Investigating the effectiveness of self-control and habit strength based interventions in reducing discretionary food consumption Open Access

More than two-thirds (65.8%) of adults in Australia are living with overweight or obesity (Australian Bureau of Statistics, 2022). Discretionary foods are defined as foods that are not nutritionally beneficial but can add variety to one’s diet (National Health and Medical Research Council, 2013). This study will explore the use of behaviour change techniques aiming to reduce consumption of discretionary food. These foods are high in saturated fats, added sugars and salt (Rangan et al., 2009), and are most often, but not exclusively, consumed outside of meal times as snacks (Ovaskainen et al., 2006). These foods include crisps (chips), processed meats, lollies (candy/sweets) and pastries (National Health and Medicine Research, 2013). Consuming high amounts of discretionary food (more than one serving size a day) can contribute to overweight and obesity and can put individuals at risk of non-communicable diseases such as cardiovascular disease, neurological diseases, chronic respiratory diseases and digestive disorders (Australian Institute of Health and Welfare, 2023; World Health Organization, 2024). Despite awareness of risks of overconsumption, discretionary foods make up more than one-third (35%) of the average Australian’s total daily energy intake (Australian Bureau of Statistics, 2022). Given overweight and obesity in Australia exceeds the global average, research specific to the Australian population is needed to identify strategies for reducing discretionary food consumption (Australian Bureau of Statistics, 2022).

Current theoretical models such as Ajzen’s theory of planned behaviour (1991) posit that psychological factors influencing engaging in health behaviours currently focus on the rational aspects of behaviour, which assumes health behaviours are primarily driven by intention (Lippke and Ziegelmann, 2008). However, intention alone may not be sufficient to change behaviour (Hall and Fong, 2007) evidenced by Collins and Mullan (2011) who found various psychological constructs accounted for 42% of the variance in the intention to snack, however, only 28% of the variance in snacking behaviour. This, and other similar studies (McDermott et al., 2015; Sniehotta et al., 2005) shows a noticeable gap between the intention (conscious process) to snack and actual snacking behaviour.

Temporal self-regulation theory was proposed by Hall and Fong (2007) to close this intention-behaviour gap by considering the role of non-conscious processes (such as habit) and additional conscious processes (such as self-control) in predicting behaviour (Hall and Fong, 2007). Temporal self-regulation theory proposes two post-intentional variables. The first is behavioural prepotency, which is the automatic response to a behaviour, consisting of past behaviour and habit strength (Adriaanse et al., 2014). The second is self-regulatory capacity which is an individual’s ability to steer attention away from undesired tendencies and toward behaviour aligned with their goals (Hofmann et al., 2008). According to the theory, these two variables directly influence behaviour and moderate the intention behaviour gap, as shown in Figure 1 (Hall and Fong, 2007).

Temporal self-regulation theory has been applied to predicting various behaviours, such as snacking, sugar-sweetened beverage consumption, alcohol intake and adherence behaviours (Black et al., 2017; Evans et al., 2017; Liddelow et al., 2020; McAlpine et al., 2023; Moran and Mullan, 2021; Murray et al., 2020). In adult populations, self-regulatory capacity and behavioural prepotency have explained significant variance in intention and unhealthy snack intake (Evans et al., 2017; Moran and Mullan, 2021). While the theory has been shown to predict behaviour, its role in changing behaviour has not been as extensively explored.

