Changing Deliberative and Impulsive Food Choice Through a Go/No-Go Training Paradigm
Master thesis: L. (Linda) Schmale July 1, 2016
10505075
University of Amsterdam
Faculty of Social and Behavioural Sciences Program Group Social Psychology
Supervisor: Rob W. Holland Second assessor: Michiel van Elk
Abstract
Overeating and obesity form a growing problem in today’s society. Research focusses on response inhibition training as a possible means to overcome and prevent those health-harming behaviours. The increase in overeating and obesity is partially caused by an
increased intake of high caloric food. Therefore, influencing food choice plays an important role in the prevention of these health-harming behaviours. In order to influence these choices, it is important to understand what kind of food choices are influenced by response inhibition training. Across two experiments, we examined whether it is possible to influence impulsive and deliberative food choice through GNG training. The effect of GNG training on food choice was assessed by including a choice task directly after the GNG training. Results showed that there was an effect of GNG training on food choice when participants made an impulsive choice, this effect was found for high and low value stimuli. In contrast, there was no effect found of GNG training on deliberative food choice, regardless of item value. Furthermore, we found an effect of GNG training on food evaluations within the impulsive choice experiment, but only for high value stimuli. No effect for evaluations was found in the deliberative choice experiment. These findings show that GNG training can be used as a means to influence impulsive food choice. Further research is needed to determine if the findings are applicable to healthy foods, in order to overcome and prevent people from becoming overweight and obese.
Changing Deliberative and Impulsive Food Choice Through a Go/No-Go Training Paradigm Worldwide the prevalence of overweight and obese individuals has increased over the past years (Allom, 2014; WHO 2015a, 2015b). Within 12 years’ time the prevalence has almost doubled (WHO, 2005a, 2015a). The extreme increase in overeating and obesity can be explained by an energy imbalance between the amount of calories consumed and the amount of calories expended (Stroebe, 2008; Veling, Aarts, & Stroebe, 2013a; WHO, 2005b, 2015a). There has been an increased intake of palatable high calorie food, because it is easily
available and mostly inexpensive (e.g. Allom, 2014; Veling, Aarts, & Papies, 2011; Veling, et al., 2013a; WHO, 2015a). On the other hand, there has been a decrease in physical activity due to the increase in sedentary forms of work, changing ways of transportation, and
increasing urbanization (Allom, 2014; Stroebe, 2008; WHO, 2015a). Overeating and obesity have been associated with a substantial reduction in life expectancy (Allom, 2014; Stroebe, 2008; WHO, 2005b). This reduction is due to non-communicable diseases which are associated with being overweight and obese, such as cardiovascular diseases, hypertension, diabetes and cancer (Allom, 2014; Stroebe, 2008; WHO, 2015a). Given the fact that
overeating and obesity have such severe consequences, and that those consequences are mostly preventable, it is important to determine what processes underpin eating behaviour (Allom, 2014; WHO, 2005b; WHO, 2015a). The better the nature of eating behaviour is understood, the easier it is to change people’s eating behaviour in order to overcome or prevent people from becoming overweight and obese. Our study explores one of these processes.
Reward Value of Food
One key process in guiding eating behaviour is the reward value of food items. It is suggested that the overvaluation of appetitive stimuli plays a role in health harming
Comings, 1996; Chen, Veling, Dijksterhuis, & Holland, 2016; Shin & Berthoud, 2011; Stice, Spoor, Ng, & Zald, 2009). The overvaluation of those stimuli can be caused by individual differences in the reward value from food intake, also known as food reward (Rogers & Hardman, 2015; Stice, Spoor, Bohon, Veldhuizen, & Small, 2008; Stice et al., 2009). This process comprises of the pleasure and motivation to obtain food, and determines, among other factors, how much a particular food is wanted (Berthoud, 2007; Cameron, Goldfield, Finlayson, Blundell, & Doucet, 2014; Ziauddeen, Subramaniam, Gaillard, Burke, Farooqi, & Fletcher, 2012). A factor that influences food reward is the individual’s metabolic state. That is, hunger makes food more appealing and desirable (thus, more rewarding), whereas satiety reduces the reward value (Ziauddeen et al., 2012). Also, research has shown that the effect of food reward is stronger in obese individuals, than in normal weight individuals (Stice et al., 2008; Stice et al., 2009). The prevalence of overweight and obese individuals can be
measured using Body Mass Index (BMI), an index of weight-for-height (WHO, 2015a). According to the definition of the World Health Organization (WHO), a BMI greater than or equal to 25 is overweight, and a BMI greater than or equal to 30 is obesity (WHO, 2015a). Response Inhibition Training
One way of influencing the reward value of food is by making evaluations of those stimuli more negative (Chen et al., 2016). This can be achieved by using, for example, evaluative conditioning (Hollands, Prestwich, & Marteau, 2011), or response inhibition training (Chen et al., 2016; Lawrence et al., 2015; Veling, Holland, & Van Knippenberg, 2008). Our study will focus on the latter approach. Response inhibition refers to the
suppression of motor actions, also known as motor inhibition (Mostofsky & Simmonds, 2008; Verbruggen & Logan, 2008b). According to research from Veling et al. (2013b), the
devaluation effect that is caused by response inhibition training is stronger for stimuli that are perceived as appetitive. Hence, response inhibition training may be effective in reducing the
health-harming behaviours that are caused by the overvaluation of appetitive stimuli (Allom, Mulan, & Hagger, 2015; Chen et al., 2016).
