Predicting discretionary food consumption using temporal self-regulation theory and food reward sensitivity

interventions to reduce discretionary food consumption. Past behaviour should be considered, and intention targeted in interventions to reduce discretionary food consumption.


Discretionary foods
Australian and American government dietary guidelines define discretionary foods as foods and drinks (including those consumed as part of a main meal) that have little nutritional value but are high in sugars, saturated fats and/or salt (National Health and Medical Research Council, 2013; U.S. Department of Agriculture & U.S. Department of Health and Human Services, 2020). These include crisps (chips), pastries, processed meats, lollies (candy/sweets), sugar-sweetened beverages and alcohol (National Health and Medical Research Council, 2013). Excess consumption of these foods is a risk factor for obesity, Type 2 diabetes, some cancers, and other diet related diseases (Johnson et al., 2017;Lal et al., 2020). Despite the known risks of overconsumption, discretionary foods account for over one third (35%) of the average Australian adult's total daily energy intake (Australian Bureau of Statistics, 2014) with similar dietary patterns in the United States, the United Kingdom, and Brazil (Sui et al., 2017).
Obesity is a growing problem in Australia, with almost two-thirds (63.4%) of adults and over a quarter (27.6%) of children living with overweight or obesity (Huse et al., 2018). Rates of obesity are also increasing globally with more than 124 million children estimated to be living with obesity in 2016, compared to 11 million children in 1975(Abarca-Gómez et al., 2017. In addition to the risks posed by overconsumption of discretionary foods, obesity increases the risk for many diseases including cardiovascular, liver and gallbladder disease, osteoarthritis, overall poor health and reduced quality of life (Williams et al., 2015). Even small reductions in discretionary food consumption and daily energy intake can result in major positive benefits to individuals and healthcare systems (Hall et al., 2011;Hill, 2009). It is estimated that replacing a single serving of discretionary foods with a healthier alternative per week could result in annual healthcare savings of up to AUD $793.4 million as a result of reduced rates of associated diseases (Lal et al., 2020).
Many health interventions with the aim of reducing discretionary food and/or beverage consumption have targeted conscious decisionmaking by educating the public on the health risks of overconsumption or by providing incentives for healthier behaviours (Herman & Polivy, 2011;Schubert et al., 2021). However, these interventions have had mixed results. While these interventions teach people the tools needed to make informed decisions around their food choices, health behaviours have also been shown to be influenced by unconscious processes. Therefore it is important to examine both conscious and unconscious processes of behaviour to better understand how to influence healthy eating behaviours (McAlpine & Mullan, 2022;.

Theoretical frameworks 1.2.1. Theory of planned behaviour
The theory of planned behaviour (Ajzen, 1991) has often been used to explore how individuals make food choices (McDermott et al., 2015;McKee et al., 2019) and suggests that the primary motivator for human behaviour is intention, where stronger behavioural intentions generally predict greater likelihood of behavioural engagement. However, while this theory has been effective at predicting behavioural intentions, intentions do not always predict behaviour (Sniehotta et al., 2014). This discrepancy between intentions and observed behaviour is referred to as the "intention-behaviour gap" and the theory of planned behaviour is unable to explain why this gap occurs or why it varies so greatly between different health behaviours (Sniehotta et al., 2005). Past research using the theory of planned behaviour suggests that intention is a strong predictor of healthy food choice, however intention is a weak predictor for limiting discretionary food consumption (McDermott et al., 2015). Therefore, it is important to explore which other factors drive discretionary food consumption to better understand why intention often does not result in performance of the intended behaviour.