Temporal self-regulation theory may be practically applied to change behaviour through interventions that target it’s constructs known as behaviour change techniques (Michie et al., 2013). Behaviour change techniques are specific methods that are part of interventions to change behaviour (Michie et al., 2011). They are considered the smallest possible intervention to observe a change in behaviour and therefore, theoretically the most convenient to apply to one’s life (Michie et al., 2011). There is inconsistency in the literature regarding which behaviour change techniques are effective in changing behaviour. One potential reason for this is that studies commonly utilise multiple behaviour change techniques to target pre and post-interventional predictors (Mairs and Mullan, 2015). Consequently, a limitation arises in determining what single or combination of behaviour change techniques are responsible for the observed change in behaviour (Mairs and Mullan, 2015). Charlesworth et al. (2023) found that exposing participants to four behaviour change techniques in the context of safe food-handling did not lead to more significant behaviour change compared to merely participating in the study. However, since all participants were exposed to all four behaviour change techniques, it remains unclear what particular behaviour change techniques were not effective (Charlesworth et al., 2023). Additionally, there is conflicting evidence suggesting which behaviour change techniques target habit strength and which target self-control. Thus, it would be beneficial to investigate these constructs and behaviour change techniques separately in regard to their effectiveness in changing behaviour.

Evidence suggests that a behaviour change technique known as implementation intentions, which targets setting goals and planning, should be considered in increasing self-control (Cross and Sheffield, 2019). Implementation intentions involve creating an “if-then” plan to achieve a desired goal (Cross and Sheffield, 2019). For example, an implementation intention targeting discretionary food consumption could be “if I am at a social gathering where there is discretionary food, then I will get a glass of water and situate myself away from the table.” Implementation intentions have been effective in changing behaviours such as time management (Oettingen et al., 2015), physical activity (Robinson et al., 2019) and eating behaviours (Carrero et al., 2019). Understanding the role implementation intentions have in increasing self-control is essential, as if effective, they offer a practical strategy that could be applied to the population, potentially leading to significant behaviour change.

Restructuring the environment is another behaviour change technique that is effective in changing health behaviours including food safety behaviour (Mullan et al., 2014), reducing sedentary behaviour (White et al., 2017) and weight management (Lally et al., 2008). Restructuring the environment involves avoiding exposure to specific contextual cues for the undesired behaviour (Gardner et al., 2020). Restructuring the environment is effective for habit discontinuity, which is ceasing exposure to cues that elicit habitual responses (Gardner et al., 2020; Verplanken and Roy, 2016). Furthermore, environment restructure identifies the context-specific cues that trigger the habitual response (e.g. seeing snacks in the cupboard) and can consequently interfere with the execution of the habitual cue-triggered responses thereby aiding in habit disruption over time (Gardner et al., 2020). Investigating the effectiveness of restructuring the environment may provide a practical strategy to address automatic responses to cues which may be driving of the consumption of discretionary food.

Whilst temporal self-regulation theory has shown to predict health behaviour, it has been less explored in changing health behaviours. Given that research has found both self-control and habit strength to be important in predicting health behaviours (Dominguez Garcia et al., 2023; McAlpine and Mullan, 2022), interventions that may target these constructs should be refined and explored. The aim of the current study was to determine if individual behaviour change techniques targeting self-control and habit (i.e. implementation intentions or restructuring the environment) were effective for decreasing discretionary food consumption.

The inclusion criteria for our study were people over the age of 18 who were currently residing in Australia who regularly consume discretionary food (more than one serve a day). There were no specific exclusion criteria. The final sample size of 152 participants consisted of 51 participants in the self-control intervention group, 48 participants in the habit strength intervention group and 53 participants in the control group. A sensitivity power analysis was run which indicated that with a sample size of 152, an alpha level of 0.05 and power of 80%, this study is sensitive enough to detect an effect of 0.15 (Faul et al., 2007). The majority of participants identified as women (61.84%) with a small proportion identifying as non-binary (1.3%). The age range of participants was 19–67 years (M = 34 SD = 11.66 years). A large portion of the sample were living with partners (48.68%) and had a bachelor’s degree (42.11%). The highest proportion of participants were from Victoria (36.18%) and New South Wales (26.32%) with the rest of participants being relatively evenly split across states. Means and standard deviations for variables at both time points can be seen in Table 1.

Ethics approval was obtained from Curtin University Human Research Ethics Committee. Participants were recruited through Prolific (https://www.prolific.com/), a UK-based crowd sourcing survey platform (Palan and Schitter, 2018). Data were collected in April 2024.