Out of the several response inhibition tasks generally used, our study focuses on the Go/No-Go training task (GNG). This is because research has suggested that GNG training is more likely to result in automatic response inhibition than other response inhibition tasks, such as the Stop-Signal Task (SST), because GNG stimuli are consistently associated with going or stopping, whereas in SST stimuli are inconsistently paired with going or stopping (Lawrence, Verbruggen, Morrison, Adams, & Chambers, 2015; Verbruggen & Logan, 2008a, 2008b). This is also illustrated by the greater effect sizes found in GNG training paradigms, in comparison to SST paradigms (Allom et al., 2015).
The GNG task is a response inhibition task that has been shown to be a powerful tool to devaluate appetitive stimuli (Houben, Nederkoorn, Wiers, & Jansen, 2011; Houben, Havermans, Nederkoorn, & Jansen, 2012; Veling, Aarts, & Stroebe, 2013a). In the GNG tasks participants are required to respond as rapidly as possible to one set of stimuli (go stimuli, e.g. pressing a key), while withholding responses to another set of stimuli (no-go stimuli, e.g. not pressing a key; Allom, Mullan, & Hagger, 2015; Chen et al., 2016). Within the eating domain, the GNG-training task has been shown to affect food evaluations (Chen et al., 2016; Lawrence et al., 2015), weight loss (Lawrence et al., 2015; Veling, Van
Koningsbruggen, Aarts, & Stroebe, 2014), food intake (Houben & Jansen, 2011, 2015; Lawrence et al., 2015; Veling et al., 2011), and desire to eat (Houben & Jansen, 2015).
Because the increase in overeating and obesity is partially caused by an increased intake of high caloric food, influencing food choice plays an important role in the prevention of these health-harming behaviours. However, research on the effect of GNG training on food choice is rare. Of course, it could be assumed that food choice mediates effects of the studies on weight loss, but food choice was not directly measured in those studies. A couple of
studies have investigated food choice more directly by including a choice task (Veling et al., 2013a, 2013b), but in those studies the participants only had to make one food choice (e.g., choosing 3 out of 7 food pictures; Veling et al., 2013a). Although this study provides valuable insights in the field of choice, making only one choice does not provide a reliable measure of food choice.
Furthermore, in order to influence food choice, it is important to understand what kind of food choices are influenced by GNG training. On the one hand, people can make a choice based on a conscious initiated process (Friese, Wänke, & Plessner, 2006). These deliberative choices are of a reasoned and reflective nature (Marteau, Hollands, & Fletcher, 2012). That is, consequences are weighed and integrated before a decision is made (Strack & Deutsch, 2004). On the other hand, impulsive choices are made when no conscious monitoring occurs (Friese et al., 2006; Fujita & Han, 2009). These choices are guided by associative processes, which facilitate choices (Strack & Deutsch, 2004).
A factor that influences whether choices are made under more (i.e., deliberative) or less control (i.e., impulsive), is the opportunity to deliberate (Fazio, 1990; Friese et al., 2006). That is, time pressure influences the opportunity to deliberate, and therefore influences the kind of choice people make. Deliberative choices are made when there is ample time to make a rational choice (Young, Goodie, Hall, & Wu, 2012). On the other hand, impulsive choices mostly take place when little time is available, in that case a speed-accuracy trade-off occurs, resulting in impulsive choices (Young et al., 2012).
The Present Research
The goal of our study was two folded. First, given the fact that food choice has not yet been examined in a reliable way, the goal of this study was to investigate the influence of GNG training on food choice in a systematic and reliable manner. This was achieved by using repeated choices, instead of only one choice, after a GNG training. Our systematic
approach is based on previous research (Schönberg, Bakkour, Hover, Mumford, Nagar, Perez, & Poldrack, 2014), and starts with an auction in which willingness to pay (WTP) is measured for different snack items, based on those ratings several high and low value items are selected for usage in the next part: the GNG training. After the training, participants are presented with pairs of items, and are asked to choose one item per trial. In order to test whether the effect of response inhibition training is stronger for appetitive stimuli, a distinction is made between high and low value items (i.e., appetitive and less appetitive stimuli). This results in four sorts of pairs (i.e., high go vs. high no-go, low go vs. low no-go, high go vs. low go, and high no-go vs. low no-go). Given the fact that the devaluation effect of response inhibition training is stronger for appetitive stimuli (Veling et al., 2013b), one would expect a training effect on high value items and not on low value items. Afterwards, one trial is selected and the choice on that trial is honoured. This is done in order to make the participants’ choice realistic, and motivate them to take the choices within the experiment seriously. After the choice task, the initial auction is repeated to test whether the WTP of the individual items has changed as a function of the GNG training.