Temporal self-regulation theory
Temporal self-regulation theory (Hall & Fong, 2007) was developed to explain the intention-behaviour gap and suggests that while intention is important, non-intentional factors such as behavioural pre-potency (how automatic that behaviour is) and self-regulatory capacity (how much self-control an individual has over the intended behaviour) are also influential predictors of behaviour (see Fig. 1). Hall and Fong (2007) suggest that the more automatic a behaviour is, the greater the self-control required to change it. Conversely, the more conscious thought a behaviour requires, the less self-control is needed to alter it. They also propose that there is a temporal component to motivation where intentions are stronger for behaviours with more immediate anticipated outcomes than behaviours with delayed anticipated outcomes. For example, as discretionary foods are often tastier than non-discretionary foods (Nansel et al., 2016), they may provide an immediate reward and individuals may be more tempted to consume them, whereas the health risks posed by overconsuming these foods are unlikely to be felt for some time so individuals may be less motivated to avoid eating them.
Behavioural pre-potency considers the frequency of that behaviour being executed in the past, habit strength and the effects of environmental cues (Hall & Fong, 2007). Habit strength is a strong predictor of unhealthy eating behaviours, and the influence of intention appears to weaken as habit strength increases (Evans et al., 2017;Mullan et al., 2016;Verhoeven et al., 2012). Similarly, increased frequency of past behaviour and environmental cues have also been associated with increased rates of unhealthy eating behaviours (Evans et al., 2017). Self-regulatory capacity has often been assessed using measures of self-control as both terms have often been used interchangeably (Elliston et al., 2017;Murray & Mullan, 2019). Higher self-control has been associated with lower unhealthy snack consumption in past studies (Adriaanse et al., 2014). Hall and Fong (2007) also propose that behavioural pre-potency and self-regulatory capacity moderate the relationship between intention and observed behaviour such that weaker behavioural pre-potency and stronger self-regulatory capacity should result in a stronger intention-behaviour relationship. They suggest that as highly automatic behaviours require less conscious thought, it should require less intention to perform. Conversely, if that behaviour has low automaticity, then it will require higher levels of intention to perform (Ouellette & Wood, 1998;Verplanken & Orbell, 2003). For example, unhealthy snacking is often a convenient behaviour that most people do not think about, mainly driven by automatic mechanisms such as past behaviour, habit, and cues. Conversely, eating fruit and vegetables is a more effortful behaviour, requiring knowledge, planning and the ability and resources to prepare them (Allom & Mullan, 2012;Evans et al., 2017).

Food reward sensitivity and discretionary foods
Food reward sensitivity is based on reinforcement sensitivity theory (Corr, 2004) which suggests that whether individuals engage in or avoid a behaviour is neurologically driven and is a result of appetitive or aversive stimuli respectively. Food reward sensitivity is another non-intentional factor that has been associated with greater consumption of discretionary foods, as these foods are designed to be maximally appetising through increased levels of sugars, fats, and/or salts (Nansel et al., 2016). Discretionary foods are so rewarding that exposure to pictures of these foods is enough to activate the reward centres of the brain (Stoeckel et al., 2008). The rewarding nature of discretionary foods has also been reported in qualitative interviews, with some of the largest motivators for eating discretionary foods being their taste, convenience and social influences (Farringdon et al., 2018). Increased sensitivity to the rewards these foods provide might explain the variations in hedonic hunger (the motivation to eat for pleasure above normal energy needs) between different individuals (Cappelleri et al., 2009;Nansel et al., 2016).
While studies have explored the influence of food reward sensitivity on discretionary food consumption (Nansel et al., 2016) and snacking (Schüz et al., 2015), it does not appear to have been studied in combination with other theories such as temporal self-regulation theory (Hall & Fong, 2007). As both temporal self-regulation theory and food reward sensitivity explore intentional and non-intentional motivators of behaviour, it was expected that assessing the interactions between these factors may explain more variance in behaviour than intention alone (see Fig. 2).

Research aims
This study aimed to examine the effectiveness of temporal selfregulation theory in predicting discretionary food consumption and explore whether including food reward sensitivity will offer additional predictive utility to the model, over and above intention. As temporal self-regulation theory has had partial success in predicting similar behaviours such as unhealthy snacking (Evans et al., 2017) and sugar-sweetened beverage consumption (McAlpine & Mullan, 2022; it was expected that the model can also predict discretionary food consumption. Further, food reward sensitivity may improve the utility of the temporal self-regulation theory model as this non-intentional motivator is associated with greater consumption of discretionary foods (Nansel et al., 2016).

Hypotheses
It was hypothesised that: H1. Intention to limit discretionary food consumption will account for significant variance in behaviour and be negatively correlated with discretionary food consumption.
H2. Behavioural pre-potency and self-regulatory capacity will account for significant variance in behaviour where behavioural pre-potency will be positively correlated with discretionary food consumption, while self-regulatory capacity will be negatively correlated with discretionary food consumption.