Participants completed a baseline questionnaire including measures of intention, habit strength, self-control and discretionary food consumption over the last two days. Participants were then randomly assigned into one of three groups: (1) self-control intervention task, (2) habit strength intervention task or (3) control group. The control group did not receive an intervention task and only completed the baseline measures. At time two, one week later, participants repeated the same questionnaire as they did at time one, excluding the intervention task.

Participants reported their age, gender, current living situation, highest level of education and where they currently reside (i.e. Australian State).

Intention to decrease discretionary food intake was evaluated using a single item based on the theory of planned behaviour (Ajzen, 1991). Using a seven-point Likert scale ranging from “strongly disagree” to “strongly agree” (1 = strongly disagree, 7 = strongly agree), participants responded to the prompt “I intend to limit my intake of discretionary snack foods”. Higher scores on the item indicated stronger intention to limit discretionary food consumption.

Habit strength was assessed using the self-report behavioural automaticity index (Gardner et al., 2012) which had four items rated on a seven-point Likert scale ranging from “strongly disagree” to “strongly agree” (1 = strongly disagree, 7 = strongly agree), participants responded to the prompt “eating discretionary foods is something” with an example item being “I do automatically.” Higher scores indicated a stronger habit for consuming discretionary food. The habit strength score is obtained as an average of the four items. This measure demonstrates strong internal consistency (α = 0.90; Evans et al., 2017) which is consistent in this sample (time-one Cronbach’s α = 0.92, time-two Cronbach’s α = 0.94). Good convergent validity has been shown when comparing the measure to the validated self-report habit index (Gardner et al., 2012; Verplanken and Roy, 2016). Furthermore, it has demonstrated strong predictive validity, showing high correlations with behaviours such as alcohol consumption and unhealthy snacking (Gardener et al., 2012; Gardner et al., 2020).

Self-control was measured using a modified version of the self-regulation of eating behaviour questionnaire (Kliemann et al., 2016). The measure was modified from general eating self-control to be specific to discretionary snack food consumption. There was one prompt; “please indicate how often the below applies to you in relation to eating discretionary snack foods.”, followed by five items (e.g. “I am good at resisting discretionary snack foods”). Participants used a five-point Likert scale ranging from “never” to “always” (1 = never, 5 = always) to respond to items. Higher scores indicated a stronger level of self-control to resisting discretionary food consumption. The self-control score is obtained as an average of the four items. This measure displays acceptable internal consistency (α = 0.75; Kliemann et al., 2016) which was consistent in this study across time points (time-one Cronbach’s α = 0.69, time-two Cronbach’s α = 0.69). The measure shows good discriminant validity, indicated by weak associations with food selectiveness, satiety and eating speed (Kliemann et al., 2016).

Discretionary food consumption over the preceding two days was measured by participants completing an adapted version of the timeline follow-back questionnaire (Sobell and Sobell, 1992). Participants recorded their serves of discretionary foods over the last two days in a calendar format. Participants were also asked to recall and record what was considered a “usual” and “not usual” day that may have changed their normal eating routine and classify anything unusual as a “special event” for example, feeling sick on a particular day or attending a birthday party, to encourage recall and help to report behaviour more accurately. A mean score of their servings over the preceding two days was then calculated. The timeline follow-back is a reliable self-report tool often used in health behaviour research showing good test-retest reliability (r = 0.85; Sobell and Sobell, 1992).

The self-control intervention group was provided with a self-control intervention. The implementation intention was chosen based on the behaviour change taxonomy that aims to change self-control (Michie et al., 2013). The task required participants to create an implementation intention to reduce their discretionary food consumption over the next week in the form of an “if, then” statement. They were provided with an example for someone whose intention it is to spend less time sitting, “If I have been sitting for longer than 1 h, then I will get up and walk around the office.” Participants were asked to write their implementation intention in a table and were encouraged to keep a reminder of it for the next seven days such as, writing it in their notes app or setting it as their wallpaper.