Secondly, in addition to systematically and reliably measuring food choice as a function of GNG training, our study also aims to explore the effect of GNG training as a function of type of food choice. Since we investigate direct behaviour (i.e., making a choice), food choice type is operationalised as time pressure. That is, whether people make a
deliberative or an impulsive choice. We examined deliberative and impulsive food choice within two separate experiments. Within the impulsive choice experiment participants have to make a choice within a short amount of time, whereas in the deliberative choice experiment there is ample time.
Because response inhibition training has been shown to affect explicit evaluations (Chen et al., 2016), one would predict GNG training to have influence on deliberative choices.
With regard to the effect on impulsive choices, the picture is less clear and more exploratory. It is possible to find the same pattern for impulsive choice, because evaluations also have been shown to influence implicit choices (Fazio, 1990; Friese et al., 2006; Strack & Deutsch, 2004). However, this has not yet been investigated.
Hence, within the choice task, with regard to the impulsive choice experiment, we do not have a strong hypothesis. However, within the deliberative choice experiment, we expect an effect of GNG on food choice for high value stimuli, but we expect no effect of GNG on food choice for low value stimuli.
With regard to the auction task, we expect in both the deliberative and in the
impulsive choice experiment an effect on the WTP. That is, the WTP for stimuli paired with a no-go cue will be lower, than the WTP for stimuli paired with a go cue. This effect is only expected for items with high value on the pre-auction, but not for items with low value on the pre-auction.
As mentioned above, hunger increases the reward value of food (Ziauddeen et al., 2012). Therefore, participants are asked to fast for three hours before participating in the experiment, which results in hungry participants. Also, obesity is said to influence the effect of food reward (Stice et al., 2008; Stice et al., 2009). If overweight or obese individuals participate in our study, this factor is included for exploratory reasons.
Experiment 1: Deliberative Food Choice Method
Participants.
According to previous research in this domain (Chen et al., 2016), 30 participants were sufficient to obtain reliable effects (i.e., achieve power above 80%). Therefore, we recruited a minimum of 30 participants for each experiment. Participants were alternately
assigned to one of the experiments, either the deliberative (i.e., experiment 1) or the impulsive (i.e., experiment 2) experiment.
Participants were recruited through the lab website of the University of Amsterdam, by sharing this website via social media, and flyers asking for participants were distributed on campus. As a reward for participation they received either course credits (1.5 Psychology Research Credits) or a payment of €15 worth of VVV vouchers. We announced sample sizes, hypotheses, exclusion criteria, and main analyses before data analysis via the Open Science Framework (OSF) (see
https://osf.io/ygp4w/?view_only=d040373d85574fa2b5000db17db5e87c). Exclusion criteria were participants:
1. Who were deaf and not able to read. 2. Other than between 18 and 26 years.
3. Who ate less than three hours before the start of the experiment.
4. Who bid less than 25 cents on more than 40 food items in the first auction task (based on Schönberg et al., 2014).
5. With accuracies (either go or no-go) in the go/no-go task that were 3SD below the sample mean were inspected: if they were below 90%, the corresponding participants were excluded.
Materials.
The materials used in the study are previously used materials. The pictures used in the experiment were used before in related research (Veling et al., 2016) and consist of 60 high-energy dense snacks. Pictures were taken in accordance with previous work (Schönberg et al., 2014), such that the package as well as the inside of the product were visible. The pictures were taken against a black background.
Auction task. The auctions were based on the Becker-DeGroot-Marschak (BDM) procedure. This procedure has been shown to measure willingness to pay (WPT) per item in a reliable way (Schönberg et al., 2014). The purpose of the first auction was (1) to establish a baseline of what value participants ascribe to the different snack items, and (2) to create go and no-go stimulus sets with similar initial WTP to include in the GNG task. To do so, items were individually ranked based on WTP, in which items 1 had the highest WTP and item 60 the lowest WTP. Based on the ranking 30 items were selected as high value items (ranked 1-30), and 30 items as low value items (ranked 31-60). These 60 items were used in the GNG training.
There was a second identical auction included to measure differences in the value participants ascribe to the different snack items before and after GNG.
GNG training. The GNG-training task was derived from Chen et al. (2016) and was programmed and presented using PsychoPy (version 1.83.04; Peirce, 2007). Before the start of the training task, there were 44 items (22 high- and 22 low value items) randomly assigned to the go trials, and 16 items (8 high- and 8 low value items) to the no-go trials. In each experimental block, each of the 60 selected items was randomly presented once, and the whole training consisted of 8 blocks, resulting in 480 trials in total. In all trials, the item stayed on the screen for 1000ms. The two tones (1000Hz and 400Hz, duration 300ms) used as go and no-go cues were counterbalanced across participants. Before the experimental blocks, the participants received a practice block of 16 trials in which we used filler picture items.
