Influence of sedentary, social and physical alternatives on food reinforcement

Abstract

Objective

This study examined the potential for non-food alternative activities to compete with the reinforcing value of food. Participants rated the frequency and pleasantness of engaging in a variety of activities and made hypothetical choices between food and four types of alternatives; cognitive-enriching (reading, listening to music), social (attending a party or event), sedentary (watching television) and physically active (running, biking).

Methods

Two-hundred and seventy-six adults completed an online survey using a crowdsourcing platform.

Results

Adults with higher BMI reported engaging in fewer activities within the cognitive-enriching, social and physically active categories. When examining how well each alternative activity type was able to compete with food, sedentary alternatives were associated with the highest food reinforcement, or were least able to compete with food reinforcers, as compared to cognitive-enriching, social and physical. Social activities were associated with the lowest food reinforcement, or the best able to compete with food reinforcers.

Conclusion

These results suggest that increasing the frequency and range of non-food alternative activities may be important to obesity. This study also suggests that the class of social activities may have the biggest impact on reducing food reinforcement, and the class of sedentary may have the smallest effect on food reinforcement. These tools have relevance to clinical interventions that capitalize on increasing access to behaviors that can reduce the motivation to eat in clinical interventions for obesity.

Daily life involves hundreds of choices between many different options. Some options are more likely or more frequently chosen than others. Reinforcing value, or the motivation to obtain a specific reinforcer, is likely to be important to the frequency of choosing reinforcers and likely guides choices between reinforcers. This is especially relevant for eating behaviors, as there are many choices about food that occur throughout a day. Relative reinforcing value describes the motivation to obtain a reinforcer within a choice situation that may include alternatives that do not involve eating Inherent in the concept of relative reinforcing value of food is that the context in which food choices are made is critical to understanding why people choose to eat rather than engage in alternative behaviors. The relative reinforcing value of food versus non-food alternative reinforcers is a reliable predictor of obesity and weight gain in both children (Epstein et al., 2015; Hill, Saxton, Webber, Blundell, & Wardle, 2009; Kong, Feda, Eiden, & Epstein, 2015; Temple, Legierski, Giacomelli, Salvy, & Epstein, 2008) and adults (Carr, Lin, Fletcher, & Epstein, 2014; Epstein, Carr, Lin, Fletcher, & Roemmich, 2012; Giesen, Havermans, Douven, Tekelenburg, & Jansen, 2010; Saelens & Epstein, 1996).

The relative reinforcing value of food depends on the availability of enjoyable alternatives to eating (Epstein, Bulik, Perkins, Caggiula, & Rodefer, 1991; Epstein, Roemmich, Saad, & Handley, 2004; Goldfield & Epstein, 2002). Children with obesity work harder for food and work less for concurrent non-food reinforcers than leaner peers (Temple et al., 2008). Infants who are overweight find food relatively more reinforcing than a variety of non-food alternatives. Interestingly, all infants respond similarly for food, but infants who are overweight respond significantly less for non-food alternatives than their leaner peers (Kong et al., 2015). Related to the idea that alternatives to food can modify the choice to eat and may be protective factors against obesity. Studies have shown living in an environment that included a variety of cognitively stimulating activities, including reading materials, access to music and musical instruments and encouragement of hobbies, was associated with a reduced risk for subsequent obesity in infants and young children, controlling for socioeconomic status of the family (Strauss & Knight, 1999), and was associated with healthier eating behaviors in children (Pieper & Whaley, 2011). The same principles also apply to addictive substances, and research has found that access to alternatives to smoking is a protective factor to substance use (Andrabi, Khoddam, & Leventhal, 2017) and having overall lower frequency and enjoyability of substance-free alternatives was related to increased risk of substance use (Leventhal et al., 2015). These data strongly suggest that the process of choice, and the study of concurrent reinforcers, is critical to understanding why people may overeat.