The habit strength intervention group were provided with the habit strength intervention task. Restructuring the environment was chosen based on the behaviour change technique taxonomy that aims to decrease habit strength surrounding discretionary food consumption (Michie et al., 2013). This included a prompt for participants to think about how to restructure their environment to reduce their discretionary food consumption over the next week. They were provided with an example of someone restructuring their environment to reduce their consumption of alcohol, “When I am at home during the week, I will keep alcohol at the back of the cupboard where I can’t see it.” Participants were asked to write their statement in a table and were encouraged to keep a reminder of it for the next seven days such as, writing it in their notes app or setting it as their phone wallpaper.

All assumption testing and data analysis was conducted in IBM SPSS (version 29). Data were initially screened for missing values and outliers. Normality was tested through visual inspection of boxplots and Q–Q plots. To check normality of residuals, Shapiro-wilk’s tests were examined for all variables. Box’s Test of Equality of Covariances and Levene’s Test of Equality was used to test homogeneity of variance, which are assumptions of a mixed between-within subjects’ analysis of variance (ANOVA). A mixed between-within ANOVA was run to determine changes in intention, self-control, habit strength and discretionary food consumption following the intervention.

There was no significant main effect for time F (1, 149) = 2.10, p = 0.149, η_p² = 0.01, nor for group F (1, 149) = 0.70, p = 0.498, η_p² = 0.01. Neither was there a significant interaction between time and group F (1, 149) = 0.70, p = 0.501, η_p² = 0.01. As behaviour did not significantly change over time, follow-up mediation analyses examining whether behaviour changed as a result of changes in habit strength and self-control were not conducted (Mergelsberg et al., 2021). Results are show in Table 2.

There was no significant main effect for time F (1, 149) = 2.94, p = 0.089, η_p² = 0.02, nor for group F (1, 149) = 2.94, p = 0.089, η_p² = 0.02. Neither was there a significant interaction between time and group F (1,149) = 1.58, p = 0.089, η_p² = 0.02. Results are shown in Table 3.

There was no significant main effect for time F (1, 149) = 1.59, p = 0.209, η_p² = 0.01, nor for group F (2, 149) = 2.61, p = 0.077, η_p² = 0.03. Neither was there a significant interaction between time and group F = (2, 149) = 1.21, p = 0.302, η_p² = 0.02. Results are shown in Table 4.

There was a significant main effect for time F (1, 149) = 16.25, p < 0.001, η_p² = 0.10 whereby intention scores were significantly higher at time two (M = 5.14, SD = 1.31) compared to time one (M = 4.28, SD = 1.47) among all participants. However, there was no significant main effect for group F (2, 149) = 1.49, p = 0.230, η_p² = 0.02, nor was there a significant interaction between time and group F (2, 149) = 0.25, p = 0.722, η_p² = 0.004. Results are shown in Table 5.

The current study found no change in behaviour, nor the key moderators proposed in temporal self-regulation theory over time. However, a change in intention was found over time in all groups. Although previous studies have supported the predictive utility of temporal self-regulation theory (Evans et al., 2017; Grieger et al., 2016) the results from this study do not support its use to change behaviour. To the authors knowledge, this study is the first to examine the constructs of temporal self-regulation theory in changing discretionary food consumption with an experimental design which directly addresses the limitations to previous studies adopting a predictive design.

Our results support an intention behaviour gap whereby all groups exhibited moderate to high levels of intention at both time one and time two, indicating that despite intending to the decrease discretionary food consumption, behaviour did not change (Conner and Norman, 2022). Additionally, intention increased in all groups over time, supporting a mere measurement effect, whereby a behaviour can improve as a result of completing questions relating to it (Godin et al., 2008). The mere measurement effect is commonly observed in health research and has been recognised in other interventions (Charlesworth et al., 2023; Mergelsberg et al., 2021).