Food choice task. The food choice task was derived from Schönberg et al. (2014). Two types of trials were included: in the experimental trials, participants chose between a go and a no-go item (the initial value was matched, i.e., high go vs. high no-go, low go vs. low
no-go); in the filler trials, both items were paired with go or no-go while one had high value and the other one had low value (i.e., high go vs. low go, high no-go vs. low no-go).
Memory recall task. The memory recall task was included to assess participants’ memory of the go/no-go training. This is done in order to make sure that participants were paying attention to the experiment and were looking at the presented items. The memory recall should be around 80%. The reason for including this task is for comparison with other studies in this line of research, and is beyond the scope of our study.
Procedure.
Participants were informed that the study was about food preferences, and that they would be able to eat snacks during the experiment. At their arrival, participants were asked to indicate how long ago they last ate. If they ate less than three hours ago, participants were not allowed to start the experiment. In that case, participants were asked to return at another time and it was emphasized that they could not eat for three hours before participating in the study. Before the start of the experiment, participants gave written informed consent. The
experiment was run individually for each participant on a Windows 7 computer in a cubicle, and lasted around one hour. In Figure 1 the general procedure of both experiments is shown.
First, participants took part in the auction. At the start of the auction, participants received €2 (1 euro, 50 cents, 20 cents, 10 cents, 10 cents, 5 cents, 5 cents) and were told they were able to spend it on a snack. Participants were informed that, at the end of the experiment, one trial would be drawn. If their bid was higher than the computer’s random bid, the
participant bought the snack for the offered price. During the auction participants were presented with 60 pictures of palatable snack items, one item at a time on the computer screen, and were able to place an individual bid (using the €2) on each of those items. They could place their bid by moving the mouse cursor along an analogue scale that ranged from 0
to 2 euro on the bottom of the screen. The next item was presented only after the participant had placed a bid.
The second part of the study involved the Go/No-Go (GNG) training paradigm. The training started with the presentation of a picture in the middle of the screen. If the picture was assigned to the go trial, 100ms after the onset of the picture a tone was played via a headphone, and participants were instructed to press the B key on the keyboard as fast as possible before the picture disappeared. If the picture was assigned to the no-go trial, a different tone was played 100ms after picture onset, and participants were asked not to press any key until the picture disappeared. During practice, participants received an error message if they made incorrect responses. No performance feedback was provided for the
experimental blocks. After every two blocks, participants received information regarding their accuracy in the current part, and could take a short break if they wanted to.
Third, after the GNG training, participants proceeded to the choice task. Participants were instructed to make a series of dichotomous deliberative (i.e., unlimited time) food choices. Participants were allowed to take all the time they needed to make a choice. Next, participants were told that one trial would be drawn at the end of the session and that their choice on that trial would be honoured, namely they would receive the snack item they chose in that trial.
Fourth, a second auction was included. This auction and the instructions were identical to the first auction.
Fifth, there was a memory recall task. Participants were asked to think back to the GNG task. They were shown the same snack items as in the GNG task, and were asked to indicate for each food item whether it was paired with pressing B (i.e., go) or not pressing B (i.e., no-go) in the GNG task.
At the end of the experiment, participants filled out the restrained eating scale (RS; Herman, & Polivy, 1980) and indicated whether they were on a diet, their weight, height, current hunger level and the last time of food consumption (age and gender were recorded at the beginning of the task).
In the end participants got the results from the auction and the choice task, and received the selected snacks. Although participants were shown a collection of snacks at the beginning of the experiment to convince them that we had all the snacks in the lab, for logistical reasons we included two rigged choice trials at the end of the choice task, with four snack items that we actually had in store. We also rigged the selection of the auction trial to make sure the programme selected one of these four snack items. Finally, participants were debriefed.
Results
In Experiment 1 a total of 31 subjects participated. One participant was excluded due to a programming error, and eight participants were excluded based on pre-registered
exclusion criteria (1 participant fell outside the age limit, 2 participants ate less than 3 hours ago, 3 participants have bid less than 25 cents on more than 40 items, and 2 participant had an accuracy 3SD below the sample mean and below 90%). This resulted in the inclusion of 22 subjects in the further analysis. The subjects were between 18 and 26 years old (M = 22.1, SD = 2.60), most of them were female (n = 20) and they had last eaten between 3 and 13 hours ago (M = 5.64, SD = 3.32). The analyses were conducted in SPSS 22. All the effects reported below are reported as significant at p < .05. Table 1 contains detailed information about the participants.
Preliminary analysis.