Understanding how the context of choice of food versus non-food alternatives may be important for prevention or treatment of obesity. Individuals who are overweight/obese have been shown to have fewer strong alternative reinforcers (Temple et al., 2008) and engage in fewer alternative activities (Doell & Hawkins, 1982; Pagoto, Spring, Cook, McChargue, & Schneider, 2006). Relative reinforcing value of food is associated with greater weight gain in both children (Hill et al., 2009) and adults (Carr et al., 2014). These data, in combination with data on differences in food and non-food reinforcement in infants (Kong et al., 2015), suggest that these individuals may choose food because of a deficit in access or enjoyability of non-food activities.

There has been no research on the types of alternatives that are most likely to influence the relative reinforcing value of food in adults. This study addresses this gap in the literature in two ways. First, the relationship between the frequency and enjoyability of engaging in non-food activities and BMI was studied. Building on this data, the impact of four different types of non-eating activities on the relative reinforcing value of food was examined, including: social activities, sedentary activities and physical activity. Sedentary activities were further separated into cognitive-enriching activities, including reading, listening to music, taking lessons and non-enriching activities, including watching television and shopping, for a total of four activity dimensions as alternatives to eating in a relative reinforcing value questionnaire.

Methods

Participants

Participants were adults aged 18–55 recruited through researchmatch.com and other online advertisements, including Craigslist and Facebook. After indicating interest in the study, participants were directed to an online survey hosted by SurveyMonkey.com. Researchmatch is a database of volunteers who have registered to participate in research studies. To ensure reliable data, participants were excluded if they did not complete the behavioral choice questionnaire (n = 76), or they completed the surveys in under 18 minutes which suggested they did not properly attend to the questionnaire (n = 8). In addition, participants were excluded if they scored greater than 15 on the PHQ-9 (n = 28), a standard measure of depression (Kroenke & Spitzer, 2002), as depression can reduce enjoyability of events (Bouman & Luteijn, 1986), thereby biasing their rating. Three were outliers (> 3 SD’s from the mean) on self-reported BMI (values of 65 or greater that may not have been biologically plausible and did not represent the usual range of population BMI values), and one person was excluded for reporting pregnancy, which meant that we could not obtain a usual weight. Three hundred and ninety two people signed the consent form, and data from 276 participants were used for analysis. Characteristics of participants are shown in Table 1.

Table 1.

Participant Characteristics (mean + standard deviation)

	Total
n	276
Age (years)	36.4 ± 10.6
Weight (kg)	78.0 ± 19.7
Height (cm)	171.0 ± 9.2
PHQ-9 Score	3.7 ± 3.5
BMI	26.5 ± 5.6
Percent overweight	45.7 ± 21.3
Income (dollars)	102,453 ± 94,923
Education (years)	16.4 ± 2.1
Pleasant Events Schedule
Total PES frequency	31.6 ±+ 8.1
Total social event frequency	23.4 ± 9.8
Total food event frequency	13.3 ± 7.1
Proportion Food event frequency	0.72 + 0.2
Proportion Social event frequency	0.41 + 0.2

Note – Proportion of food event frequency = Total food event frequency/(Total food event frequency + Total PES frequency). Proportion of social event frequency = Total social event frequency/(Total social event frequency + Total PES frequency)

Procedures

Eligible participants completed an online survey through surveymonkey.com. They first read a study description and consented to participate by clicking an agree button. All procedures were approved by the Institutional Review Board at the University at Buffalo. All participants completed the survey in the order presented in the measurement section.

Measurement

Height and weight

Participants self-reported height and weight. In additional, participants completed a body shape silhouette, as a validated way to estimate overweight/obesity status (Epstein, McCurley, & Murdock Jr, 1991).

Demographics

Race/ethnicity, household income, and education level were assessed using a standard questionnaire.

Depression screener

The Patient Health Questionnaire – 9 (PHQ-9) was used to screen for depression (Kroenke & Spitzer, 2002).

Food preferences

Participants rated how much they liked a series of entrée foods (burger and fries, chicken sandwich, fish and chips, pasta and sauce, pizza, pork chops and chicken tenders) on a scale from 1 (do not like at all) to 9 (like very much). Participants were instructed to assume that each dish met their dietary requirements, i.e. a burger is a veggie burger if the participant is a vegetarian. They were then instructed to imagine they were going to a casual restaurant for a meal and to choose their preferred entrée from the list. This entrée was then used as the target food in the behavioral choice questionnaire.

