome. All three models included the main effects of partner influence, couples’ average scores for HBV, couples’ difference scores for HBV, and the interactions of partner influence with the average and difference scores. The participants’ BMIs, subjective ratings of their own weight, and a dichotomous variable distinguishing weekdays from weekends were included as controls in all models. In addition, we explored their role as moderators and found one significant interaction for men (see results). All predictors were treated as fixed effects to limit the complexity of the model relative to the sample size. The models provide separate estimates for men and women and so gender is not included as a separate predictor (see Laurenceau & Bolger, 2005, for more details). Interactions were decomposed following Aiken and West (1991), with high and low values of the continuous predictors centered at 25th and 75th percentiles. For average couple HBV, these values correspond to health behaviors being “slightly important” for the 25th percentile and “moderately important” for the 75th percentile. For the difference in HBVs, these values correspond to health behaviors being somewhat more important for the female partner and health behaviors being somewhat more important for the male partner. For partner influence, low and high values corresponded to neutral or no influence and positive influence.

Results

The interaction of partner influence and couple average health values

Our first hypothesis was that couples with high-average HBV would show more effects of partner influence than low-HBV couples. Our data show support for this hypothesis for eating and exercising amounts among men. For eating amount, the interaction of couple average HBV and partner influence was significant for men, F(1, 478) = 9.10, p = .003, but not for women. As illustrated in Figure 1, men in couples with a high-average HBV ate significantly less than their normal intake on days when they reported more helpful partner influence than when they reported less helpful partner influence, b = −.15, p = .003. In contrast, men in low-HBV couples ate about the same regardless of influence, b = .12, p = n.s.

Figure 1.

Men’s eating amount predicted from partner influence and couples’ average health behavior values.

For exercise amount, the interaction between couple average HBV, partner influence, and BMI was significant for men F(1, 467) = 19.98, p < .0001, but not for women. Figure 2 shows that on days when normal weight men in high-HBV couples received positive partner influence, they engaged in significantly more physical activity relative to their usual activity than on days when partner influence was neutral (b = .51, p < .0001). However, normal weight men in low-HBV couples reported lower physical activity on days when they received positive partner influence as compared to days when the influence was neutral (b = −.33, p = .03). For overweight and obese men regardless of HBV, partner influence had no effect. No significant relationship between partner influence, HBV, and the amount of physical activity was found for women.

Figure 2.

Men’s physical activity predicted from partner influence, couples’ average health behavior values, and body mass index.

The interaction of partner influence and partner differences in health values

Our second hypothesis was that individuals who valued health behaviors more than their partners would not be as affected by partner influence as would those who valued health behaviors less than their partners. This hypothesis received some support. For eating quality, the interaction between partner differences in HBVs and partner influence was significant for women F(1, 470) = 5.84, p = .02. As shown in Figure 3, women who reported a higher HBV than their partners showed no change in their dietary quality in response to their partners’ influence (b = .05, p = n.s.), but women who reported a lower HBV than their partner reported eating healthier foods on days when they received positive partner influence (b = .33, p = .02).

Figure 3.

Women’s eating quality predicted from partner influence and partner’s differences in health behavior values.

For eating amount, the interaction between partner differences in HBVs and partner influence was significant for men F(1, 478) = 4.55, p = .03. Figure 4 demonstrates that the slope for men with higher HBV than their partners is in opposite direction from the slope for men with lower HBV than their partners (b = −.14, p = .04). Men with lower HBV than partners tended to eat less on days when they experienced positive partner influence and men with higher HBV than partners tended to eat more on such days, although at the centering values used for the analysis, neither slope was significantly different from zero.

Figure 4.

Men’s eating amount predicted from partner influence and partner’s differences in health behavior values.

Discussion

Our data suggest that partners influence eating and physical activity under certain conditions. For couples with high-average HBV, on days when men received helpful influence from their partners, they also reported eating less than normal. Men in high-HBV couples also reported engaging in more exercise in response to positive partner influence, an effect seen in normal weight, but not in overweight or in obese men. Normal weight men in low-HBV couples reported exercising less on days when they received positive partner influence. Women who did not value health behaviors as much as their partners responded to helpful influence by eating healthier that day than normal, but women who were more committed to health behavior than their partners were less susceptible to partner influence. A similar trend was observed among men in regard to eating amounts.

