bs_bs_banner
APPLIED PSYCHOLOGY: HEALTH AND WELL-BEING, 2014, 6 (1), 119–134 doi:10.1111/aphw.12020
Intrinsic Rewards, Fruit and Vegetable Consumption, and Habit Strength: A Three-Wave Study Testing the Associative-Cybernetic Model Amelie U. Wiedemann1* Freie Universitaet Berlin, Germany
Benjamin Gardner1 University College London, UK
Nina Knoll Freie Universitaet Berlin, Germany
Silke Burkert Universitaetsmedizin Berlin, Germany
Background: Habit formation is thought to lead to long-term maintenance of fruit and vegetable consumption. Habits develop through context-dependent repetition, but additional variables such as intrinsic reward of behaviour may influence habit strength. Drawing upon the Associative-Cybernetic Model, this exploratory study tested different pathways by which intrinsic reward may influence fruit and vegetable consumption habit strength. Methods: In a threewave study of fruit and vegetable intake in adults (N = 127) from the general population, intrinsic reward, intention, and self-efficacy were assessed at baseline, fruit and vegetable consumption and intrinsic reward two weeks later, and habit strength another two weeks later. Direct, indirect, and moderation effects of intrinsic reward on habit strength were tested simultaneously in a moderated mediation model. Results: Intrinsic reward had a positive indirect effect on habit strength through its influence on the frequency of fruit and vegetable consumption. Further, the relationship between fruit and vegetable consumption and habit was stronger where consumption was considered more intrinsically rewarding. Conclusions: Findings highlight the potential relevance of intrinsic reward to habit. We suggest that intrinsic rewards from behaviour may
* Address for correspondence: Amelie U. Wiedemann, Department of Education and Psychology, Division of Health Psychology, Freie Universitaet Berlin, Habelschwerdter Allee 45, 14195 Berlin, Germany. Email: [email protected] 1 Amelie U. Wiedemann and Benjamin Gardner share first authorship. © 2013 The International Association of Applied Psychology
120
WIEDEMANN ET AL.
not only facilitate habit via behaviour frequency, but also reinforce the relationship between behavioural repetition and habit strength. Keywords: fruit and vegetable consumption, habit, intrinsic reward
INTRODUCTION The promotion of long-term maintenance of fruit and vegetable (FV) consumption requires an understanding of the psychological variables that underpin continuation of behaviour over time (Rothman, Baldwin, & Hertel, 2004). One variable that has been attracting interest as a mechanism for behaviour maintenance is “habit” (e.g. Lally & Gardner, 2013; Rothman, Sheeran, & Wood, 2009). Habit has been defined as a disposition to automatically repeat actions in settings in which they have been previously performed (Wood & Neal, 2007). For example, when sitting at the dinner table in the evening, individuals may automatically repeat their daily dietary habit of eating a piece of fruit. Habits are acquired through a process of “context-dependent repetition” (Lally, van Jaarsveld, Potts, & Wardle, 2010): repeating behaviour in stable contexts reinforces a mental association between context (e.g. at dinner table in the evening) and behaviour (e.g. eating a piece of fruit). As this association is strengthened, alternative behavioural options (e.g. other foods) become less accessible within memory (Wood & Neal, 2009). A habit is said to have formed when merely encountering the context is sufficient to activate the associated behaviour directly and automatically, with minimal forethought (Orbell & Verplanken, 2010). Automaticity is the “active ingredient” of habit (Sniehotta & Presseau, 2012). Regarding FV consumption, for instance, acting on the intention to consume an apple at breakfast may require conscious and effortful deliberation. In contrast, the habitual consumption of a piece of fruit for dinner may proceed quickly, outside of awareness, and without conscious intention (Bargh, 1994). Impulsive habitual tendencies can thereby override deliberative counter-habitual intentions in determining action (Gardner, de Bruijn, & Lally, 2011). As automaticity strengthens, behaviours such as FV consumption become less reliant on conscious motivation, memory, or planning, so easier to enact, and harder to forget to perform (Wood & Neal, 2007). For this reason, it has been suggested that habit formation be treated as a goal for healthful dietary interventions (Lally & Gardner, 2013; Rothman et al., 2009). Promoting FV consumption or other health-related habits necessitates understanding of the habit development process. Decades of animal and human learning research point to habit forming through repeated contextspecific performance (Adams, 1982; Judah, Gardner, & Aunger, 2013; Lally et al., 2010; Wood & Neal, 2007). A study of participants forming healthy dietary or physical activity habits showed the habit formation curve to be © 2013 The International Association of Applied Psychology
INTRINSIC REWARD PREDICTS FRUIT AND VEGETABLE HABIT
121