To specifically check the selection procedure, the average WTP for the go and no-go items used in the experimental trials of the choice task were computed. A repeated-measures
ANOVA with food item condition (i.e., go vs. no-go) and food value (i.e., high vs. low value) as the within-subject factors and the average WTP as the dependent variable was used. There was a main effect of food value on WTP, F(1,21) = 181.80, p < .001, ηp2 = .90. WTP for
high value items was significantly higher, than WTP for low value items. Furthermore, there was no significant main effect of food item condition (p = .462), and there was no interaction effect between food item condition and food value (p = .325). These findings confirm that the selection procedure of high and low value items worked. Also, these findings illustrate that there are no differences in WTP and food value between go and no-go items. Table 2 gives an overview of the mean WTP and standard deviations for the high and low value go and no-go items, illustrating the difference in WTP between high and low value items. This table also illustrates that there is virtually no difference between the WTP for go and no-go items. Table 3 shows the performance of the participants in the GNG training.
Effect of GNG training on food choice.
Table 4 shows the participants’ average reaction times during the choice task. This table illustrates that participants in this experiment (deliberative choice) did indeed chose slower, than participants in the second experiment (impulsive choice). A repeated-measures logistic regression on the experimental trials (i.e., choose between a go and a no-go item), with whether the chosen item is a go or a no-go item as the dependent variable, revealed that there was no significant difference (odds ratio = 1.14, p = .342; overall on 53.2% of the trials participants chose go items). Participants did not chose go items more frequently than no-go items. This suggests that there is no effect of GNG training on deliberative food choice.
It was explored whether the effect differs between the two types of trials (i.e., with two high value items or two low value items). The effect was not significant (p = .724), indicating that no difference in choice of high versus low value items was observed between
go and no-go pairs. Figure 2 gives a visual representation of the mean percentage of the choices for go items for high and low value items.
Effect of value.
A repeated-measures logistic regression with whether the chosen item has high value (between the two) or not as the dependent variable was performed on the filler trials (i.e., choose between a high and low value item). Participants chose significantly more high value items, compared to low value items (odds ratio = 5.31, p < .001; overall on 84.2% of the trials). This confirms that high value items are preferred over low value items. It was explored whether the effect differs between the two types of filler trials (i.e., with two Go items or two No-Go items). The effect was not significant (p = .458), indicating that no difference in choices of go versus no-go items was observed between high value and low value pairs.
Effect GNG training on food evaluation.
A 2 (time: pre-training vs. post-training) x 2 (value: high vs. low) x 2 (item class: go vs. no-go) repeated measures ANOVA with WTP as the dependent variable was conducted in order to explore whether the bids for the go items are higher than no-go items after the
training.
There was no main effect of item-class on WTP, p = .072. However, there was a significant main effect of value on WTP, F(1,21) = 140.79, p < .001, ηp2 = .87. WTP for high
value items was significantly higher than WTP for low value items.
There was also a significant interaction effect between time and value, F(1,21) = 11.55, p = .003, ηp2 = .36. WTP for high value items was significantly higher pre-training
than post-training. For low value items the effect is reversed; WTP for low value items is significantly higher post-training than pre-training.
Exploratory analysis.
The exploratory analyses are presented in Appendix A. Discussion
This experiment examined deliberative food choice and revealed that there was no significant effect of GNG training on food choice. Furthermore, the effects were not obtained when we separately tested for high and low value items. Since we expected an effect of GNG on food choice for high value items, this is not in line with our predictions.
Regarding the effect of GNG training on food evaluation, we also found no effect. This is not in line with our prediction, because we expected the WTP for stimuli paired with a no-go cue to be lower than the WTP for stimuli paired with a go cue. Although not significant, the trend of the effect on WTP was in line with our predictions.
Now that the effect of GNG training deliberative choice has been examined, the effect on impulsive choice remains. In the second experiment, the effect of GNG training is tested for impulsive choice.
Experiment 2: Impulsive Food Choice Method
Participants.
Just as in experiment 1, we used a minimum of 30 participants in the second
experiment. Participants were recruited and excluded the same way as in the first experiment. Materials.
The materials used were the same as in experiment 1. Procedure.
The procedure for experiment 2 was identical to the procedure of experiment 1, with the exception of the food choice task. Within the food choice task, participants were instructed to make a choice within 1500ms (i.e., impulsive food choice), instead of ample
time (i.e., deliberative food choice). If participants did not react in time, the current choice pair would disappear from the screen and be repeated at a later time, until they were able to make a choice in time.
Results
A total of 31 subjects participated in experiment 1. One participant was excluded due to a programming error, and four other participants were excluded based on pre-registered exclusion criteria (2 participants ate less than 3 hours ago, and 2 participants have bid less than 25 cents on more than 40 items). This resulted in the inclusion of 26 subjects in the further analyses. The subjects were between 18 and 26 years old (M = 21.5, SD = 2.18), most of them were female (n = 19) and they had last eaten between 3 and 16 hours ago (M = 6.65, SD = 4.37). All the effects reported below are reported as significant at p < .05. Detailed information about the participants is shown in Table 1.
Preliminary analysis.