H3.
Food reward sensitivity will account for significant variance in behaviour and be positively correlated with discretionary food consumption.
H4. Behavioural pre-potency and self-regulatory capacity will significantly moderate the relationship between intention and behaviour, such that lower levels of behavioural pre-potency and higher levels of selfregulatory capacity will strengthen the association between intention and behaviour.

Participants and procedure
Convenience sampling was used to recruit 273 participants through social media (N = 225) and a university undergraduate participant pool (N = 48). An a-priori analysis using G*Power (Faul et al., 2009) showed that a minimum sample size of 131 participants was required (power = .8, α = 0.05, predictors = 13) to detect a moderate effect size (ƒ 2 = 0.15).
Participants recruited through the undergraduate participant pool received course credit points for completing both parts of the study.
The study was advertised through social media (Facebook, Twitter, Reddit and Instagram) and a university undergraduate participant pool conducted via Qualtrics, a survey hosting website. Participants provided informed consent by checking a box prior to starting the questionnaire. Participants then completed the first questionnaire asking about their past discretionary food intake, intention to limit intake of discretionary foods over the following week, habit strength, environmental cues, selfregulatory capacity, food reward sensitivity and the demographic questions. One week after completing the first questionnaire, participants were emailed the link to the second questionnaire which asked about their behaviour over the past two days. This study was approved by the Curtin University Human Research Ethics Committee (HRE2021-0463).

Intention
Intention to reduce discretionary food intake was measured using one item based on the theory of planned behaviour (Ajzen, 1991), "I intend to limit my intake of discretionary foods over the next week". Past studies have recommended using a single item to measure intention due to the high correlation between each of the original intention items, and to reduce participation burden (Charlesworth et al., 2021;McAlpine & Mullan, 2022). The item was answered on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree), with higher scores indicating stronger intention to limit discretionary foods.

Behavioural pre-potency
Past Behaviour. Participants were provided with a definition of discretionary foods from the Australian Dietary Guidelines: foods and drinks that are not an essential part of our diet and are high in saturated fats, sugar, salts or alcohol but low in nutrients (National Health and Medical Research Council, 2013). They were also provided with examples of discretionary foods and serving sizes to assist with conceptualising the serving sizes of the foods and drinks they had consumed (see Supplementary Material 1). An adapted version of the timeline follow-back questionnaire (Sobell & Sobell, 1992) was used to record participants' past discretionary food intake. Participants were asked to recall and record, presented in a calendar format, any special events on each day to assist with memory recall, and then to record the number of discretionary foods they consumed each day, over the previous two days. The timeline follow-back questionnaire is a valid self-report tool with good test-retest liability (r = 0.85; Sobell & Sobell, 1992) and has often been used in health behaviour research. Although researchers have traditionally used five days in past uses of this measure, the current study only recorded two days of behaviour due to memory recall errors inherent in dietary recall over longer periods of time (Jackson et al., 2008) to ensure participants could more accurately recall the discretionary foods they had consumed. While 24-h dietary recall interviews are often considered one of the most effective ways to reduce these memory errors, they are not practical for large cohort studies aiming to recruit large numbers of participants (Jackson et al., 2008;Schatzkin et al., 2003). A mean number of serves of discretionary foods per day was calculated.
Habit Strength. Two measures were used to measure habit strength. . This shortened version of the Self-Report Habit Index (Verplanken & Orbell, 2003) was adapted for this study to refer to discretionary foods and indicated discretionary food consumption-related habit strength. This measure has one central statement "Eating discretionary foods is something …" followed by four items (e.g., "That I do automatically") answered on a 7-point Likert scale (0 = strongly disagree, 6 = strongly agree). Scores were averaged with higher average scores indicating stronger habit strength to consume discretionary foods. This measure has good internal consistency (α = 0.90, Evans et al., 2017), which is similar to what was found in the current study (α = 0.94) and has also shown good convergent validity when compared to the Self-Report Habit Index Verplanken & Orbell, 2003). The measure has also shown good predictive validity, with high correlations with the frequency of behaviours such as, unhealthy snacking and alcohol consumption . (Ersche et al., 2017). This measure assesses individuals' general habitual tendencies and consists of two subscales. The first subscale measures 'routine' and contains 16 items (e.g., "I tend to go to bed at roughly the same time every night"). The second subscale measures 'automaticity' and consists of 11 items (e.g., "When walking past a plate of sweet biscuits, I can't resist taking one"). Items were answered on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree) and summed with higher total scores indicating stronger habitual tendencies. This measure has good internal consistency (α = 0.86-0.89; Ersche et al., 2017), which is similar to what was found in the current study (α = 0.86), and has also shown good convergent validity and discriminant validity, with high correlations with personality traits associated with differences in habit strength and weak correlations with other personality traits (Ersche et al., 2017).