Pleasant events schedule

The pleasant events schedule has been used in previous studies to assess ratings of frequency of engaging in and pleasantness of activities (Correia, Carey, & Borsari, 2002; Pagoto et al., 2006). Participants were asked to indicate how often each event happened in the last month (0 = never happened; 1 = happened a few times (1 to 6); 2 = happened often (7 or more times)). Participants then completed pleasantness ratings of the events in which they participated (1, not pleasant to 5 extremely pleasant) and were recoded to fit a 0–2 scale with half points.

Participants then completed two additional versions of the pleasant events schedule based on previous substance abuse research examining the relative reinforcement obtained from substance-free and substance-related activities (Correia et al., 2002). First, participants were asked to indicate how often they participated in events in a social setting (0 = never, 1 = sometimes/most of the time, 2 = always) and how much they enjoyed these events (1 to 5, recoded to a 0–2 scale) when participation was in a social setting. Second, they were asked how often they ate while participating in these events and how much they enjoyed these events when participation included eating. The schedule was limited to 45 items that were categorized into the following event types; social (e.g. dating, helping someone, going to a party, etc n = 11), sedentary non-enriching (watching television, shopping, taking a nap, etc, n = 10), sedentary enriching (reading, solving a puzzle, playing an instrument, teaching, etc, n = 12), and physically active (e.g. dancing, running, swimming, recreational sports, etc, n = 12). The items that were specifically about eating, i.e. going to lunch with a friend, cooking, or substance abuse were removed from the list of activities.

The pleasant events schedule were scored on a 0 – 2 scale for frequency and pleasantness and included a cross-product of frequency x pleasantness that represents reinforcement (scale 0 – 4). The sum total for frequency and reinforcement were calculated, in addition for each individual event category. Finally, a proportion measure was calculated for each of the modified events, for a proportion of frequency and reinforcement of social-related events and food related events (social (or food) -related frequency total/(social (or food)-related frequency total + all events frequency total). Frequency of engaging in activities and pleasantness of activities were also calculated for each activity class.

Behavioral choice questionnaire

Participants first indicated their favorite activity in each of the following categories; social, sedentary, cognitive-enriching and physically active. Participants were then given the following instructions:

“Imagine that you have the option of going to a casual dining restaurant for the meal selected below or doing another activity that you enjoy. Please assume these are your only choices of activity you have at this time.”

Participants completed four choice questionnaires, one for each alternative. On the questionnaire, participants were given a choice between their preferred entrée food and their preferred alternative. Each choice was associated with a response requirement, asking the participant to indicate whether they preferred to click a mouse X times for food or 80 times for the alternative. The food response requirements started at 40 clicks and increased to 650 clicks over 18 questions (40, 60, 80, 100, 120, 140, 160, 180, 200, 250, 300, 350, 400, 450, 500, 550, 600, 650). The response requirements were chosen based on previous research with an in-laboratory behavioral task in which participants completed progressive ratio response schedules that doubled each schedule, starting with 20 responses, 40, 80, 160, 320, 640, etc (Epstein et al., 2007; Temple et al., 2008). The response requirements for the alternative did not change. As soon as the participant indicated they preferred the alternative to food, the survey moved onto the next alternative category. The Breakpoint_FOOD, or maximum responses a participant would make for food before switching to alternative, was used as the main dependent measure.

Analytic Plan

First, correlations between BMI and frequency of events were examined.

Next, mixed effect analysis of variance was used to examine differences in Breakpoint_FOOD when the choice was food or each type of alternative activity. Sex, age, BMI and PHQ-9 score were used as covariates, in addition to the frequency and pleasantness of engaging in alternative activities by activity class (i.e. frequency of social, cognitive-enriching, sedentary and physical activities). The effect of BMI on the effect of alternative activities on food reinforcement was examined by interacting alternative type with BMI. Significant differences in the ANOVA between types of alternatives were examined using linear contrasts. To correct for skewness in the breakpoint data (Shaprio-Wilks test of normality for social = 0.33, sedentary = 0.40, cognitive-enriching = 0.33, physical = 0.46, all p’s < 0.001), these data were log transformed prior to analysis.