asymptotic: initial repetitions of action led to greatest gains in automaticity, and subsequent repetitions had lesser contributions, until an automaticity plateau was reached (Lally et al., 2010). There was, however, considerable variation across participants and behaviours in both the time required for automaticity to plateau and the plateau point, despite equal repetitions over the study period. Variables additional to repetition may therefore influence the process of habit formation and habit strength. It has been suggested that habit learning can be further aided by rewards (Lally & Gardner, 2013). The Associative-Cybernetic Model (De Wit & Dickinson, 2009) proposes that when performing an action in a given context (e.g. eating fruit at dinner table in the evening), the experience of a positive outcome of behaviour (e.g. the pleasurable taste of the fruit) facilitates learning of the context–behaviour relationship (Wood & Neal, 2007). The model suggests that the reward value of behaviour (or its outcomes) should strengthen habit via two routes. First, reward should have an impact on habit that is mediated by behaviour repetition: repeated sequential presentation of context, behaviour, and reward is hypothesised to lead to the context signalling both an opportunity for action and an incentive to act, so increasing its value as a cue to action above that imbued by the simple pairing of context and behaviour regardless of consequences (De Wit & Dickinson, 2009). Thus, greater rewards should provide greater incentive to repeat the behaviour, thereby increasing the likelihood of context-dependent performance on subsequent occasions. Second, reward should moderate the relationship between repetition and habit strength. Experiencing rewarding outcomes each time the behaviour is performed in the context is predicted to strengthen the context–behaviour association more than it would were the behaviour unrewarded. Reward should thereby facilitate quicker learning of the underlying context–behaviour associations, and so rewarding behaviour should lead to stronger habits than repetition of less rewarding behaviours (De Wit & Dickinson, 2009). When considering the impact of reward on repetition and habit strength, a distinction must be drawn between extrinsic and intrinsic rewards. Tangible extrinsic rewards (e.g. financial incentives) may not be conducive to predict habit strength (Dickinson, 1985): external rewarding can lead performance to be dependent on the provision of such a reward (Colwill & Rescorla, 1985). Thus, discontinuing the provision of rewards may prompt disengagement from action (Wood & Neal, 2009) and consequently impede habit formation. Intrinsic rewards, i.e. retrospective evaluations of the reward value of a certain action derived from past performance of the behaviour itself (e.g. satisfaction or pleasure), may, however, have a positive impact on habit. In Lally et al.’s (2010) study, habits formed despite no tangible behaviour-contingent rewards being offered to participants, and in another study, physical activity habits developed among rehabilitation © 2013 The International Association of Applied Psychology
122
WIEDEMANN ET AL.
patients without extrinsic rewards (Fleig, Lippke, Pomp, & Schwarzer, 2011). In both studies, participants pursued self-chosen behavioural goals, and so may have derived intrinsic rewards (e.g. pleasure or satisfaction) from performing a desired action. Elsewhere, a study of dental flossing habit formation found that positive outcome expectancies not only predicted greater behavioural repetition, but also stronger habit when controlling for repetition frequency (Judah et al., 2013). Intrinsic reward can be a special case of a positive outcome expectancy when the outcome is defined as anticipated reward inherent within behavioural performance. Accordingly, Judah et al. suggested that rewards may have aided habit development, because outcome expectancies provided a proxy indicator of the intrinsic reward value of flossing. The direct effect of outcome expectancies on habit strength was unexpected, and no moderating effect of expected outcomes on the repetition–habit strength link was found. These findings may have been due to methodological limitations arising from the small study sample (N = 50), and limited variation in behaviour frequency (Judah et al., 2013). Elsewhere, a cross-sectional study of physical activity habits found that, of participants who were frequently physically active, those also motivated by intrinsic rewards reported stronger habits (Gardner & Lally, 2013). Gardner and Lally suggested that this finding may reflect intrinsic rewards having more absolute value than extrinsic rewards, and so, in line with the Associative-Cybernetic Model, greater reward may reinforce the behaviour–habit strength relationship.
The Present Study This study offers a test of the Associative-Cybernetic Model (De Wit & Dickinson, 2009) in relation to FV consumption. The model predicts that intrinsic rewards will determine habit strength in two ways. First, intrinsic rewards may increase the likelihood of repetition, such that repetition will mediate the influence of intrinsic reward on habit. Second, intrinsic rewards may enhance the effect of each repetition on habit development, thus moderating the relationship between repetition and habit strength. There is some tentative empirical evidence that positively valued behavioural outcomes lead to stronger habits (Gardner & Lally, 2013; Judah et al., 2013). To our knowledge however, no study has explicitly explored the relationship between intrinsic rewards and acquired behavioural automaticity (i.e. habit) of health behaviours. Following the Associative-Cybernetic Model, we hypothesised that: Hypothesis 1: Perceived intrinsic reward (measured at Time 1 [T1]) will influence habit strength of FV consumption (T3) indirectly, through its influence on (T2) FV consumption frequency (i.e. mediation). © 2013 The International Association of Applied Psychology
INTRINSIC REWARD PREDICTS FRUIT AND VEGETABLE HABIT
123
Hypothesis 2: Perceived intrinsic reward (T2) will strengthen the relationship between FV consumption frequency (T2) and habit strength (T3), such that where FV consumption is viewed as more rewarding, behaviour will be more predictive of habit strength (i.e. moderation).
