Interplay between automatic and controlled processes

One of the key predictions of the dual-process models is that automatic and controlled processes interact to predict behaviour and there is considerable empirical evidence to support this claim. A prominent example of this interaction is that automatic affective reactions or implicit evaluations (attitudes) towards unhealthy foods predict food choices and consumption only when individuals have poor trait and/or state inhibitory control or other low cognitive resources (e.g., working memory capacity),

as measured or manipulated in laboratory settings. Hofmann et al. (2009a) measured inhibitory control via an adapted SST and automatic affective reactions towards a specific food item (peanut m&m’s) using an SC-IAT. Food consumption was examined in a product test procedure, where participants are provided with 125g of candy and are asked to taste and rate them on several characteristics. The weight of consumed candy is calculated at the end as a measure of food consumption. The authors reported that automatic affective reactions towards candy had a stronger influence on subsequent consumption in individuals that showed low inhibitory control. Differences in controlled processing, such as inhibitory control capacity, may therefore explain individual differences in how strongly automatic processing affects behaviour. Nederkoorn, Houben, Hofmann, Roefs, & Jansen (2010) measured response inhibition via an SST (i.e., top-down inhibitory control) whereby higher SSRTs can reflect lower inhibitory control (Logan et al., 1997) and examined implicit attitudes (‘preferences’6) towards snack food via an SC-IAT. Instead of food consumption in the laboratory, Nederkoorn et al. (2010) provided a longitudinal measurement of weight after a one-year period to examine whether response inhibition and implicit snack food attitudes predicted actual weight change. They found that response inhibition efficacy predicted weight change only in individuals with strong implicit attitudes. Specifically, participants who showed a greater ‘preference’ for snack foods in the SC-IAT and had less effective response inhibition gained more weight after one year compared to participants with more effective response inhibition. A limitation of this study, however, was that although dietary intake was assumed to drive the potential weight-gain after the one-year period, a measure of dietary intake was not employed.

Other studies have directly manipulated cognitive control resources in the reflective system, such as working memory capacity and other self-regulatory resources (Friese, Hofmann, & Wänke, 2008; Hofmann, Rauch, & Gawronski, 2007). Friese et al. (2008) employed a food choice task where participants had to choose five items from a variety of fruit and chocolate categories (e.g., tangerines, apples, Snickers bars, Twix bars) which were placed on the table (Study 1). Working memory capacity was manipulated by instructing participants to remember a number they saw in the beginning of the task, which they would need to recall upon task completion. In the low-capacity condition participants were given a one-digit number to remember, whereas in the high-capacity condition the number consisted of eight digits. The behavioural outcome reflected the

6_{I purposefully avoid the term preferences as it is currently being used in the literature because}

food preferences incorporate an element of choice behaviour and not only evaluations or affect (e.g., see Chen, Holland, Quandt, Dijksterhuis, & Veling, 2019).

number of chocolates that were chosen out of the presented food items. The authors found that explicitly measured food attitudes (Likert scale) predicted food choices in the high- but not low- capacity condition, while implicit food attitudes (IAT variant) predicted choice behaviour very well in the low- but not high- capacity condition (Friese et al., 2008). In Study 2 (Friese et al., 2008), self-regulatory resources (Muraven & Baumeister, 2000) were manipulated via an emotion suppression task and food consumption (potato crisps) was measured using the product test procedure. Explicit and implicit attitudes (SC-IAT) were also assessed, while the explicit attitude index was merged from two types of ratings that incorporated affective valence (‘negative’ vs ‘positive’) and taste (‘not delicious at all’ vs ‘very delicious’). In the control condition where participants were instructed to watch the short movie while expressing freely any emotions they had (no depletion), a good predictor of food consumption was the explicit attitude measure. On the contrary, when participants were required to suppress their emotions while watching the movie (depletion), the implicit attitude measure was the significant predictor of food consumption. Authors also reported that the interaction between explicit and implicit measures was non-significant, which indicates that explicit food attitudes that are based on controlled processes within the reflective system may differ from automatic affective reactions, or implicit attitudes, in the impulsive system. Similar results have been reported by Hofmann et al. (2007) who found that positive implicit attitudes towards candy as measured via an IAT significantly predicted increased candy consumption in the depletion condition, but not in the control condition. Both studies make references to previous research and theory on self-regulation, such as ego depletion (Muraven & Baumeister, 2000; Vohs & Baumeister, 2017) and these concepts are not discussed further due to recent findings that raised concerns about the replicability and size of the ego-depletion effect (Hagger et al., 2016). All the studies mentioned thus far measured general inhibitory control and it remains to be seen whether the interactive effects of automatic and controlled process on eating behaviour can be supported in studies that measure food-specific inhibitory control.

The interplay between automatic and controlled processing in the context of eating behaviours has also been examined for approach bias and inhibitory control. Kakoschke et al. (2015) measured approach bias via an AAT with positive food and non-food stimuli. Food items were high in sugar, salt and/or fat and non-food pictures included animal categories (e.g., koala) and all stimuli were selected from a pilot study where ratings of pleasure and arousal were provided. Inhibitory control was not measured via an SST, but a food-specific go/no-go task where a higher proportion of commission

errors (i.e., not inhibiting a response on no-go trials) reflects lower, or poorer, inhibitory control. The authors report that cognitive bias7 and inhibitory control did not predict the intake of unhealthy foods in the laboratory taste test when considered in isolation. However, there was an interactive effect of approach bias and inhibitory control on food intake, whereby participants with high approach bias towards unhealthy foods and lower inhibitory control showed increased consumption compared to participants with high approach bias and higher inhibitory control. This interaction was not found for participants who had low approach bias towards unhealthy foods. This study provides additional evidence for the interaction between automatic and controlled processing in predicting eating behaviours when food-specific inhibitory control is examined, which may be more important for unhealthy eating compared to general inhibitory control, as previously discussed in section 1.5.2.