Discretionary food consumption did not change in our study. Previous studies have found support for effective strategies to reduce discretionary food consumption such as reducing portion size (Grieger et al., 2016). However, studies that attempt to target the underlying construct driving consumption are less explored. An explanation for behaviour not changing in the current study may be that changing eating behaviours is complex due to their multistep nature and the balance between distal and proximal rewards associated with snacking. Changing eating behaviours may be more complex whereby it involves not only reducing discretionary food consumption, but decision making about healthier alternatives, for example, planning and preparing food and resisting readily available tempting foods. Additionally, the delayed gratification of implementing a healthy diet such as reducing disease risk (Australian Institute of Health and Welfare, 2023) does not offer immediate reward and hence may decrease motivation to sustain long term behaviour change (Dorina et al., 2024; Kaushal and Rhodes, 2015). The aim of the study was to target snack consumption with single behaviour change techniques that have been shown to be effective in other behaviours (Oettingen et al., 2015) however, based on the results of the current study, future studies may wish to explore strategies for targeting the multistep nature of eating behaviours and both distal and proximal rewards in this context.

Habit strength did not change in any group. This is unexpected and diverges in the literature in that habit strength has shown to be a predictor of discretionary food consumption (Dominguez Garcia et al., 2023) hence it was hypothesised that habit strength would decrease in the habit strength group. Despite the predictive support for habit strength, decreasing discretionary food consumption via targeting habit strength may be more difficult as it requires both the suppression of the undesired response (reducing snacking) and the substitution of the unwanted response with a healthier choice (eating an apple in place of a chocolate bar) in response to a hunger cue (Gardner et al., 2020).

The habit intervention in our study focused on breaking the habit of snacking through environmental restructuring by removing automatic exposure to discretionary food, which is grounded in habit theory (Gardner et al., 2023). However, the lack of change in behaviour and habit strength in this study diverges from what is proposed in habit theory. Behaviour change techniques that focus on replacing an unwanted behaviour with a new behaviour and in turn discontinuing the unwanted habit such as substitution may be more applicable to eating behaviours (Gardner et al., 2020; Michie et al., 2013). In the context of discretionary food consumption, this approach would involve replacing unhealthy snacks with healthier alternatives. Previous research has been successful in building habit to consume more fruits and vegetables (Rompotis et al., 2014). Future research may consider comparing a sample focused on decreasing discretionary food consumption with two subsets: one subset focusing on avoiding the behaviour (environment restructure) and the other replacing the behaviour (substitution) to see if this is more effective in reducing discretionary food consumption.

The results showed no change in self-control in any group. Past research has been mixed in whether self-control is a significant predictor of health behaviour. McAlpine and Mullan (2022) and Evans et al. (2017) found self-control was not a significant predictor of sugar sweetened beverage consumption and eating behaviours respectively. However, self-control has shown to be significant in increasing physical activity (Robinson et al., 2019) and promoting healthy eating behaviours (Carrero et al., 2019).

The lack of change in self-control in this study may be due to what aspect of self-control was targeted in the chosen intervention. Self-control remains hard to define, most likely due to its multifaceted nature (Duckworth et al., 2018). Self-control comprises of the ability to inhibit immediate impulses and planning to initiate and maintain behaviours hence, different elements of self-control may be more relevant in different behaviours. Inhibition may be more relevant for simple and hedonic behaviours, and planning is likely to be important in enacting complex and distal behaviours, requiring greater executive functioning and forethought to cease engagement in complex behaviours (McAlpine et al., 2024). In the current study, planning was targeted through implementation intentions as eating behaviour relies on forethought and goal-aligned decision making. However, it may be justified to target inhibition to change snacking as a means to target temptation when decreasing discretionary food consumption, known as inhibitory control. Inhibitory control training may be beneficial as it includes cognitive tasks aiming to improve regulation involved with a health behaviour (Allom et al., 2016). Studies have shown mixed results for inhibitory control training in its ability to change behaviour in the long term. Nonetheless, inhibitory control training has been effective for decreasing discretionary food consumption, alcohol consumption and sweet food cravings (Allom et al., 2016; Memarian et al., 2022). In isolation, both implementation intentions and inhibitory control training have their limitations as shown by our study and previous studies in achieving lasting behaviour change (Allom et al., 2016). Consequently, for future research, it may be beneficial to investigate a combined approach, targeting both inhibition and planning through inhibitory control training and implementation intentions to best enhance impulse control and planning for long-term behaviour change.

