The dynamics of self-control: within-participant modeling of binary food choices and underlying decision processes as a function of restrained eating

Consumers face a daily struggle between maintaining a healthy eating style propagated by nutritionists and medical experts, and giving into immediate food temptations. Resisting tempting but energy dense foods is considered to require constant successful self-control (Baumeister, Vohs, & Tice, 2007; Hofmann, Friese, & Strack, 2009), defined as a preference for larger, but delayed, rewards (e.g., weight loss) over smaller, but immediate, rewards (e.g., eating chocolate; Mischel, Shoda, & Rodriguez, 1989). However, the pandemic rates of overweight and obesity (Haftenberger et al., 2016; Schienkiewitz, Mensink, Kuhnert, & Lange, 2017) and the low long-term success of weight-loss diets (Mann et al., 2007) indicate that self-control is prone to failures.

In search of the mechanisms underlying this varying success in self-control, one eating style has been studied intensely: Restrained eating describes a pattern of restricted food intake and weight watching to reduce or maintain weight (Schaumberg, Anderson, Anderson, Reilly, & Gorrell, 2016). Yet, the literature is mixed as to whether restrained eaters are actually successful in cutting down on intake: Laboratory food intake is often reduced in restrained eaters (Robinson et al., 2017). However, in several studies in naturalistic settings, psychometric measures of restrained eating do not consistently relate to actual calorie intake (e.g., Stice, Cooper, Schoeller, Tappe, & Lowe, 2007; Stice, Fisher, & Lowe, 2004; Stice, Sysko, Roberto, & Allison, 2010). Furthermore, a hallmark finding is that restrained eaters overeat after a perceived breach of their diet (Herman & Polivy, 1984).

One prominent theoretical explanation of why restrained eaters may be unsuccessful in exerting self-control in eating—despite their explicit intention to do so—is the need for effortful and conscious inhibition of temptation impulses (for a critical overview see Fujita, 2011). This approach states that in order to avert self-control failures, tempting impulses need to be consciously recognized as undesirable and then need to be inhibited. Thus, self-control failures occur due to the inability to inhibit such impulses, e.g., due to depleted cognitive resources, reduced motivation to exert self-control/attention to gratification (Inzlicht & Schmeichel, 2012; Inzlicht, Schmeichel, & Macrae, 2014) or particularly strong temptations (Kotabe & Hofmann, 2015; Stroebe, Mensink, Aarts, Schut, & Kruglanski, 2008). However, as proposed by Fujita (2011), viewing effortful impulse inhibition as the defining criteria for self-control neglects people’s capacity to monitor and process environmental information in a cognitively efficient way. The routinization and automatization of goal-striving behaviors, which would be less resource-demanding, would enable restrained eaters to enact self-control without effortful and conscious inhibition of temptation impulses (Bargh & Chartrand, 1999; Fishbach, Friedman, & Kruglanski, 2003; Papies, Stroebe, & Aarts, 2008). Taken together, it remains unclear whether restrained eaters do actually reduce food intake in line with their intentions and which type of self-control processes (conscious/effortful vs. non-conscious/effortless) are enacted to produce goal-consistent behavior.

Food choice is central to successful self-control but represents a rather challenging task: Average grocery stores host thousands of products, so how does one choose the small number of foods needed, when choices are affected by several, potentially conflicting, motivational dimensions? Besides economic, ethical, and cultural reasons, food choices are determined by palatability, calorie density, and healthiness (Köster, 2009; Leng et al., 2016; Mela, 2001; Steptoe, Pollard, & Wardle, 1995), the latter three being most relevant for weight- and self-control and often in conflict. For example, van der Laan, de Ridder, Charbonnier, Viergever and Smeets (2014) contrasted one condition with a maximized choice conflict between palatability and calorie density and another condition without this conflict, to examine the need to exert self-control through inhibition of temptation impulses. Surprisingly, in the self-control condition, in which palatability had to be discounted to choose low energy foods, weight-concerned women showed less experienced conflict (shorter reaction times and decreased brain activity in conflict monitoring regions). The authors concluded that effortful inhibition of temptation impulses is absent in their sample of weight-concerned women, possibly because their weight-control goals and respective self-control processes were not activated. Yet, even in weight-concerned individuals with high levels of (self-reported) self-control, Stillman, Medvedev, and Ferguson (2017) did not find an indication of effortful inhibition of temptation impulses that may arise from a conflict between food-enjoyment goals and weight-watching goals. This raises the question of whether less experienced conflict during food choice indicates an absence of self-control through effortful and conscious inhibition of temptation impulses or points to less resource-demanding mechanisms of self-control without conscious deliberation.