METHODS
Participants and Procedure A study design with three measurement points with two-week intervals was used. Recruitment of participants took place in adult physical education classes in Germany (e.g. yoga, spinal exercises; no diet or weight loss programmes). This sample was chosen because we expected that their motivation to engage in health-promoting behaviour would be strong and so fruit and vegetable consumption would be likely. Research assistants approached individuals at the education centre within the first class, and explained the study. Participants were eligible for inclusion only where they gave informed consent, were aged 18 years or above, and had no self-reported medical conditions conflicting with health recommendations for dietary behaviour. Participants received entry into a prize draw for health-related products (e.g. cookbooks). A total of 127 individuals completed baseline paper-and-pencil questionnaires on-site after class and were invited to complete follow-up measures two weeks (T2) and four weeks (T3) later in the same course. The baseline sample had a mean age of 31.7 years (SD = 10.1; range 20–66 years), comprised 74.0 per cent women, and 37.1 per cent were in a long-lasting relationship. The majority of the participants reported 10 or more years of schooling (91.4%). Between 0 and 3 participants chose not to relay specific socio-demographic information. On average, individuals in this study had moderate intentions to consume the recommended number of servings of FV intake at baseline (M = 3.50; SD = 1.38). Follow-up data were available from 53.5 per cent (n = 68) of the participants at T2 and of 48 per cent (n = 61) of the participants at T3. The study was approved by an Institutional Review Board and conducted in line with the German Psychological Society ethical guidelines.
Measures At the top of the first page of all questionnaires, one serving of fruit or vegetables was defined as the amount of food that fits into the palm of the hand. A statement was also provided that rice and products made of potatoes should not be regarded as fruit or vegetables. Unless otherwise stated, response options were on 6-point Likert scales, ranging from “completely disagree” (1) to “completely agree” (6). Item examples reported below are translations from German. © 2013 The International Association of Applied Psychology
124
WIEDEMANN ET AL.
Intrinsic reward at T1 and T2 was assessed with the item: “Thinking about last week’s fruit and vegetable intake, please indicate how much you agree with the following statement: The consumption of fruit and vegetables in itself is rewarding for me.” Intention at T1 was measured with a single item: “I intend to consume 5 servings of fruit or vegetables a day.” Similar items were used in previous studies as single items or within multi-item scales with high internal consistencies and predictive validity (Schwarzer et al., 2007; Wiedemann, Lippke, Reuter, Ziegelmann, & Schüz, 2011a). Self-efficacy for FV consumption at T1 was measured with four items, capturing different facets of the perceived ability to consume 5 servings of fruit or vegetables a day, namely task-specific, initiation, maintenance, and recovery self-efficacy (Cronbach’s α = .75; Ochsner, Scholz, & Horning, 2013; Wiedemann, Lippke, Reuter, Ziegelmann, & Schwarzer, 2011b), e.g. “I am confident that I can consume 5 servings of fruit or vegetables a day for a prolonged period of time even if it is difficult sometimes.” Fruit and vegetable consumption at T2 was assessed with the item “How many servings of fruit and vegetables did you eat on an average day in the last week?” with an open-ended format. A similar single-item measure of FV intake has been validated against dietary biomarkers (Steptoe et al., 2003). Habit strength at T3 was assessed with the Self-Report Behavioural Automaticity Index (SRBAI), an automaticity-specific abbreviation of the SelfReport Habit Index (Verplanken & Orbell, 2003), which has been shown to have good content and predictive validity (Gardner, Abraham, Lally, & de Bruijn, 2012). The stem “Consuming 5 servings of fruit or vegetables is something . . .” was followed by the four SRBAI items: “. . . I do automatically”, “. . . I do without having to consciously remember”, “. . . I do without thinking”, “. . . I start doing before I realise I’m doing it”, Cronbach’s α = .97.1 All analyses were conducted with missing data of the baseline sample (N = 127; ≤ 5% item-non-response) imputed longitudinally using the expectation maximisation algorithm (Enders, 2001), which has proven more robust than regression imputation (Gold & Bentler, 2000).
1 Habits are characterised by their stability over time (Rothman et al., 2009; Wood & Neal, 2007). Hence, we also measured habit strength using the SRBAI at T1 (α = .95) to permit a preliminary analysis of habit stability over two time points, and a strong correlation was found (r = .66, p < .001). We use T3 habit strength as the outcome variable in our analysis to better capture the temporal sequence of expected relationships, and so do not discuss the T1 habit strength measure further.
© 2013 The International Association of Applied Psychology
INTRINSIC REWARD PREDICTS FRUIT AND VEGETABLE HABIT
125
Statistical Methods Both hypotheses on indirect and moderating effects of reward on habit strength were tested in one integrative model: the indirect effect of intrinsic reward on habit strength, as mediated by the frequency of FV consumption (Hypothesis 1), was tested simultaneously with the moderating effects of reward on the behaviour–habit relationship (Hypothesis 2) in a single moderated mediation model. More precisely, the indirect effect of intrinsic reward (independent variable) via frequency of FV consumption on habit strength (dependent variable) was tested as a function of intrinsic reward (continuous moderator) using two ordinary least square regression analyses, controlling for intentions and self-efficacy. The conceptual model is depicted in Figure 1. Age and gender were not included as covariates in the model. No significant bivariate associations between age and any of the model variables were found (p > .05). Significant correlations between gender and model variables were of a small effect size (r = .21 to .26) for reward (T1/T2), FV consumption T2 and habit strength T3, but relationships proved non-significant in regression models. In one regression analysis (i.e. the mediator model), frequency of FV intake (T2) was predicted by baseline T1 intrinsic reward, intentions, and selfefficacy. In the second regression (i.e. the dependent variable model), habit strength (T3) was predicted by T2 intrinsic reward, FV consumption, their interaction term, and the T1 covariates intentions and self-efficacy. To investigate the conditionality of the indirect effect, the strength of the indirect effect of reward on habit strength was tested at different values of the continuous moderator observed in the data set. Bootstrapping was applied to generate bias-corrected and accelerated confidence intervals (CIBCA) from
Reward T2
FV ConsumptionT2
FV Consumption T2 * Reward T2
Reward T1
Habit Strength T3
Self-Efficacy T1 Intention T1
FIGURE 1. Conceptual Model: Conditional indirect effect of reward T1 on habit strength T3 via fruit and vegetable consumption frequency T2 as function of reward T2 as moderator. © 2013 The International Association of Applied Psychology
126
WIEDEMANN ET AL.