Results

As shown in Table 2, BMI was negatively correlated with total frequency of pleasant events (r = −0.18, p = 0.003) and social situations-related frequency (r = −0.26, p < 0.0001), but not food situations-related frequency (r = −0.06, p = 0.34). When breaking the pleasant events schedule down into the four alternative categories, BMI was negatively correlated with frequency of engaging in physical activity, social events, and cognitively enriching events (Table 2).

Table 2.

Relationships between BMI, Pleasant Events variables and Relative Reinforcing value of food versus alternatives to food

	BMI	Pleasant Events Schedule							Relative Reinforcing Value of Food
		Total Frequency Scores			Frequency by Activity Class				Alternative type
		(1) Total	(2) Social situations	(3) Food Situations	(4) Social	(5) Sedentary	(6) Cognitive -enriched	(7) Physical Activity	(8) Sedentary	(9) Social	(10) Cognitive -enriched	(11) Physical Activity
(1)	−0.18**
(2)	−0.26***	0.82***
(3)	−0.06	0.61***	0.64***
(4)	−0.15*	0.81***	0.73***	0.57***
(5)	0.10	0.55***	0.39***	0.34***	0.37***
(6)	−0.17**	0.74***	0.57***	0.37***	0.40***	0.25***
(7)	−0.22***	0.71***	0.60***	0.45***	0.47***	0.16**	0.33***
(8)	−0.07	−0.11	−0.03	−0.11	−0.10	−0.23***	−0.03	0.004
(9)	0.10	−0.23***	−0.18**	−0.10	−0.21***	−0.13*	−0.17**	−0.13*	0.20**
(10)	0.06	−0.18**	−0.15*	−0.05	−0.12	−0.08	−0.15*	−0.16**	0.30***	0.30***
(11)	0.13*	−0.21***	−0.16**	−0.03	−0.16**	−0.001	−0.14*	−0.27***	0.25***	0.33***	0.48***	1.0***

^*p<0.05

^**p<0.01

^***p<0.001

Mixed effects analysis of variance was used to examine the effect of alternative activity class on relative food reinforcement. Sex had a significant main effect on breakpoint for food (F(1,271) = 12.0, p = 0.0006), with men having significantly higher breakpoint than women (62.3 ± 1 breakpoint versus 52.7 ± 1 breakpoint). Frequency and pleasantness of engaging in activities were highly correlated (r = 0.90, p < 0.0001) and were examined as covariates in separate models. Both frequency and pleasantness had main effects on Breakpoint_FOOD (Frequency (F(1,271) = 31.34, p < 0.0001; Pleasantness (F(1,271) = 43.04, p < 0.0001). Individuals with more frequent engagement of activities had an overall lower breakpoint_FOOD (18.6 ± 1 versus 32.9 ± 1) and higher pleasantness ratings had an overall lower breakpoint_FOOD (16.2 ± 1 versus 33.4 ± 1). Results for the mixed effect analysis of variance were not different between models, so frequency is the covariate in the models described.

Mixed effects analysis of variance showed significant differences in food reinforcement as a function of effect of alternative reinforcer type (F(3,271) = 6.47, p = 0.0003, ω² = 0.013). Contrast tests showed the reinforcing value of food was lowest when social activities were the alternative (52.8 ± 1 breakpoint (least squared mean + SE)) choice, which was significantly lower than the other three activities, (t(271) = −2.86, p = 0.005). Food reinforcement was significantly different between social and sedentary alternatives (t(271) = 3.61, p = 0.0004), and marginally significant between social and physical activity alternatives (t(271) = 1.94, p = 0.053) (Figure 1). The reinforcing value of food was highest when sedentary activity was the alternative (64.0 ± 1 breakpoint) in comparison to all three other activities (t(271) = 3.15, p = 0.002) and was significantly higher than cognitive-enriching alternatives (t (271) = 3.52, p = 0.001). There was not a significant interaction between BMI and alternative activity class on food reinforcement (F(3,271) = 1.87, p = 0.13) or an interaction of BMI with frequency (F(1,271) = 1.17, p = 0.28) or pleasantness (F(1,271) = 1.55, p = 0.21).