To specifically check the selection procedure of the go and no-go items, the average WTP for the go and no-go items used in the experimental trials of the choice task were computed. This was done by using a repeated-measures ANOVA with food item condition (i.e., go vs. no-go) and food value (i.e., high vs. low value) as the within-subject factors and the average WTP as the dependent variable. There was a main effect of food value on WTP, F(1,25) = 172.64, p < .001, ηp2 = .87. WTP for high value items was significantly higher than
WTP for low value items. Furthermore, there was no significant main effect of food item condition (p = .845) and there was no interaction effect between food item condition and food value (p = .303). These findings confirm that the selection procedure of low and high value items worked. Also, these findings illustrate that there are no differences in WTP and food value between go and no-go items. Table 2 gives an overview of the means and standard
deviations for the high and low value go and no-go items. In Table 3 the performance of the participants in the GNG training is shown.
Effect of GNG training on food choice.
A repeated-measures logistic regression on the experimental trials (i.e., choose between a go and a no-go item) with whether the chosen item is a go or a no-go item as the dependent variable, revealed that participants chose significantly more go items compared to no-go items (odds ratio = 1.28, p = .025; overall on 56.2% of the trials). Participants chose go items more frequently than they chose no-go items. This suggests that there is an effect of GNG training on impulsive food choice. Furthermore, it was explored whether the effect differs between the two types of trials (i.e., with two high value items or two low value items). The effect was not significant (p = .108), indicating that the effect of GNG training is
significant for high and low value items. Figure 3 gives a visual representation of the effects for high and low value go items.
Effect of value.
A repeated-measures logistic regression with whether the chosen item has high value (between the two) or not as the dependent variable was performed on the filler trials (i.e., choose between a high and low value item). Participants chose significantly more high value items compared to low value items (odds ratio = 3.62, p < .001; overall on 78.4% of the trials). This confirms that high value items are preferred over low value items. It was explored whether the effect differs between the two types of filler trials (i.e., with two Go items or two No-Go items). The effect was not significant (p = .558), indicating that no difference in choices of go versus no-go items was observed between high value and low value pairs.
Effect GNG training on food evaluation.
vs. no-go) repeated measures ANOVA was conducted with WTP as the dependent variable. There was a main effect of item-class on WTP, F(1,25) = 5.56, p = .027, ηp2 = .18.
WTP for go-items was significantly higher than WTP for no-go items.
There was also a significant main effect of value on WTP, F(1,25) = 155.64, p < .001, ηp2 = .86. WTP for high value items was significantly higher than WTP for low value items.
There was a significant interaction effect between time and value, F(1,25) = 24.83, p < .001, ηp2 = .50. WTP for high value items was significantly higher pre-training than
post-training. For low value items the effect is reversed; WTP for low value items is significantly higher post-training than pre-training.
There was also a significant interaction effect between time and item class, F(1,25)= 5.95, p = .022, ηp2 = .19. WTP for go and no-go items was significantly higher post-training,
than pre-training.
Exploratory analysis.
The exploratory analyses are presented in Appendix B. Discussion
In the second experiment impulsive food choice was examined. This experiment revealed that there was a significant effect of GNG training on food choice. The effect was not moderated by the value of stimuli, effects were found for both high and low valued items.
Regarding the effect of GNG training on food evaluations, we expected an effect on food evaluations. In line with our expectations, we found an effect of GNG training on food evaluations. The WTP for stimuli paired with a no-go cue was lower, than the WTP for stimuli paired with a go cue. Although there was no significant interaction, the effect was, as predicted, only found for high value items and not for low value items.
General discussion
overeating and consequently obesity, that are caused by the overvaluation of appetitive
stimuli. Therefore, we examined whether it is possible to influence impulsive and deliberative food choice through GNG training. The results showed, in contrast to our predictions, that there was no significant effect of GNG training on deliberative food choice, neither for high or low value items. On the other hand, there was an effect of GNG training on impulsive food choice. This effect did not differ between high and low value items.
For food evaluations, a different pattern of results was obtained. We found no significant effect of GNG training on food evaluations within the deliberative experiment, neither for high or low value items. However, within the impulsive experiment, there was an effect of GNG training food evaluations, but only for high value items.
The fact that there was no effect of GNG training on deliberative food choice is surprising. We expected this effect to occur, because previous research has found an effect of GNG training on deliberative food evaluations (Chen et al., 2016; Veling et al., 2008) and food choice (Veling et al., 2013a, 2013b).
The idea behind the effect of the GNG training on food choice, was that the response inhibition training task would result in devaluation of the no-go stimuli. This leads to more negative evaluations of those stimuli, which makes participants less likely to choose those stimuli. Because previous research showed that the devaluation effect of response inhibition training was stronger for appetitive stimuli (Veling et al., 2013a, 2013b), the effect was expected to be influenced by value. Based on this, one would expect the effect of GNG training on food choice, within the deliberative choice experiment, to occur only for high value stimuli. However, this theory does not seem to apply to our study, because we found no effect of GNG training on food choice for high and low value items within the deliberative choice experiment. Therefore, the effect of GNG training on food choice in our study does not seem to be driven by evaluations.