5,000 resamples with α = .05 as indirect effects tend to have a non-normal sampling distribution (Preacher & Hayes, 2004). Analyses were conducted using the PROCESS macro for SPSS (model 14; Hayes, 2012).
RESULTS
Descriptive Results Sample Characteristics. On average, individuals in this study consumed 2.99 servings of fruit or vegetables (SD = 1.17), and perceived their FV intake as quite intrinsically rewarding on average (M = 3.66; SD = 1.43). Selfefficacy for FV intake was quite high (M = 4.30; SD = 0.89). Habit strength of FV consumption averaged 3.43 (SD = 1.52). At T3, individuals consumed 3.61 servings (SD = 0.95) of fruit or vegetables and reported a habit strength of FV consumption of 3.52 (SD = 1.10).
Attrition Analysis Attrition analyses compared participants retained until T3 and those lost after T1 and T2, respectively, using analyses of variance (ANOVAs) for continuous measures, and χ2 tests for categorical measures. No significant baseline differences were identified regarding FV consumption, intentions and self-efficacy for FV intake, habit strength regarding FV intake, reward, age, education status, marital status, and gender (p > .05).
Correlations between Model Variables and Covariates Intercorrelations between intrinsic reward, FV consumption, habit strength, as well as potential covariates (intentions, self-efficacy) were tested as a precondition for the main model test. Reward at T1 was positively associated with FV consumption (T2) and habit strength (T3), and FV consumption and habit strength correlated positively (see Table 1). Reward at T2 was also positively associated with FV consumption and habit strength. Intentions and self-efficacy correlated with all model variables, except a non-significant association between self-efficacy and reward (T1). Thus, potential covariates were controlled for in model tests.
Direct, Indirect, and Moderating Effects of Intrinsic Reward on Habit Strength First, in the mediator model, frequency of FV consumption T2 was predicted by T1 intrinsic reward, intentions, and marginally self-efficacy (see Table 2). © 2013 The International Association of Applied Psychology
INTRINSIC REWARD PREDICTS FRUIT AND VEGETABLE HABIT
127
TABLE 1 Intercorrelations between Reward, Fruit and Vegetable Intake, Habit Strength, and Covariates
1. 2. 3. 4. 5. 6.
Self-efficacy T1 Intention T1 Reward T1 Reward T2 FV Intake (no. of servings) T2 Habit Strength T3
1.
2.
3.
4.
5.
— .40*** .13ns .29** .33*** .32***
— .29*** .35*** .50*** .35***
— .75*** .42*** .49***
— .51*** .53***
— .72***
Note: N = 127. * p < .05; ** p < .01; *** p < .001;
ns
= non-significant.
Variables in the model explained 35.0 per cent of the variance in FV consumption, F(3, 123) = 22.52, p < .001.2 In the subsequent dependent variable model, strength of FV consumption habits was predicted by the frequency of FV consumption, baseline intrinsic reward, the interaction term between T2 FV consumption and T2 intrinsic reward, and, marginally, T2 self-efficacy, but not by main effects of T2 intentions and T2 reward (see Table 2). The direct effect of baseline reward on T3 habit strength was significant (B = .18, p = .01). Significance tests at all values of the reward scale supported intrinsic reward having a significant indirect effect on habit strength through the frequency of FV intake, supporting Hypothesis 1 (see Figure 2). A probe of the significant interaction indicated that the influence of baseline reward on habit strength via behaviour was amplified by the perceived reward value of FV consumption at T2 (see Table 2): The more consumption was perceived as intrinsically rewarding, the larger its indirect effect on habit strength (see Figure 2). For example, at low reward values (reward = “2” on a 6-point Likert scale), the indirect effect on habit strength was B = 0.09 (CI 0.03, 0.21), at moderate values (reward = “4”) it was B = 0.15 (CI 0.07, 0.26), and at high values on the reward measure (reward = “6”) the indirect effect was B = 0.20 (CI 0.08, 0.37). In total, 60.0 per cent of the variance in habit strength regarding FV consumption was explained by the model, F(6, 120) = 30.24, p < .001. Hypothesis 2 was supported.
DISCUSSION This three-wave study investigated predictions from the AssociativeCybernetic Model (De Wit & Dickinson, 2009). Results supported the 2 Analyses based on the longitudinal sample (i.e. excluding dropouts) led to the same pattern of results with only small changes in coefficients, and thus are not reported here.