Overall, we found that women responded to partner influence by changing the quality of their dietary intake and men responded by changing the amounts of food intake and exercise. These gender differences are consistent with previous findings showing that men are less aware of the content of foods (e.g., Beardsworth et al., 2002) and are less likely to try new foods (e.g., Bonds-Raacke, 2006; Savoca & Miller, 2001), whereas women have higher awareness of the healthiness of foods than men (e.g., Beardsworth et al., 2002) and are more likely to change their diets in general (e.g., Savoca & Miller, 2001; Timperio, Cameron-Smith, Burns, & Crawford, 2000). These results highlight that women may be more informed of food content and more willing to change it in order to maintain good health. Consequently, findings based on measures that involve knowledge of food content should be interpreted with caution because the lack of reported changes in eating quality among men may be reflecting differences in knowledge rather than differences in food consumption.

Similarly, studies on the role of gender in exercise suggest that men and women have different attitude toward and engagement in physical activity (e.g., Shifren et al.,1993), with men engaging in more physical activity than women (e.g., Darlow & Xu, 2011; Treiber et al., 1991). Considering these gender differences in health behaviors, it is not surprising that men and women would respond differently to partner influence. Specifically, as we observed, women respond to partner’s supportive influence using their knowledge about the content of foods and choosing the ones that are healthier for them (at least according to the information they have). On the contrary, men, who may be less knowledgeable about the healthiness of certain foods, may respond to partner encouragement by choosing to eat less or exercise more. The results indicating that men from low-HBV couples respond to helpful partner influence by exercising less may appear counterintuitive. However, such men may have used exercise as a way to counterbalance unhealthy eating. With partner influence assessed globally (i.e., it was not specific to either exercise or eating), these men may have simply bypassed the exercise on days when their partners were successful in helping them stay on track with their diet. On the contrary, men in couples with high HBVs may see merit in exercising per se and engage in exercise regardless of their dietary intake.

Our finding that couples’ average HBVs moderated the impact of partner influence on dietary and activity fluctuations in men but not in women is intriguing. On the one hand, this contradicts the general finding that relationship factors such as marital quality and relationship satisfaction are more significant for the health of women than of men (Coyne et al., 2001; Kiecolt-Glaser & Newton, 2001; Rohrbaugh, Shoham, & Coyne, 2006). We expected that if there were a gender difference, it would be in the opposite direction, with women’s health behaviors being more affected by partner influence than men’s. On the other hand, an important factor to consider may be not who gets affected by partner influence, but who attempts partner influence and who is more effective at eliciting change in partner’s behavior. At least two studies showed that women were significantly more likely to attempt influencing their partners’ health behavior and used a wider repertoire of social control strategies that potentially allowed them to be more effective in facilitating health-related change (Tucker & Mueller, 2000; Umberson, 1992). Consistent with these studies, women in our sample had more helpful influence than did men. Considering that in high-average HBV couples, both partners value their health and are about equally likely to respond to partner influence, if women make more helpful influence attempts than men, it follows that men would make more changes than women. Similar results were reported by Markey et al. (2008) who showed that women’s, but not men’s, attempts to regulate their partners’ eating behaviors were associated with their partners’ healthy dieting behaviors. Furthermore, although people of both genders acknowledge partner influence on their diets, men believe that their diets are influenced more so than women do (Bock et al., 1998), suggesting that men might be more amenable to direct partner influence. It is also possible that because of women’s greater concern about diet and appearance, men may be reluctant to bring up these topics with their female partners, whereas women may be in a better position to do so.²

The results of the present study also suggest that people with higher HBV than their partners may remain unresponsive to partner influence, while those with lower HBV may be more susceptible to influence. Those who valued health less than their partners were more likely to eat healthier than normal in response to helpful partner influence. These results may be reflective of the notion that those with strong HBV and deep involvement in a matter are unlikely to change their behavior based on influence attempted by those who do not share those values. This is in accord with the idea that people tend to favor the opinions of those who share their views and dismiss opinions that do not support their values or identity. In the present study, women who highly valued health behaviors were therefore unlikely to respond to the influence of their partners who did not share their commitment to health. However, women with low HBVs may have been less likely to have strong opinions about health and may have lacked any strong health-related identity. Such women may be more likely to be susceptible to partner influence, especially coming from a partner who is determined and knowledgeable.