Creature of Habit Scale
Environmental Cues. The Cues to Action Scale (Booker & Mullan, 2013) was adapted for this study to measure environmental cues which trigger discretionary food consumption. This measure consists of five subscales, each assessing one domain of cues (physical, sensory, internal, social, and emotional). Participants were given a description and example for each domain and were asked if there are any cues within that domain that trigger them to consume discretionary foods. If answering "yes", they were asked to answer one item (e.g., "How often do you experience these cues?"), on an 8-point Likert scale (0 = never, 7 = all the time) and a second item (e.g., "How much is experiencing these cues likely to make you eat discretionary foods?") answered on a 7-point Likert scale (0 = not at all likely, 6 = every time. Domain scores were created by multiplying the scores from each item within that domain. Domain scores were then summed to create an overall score, with higher scores indicating a stronger influence of environmental cues on the consumption of discretionary foods.

Self-regulatory capacity
Two measures were used to measure self-regulatory capacity. Self-Regulation of Eating Behaviour Questionnaire (Kliemann et al., 2016). This measure indicates food-related self-regulatory capacity. Participants were given written definitions for 'tempting foods' and 'eating intentions' and were then asked to answer five items (e.g., "I am good at resisting tempting food.") on a 5-point Likert scale (1 = never, 5 = always). Scores were summed, with higher total scores indicating more self-control around eating behaviours. This measure has good internal consistency (α = 0.75; Kliemann et al., 2016), which is similar to what was found in the current study (α = 0.75), and has also shown good convergent validity with positive correlations with behavioural automaticity and intention, and negative correlations with emotional over-eating and responsiveness to food cues (Kliemann et al., 2016). The measure has also shown good discriminant validity, with only weak associations with eating speed, food selectiveness and satiety (Kliemann et al., 2016).
Brief Self Control Scale (Tangney et al., 2004). This measure assesses general self-control using 13 items (e.g., "I am good at resisting temptation") answered on a 5-point Likert scale (0 = not at all, 4 = very much). Scores were averaged with higher scores indicating stronger general self-control. This measure has shown good internal consistency (α = 0.82; Evans et al., 2017), which is similar to what was found in the current study (α = 0.84), and has also shown good convergent validity, with strong correlations with the Total Self-Control Scale (Tangney et al., 2004). The measure has also shown good discriminant validity, with no significant associations with emotional exhaustion (Maloney et al., 2012).

'Food reward sensitivity'
'Food reward sensitivity' was measured using the Power of Food Scale (Cappelleri et al., 2009). This measure assesses individuals' responsiveness to nearby appetitive foods and has previously been used as a measure of food reward sensitivity (Appelhans et al., 2011;Nansel et al., 2016). This scale consists of three domains (Food Available, Food Present, and Food Tasted). Fifteen items (e.g., "I find myself thinking about food even when I am not physically hungry") were answered on a 5-point Likert scale (1 = do not agree at all, 5 = strongly agree). Scores for each domain were calculated by taking the average score of all the items within that domain, then a total score was created by averaging those domain scores. Higher total scores indicate increased food reward sensitivity. This measure has good internal consistency (α = 0.81-0.91; Cappelleri et al., 2009) which is similar to what was found in the current study (α = 0.91), and has also shown good convergent validity, with strong correlations with measures of overeating (Cappelleri et al., 2009). This measure has also shown great predictive utility, with higher scores correlating with stronger neurological responses (Yoshikawa et al., 2014) and higher self-reported food cravings (Bullins et al., 2013) when individuals were exposed to food stimuli.

Demographic questions
Participants answered demographic questions about their age, gender and place of residence.

Behaviour
Participants were asked to complete a timeline follow-back questionnaire (Sobell & Sobell, 1992) one week after the initial questionnaire to record their discretionary food consumption over the past two days. Participants were provided with a definition of discretionary foods from the Australian Dietary Guidelines: foods and drinks that are not an essential part of our diet and are high in saturated fats, sugar, salts or alcohol but low in nutrients (National Health and Medical Research Council, 2013). They were also provided with examples of discretionary foods and serving sizes to assist with conceptualising the serving sizes of the foods and drinks they had consumed (see Supplementary Material 1).