© 2013 The International Association of Applied Psychology
128
WIEDEMANN ET AL. TABLE 2 Moderated Mediation Analysis with Unstandardised Parameter Estimates, Using Reward as Moderator of the Reward–Behaviour Frequency–Habit Relationship Prediction of FV Consumption T2 (mediator)
Predictor
B
SE
Reward T1 0.19 0.05 Intention T1 0.24 0.06 Self-Efficacy T1 0.16 0.08 Model summary: R2 = .35; F(3, 123) = 22.52, p < .001
T
p
3.91 4.34 1.92
APPLIED PSYCHOLOGY: HEALTH AND WELL-BEING, 2014, 6 (1), 119–134 doi:10.1111/aphw.12020
Intrinsic Rewards, Fruit and Vegetable Consumption, and Habit Strength: A Three-Wave Study Testing the Associative-Cybernetic Model Amelie U. Wiedemann1* Freie Universitaet Berlin, Germany
Benjamin Gardner1 University College London, UK
Nina Knoll Freie Universitaet Berlin, Germany
Silke Burkert Universitaetsmedizin Berlin, Germany
Background: Habit formation is thought to lead to long-term maintenance of fruit and vegetable consumption. Habits develop through context-dependent repetition, but additional variables such as intrinsic reward of behaviour may influence habit strength. Drawing upon the Associative-Cybernetic Model, this exploratory study tested different pathways by which intrinsic reward may influence fruit and vegetable consumption habit strength. Methods: In a threewave study of fruit and vegetable intake in adults (N = 127) from the general population, intrinsic reward, intention, and self-efficacy were assessed at baseline, fruit and vegetable consumption and intrinsic reward two weeks later, and habit strength another two weeks later. Direct, indirect, and moderation effects of intrinsic reward on habit strength were tested simultaneously in a moderated mediation model. Results: Intrinsic reward had a positive indirect effect on habit strength through its influence on the frequency of fruit and vegetable consumption. Further, the relationship between fruit and vegetable consumption and habit was stronger where consumption was considered more intrinsically rewarding. Conclusions: Findings highlight the potential relevance of intrinsic reward to habit. We suggest that intrinsic rewards from behaviour may
* Address for correspondence: Amelie U. Wiedemann, Department of Education and Psychology, Division of Health Psychology, Freie Universitaet Berlin, Habelschwerdter Allee 45, 14195 Berlin, Germany. Email: [email protected] 1 Amelie U. Wiedemann and Benjamin Gardner share first authorship. © 2013 The International Association of Applied Psychology
120
WIEDEMANN ET AL.
not only facilitate habit via behaviour frequency, but also reinforce the relationship between behavioural repetition and habit strength. Keywords: fruit and vegetable consumption, habit, intrinsic reward
INTRODUCTION The promotion of long-term maintenance of fruit and vegetable (FV) consumption requires an understanding of the psychological variables that underpin continuation of behaviour over time (Rothman, Baldwin, & Hertel, 2004). One variable that has been attracting interest as a mechanism for behaviour maintenance is “habit” (e.g. Lally & Gardner, 2013; Rothman, Sheeran, & Wood, 2009). Habit has been defined as a disposition to automatically repeat actions in settings in which they have been previously performed (Wood & Neal, 2007). For example, when sitting at the dinner table in the evening, individuals may automatically repeat their daily dietary habit of eating a piece of fruit. Habits are acquired through a process of “context-dependent repetition” (Lally, van Jaarsveld, Potts, & Wardle, 2010): repeating behaviour in stable contexts reinforces a mental association between context (e.g. at dinner table in the evening) and behaviour (e.g. eating a piece of fruit). As this association is strengthened, alternative behavioural options (e.g. other foods) become less accessible within memory (Wood & Neal, 2009). A habit is said to have formed when merely encountering the context is sufficient to activate the associated behaviour directly and automatically, with minimal forethought (Orbell & Verplanken, 2010). Automaticity is the “active ingredient” of habit (Sniehotta & Presseau, 2012). Regarding FV consumption, for instance, acting on the intention to consume an apple at breakfast may require conscious and effortful deliberation. In contrast, the habitual consumption of a piece of fruit for dinner may proceed quickly, outside of awareness, and without conscious intention (Bargh, 1994). Impulsive habitual tendencies can thereby override deliberative counter-habitual intentions in determining action (Gardner, de Bruijn, & Lally, 2011). As automaticity strengthens, behaviours such as FV consumption become less reliant on conscious motivation, memory, or planning, so easier to enact, and harder to forget to perform (Wood & Neal, 2007). For this reason, it has been suggested that habit formation be treated as a goal for healthful dietary interventions (Lally & Gardner, 2013; Rothman et al., 2009). Promoting FV consumption or other health-related habits necessitates understanding of the habit development process. Decades of animal and human learning research point to habit forming through repeated contextspecific performance (Adams, 1982; Judah, Gardner, & Aunger, 2013; Lally et al., 2010; Wood & Neal, 2007). A study of participants forming healthy dietary or physical activity habits showed the habit formation curve to be © 2013 The International Association of Applied Psychology
INTRINSIC REWARD PREDICTS FRUIT AND VEGETABLE HABIT
121
asymptotic: initial repetitions of action led to greatest gains in automaticity, and subsequent repetitions had lesser contributions, until an automaticity plateau was reached (Lally et al., 2010). There was, however, considerable variation across participants and behaviours in both the time required for automaticity to plateau and the plateau point, despite equal repetitions over the study period. Variables additional to repetition may therefore influence the process of habit formation and habit strength. It has been suggested that habit learning can be further aided by rewards (Lally & Gardner, 2013). The Associative-Cybernetic Model (De Wit & Dickinson, 2009) proposes that when performing an action in a given context (e.g. eating fruit at dinner table in the evening), the experience of a positive outcome of behaviour (e.g. the pleasurable taste of the fruit) facilitates learning of the context–behaviour relationship (Wood & Neal, 2007). The model suggests that the reward value of behaviour (or its outcomes) should strengthen habit via two routes. First, reward should have an impact on habit that is mediated by behaviour repetition: repeated sequential presentation of context, behaviour, and reward is hypothesised to lead to the context signalling both an opportunity for action and an incentive to act, so increasing its