Limitations and future directions

A central limitation of this study is that single-item measures were used to represent each of the daily constructs. Although this makes it impossible to assess reliability, we believe that the advantages of minimizing participant burden by not using multiple-item scales outweighed the downside. Another potential limitation of this study is that we measured perceived valence of partner influence without objectively assessing the exact content or form of the influence. This gave us a good measure of the impact the influence had on the respondent, but did little to increase our insight into which exact influence behaviors are more likely to be perceived as supportive versus unhelpful. Nevertheless, the measure we used allowed us to show that helpful partner influence—regardless of what exactly that means for each given couple—is an important component of maintaining a healthy lifestyle. In addition, this appears to be especially true for those who do not have strong HBVs themselves, but who have more health-conscious partners. One way to obtain more in-depth information on partner influence without increasing participant burden would be to collect data on provision of support in addition to receipt of partner support.

Another potential shortcoming of this study is that we used the relative quality and amount of food consumed by the participants and the relative amount of exercise they engaged in and did not use objective measures reflecting the actual content of participants’ diets or physical activity. Although such measures may have allowed us to assess the accuracy of participants’ perceptions, they would have also raised the problem that daily monitoring of actual dietary intake has a documented effect on peoples’ eating behavior (e.g., Hollis et al., 2008). In other words, recording details of food eaten during the day usually reduces the overall intake and makes it likely that people would eat healthier foods. Therefore, a detailed food diary for a week may poorly reflect what participants normally eat, but having the participants report briefly on their relative dietary quantity and quality allowed us to minimize this effect. The fact that we did not observe any linear changes in health behaviors over the duration of the study suggests we were successful in minimizing any monitoring effects.

Another consideration is that the relatively small sample size used in this study may have limited the statistical power. As completion rates in our study were correlated with educational levels and relationship duration, our results may be more representative of longer relationships between those with higher education, thus affecting the generalizability of the results. Additional results may have emerged with more participants or more complete data, so using a larger sample is advisable for future studies. Similarly, although our analyses revealed no connection between relationship duration and any of the key study variables, we acknowledge that relationship dynamics might be different in long-term married couples versus those just beginning to date. The relatively small sample size in our study precluded us from obtaining significant results from married versus dating subsamples; a larger sample would allow for a meaningful comparison.

Furthermore, the present sample had a lower percentage of overweight and obesity than the general U.S. population. If partner influence indeed plays a role in shaping one’s diet and exercise, a sample with below-average BMI may also have above-average success rate of partner influence or above-average rates of beneficial responses to partner influence. Future studies would therefore benefit from using a more representative sample in regard to overweight and obesity. Research would also benefit from examining measures of relationship quality as potential moderators of the link between partner influence and health behaviors. Although in the present study relationship duration did not have any effect on the outcomes, other measures of relationship quality should be included in future studies. Finally, we acknowledge the limitations of the BMI measure as the sole determinant of one’s weight-related health (e.g., Nevill, 2006). However, we chose to use BMI because it is an easily obtainable measure with widely demonstrated connection with various aspects of health (e.g., Stommel & Schoenborn, 2010).

Conclusions

Our findings suggest that partner influence plays a role in the maintenance of a healthy diet and physical activity and highlight the importance of dyadic analysis in capturing the dynamics of health maintenance. Partner influence concerning one’s health behaviors may be perceived as helpful or unhelpful and peoples’ behaviors are affected accordingly. Men and women differ both in the helpfulness of their influence and in the changes they themselves implement when influenced. Consistent with previous research, women had more helpful influence on their male partners than vice versa and responded to helpful partner influence by eating healthier foods. Men’s influence was perceived as less helpful and they responded to helpful partner influence by eating less and changing their amount of exercise. In couples that highly valued health behaviors, men responded to partner influence more than did women, perhaps due to women making more helpful influence attempts overall. Those who valued health less than their partners were more likely to be influenced than their more health-conscious partners.