Figure 1.

Alternative reinforcer types influence food reinforcement. Natural log values were converted to breakpoint to visualize the data more clearly (mean ± standard error). All data is calculated from the mixed ANOVA models, co-varying age, sex, BMI and frequency of activities. There was a significant effect of type of alternative on the natural log of Breakpoint_FOOD, with social activities having the lowest reinforcement and sedentary activities having the highest food reinforcement. Between group differences are shown by contrast tests as indicated by the marked lines. *p < 0.01

Discussion

These results replicate the finding that persons with obesity engage in fewer, and less enjoyable activities that do not involve eating (Pagoto et al., 2006). Higher levels of BMI were associated with fewer cognitive-enriching, social activities, and sedentary activities. The pattern of the negative relationship between BMI and activities are consistent with the observations that growing up in an environment with a variety of cognitively enhancing activities is associated with less weight gain in children (Strauss & Knight, 1999), and that an enriched environment, or greater access to non-food activities can improve weight loss in a behavioral modification program for some children (Best et al., 2012; Buscemi, Murphy, Berlin, & Raynor, 2014).

Examining the relationship between the frequency of engaging in activities and BMI also suggest that individuals with obesity are not engaging in more food related events, but fewer alternative activities, in addition to a smaller proportion of social-related activities. In a study in infants, those with greater weight-for-length scores did not have higher food reinforcement, but lower reinforcing value of non-food alternatives (Kong et al., 2015), similar to the results presented here. The current study is a step in developing methods to identify alternative reinforcers to eating in adults.

Understanding the characteristics of a strong alternative to may lead to a determination of which alternatives are added to one’s environment to compete with food. The relative reinforcing value data begin to reveal classes of behaviors that may be relevant for reducing the reinforcing value of food. Alternatives to food can reduce the motivation to eat, and being able to individualize a behavior prescription is a key step in using alternatives as a tool for behavior change (Epstein, Roemmich, Stein, Paluch, & Kilanowski, 2005). Social activities were associated with the lowest reinforcing value of food, while sedentary were associated with the highest reinforcing value of food. This was true for participants whether they were considered obese or lean.

The common approach to modifying energy balance in treating obesity is to reduce energy intake while increasing physical activity. Reducing energy intake may have the unwanted effect of making food more reinforcing, as food deprivation is reliably increases food reinforcement (Epstein, Truesdale, Wojcik, Paluch, & Raynor, 2003; Perkins, Epstein, Fonte, Mitchell, & Grobe, 1995; Raynor & Epstein, 2003). However, these data suggest that increasing the frequency of alternatives to food, in particular social and cognitive-enriching activities, may have a complementary effect on a dietary intervention, making it easier for people to adhere to a diet. Research showing greater weight loss for those who engaged in fewer eating related activities in relationship to non-eating related activities supports this hypothesis (Best et al., 2012; Buscemi et al., 2014). However, this study did not examine specific classes of alternatives to food that could serve to make food less reinforcing.

Choices are usually made in an environment with many alternative options and the field of behavioral economics focuses on the context of choice. Which alternatives are simultaneously available will influence the choice for your target activity. When offering both coffee and tea, one assumes that individuals will choose their preferred beverage. If the coffee runs out, and individuals are asked to go to the next floor to get coffee, many people will switch to drinking tea. This is described as a substitutable relationship, in which one choice can be exchanged for the other if the preferred alternative becomes unavailable or less accessible. On the other hand, if you were offering coffee and pastries, your guests may choose to take one of each offered food. Running out of coffee would likely result in less guests eating pastries, as these foods may be thought of as complements, or likely to be chosen together. The idea of providing alternatives to food then is twofold; reinforcers that are complements to food should be decreased, while those that are substitutes should be increased.