Data analysis
Hypotheses were specified before the data were collected. The analytic plan was pre-specified, and any no data-driven analyses were conducted. Descriptive statistics, internal reliability analyses, bivariate correlations and a hierarchical multiple regression were conducted using IBM SPSS Statistics 29 (IBM, 2022). The hierarchical multiple regression analysis was used to determine the variance in discretionary food consumption that could be explained by the temporal self-regulation theory variables and food reward sensitivity, and to determine if behavioural pre-potency and self-regulatory capacity will moderate the relationship between intention and discretionary food consumption.

Preliminary analyses
A missing values analysis was conducted after removing seventeen cases with at least one missing value. Missing values were imputed using expectation maximisation (Allen et al., 2018) as Little's MCAR test showed that the data was missing completely at random (χ 2 = 516.75, df = 594, p = .990).
A Pearson's Bivariate Correlation analysis (r) was conducted to investigate the relationships between the variables. Age did not correlate with discretionary food consumption at time two (r = − 0.02, p = .70), and thus was not controlled for in the analyses. However, all predictor variables had significant correlations with discretionary food consumption at time two. Only past behaviour had a strong statistically significant correlation with discretionary food consumption at time two (r = 0.60, p < .01), while other predictors had small to moderate correlations with the behaviour. See Table 1 for a summary of bivariate correlations between the variables.
A Pearson's Chi-square test was then performed to determine if gender was significantly correlated with discretionary food consumption. However, the results were non-significant, χ 2 (105, 273) = 116.49, p = .21. Hence, gender was also not controlled for in the hierarchical multiple regression analysis.

Participant demographics
A total of 406 participants completed the first part of the study, and 274 completed the second part of the study (32.5% attrition rate). Participants were excluded from the final sample if they did not complete the second part of the study (N = 132), and one case was removed due to being an extreme multivariate outlier, resulting in a final sample size of 273 participants. Independent samples t tests were conducted for each predictor variable to determine any differences between participants who completed both parts of the study and those who only completed the first part. Statistically significant differences were found for age and general self-control. Participants that completed both parts of the study (M = 42.51, SD = 17.05) were, on average, 7.36 years older than those that only completed the first part (M = 35.15, SD = 16.01), t (404) = 4.15, p < .001, two-tailed, d = 0.44. Participants that completed both parts of the study (M = 2.10, SD = 0.69) also scored slightly higher on the general self-control measure (by 0.19) compared to participants that only completed the first part of the study (M = 1.90, SD = 0.74), t (404) = 2.60, p = .01.
Participants in the final sample were aged between 18 and 80 years (M = 42.55, SD = 17.07), with most participants identifying as female (79.5%), 16.1% identifying as male, and 4% identifying as non-binary. Most participants (93.4%) currently resided in Australia, with 3.3% residing in Europe, 1.1% in North or Central America, 1.1% in Africa, 0.4% in Asia and 0.7% indicated their country of origin as Other.

Hierarchical multiple regression analysis
Each of the predictor variables was centred and standardised before being entered into the regression. These standardised variables were used to create interactions between discretionary food-related habit and intention, general habitual tendencies and intention, environmental cues and intention, food related-self regulatory capacity and intention, and self-regulatory capacity and intention. Past behaviour was entered into the first step of the regression as a control variable. The remaining variables were then entered in the regression according to temporal selfregulation theory (Hall & Fong, 2007).

Discussion
The results of this study show partial support for the utility of temporal self-regulation theory to predict discretionary food consumption, as only intention and past behaviour were unique significant predictors. Of interest was the non-significant correlation between age and discretionary food consumption, as total daily energy intake is expected to decrease with decreased energy requirements (Wakimoto & Block, 2001). However, Australian government statistics show that daily energy intake only decreases by around 600 kJ between the age groups of 19-30, 31-50 and 51-70 years (Australian Bureau of Statistics, 2014). While this appears to be a large difference, 600 kJ is only a small amount of food consumed (e.g., half of a standard 50-g chocolate bar), which could partially explain the current study's lack of statistically significant correlation. Also of interest was that older participants were more likely to complete both parts of the study than those who only completed the first part. As older age groups consume fewer discretionary foods than younger groups, they may be more comfortable sharing their food intake and therefore participate to the end of the study. This could also contribute to the current study's lack of statistically significant correlation.