Considering this interplay between both partners’ HBVs and partner influences might sharpen couple interventions by allowing for a more idiosyncratic approach to assessing and treating couples. The results of studies on couple interventions for weight loss are mixed. Some studies demonstrate greater weight loss when both partners are treated together or recruited to support one of the partners’ weight loss (Black, Gleser, & Kooyers, 1990; McLean, Griffin, Toney, & Hardeman, 2003; Murphy, 1982; Pearce, LeBow, & Orchard, 1981), although others have found no differences between couple and individual interventions (Black & Lantz, 1984; Brownell & Stunkard, 1981; Dubbert & Wilson, 1984; Rosenthal, Allen, & Winter, 1980). In one study, men were found to lose more weight when treated alone (Wing, Marcus, Epstein, & Jawad, 1991). Our finding that partner influence interacts with both partners’ HBVs and has associated fluctuations in diet and activity suggests that partner participation in interventions targeting health behavior changes might be a crucial component in achieving lasting outcomes. To realize this potential, however, it is important to identify the combination of the partners’ health beliefs and how that interacts with partner influence to determine health behavior. Doing so might enable us to better recognize effective strategies for a given couple based on match or mismatch in their HBVs. We might consequently be better able to assist these couples in identifying the influence strategies that are effective for them.

Footnotes

Funding

This research was supported in part by the Frances McClelland Institute for Children, Youth, and Families in the Norton School of Family and Consumer Sciences at The University of Arizona. Information about the Frances McClelland Institute is available at: http://McClellandInstitute.arizona.edu.

Notes

References

Aiken

L. S.

West

S. G.

(1991). Multiple regression: Testing and interpreting interactions. Thousand Oaks, CA: Sage Publications.

Beardsworth

Bryman

Keil

Goode

Haslam

Lancashire

(2002). Women, men and food: The significance of gender for nutritional attitudes and choices. British Food Journal, 104, 470–491.

Black

D. R.

Gleiser

L. J.

Kooyers

K. J.

(1990). A meta-analytic evaluation of couples weight-loss programs. Health Psychology, 9, 330–347.

Black

D. R.

Lantz

C. E.

(1984). Spouse involvement and a possible long-term follow-up trap in weight loss. Behaviour Research and Therapy, 22, 557–562.

Bock

Read

Bruhn

Auld

Gabel

Lauritzen

G.,

… Sheehan,

(1998). Gender and ethnic differences in factors that influence food intake. Journal of Consumer Studies & Home Economics, 22, 25–37.

Bonds-Raacke

J. M.

(2006). Using cluster analysis to examine husband–wife decision-making. The Psychological Record, 56, 521–550.

Brownell

K. D.

Stunkard

A. J.

(1981). Couples training, pharmacotherapy, and behavior therapy in the treatment of obesity. Archives of General Psychiatry, 38, 1224–1229.

Conner

Kirk

S. L.

Cade

J. E.

Barrett

J. H.

(2001). Why do women use dietary supplements? The use of the theory of planned behaviour to explore beliefs about their use. Social Science & Medicine, 52, 621–633.

Coyne

J. C.

Rohrbaugh

M. J.

Shoham

Sonnega

J. S.

Nicklas

J. M.

Cranford

J. A.

(2001). Prognostic importance of marital quality for survival of congestive heart failure. American Journal of Cardiology, 88, 526–529.

10.

Darlow

S. D.

(2011). The influence of close others’ exercise habits and perceived social support on exercise. Psychology of Sport & Exercise, 12, 575–578.

11.

De Castro

J. M.

(1991). Social facilitation of the spontaneous meal size of humans occurs on both weekdays and weekends. Physiology & Behavior, 49, 1289–1291.

12.

Dubbert

P. M.

Wilson

(1984). Goal-setting and spouse involvement in the treatment of obesity. Behaviour Research and Therapy, 22, 227–242.

13.

Enders

C. K.

Tofighi

(2007). Centering predictor variables in cross-sectional multilevel models: A new look at an old issue. Psychological Methods, 12, 121–138.

14.

Freedman

M. R.

Rubinstein

R. J.

(2010). Obesity and food choices among faculty and staff at a large urban university. Journal of American College Health, 59, 205–210.