value as a cue to action above that imbued by the simple pairing of context and behaviour regardless of consequences (De Wit & Dickinson, 2009). Thus, greater rewards should provide greater incentive to repeat the behaviour, thereby increasing the likelihood of context-dependent performance on subsequent occasions. Second, reward should moderate the relationship between repetition and habit strength. Experiencing rewarding outcomes each time the behaviour is performed in the context is predicted to strengthen the context–behaviour association more than it would were the behaviour unrewarded. Reward should thereby facilitate quicker learning of the underlying context–behaviour associations, and so rewarding behaviour should lead to stronger habits than repetition of less rewarding behaviours (De Wit & Dickinson, 2009). When considering the impact of reward on repetition and habit strength, a distinction must be drawn between extrinsic and intrinsic rewards. Tangible extrinsic rewards (e.g. financial incentives) may not be conducive to predict habit strength (Dickinson, 1985): external rewarding can lead performance to be dependent on the provision of such a reward (Colwill & Rescorla, 1985). Thus, discontinuing the provision of rewards may prompt disengagement from action (Wood & Neal, 2009) and consequently impede habit formation. Intrinsic rewards, i.e. retrospective evaluations of the reward value of a certain action derived from past performance of the behaviour itself (e.g. satisfaction or pleasure), may, however, have a positive impact on habit. In Lally et al.’s (2010) study, habits formed despite no tangible behaviour-contingent rewards being offered to participants, and in another study, physical activity habits developed among rehabilitation © 2013 The International Association of Applied Psychology
122
WIEDEMANN ET AL.
patients without extrinsic rewards (Fleig, Lippke, Pomp, & Schwarzer, 2011). In both studies, participants pursued self-chosen behavioural goals, and so may have derived intrinsic rewards (e.g. pleasure or satisfaction) from performing a desired action. Elsewhere, a study of dental flossing habit formation found that positive outcome expectancies not only predicted greater behavioural repetition, but also stronger habit when controlling for repetition frequency (Judah et al., 2013). Intrinsic reward can be a special case of a positive outcome expectancy when the outcome is defined as anticipated reward inherent within behavioural performance. Accordingly, Judah et al. suggested that rewards may have aided habit development, because outcome expectancies provided a proxy indicator of the intrinsic reward value of flossing. The direct effect of outcome expectancies on habit strength was unexpected, and no moderating effect of expected outcomes on the repetition–habit strength link was found. These findings may have been due to methodological limitations arising from the small study sample (N = 50), and limited variation in behaviour frequency (Judah et al., 2013). Elsewhere, a cross-sectional study of physical activity habits found that, of participants who were frequently physically active, those also motivated by intrinsic rewards reported stronger habits (Gardner & Lally, 2013). Gardner and Lally suggested that this finding may reflect intrinsic rewards having more absolute value than extrinsic rewards, and so, in line with the Associative-Cybernetic Model, greater reward may reinforce the behaviour–habit strength relationship.
The Present Study This study offers a test of the Associative-Cybernetic Model (De Wit & Dickinson, 2009) in relation to FV consumption. The model predicts that intrinsic rewards will determine habit strength in two ways. First, intrinsic rewards may increase the likelihood of repetition, such that repetition will mediate the influence of intrinsic reward on habit. Second, intrinsic rewards may enhance the effect of each repetition on habit development, thus moderating the relationship between repetition and habit strength. There is some tentative empirical evidence that positively valued behavioural outcomes lead to stronger habits (Gardner & Lally, 2013; Judah et al., 2013). To our knowledge however, no study has explicitly explored the relationship between intrinsic rewards and acquired behavioural automaticity (i.e. habit) of health behaviours. Following the Associative-Cybernetic Model, we hypothesised that: Hypothesis 1: Perceived intrinsic reward (measured at Time 1 [T1]) will influence habit strength of FV consumption (T3) indirectly, through its influence on (T2) FV consumption frequency (i.e. mediation). © 2013 The International Association of Applied Psychology
INTRINSIC REWARD PREDICTS FRUIT AND VEGETABLE HABIT
123
Hypothesis 2: Perceived intrinsic reward (T2) will strengthen the relationship between FV consumption frequency (T2) and habit strength (T3), such that where FV consumption is viewed as more rewarding, behaviour will be more predictive of habit strength (i.e. moderation).
METHODS
Participants and Procedure A study design with three measurement points with two-week intervals was used. Recruitment of participants took place in adult physical education classes in Germany (e.g. yoga, spinal exercises; no diet or weight loss programmes). This sample was chosen because we expected that their motivation to engage in health-promoting behaviour would be strong and so fruit and vegetable consumption would be likely. Research assistants approached individuals at the education centre within the first class, and explained the study. Participants were eligible for inclusion only where they gave informed consent, were aged 18 years or above, and had no self-reported medical conditions conflicting with health recommendations for dietary behaviour. Participants received entry into a prize draw for health-related products (e.g. cookbooks). A total of 127 individuals completed baseline paper-and-pencil questionnaires on-site after class and were invited to complete follow-up measures two weeks (T2) and four weeks (T3) later in the same course. The baseline sample had a mean age of 31.7 years (SD = 10.1; range 20–66 years), comprised 74.0 per cent women, and 37.1 per cent were in a long-lasting relationship. The majority of the participants reported 10 or more years of schooling (91.4%). Between 0 and 3 participants chose not to relay specific socio-demographic information. On average, individuals in this study had moderate intentions to consume the recommended number of servings of FV intake at baseline (M = 3.50; SD = 1.38). Follow-up data were available from 53.5 per cent (n = 68) of the participants at T2 and of 48 per cent (n = 61) of the participants at T3. The study was approved by an Institutional Review Board and conducted in line with the German Psychological Society ethical guidelines.