Past behaviour
Past behaviour was a significant predictor of discretionary food consumption, consistent with previous research on unhealthy snacking (Evans et al., 2017). Although past behaviour is often a strong predictor of future behaviour, it does not cause future behaviour, and is likely a combination of all the factors that motivate that behaviour (e.g., attitudes, beliefs and habit strength) (Ajzen, 1991). Hence, findings indicate that the main causal factors are likely missing from the model currently tested. Future research should continue to explore how factors such as taste, convenience and social influences motivate discretionary consumption to understand whether these factors influence habit strength in discretionary food consumption. For example, if discretionary food is eaten regularly in a social setting, stronger habit strength might be created if the social event occurs at the same location with the same types of foods. However, as past behaviour accounted for 23% of the variance in discretionary food in the current study, future interventions should still consider the influence of past behaviour and how people perceive their past behaviour when aiming to reduce discretionary food intake. For example, an experimental study by Reynolds et al. (2018) found that participants, when asked to reflect on their healthy past behaviour choices, reported stronger intentions to eat healthily in the future.

Intention
Intention accounted for significant, albeit small (2%), variance in discretionary food consumption. Findings are in line with McAlpine and Mullan (2022) where intention only accounted for a small variance in sugar-sweetened beverage consumption, but in contrast with Moran and Mullan (2021) where intention explained a large variance in sugar-sweetened beverage consumption. Conflicting findings may be due to the framing of the intention measure, where in the current study, it was framed in terms of avoiding the behaviour (e.g., "I intend to limit my intake of discretionary foods every day for the next week"). Some research suggests that avoidance-framed intention measures are unlikely to significantly predict behaviours that are immediately rewarding with no short-term consequences (McAlpine & Mullan, 2022). More generally, a meta-analysis found that relationships between intention and behaviour are stronger for behaviours which individuals tend to approach, rather than avoid (e.g., aiming to eat healthy rather than avoiding unhealthy eating) (Dorina et al., 2022). Hence, future research should explore the most appropriate method to frame intention measures for different behaviour types to more accurately assess the influence of intention in health behaviours (e.g., by measuring intention framed positively as well as negatively to predict behaviour).

Habit and environmental cues
Habit was not a significant predictor of discretionary food consumption, which is inconsistent with previous studies on unhealthy snacking (Evans et al., 2017) and sugar-sweetened beverage consumption (McAlpine & Mullan, 2022). It may be that both habit and intention are motivators for behaviours which people perform regularly, whereas intention tends to be more influential for behaviours that people do not regularly perform (Ouellette & Wood, 1998). While past studies indicate that discretionary food consumption in between meals, (e.g., unhealthy snacking and sugar-sweetened beverage consumption) are automatic and convenient (Evans et al., 2017;McAlpine & Mullan, 2022), it could be that discretionary food consumption as part of a main meal (also considered under the behaviour conceptualisation in the current study) is more consciously chosen, hence intention being more influential in the current findings. Future studies should explore alternative ways of measuring individuals' habit strength for discretionary food consumption. For example, through the use of ecological momentary assessment study designs to record participant dietary intake to assess variances in dietary behaviour, in combination with a journal so that participants can record how they are feeling and why they are consuming discretionary foods (e.g., hungry, bored, tired, stressed, social situation).
Environmental cues were also not a significant predictor of discretionary food consumption, which is inconsistent with previous studies on unhealthy snacking (Elliston et al., 2017) and sugar-sweetened beverage consumption (McAlpine & Mullan, 2022). However, the mean score for environmental cues in the current study was low (M = 37.74 of a total possible score of 210), which may reflect difficulties in measuring environmental cues that are predominantly unconscious influencers of behaviour which individuals may not be aware of (Hall & Fong, 2007;McAlpine & Mullan, 2022). Future research should explore alternative ways of assessing the influence of environmental cues to behaviour to assess these unconscious processes more accurately (e.g., via ecological momentary assessment study designs using location tracking to assess locational physical cues such as proximity to fast food restaurants, in combination with time-based questions to detect other physical, sensory, internal, social, and emotional cues and instances of behaviour).