15.

Hollis

J. F.

Gullion

C. M.

Stevens

V. J.

Brantley

P. J.

Appel

L. J.

Ard

J. D.,

… Svetkey

L.P.

(2008). Weight loss during the intensive intervention phase of the weight-loss maintenance trial. American Journal of Preventive Medicine, 35, 118–126.

16.

Homish

Leonard

(2008). Spousal influence on general health behaviors in a community sample. American Journal of Health Behavior, 32, 754–763.

17.

Jackson

E. S.

Tucker

C. M.

Herman

K. C.

(2007). Health value, perceived social support, and health self-efficacy as factors in a health-promoting lifestyle. Journal of American College Health, 56, 69–74.

18.

Jeffery

Rick

(2002). Cross-sectional and longitudinal associations between body mass index and marriage-related factors. Obesity, 10, 809–815.

19.

Kenny

D. A.

Kashy

D. A.

Cook

W. L.

(2006). Dyadic data analysis. New York, NY: The Guilford Press.

20.

Kiecolt-Glaser

Newton

(2001). Marriage and health: His and hers. Psychological Bulletin, 127, 472–503.

21.

Kovalchik

(2009). Validity of adult lifetime self-reported body weight. Public Health Nutrition, 12, 1072–1077.

22.

Laffrey

Isenberg

(2003). The relationship of internal locus of control, value placed on health, perceived importance of exercise, and participation in physical activity during leisure. International Journal of Nursing Studies, 40, 453–459.

23.

Lake

Townshend

Alvanides

Stamp

Adamson

(2009). Diet, physical activity, sedentary behaviour and perceptions of the environment in young adults. Journal of Human Nutrition and Dietetics: The Official Journal of the British Dietetic Association, 22, 444–454.

24.

Laurenceau

J. P.

Bolger

(2005). Using diary methods to study marital and family processes. Journal of Family Psychology, 19, 86–97.

25.

Lewis

Butterfield

(2007). Social control in marital relationships: Effect of one’s partner on health behaviors. Journal of Applied Social Psychology, 37, 298–319.

26.

Lewis

Rook

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

27.

Lewis

McBride

Pollak

Puleo

Butterfield

Emmons

(2006). Understanding health behavior change among couples: an interdependence and communal coping approach. Social Science & Medicine, 62, 1369–1380.

28.

Lichtman

S. W.

Pisarska

Berman

E. R.

Pestone

(1992). Discrepancy between self-reported and actual caloric intake and exercise in obese subjects. The New England Journal of Medicine, 327, 1893–1898.

29.

Lyons

R. F.

Mickelson

K. D.

Sullivan

M. J.

Coyne

J. C.

(1998). Coping as a communal process. Journal of Social and Personal Relationships, 15, 579–605.

30.

Markey

C. N.

Gomel

J. N.

Markey

P. M.

(2008). Romantic relationships and eating regulation: An investigation of partners’ attempts to control each others’ eating behaviors. Journal of Health Psychology, 13, 422–432.

31.

Markey

C. N.

Markey

P. M.

Gray

(2007). Romantic relationships and health: An examination of individuals’ perceptions of their romantic partners’ influences on their health. Sex Roles, 57, 435–445.

32.

Maurer

Taren

Teixeira

Thomson

Lohman

Going

Houtkooper

(2006). The psychosocial and behavioral characteristics related to energy misreporting. Nutrition Reviews, 64, 53–66.

33.

McLean

N. N.

Griffin

S. S.

Toney

K. K.

Hardeman

W. W.

(2003). Family involvement in weight control, weight maintenance and weight-loss interventions: a systematic review of randomized trials. International Journal of Obesity & Related Metabolic Disorders, 27, 987–1005.

34.

Merrill

R. M.

Friedrichs

Larsen

(2002). Perceptions of healthy behaviors versus health practices. Health Promotion Practice, 3, 497–500.

35.

Morse

Driskell

(2009). Observed sex differences in fast-food consumption and nutrition self-assessments and beliefs of college students. Nutrition Research, 29, 173–179.

36.

Mozumdar

Liguori

(2011). Correction equations to adjust self-reported height and weight for obesity estimates among college students. Research Quarterly For Exercise & Sport, 82, 391–399.