Distinguishing between these two types of self-control would require a measure of the effort that needs to be invested in aligning one’s behavior with overarching goals (e.g., weight reduction) in the face of several, potentially conflicting motivations. One promising methodological approach to measure continuous competition between various motivational forces during binary choice is afforded by the mouse-tracking technique (Freeman & Ambady, 2010; Freeman, Dale, & Farmer, 2011; Stillman, Shen, & Ferguson, 2018; Sullivan, Hutcherson, Harris, & Rangel, 2015). In contrast to traditional self-report-based metrics, which are prone to memory and other biases (Gorin & Stone, 2001) and metrics as reaction time (Stillman et al., 2017), it is assumed that mouse-tracking continuously measures real-time motor traces of cognitive processes and that less direct mouse traces toward a preferred choice option is indicative of a stronger underlying motivational conflict. Thus, mouse trajectories allow a deeper understanding of how different types of self-control facilitate healthy food choices (Lopez, Stillman, Heatherton, & Freeman, 2018).

One methodological constraint of most food choice tasks is that the expected self-control conflict has to be modeled a priori: For example, trials are artificially constructed for each participant—by selecting pre-rated food images (e.g., high palatable vs. low caloric)—to induce self-control conflicts (e.g., van der Laan et al., 2014). As other researchers have argued, the a priori construction of food pairs limits the generalizability to real-world decisions (e.g., Lopez et al., 2018). Thus, the present study took a novel approach to this methodological problem by realizing all possible food pairings of a representative set of foods during binary choice. Mixed-effects modeling was used to better characterize participants’ trial-level choice behaviors as a function of both trial-level features (subjective ratings on important choice dimensions as: palatability, health, calorie density) and person-level characteristics (i.e., restrained eating).

Using this approach, we hypothesized that choice would be primarily predicted by palatability preferences but—secondly—also by calorie density and perceived healthiness of the two food options (Raghunathan, Walker Naylor, & Hoyer, 2006; van der Laan et al., 2014), and that the latter two dimensions would be more influential in restrained eaters. Due to the inconsistent literature on food intake—as reviewed above—we did not make directional predictions as to whether restrained eaters would choose foods with lower or higher caloric density. Beyond choice behavior, we aimed to determine the types of self-control underlying food choice in restrained eaters: An conscious and effortful type of self-control would predict more conflict in restrained eaters, as manifested in less direct mouse trajectories, whereas a less conscious and effortful mechanism would predict the opposite. Due to these two contrasting theoretical accounts regarding self-control type, we anticipated additional exploratory analyses.

The current study adopted a new experimental and statistical modeling approach to investigate determinants of food choice in restrained eating. The external validity of our new binary choice task was supported by strong and consistent associations of choice with actual food intake on a test meal. In line with our hypothesis and other research (e.g., Raghunathan et al., 2006), we showed that palatability is the main driver for food choice with health and calorie density having significant but subordinate roles. Furthermore, we are able to show that individuals with higher levels of restrained eating were less likely to choose highly palatable and calorie-dense foods than their counterparts with lower levels of restrained eating. Thus, restrained eaters’ choice pattern was in line with their weight-control goal.