Self-regulatory capacity
Self-regulatory capacity was not a significant predictor of discretionary food consumption in the current study. Of interest was the slightly higher general self-regulatory capacity reported by participants who completed both parts of the study than those who only completed the first part. This difference could suggest that individuals with higher self-regulatory capacity might be more willing to exert the effort necessary to complete their participation once they have committed. Self-regulatory capacity as a predictor of dropout has been explored in the context of academic (Nemtcan et al., 2022) and intervention settings (Keshen et al., 2017). However, future research should examine how self-regulatory capacity influences participation in other common study designs (e.g., online questionnaires, university participation pools) and whether recruitment methods (e.g., volunteers, compensated participants) affect the influence of self-regulatory capacity on participation.
Previous research has found mixed results regarding the influence of self-regulatory capacity on discretionary food consumption. For Note. **p < .01, ***p < .001. A. Dominguez Garcia et al. example, Evans et al. (2017) found that trait-level self-regulatory capacity did not significantly predict unhealthy snacking and suggested using executive function measures to assess state-level self-regulatory capacity which might be more important for eating-related decisions. However, McAlpine and Mullan (2022) found that an executive function task did not significantly predict sugar-sweetened beverage intake, but the trait-level self-regulatory capacity measure did. It could be that various trait-level and state-level measures are assessing different aspects of self-regulatory capacity (e.g., self-control, planning, attention switching) which are important for various tasks (Allom & Mullan, 2014;Dorina et al., 2022). Therefore, future research should continue to explore the most applicable measures and aspects of self-regulatory capacity in discretionary food consumption to better understand its self-regulatory drivers (e.g., continue to explore the utility of executive function measures to assess self-regulatory capacity). While self-regulatory capacity was not a significant unique predictor, it had strong negative correlations with environmental cues, both habit strength measures and past behaviour, which was influential to behaviour. This was also found by McAlpine and Mullan (2022), and suggests that self-regulatory capacity might affect discretionary food consumption indirectly by influencing habit formation and/or vice versa. It is suggested that people with stronger habits and routines, tend to have higher self-control, not because of increased willpower, but due to having to rely less on self-control to resist temptations and impulsive behaviour (Adriaanse et al., 2014;Gillebaart & De Ridder, 2015). For example, strong self-control might make it easier to form healthy habits, such as eating more fruits and vegetables to reduce urges to eat discretionary foods, or by avoiding the formation of unhealthy habits, such as refraining from purchasing discretionary food items to remove temptation to consume them entirely. Hence, future research should explore the interactions between habit and self-control to better understand the underlying processes of discretionary food consumption.

Food reward sensitivity
Food reward sensitivity was not a significant predictor of discretionary food consumption. Findings are supported by Murray and Mullan (2019) on binge drinking behaviours. In the current study, food reward sensitivity had strong correlations with habit strength, environmental cues and self-regulatory capacity which suggests that factors of temporal self-regulation theory account for at least some of the effects proposed by reinforcement sensitivity theory (Corr, 2004). It is possible that discretionary food consumption in the current study was motivated for reasons other than the rewarding nature of these foods, for example, social factors, accessibility and beliefs have also been identified as motivators of discretionary food consumption (Kombanda et al., 2022). Future research could further explore these other motivators by in addition with the temporal self-regulation theory framework to better understand to intentional and non-intentional processes of discretionary food consumption.