Measures At the top of the first page of all questionnaires, one serving of fruit or vegetables was defined as the amount of food that fits into the palm of the hand. A statement was also provided that rice and products made of potatoes should not be regarded as fruit or vegetables. Unless otherwise stated, response options were on 6-point Likert scales, ranging from “completely disagree” (1) to “completely agree” (6). Item examples reported below are translations from German. © 2013 The International Association of Applied Psychology
124
WIEDEMANN ET AL.
Intrinsic reward at T1 and T2 was assessed with the item: “Thinking about last week’s fruit and vegetable intake, please indicate how much you agree with the following statement: The consumption of fruit and vegetables in itself is rewarding for me.” Intention at T1 was measured with a single item: “I intend to consume 5 servings of fruit or vegetables a day.” Similar items were used in previous studies as single items or within multi-item scales with high internal consistencies and predictive validity (Schwarzer et al., 2007; Wiedemann, Lippke, Reuter, Ziegelmann, & Schüz, 2011a). Self-efficacy for FV consumption at T1 was measured with four items, capturing different facets of the perceived ability to consume 5 servings of fruit or vegetables a day, namely task-specific, initiation, maintenance, and recovery self-efficacy (Cronbach’s α = .75; Ochsner, Scholz, & Horning, 2013; Wiedemann, Lippke, Reuter, Ziegelmann, & Schwarzer, 2011b), e.g. “I am confident that I can consume 5 servings of fruit or vegetables a day for a prolonged period of time even if it is difficult sometimes.” Fruit and vegetable consumption at T2 was assessed with the item “How many servings of fruit and vegetables did you eat on an average day in the last week?” with an open-ended format. A similar single-item measure of FV intake has been validated against dietary biomarkers (Steptoe et al., 2003). Habit strength at T3 was assessed with the Self-Report Behavioural Automaticity Index (SRBAI), an automaticity-specific abbreviation of the SelfReport Habit Index (Verplanken & Orbell, 2003), which has been shown to have good content and predictive validity (Gardner, Abraham, Lally, & de Bruijn, 2012). The stem “Consuming 5 servings of fruit or vegetables is something . . .” was followed by the four SRBAI items: “. . . I do automatically”, “. . . I do without having to consciously remember”, “. . . I do without thinking”, “. . . I start doing before I realise I’m doing it”, Cronbach’s α = .97.1 All analyses were conducted with missing data of the baseline sample (N = 127; ≤ 5% item-non-response) imputed longitudinally using the expectation maximisation algorithm (Enders, 2001), which has proven more robust than regression imputation (Gold & Bentler, 2000).
1 Habits are characterised by their stability over time (Rothman et al., 2009; Wood & Neal, 2007). Hence, we also measured habit strength using the SRBAI at T1 (α = .95) to permit a preliminary analysis of habit stability over two time points, and a strong correlation was found (r = .66, p < .001). We use T3 habit strength as the outcome variable in our analysis to better capture the temporal sequence of expected relationships, and so do not discuss the T1 habit strength measure further.
© 2013 The International Association of Applied Psychology
INTRINSIC REWARD PREDICTS FRUIT AND VEGETABLE HABIT
125
Statistical Methods Both hypotheses on indirect and moderating effects of reward on habit strength were tested in one integrative model: the indirect effect of intrinsic reward on habit strength, as mediated by the frequency of FV consumption (Hypothesis 1), was tested simultaneously with the moderating effects of reward on the behaviour–habit relationship (Hypothesis 2) in a single moderated mediation model. More precisely, the indirect effect of intrinsic reward (independent variable) via frequency of FV consumption on habit strength (dependent variable) was tested as a function of intrinsic reward (continuous moderator) using two ordinary least square regression analyses, controlling for intentions and self-efficacy. The conceptual model is depicted in Figure 1. Age and gender were not included as covariates in the model. No significant bivariate associations between age and any of the model variables were found (p > .05). Significant correlations between gender and model variables were of a small effect size (r = .21 to .26) for reward (T1/T2), FV consumption T2 and habit strength T3, but relationships proved non-significant in regression models. In one regression analysis (i.e. the mediator model), frequency of FV intake (T2) was predicted by baseline T1 intrinsic reward, intentions, and selfefficacy. In the second regression (i.e. the dependent variable model), habit strength (T3) was predicted by T2 intrinsic reward, FV consumption, their interaction term, and the T1 covariates intentions and self-efficacy. To investigate the conditionality of the indirect effect, the strength of the indirect effect of reward on habit strength was tested at different values of the continuous moderator observed in the data set. Bootstrapping was applied to generate bias-corrected and accelerated confidence intervals (CIBCA) from
Reward T2
FV ConsumptionT2
FV Consumption T2 * Reward T2
Reward T1
Habit Strength T3
Self-Efficacy T1 Intention T1
FIGURE 1. Conceptual Model: Conditional indirect effect of reward T1 on habit strength T3 via fruit and vegetable consumption frequency T2 as function of reward T2 as moderator. © 2013 The International Association of Applied Psychology
126
WIEDEMANN ET AL.