Importantly, using the mouse-tracking technique we were able to investigate how restrained eaters executed such successful choices. According to accounts that equate self-control with effortful and conscious inhibition of temptation impulses, it would be expected that impulses associated with tempting foods are in conflict with restrained eaters’ health/weight goals. Such impulses would thus need to be inhibited—through slow and controlled processes—to support successful choice outcome. Yet, we found no indication of such effortful inhibition (or choice conflict) in our mouse-tracking data: instead individuals with higher levels of restrained eating demonstrated less conflict when choosing the healthier food option, as illustrated by fewer x-flips (direction reversals) and smaller AUCs (overall less strait decision path). This pattern bears similarity with the absence of choice conflict in weight-concerned individuals during binary food choice trials in the study by van der Laan et al. (2014). Such pattern could either be due to a reduced impulse strength (Hofmann et al., 2009; Kotabe & Hofmann, 2015), a lack of activation of weight-control goals (as hypothesized by van der Laan et al., 2014) or to the operation of non-conscious and effortless self-control mechanisms. In our search for reasons for the absence of decision conflicts in our data, we found that restrained eaters rated healthier and less energy dense foods as more palatable. Thus, we ran a second study to replicate this latter result in an independent sample, suggesting that the finding was not specific to our sample or that palatability ratings were not influenced by previous food choices (e.g., according to dissonance reduction; Izuma et al., 2010). Results across both studies showed that such an alteration or ‘shift’ in palatability preference from high palatable/caloric to healthy/less energy dense foods aligns restrained eaters’ food liking (or impulses) with their weight-control goal (similar results obtained by Buckland et al., 2015). Importantly, this ‘palatability shift’—potentially reflecting a more mid-to long-term attitude change—obviates the need for regulatory efforts to inhibit tempting impulses driven by attractive yet unhealthy foods.

More generally, as indicated above, much of eating behaviour research has explicitly or implicitly operated under the ‘effortful inhibition of impulses’ account. However, a simple equation of self-control with a slow, conscious, and effortful process has been repeatedly criticized (e.g., Fishbach et al., 2003; Fujita, 2011; Galla & Duckworth, 2015; Haynes, Kemps, & Moffitt, 2016; Neal, Wood, & Drolet, 2013). In fact, there has been growing awareness that information processing below the level of consciousness may have a stronger impact on choices and decision making than previously assumed (Bargh & Chartrand, 1999; Galla & Duckworth, 2015). In our view, the observed pattern of choice, process and rating data could be better contextualized within accounts that accommodate the operation of non-conscious and effortless types of self-control, reviewed and systematized in the dual motive framework by Fujita (2011). These accounts include goal priming, which refers to the establishment of facilitative links from temptations (high energy foods) to overarching goals through repeated successful goal pursuit (Fishbach et al., 2003). Relatedly, an initially effortful act (choosing a healthy instead of a palatable food) can become more efficient over time and practice until it proceeds without conscious guidance (Bargh & Chartrand, 1999; Hagger, Wood, Stiff, & Chatzisarantis, 2010), representing a process similar to skill acquisition or the development of habits (Verplanken, 2018). The palatability shift observed here could be related to either of these mechanisms, but may also constitute a strategy of its own. Future research could study such palatability changes longitudinally to determine when and how such changes take place.

These conclusions have to be seen in the light of some limitations. Generalization is limited to predominantly healthy-weight female individuals, given differences in health beliefs and dieting between women and men (e.g., Wardle et al., 2004) and between healthy and eating disordered samples (Foerde, Steinglass, Shohamy, & Walsh, 2015; Steinglass, Foerde, Kostro, Shohamy, & Walsh, 2015). Further, despite observing goal-consistent food choice behavior, we did not assess whether restrained eaters in this sample were actually successful in terms of every day dieting (see discussion around validity or restrained eating questionnaires; Ahern, Field, Yokum, Bohon, & Stice, 2010; Stice et al., 2004, 2007) or whether they had higher levels of self-control in general (as observed by Stillman et al., 2017). Thus, investigating objective dieting success (and maybe general self-control) in various populations with altered eating behavior would offer promising future directions.

To conclude, individuals with higher levels of restrained eating showed successful self-control in a binary food choice task, and they did so using a rather effortless and automatic mechanism, which might be related to a change in their palatability preferences (‘palatability shift’). This palatability change comprises the devaluation of temptation (i.e., less liking for more calorie-dense/unhealthy foods) as well as an increased valuation of goal-congruent foods (i.e., increased liking for less calorie-dense/healthy foods) and ultimately brings food preferences and long-term goals (i.e., weight reduction) into alignment. Such mechanisms would have been hard to detect without the current statistical modeling approach that employs individual image ratings as predictors of binary choice and associated process data. Thus, this approach might be applicable to other fields of decision-making research that study conflicts between multiple choice motives. Last, current weight loss treatments heavily emphasize effortful impulse inhibition, which might explain their vulnerability for failure during times of stress and limited cognitive control resources. To replace or at least complement these approaches, the discovery of ‘effortless’ mechanisms in food choice in the present study might fuel the development of corresponding interventions in more naturalistic dieting studies.