This study suggests that cognitive-enriching, social and physical activity frequency are negatively correlated with BMI, so may act as alternatives or substitutes for eating, while sedentary activity frequency was not associated with BMI in this data. In addition, sedentary alternatives were associated with the highest relative reinforcing value of food. The activity class of non-enriching sedentary activities as defined in this study includes alternatives such as watching television, using social media and playing video games. These are activities that are thought of as complements to food, in which food is eaten while engaging in these activities. Previous research in children shows that food is more reinforcing when combined with watching television (Temple, Giacomelli, Kent, Roemmich, & Epstein, 2007) and reducing screen time was led to reduced energy intake (Epstein et al., 2008). This is a potential reason why sedentary activities were the weakest alternatives, as these activities are complements to eating. This difference should be carefully examined, as this study did not measure food complements within any category.

A challenge to implementing these ideas may be barriers to increasing the frequency of engaging in alternatives to eating. Individuals with high BMI’s engage in fewer alternative activities, including cognitive-enriching, physical and social, and engage in sedentary activities as often as individuals with lower BMIs. In addition, higher frequency and pleasantness of activities were associated with lower food reinforcement. For someone who has a rich history of engaging in activities complementary to eating, and less frequent engagement in alternative substitute activities, may have a more difficult time developing new alternatives. Building that repertoire will require the person to repeatedly engage in behaviors that are novel and may not be very reinforcing.

One option is to use the same principles that have been use to increase food reinforcement (Clark, Dewey, & Temple, 2010; Temple et al., 2009) to sensitize the reinforcing value of non-food behaviors. Repeated exposure to snack foods increased the reinforcing value of that snack food for women who are obese (Temple & Epstein, 2011; Temple et al., 2009). Finding ways to increase the frequency of engaging in alternative activities may sensitize reinforcement of alternative activities. Both frequency and pleasantness of activities predicted the relative reinforcing value of food, so strategies can be developed to target either aspect of an activity. Increasing access to alternatives may increase frequency of use (Epstein, Paluch, Kilanowski, & Raynor, 2004). This can involve decreasing the response requirements, or barriers to use of different activities. Stimulus control suggests making unhealthy activities difficult to access and healthy ones easier to access, including putting food in a high cabinet and placing running shoes by the door. Another strategy may be to increase the reinforcing value of alternatives by pairing them with other reinforcers. Social activities have been shown to amplify the pleasantness of activities (Boothby, Clark, & Bargh, 2014), suggesting that pairing less reinforcing alternatives with a social activity may increase the reinforcing value of that activity.

While this study represents a first step in research on alternative reinforcers to food, there are limitations. This was an online questionnaire that had twenty percent non-completion rate. Participants self-reported BMI, which may result in underreporting of BMI (Shields, Gorber, & Tremblay, 2008). People who reported they were overweight or obese were likely to be overweight, but many persons who are overweight or obese may have reported a lower BMI. Thus, these data may be conservative, and with objective measurement of BMI the relationships may have been stronger. The relative food reinforcement measure was collected using a questionnaire, validated against the behavioral reinforcement task (Goldfield, Epstein, Davidson, & Saad, 2005), but future research should examine behavioral responding for food against multiple alternatives.

The relevance of building a repertoire of alternatives to eating may be different for preventing versus treating obesity. To prevent obesity it is not necessary to reduce energy intake, but rather to prevent it from increasing. Identifying and planning to engage in reinforcing alternatives to eating may shift allocation of time and effort from eating to non-eating activities, which could prevent developing obesity. Treatment of obesity involves changing energy balance, which could benefit from building a repertoire of alternatives to eating in addition to successful dieting and learning behavioral weight loss skills. Identifying alternatives to eating may be particularly relevant to reduce the increases in food reinforcement that accompany food restriction and many diet programs.

Considering eating as a choice provides a new conceptual approach to how to regulate eating. More research is needed to identify what might be individual alternatives to eating, and determine how to increase the frequency of these behaviors. Research with infants has shown that a structured program to increase non-food alternatives can reduce food reinforcement in infants high in food reinforcement (Kong et al., 2016). Research is needed in older children, adolescents and adults to identify how to build a repertoire of behaviors that are alternatives to eating.