A possible alternative explanation might be that the effect of GNG training on food choice is driven by attention. This explanation stems from the attentional Drift-Diffusion-Models (aDDM, Ratcliff, 1978) on binary choice. Those models suggest that the binary choice process is guided by visual attention. In the GNG-training task people learn to respond to some stimuli, while withholding responses towards other stimuli, and therefore their attention might be direct more to the stimuli they learned to respond towards. This could explain why people tend to choose more go items, over no-go items. Simply because they focus their attention more on the items they learned to respond towards, in comparison to the items they learned to withhold their responses. The the lack of effect of GNG training within the deliberative choice experiment can also be explained by attentional mechanisms. This can be explained by the fact that the effect of attention decreases over time (Krajbich, Lu,
Camerer, & Rangel, 2012). That is, when people have unlimited time to make a decision, they keep on focussing back and forward on both options and consequently the effect of GNG training disappears. In contrast, within the impulsive condition, attentional differences are more likely to affect choice. This is because the effect of attentional drift becomes stronger over time (Krajbich et al., 2012), and when little time is available this effect might not be strong enough to rule out the effect of the GNG training. This also explains why we did find an effect of GNG training on food choice within the impulsive choice experiment. Because of the limited amount of time, attention is focussed on the trained stimulus (i.e., the go item) and there is no time to shift attention, so the trained item is more likely to be chosen.
GNG training has been found to cause devaluation of food items (Chen et al., 2016; Veling et al., 2008). However, as illustrated above it is unlikely that the results of our study are explained by evaluations. That raises the question how it is possible that we did not replicate the findings of previous studies. If the results we found are indeed driven by
namely not included in the studies that found a devaluation effect of GNG training. However, there were a few studies that did include a choice task, but those studies did not use a binary choice (e.g., Veling et al., 2013a, 2013b). It is possible that, because the choice task was placed between the GNG training and the second auction, the choice task influenced the WTP in the second auction. This would explain why we did find a devaluation effect of GNG training within the impulsive experiment. That is, the choice made in the choice task is driven by attention (which explains the results we found), and the choice that is made subsequently influences the evaluation (i.e., WTP) in the second auction. The influence of the choice on the subsequent evaluation can possibly be explained by, for example, cognitive dissonance theory. In short, the theory suggests that we have an inner drive to hold our attitudes and beliefs in harmony and avoid disharmony, also known as dissonance (Festinger, 1957).
Participants choose one of the items in the choice task, which is driven by attention. In the subsequent auction, they evaluate the items they chose more positive than the items they did nog choose. This is because of the drive to act in line with our attitudes and beliefs. So, because they chose the item in the choice task, they evaluate is more positive in the second auction because of the drive to avoid dissonance between their choice and their evaluation.
It is also surprising that we found no effect of GNG training on food evaluations within the deliberative choice experiment, since we did expect this based on previous
research (Chen et al., 2016; Veling et al., 2008). However, this could also be explained by the fact that the food choice task is prior to the second auction task. Moreover, this explanation also accounts for the effect we found of GNG training on food evaluations in the impulsive choice experiment.
Limitations and Future Directions
Our study used snack items to change food choice. This may seem rather paradoxical since the introduction of this study focusses on the prevention of overeating and obesity.
However, the reason we used snack items in this study was because previous research also used snacks (e.g. Schönberg et al., 2014). In order to investigate the effect of GNG training on food choice in a systematic manner, it was important not to change too many
characteristics of the previous studies. Otherwise, it would be impossible to explain a lack of effect. Since we found an effect of GNG training on impulsive food choice, it would be a logic next step to investigate the effect using healthy food instead of snack items (or a combination of the two).
Furthermore, we excluded a lot of participants (especially in the first experiment), which could have had an impact on the power of our study. For future research it would be wise to set the sample size slightly above the minimum, to make sure you still have enough participants left after exclusion. Given the sample size of our study, we were not able to compare the two experiments. If possible, it is recommended to include more participants in order to compare the experiments.
Also, in order to check for the effects of the choice task on the WTP, it is
recommended to include several experiments in which the order of the tasks varies. Another option is to include an experiment with the choice task, and one without the choice task. Conclusion
In the present research, we examined if GNG training could influence impulsive and deliberative food choice. We showed that GNG training only affects impulsive choice (i.e., when little time is available to make the choice), and we illustrated that the effect of GNG training on choice in ours studies were not driven by evaluation, but possibly by differences in attention. Given the fact that our study contradicts a lot of previous findings on the effect of GNG training in the eating domain, more research is needed to disclose the effect of response inhibition training on choice. However, our study shows that GNG training can be
used as a means to influence impulsive food choice. With our study, we contributed to the knowledge on the effect of GNG training on food choice.
References
Allom, V. (2014). The role and trainability of executive function in the context of healthy eating. Sydney: University of Sydney.