5,000 resamples with α = .05 as indirect effects tend to have a non-normal sampling distribution (Preacher & Hayes, 2004). Analyses were conducted using the PROCESS macro for SPSS (model 14; Hayes, 2012).
RESULTS
Descriptive Results Sample Characteristics. On average, individuals in this study consumed 2.99 servings of fruit or vegetables (SD = 1.17), and perceived their FV intake as quite intrinsically rewarding on average (M = 3.66; SD = 1.43). Selfefficacy for FV intake was quite high (M = 4.30; SD = 0.89). Habit strength of FV consumption averaged 3.43 (SD = 1.52). At T3, individuals consumed 3.61 servings (SD = 0.95) of fruit or vegetables and reported a habit strength of FV consumption of 3.52 (SD = 1.10).
Attrition Analysis Attrition analyses compared participants retained until T3 and those lost after T1 and T2, respectively, using analyses of variance (ANOVAs) for continuous measures, and χ2 tests for categorical measures. No significant baseline differences were identified regarding FV consumption, intentions and self-efficacy for FV intake, habit strength regarding FV intake, reward, age, education status, marital status, and gender (p > .05).
Correlations between Model Variables and Covariates Intercorrelations between intrinsic reward, FV consumption, habit strength, as well as potential covariates (intentions, self-efficacy) were tested as a precondition for the main model test. Reward at T1 was positively associated with FV consumption (T2) and habit strength (T3), and FV consumption and habit strength correlated positively (see Table 1). Reward at T2 was also positively associated with FV consumption and habit strength. Intentions and self-efficacy correlated with all model variables, except a non-significant association between self-efficacy and reward (T1). Thus, potential covariates were controlled for in model tests.
Direct, Indirect, and Moderating Effects of Intrinsic Reward on Habit Strength First, in the mediator model, frequency of FV consumption T2 was predicted by T1 intrinsic reward, intentions, and marginally self-efficacy (see Table 2). © 2013 The International Association of Applied Psychology
INTRINSIC REWARD PREDICTS FRUIT AND VEGETABLE HABIT
127
TABLE 1 Intercorrelations between Reward, Fruit and Vegetable Intake, Habit Strength, and Covariates
1. 2. 3. 4. 5. 6.
Self-efficacy T1 Intention T1 Reward T1 Reward T2 FV Intake (no. of servings) T2 Habit Strength T3
1.
2.
3.
4.
5.
— .40*** .13ns .29** .33*** .32***
— .29*** .35*** .50*** .35***
— .75*** .42*** .49***
— .51*** .53***
— .72***
Note: N = 127. * p < .05; ** p < .01; *** p < .001;
ns
= non-significant.
Variables in the model explained 35.0 per cent of the variance in FV consumption, F(3, 123) = 22.52, p < .001.2 In the subsequent dependent variable model, strength of FV consumption habits was predicted by the frequency of FV consumption, baseline intrinsic reward, the interaction term between T2 FV consumption and T2 intrinsic reward, and, marginally, T2 self-efficacy, but not by main effects of T2 intentions and T2 reward (see Table 2). The direct effect of baseline reward on T3 habit strength was significant (B = .18, p = .01). Significance tests at all values of the reward scale supported intrinsic reward having a significant indirect effect on habit strength through the frequency of FV intake, supporting Hypothesis 1 (see Figure 2). A probe of the significant interaction indicated that the influence of baseline reward on habit strength via behaviour was amplified by the perceived reward value of FV consumption at T2 (see Table 2): The more consumption was perceived as intrinsically rewarding, the larger its indirect effect on habit strength (see Figure 2). For example, at low reward values (reward = “2” on a 6-point Likert scale), the indirect effect on habit strength was B = 0.09 (CI 0.03, 0.21), at moderate values (reward = “4”) it was B = 0.15 (CI 0.07, 0.26), and at high values on the reward measure (reward = “6”) the indirect effect was B = 0.20 (CI 0.08, 0.37). In total, 60.0 per cent of the variance in habit strength regarding FV consumption was explained by the model, F(6, 120) = 30.24, p < .001. Hypothesis 2 was supported.
DISCUSSION This three-wave study investigated predictions from the AssociativeCybernetic Model (De Wit & Dickinson, 2009). Results supported the 2 Analyses based on the longitudinal sample (i.e. excluding dropouts) led to the same pattern of results with only small changes in coefficients, and thus are not reported here.
© 2013 The International Association of Applied Psychology
128
WIEDEMANN ET AL. TABLE 2 Moderated Mediation Analysis with Unstandardised Parameter Estimates, Using Reward as Moderator of the Reward–Behaviour Frequency–Habit Relationship Prediction of FV Consumption T2 (mediator)
Predictor
B
SE
Reward T1 0.19 0.05 Intention T1 0.24 0.06 Self-Efficacy T1 0.16 0.08 Model summary: R2 = .35; F(3, 123) = 22.52, p < .001
T
p
3.91 4.34 1.92