37.

Murphy

J. K.

(1982). The long-term effects of spouse involvement upon weight loss and maintenance. Behavior Therapy, 13, 681–693.

38.

Nevill

(2006). Relationship between adiposity and body size reveals limitations of BMI. American Journal of Physical Anthropology, 129, 151–156.

39.

Pearce

J. W.

LeBow

M. D.

Orchard

(1981). Role of spouse involvement in the behavioral treatment of overweight women. Journal of Consulting and Clinical Psychology, 49, 236–244.

40.

Peltzer

(2000). Health behaviour in Black South African university students. South African Journal of Psychology, 30, 46–49.

41.

Prattala

Paalanen

Grinberga

Helasoja

Kasmel

Petkeviciene

(2007). Gender differences in the consumption of meat, fruit and vegetables are similar in Finland and the Baltic countries. The European Journal of Public Health, 17, 520.

42.

Rhodes

D. G.

Cleveland

L. E.

Murayi

Moshfegh

A. J.

(2007). The effect of weekend eating on nutrient intakes and dietary patterns. FASEB Journal, 21, A1064.

43.

Rohrbaugh

M. J.

Mehl

M. R.

Shoham

Reilly

E. S.

Ewy

G. A.

(2008). Prognostic significance of spouse we talk in couples coping with heart failure. Journal of Consulting and Clinical Psychology, 76, 781–789.

44.

Rohrbaugh

M. J.

Shoham

Coyne

J.C.

(2006). Effect of marital quality on eight-year survival of patients with heart failure. The American Journal of Cardiology, 98, 1069–1072.

45.

Rosenthal

Allen

Winter

(1980). Husband involvement in the behavioral treatment of overweight women: Initial effects and long-term follow-up. International Journal of Obesity, 4, 165–173.

46.

Savoca

Miller

(2001). Food selection and eating patterns: Themes found among people with type 2 diabetes mellitus. Journal of Nutrition Education and Behavior, 33, 224–233.

47.

Shepherd

Stockley

(1987). Nutrition knowledge, attitudes, and fat consumption. Journal of the American Dietetic Association, 87, 615–619.

48.

Shifren

Bauserman

Carter

(1993). Gender role orientation and physical health: A study among young adults. Sex Roles, 29, 421–432.

49.

Singer

J. D.

Willett

J. B.

(2003). Applied longitudinal data analysis: Modeling change and event occurrence. New York, NY: Oxford University Press.

50.

Stommel

Schoenborn

C. A.

(2010). Variations in BMI and prevalence of health risks in diverse racial and ethnic populations. Obesity, 18, 1821–1826.

51.

Timperio

Cameron-Smith

Burns

Crawford

(2000). The public’s response to the obesity epidemic in Australia: Weight concerns and weight control practices of men and women. Public Health Nutrition, 3, 417–424.

52.

Treiber

Baranowski

Braden

Strong

Levy

Knox

(1991). Social support for exercise: Relationship to physical activity in young adults. Preventive Medicine, 20, 737–750.

53.

Tucker

Mueller

(2000). Spouses’ social control of health behaviors: Use and effectiveness of specific strategies. Personality and Social Psychology Bulletin, 26, 1120–1130.

54.

Umberson

(1992). Gender, marital status and the social control of health behavior. Social Science & Medicine, 34, 907–917.

55.

Walsh

V. R.

(1985). Health beliefs and practices of runners versus nonrunners. Nursing Research, 34, 353–356.

56.

Wardle

Haase

Steptoe

Nillapun

Jonwutiwes

Bellisie

(2004). Gender differences in food choice: The contribution of health beliefs and dieting. Annals of Behavioral Medicine: A Publication of the Society of Behavioral Medicine, 27, 107–116.

57.

Watters

Satia

(2009). Psychosocial correlates of dietary fat intake in African-American adults: A cross-sectional study. Nutrition Journal, 8, 1–9.

58.

Wing

R. R.

Marcus

M. D.

Epstein

L. H.

Jawad

(1991). A “family-based” approach to the treatment of obese type II diabetic patients. Journal of Consulting and Clinical Psychology, 59, 156–162.