Strengths and limitations
While several studies have applied temporal self-regulation theory to unhealthy snacking behaviour (Elliston et al., 2017;Evans et al., 2017), the current study appears to be the first to apply temporal self-regulation theory to discretionary food consumption more broadly. This is a strength as Australian dietary guidelines do not differentiate between discretionary foods eaten as snacks or as part of a main meal (National Health and Medical Research Council, 2013). This study also explored the effects of food reward sensitivity on the temporal self-regulation theory model. However, the current study has some limitations. Notably, convenience sampling was used, which could affect the generalisability of the findings as most participants identified as female (79.6%). However, the proportion of females in the current study is similar to other health behaviour studies (Elliston et al., 2017;Evans et al., 2017), suggesting that females are more likely to volunteer for health-related research. Future studies could attempt to recruit a more balanced gender ratio to further explore discretionary food consumption differences. Additionally, most participants resided in Australia (93.4%) but discretionary food consumption rates are similar across a range of countries such as the United States, the United Kingdom, and Brazil (Sui et al., 2017) so that the results of this study may be generalisable to similar populations.
The current study did not distinguish between weekday and weekend intake. Some previous studies have shown that people consume more discretionary foods during weekends than the rest of the week (An, 2016;Lindroos et al., 2021). However, these changes varied greatly by demographic factors such as age, gender, education level, income and current BMI. For example, adolescents are more likely to increase their discretionary food intake at weekends than older adults. Future research could explore whether different motivating factors explain weekday and weekend discretionary food intake changes. For example, weekday intake may be more likely to be driven by a perceived lack of time or food preparation. In contrast, social events and expectations could be the primary motivator for weekend intake.
As the current study used self-report measures to record discretionary food intake, it is possible that participants underreported the number of daily servings of discretionary foods consumed. Steps were taken to minimise the risk of self-report bias by providing written and visual descriptions of discretionary foods and serving sizes and by only requiring participants to remember their intake over the previous two days. The mean daily number of discretionary food servings (2.93 serves per day) was consistent with previous research in similar areas, despite different measures being used to record intake (Kerr et al., 2017;Verhoeven et al., 2012). However, discretionary food intake in these studies is lower than those reported in government statistics (4.3 serves per day; Australian Bureau of Statistics, 2014). As the government statistics were derived from data from a single time point for each participant (the foods they consumed the day before they were interviewed), they might not account for normal fluctuations in diet due to factors such as special events. Therefore, the results of previous studies that recorded multiple days of intake might be more accurate representations of individuals' average daily food intake.
The government statistics also show that while the average adult consumes around 4.3 servings of discretionary foods per day, the average female only consumes around 3.4 servings per day (Australian Bureau of Statistics, 2014). As most participants in the current study identified as female, this could partially explain the slightly lower discretionary food intake. Similarly, many health behaviour studies also consist of a high percentage of female participants; therefore, this could explain why discretionary food intake in the current study was consistent with similar past studies (Kerr et al., 2017;Verhoeven et al., 2012).
The difference in intake between the current study and government statistics may also be attributed to those consuming fewer discretionary foods being more comfortable sharing this information and therefore more likely to participate in the research, or it may be that people tend to underreport their food intake in self-report studies due to recall errors (Elliston et al., 2017). Hence, future research should consider measures such as 24-h dietary recall interviews where accurate dietary intake is required (Jackson et al., 2008;Schatzkin et al., 2003). Future research should also consider the use of ecological momentary assessment study designs to measure discretionary food consumption, as participants can record their dietary intake in real-time (Maugeri & Barchitta, 2019;Spook et al., 2013). While ecological momentary assessments can reduce recall errors, they are not without limitations. They are often more expensive to conduct than questionnaires or interviews, can have low participant compliance and response rates, and could exclude potential participants that do not own a smartphone or that have low electronic literacy (Dulin et al., 2017;Maugeri & Barchitta, 2019). As 24-h dietary recall interviews and ecological momentary assessments are resource-intensive and discretionary food intake in the current study was consistent with previous research, the timeline follow-back questionnaire (Sobell & Sobell, 1992) remains a suitable alternative where large sample sizes are required. Future research could also explore how participants are completing the timeline follow-back questionnaire by asking participants if they completed the questionnaire with the assistance of written records (as instructed) or if they are relying on memory, to explore whether this affects self-reported dietary intakes.

Conclusion
The results of the current study show partial support for the use of temporal self-regulation theory in predicting discretionary food consumption. Intention and past behaviour were found to be the only significant predictors of behaviour. However, the addition of food reward sensitivity did not provide any additional utility to the temporal selfregulation theory model. As most of the unique variance in the current study was explained by past behaviour, this suggests that future research should continue to study other known motivators of discretionary food consumption related to past behaviour to gain a clearer understanding of the primary motivators of discretionary food consumption.
Results show that intention to reduce discretionary food consumption can be an effective and changeable target for intervention. But interventions aiming to reduce discretionary food intake should also consider the role of past behaviour in future discretionary food consumption. Finally, findings indicate that there is scope to explore the most appropriate method to measure intention in future studies so that researchers may better understand the role of intention in behaviour and better address the intention-behaviour gap.

Ethical statement
Ethics approval was obtained from the Human Research Ethics Committee (HREC) of Curtin University (HREC number HRE2021-0463). Informed consent was obtained from all individual participants prior to completing the study. The procedures used in this study adhere to the tenets of the Declaration of Helsinki.