A strength of our study was the self-control measure was specific to discretionary food consumption, as it focused on participants’ ability to resist discretionary food intake, rather than assessing general self-control. Past research has been criticised for using varying general measures of self-control (Suchy, 2009), which may lead to inconsistent results, as different measures may measure different aspects of self-control (Dorina et al., 2023). Additionally, since most measures of habit strength and intention are specific to discretionary food consumption, it creates consistency in the measures and more accurate representations of perceptions towards discretionary food consumption. Another strength of the current study is that it is unique in that most of the current habit strength literature has focused on building desirable habits rather than breaking unwanted habits (Gardner et al., 2020). Our study contributes to the need for research to focus on the reduction and avoidance of health-risk behaviours, that is reduction of discretionary food consumption (Gardner et al., 2023). By focusing on reducing discretionary food consumption, we focused on addressing a health-risk behaviour which often receives less targeted attention.

Despite the lack of behaviour change, the implications of our study are nonetheless important. Our findings suggest that simple interventions may not be effective for changing behaviour that occurs less often. A limitation of our study was the low reported means of behaviour. An inclusion criterion for our study was regular consumption of discretionary food however, the mean was low across all groups at baseline (M < 3). This was lower than what government statistics would indicate (M = 4.3 serves per day; Australian Bureau of Statistics, 2014) but is consistent with research in similar areas (Dominguez Garcia et al., 2023; Verhoeven et al., 2012). It is possible that it may be harder to decrease behaviour when it is not performed regularly. For example, it may be that people do not see anything wrong with consuming two serves of discretionary food per day and the consequences that it may have for their health in comparison to consuming five serves per day, when more than one serve a day is considered a health risk (Australian Bureau of Statistics, 2014). Consequently, supplementary interventions may be needed. For example, a brief intervention based on self-determination theory has been successful in smoking cessation which includes a quick conversation with individuals outlining the dangers of smoking and discussing barriers in smoking cessation (Li et al., 2020). A similar approach may be applied to discretionary food consumption whereby individuals are made aware of the risks associated with consumption, even at moderate levels and are supported in identifying any personal barriers and personalised strategies to reduce intake. With this, further research should also consider higher intakes of discretionary food and targeting lower socioeconomic populations who are more likely to consume discretionary food as such in our study, more than two-thirds of the sample had at least a bachelor’s degree (Si Hassen et al., 2018; Van et al., 2014).

Additionally, our study aimed to separate behaviour change techniques to find the simplest intervention that would be easiest to implement in one’s lifestyle. However, it is possible that they should be used in combination in that behaviour change techniques may address multiple constructs and different elements of constructs. Although, research in this context is limited. Future research may consider a full factorial design, whereby different constructs are paired with different single, or combinations of behaviour change techniques to better understand the interaction of construct and behaviour change technique. Practically, the increase in intention due to a mere measurement effect indicates that people are engaged in their health behaviours and health research, highlighting the importance of further research investigating ways to make interventions effective in the population.

The current study found no changes in discretionary food consumption nor self-control or habit strength. However, intention did increase overtime in all three groups. This provides a starting point for testing temporal self-regulation theory with an experimental design. If future studies can target self-control and habit strength, these principles may change health behaviours and disease risk associated with high consumption of discretionary food. Although, our study was unable to change the target behaviours and hence detect any change in self-control or habit strength relating to discretionary food consumption, further research could explore intervention dosage across self-control and habit interventions, as well as longer duration.