Allom, V., Mullan, B., & Hagger, M. (2015). Does inhibitory control training improve health behaviour? A meta-analysis. Health Psychology Review, 10, 168-186. doi:
10.1080/17437199.2015.1051078
Berthoud, H-R. (2007). Interaction between the “cognitive” and “metabolic” brain in the control of food intake. Psychology & Behavior, 91, 486-498.
doi:10.1016/j.physbeh.2006.12.016
Blum, K., Cull, J. G., Braverman, E. R., & Comings, D. E. (1996). Reward deficiency syndrome. American Scientist, 84, 132-145.
Cameron, J. D., Goldfield, G. S., Finlayson, G., Blundell, J. E., & Doucet, E. (2014). Fasting for 24 hours heightens reward from food and food-related cues. PLOS ONE, 9, e85970. doi:10.1371/journal.pone.0085970
Chen, Z., Veling, H., Dijksterhuis, A., & Holland, R. W. (2016). How does not responding to appetitive stimuli cause devaluation: Evaluative conditioning or response inhibition? Manuscript submitted for publication.
Fazio, R. H. (1990). Multiple processes by which attitudes guide behavior: The MODE model as an integrative framework. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 23, pp. 75‐109). New York: Academic Press.
Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press.
Friese, M., Wänke, M., & Plessner, H. (2006). Implicit consumer preferences and their influence on product choice. Psychology & Marketing, 23, 727-740.
Fujita, K., & Han, H. A. (2009). Moving beyond deliberative control of impulses: The effects of construal levels on evaluative associations in self-control. Psychological Science, 20, 799-804.
Herman, C. P., & Polivy, J. (1980). Retrained eating. In A. J. Stunkard (Ed.), Obesity (pp. 208–225). Philadelphia: Saunders.
Hollands, G. J., Prestwich, A., & Marteau, T. M. (2011). Using aversive images to enhance healthy food choices and implicit attitudes: An experimental test of evaluative conditioning. Health Psychology, 30, 195-203. doi:10.1037/a0022261
Houben, K., & Jansen, A. (2011). Training inhibitory control: A recipe for resisting sweet temptations. Appetite, 56, 345-349. doi:10.1016/j.appet.2010.12.017
Houben, K., & Jansen, A. (2015). Chocolate equals stop. Chocolate-specific inhibition training reduces chocolate intake and go associations with chocolate. Appetite, 87, 318-323. doi:10.1016/j.appet.2015.01.005
Houben, K., Havermans, R. C., Nederkoorn, C., & Jansen, A. (2012). Beer á no-go: Learning top stop responding to alcohol cues reduces alcohol intake via reduced affective associations rather than increased response inhibition. Addiction, 107, 1280-1287. doi:10.1111/j.1360-0443.2012.03827.x
Houben, K., Nederkoorn, C., Wiers, R. W., & Jansen, A. (2011). Resisting temptation: Decreasing alcohol-related affect and drinking behavior by training response inhibition. Drug and Alcohol Dependence, 116, 132-136.
doi:10.1016/j.drugalcdep.2010.12.011
Krajbich, I., Lu, D., Camerer, C., & Rangel, A. (2012). The attentional drift-diffusion model extends to simple purchasing decisions. Frontiers in Psychology, 3, 1-18. doi:
Lawrence, N. S., O’Sullivan, J., Parslow, D., Javaid, M., Adams, R. C., Chambers, C. D., Koss, K., & Verburggen, F. (2015). Training response inhibition to food is associated with weight loss and reduced energy intake. Appetite, 95, 17-28.
doi:10.1016/j.appet.2015.06.009
Lawrence, N. S., Verbruggen, F., Morrison, S., Adams, R. C., & Chambers, C. D. (2015). Stopping to food can reduce intake: Effects of stimulus-specificity and individual differences in dietary restraint. Appetite, 85, 91-103. doi:10.1016/j.appet.2014.11.006 Marteau, T. M., Hollands, G. J., & Fletcher, P. C. (2012). Changing human behavior to
prevent disease: The importance of targeting automatic processes.
Mostofsky, S. H., & Simmonds, D. J. (2008). Response inhibition and response selection: Two sides of the same coin. Journal of Cognitive Neuroscience, 20, 751-761. doi:10.1162/jocn.2008.20500
Peirce, J. W. (2007). PsychoPy – Psychophysics software in Python. Journal of Neuroscience Methods, 162, 8-13. doi:10.1016/j.jneumeth.2006.11.017
Ratcliff, R. (1978). A theory of memory retrieval. Psychological Review, 85, 59–108. Rogers, P. J., & Hardman, C. A. (2015). Food reward: What is it and how to measure it?
Appetite, 90, 1-15. doi:10.1016/j.appet.2015.02.032
Schönberg, T., Bakkour, A., Hover, A. M., Mumford, J. A., Nagar, L., Perez, J., & Poldrack, R.A. (2014). Changing value through cued approach: an automatic mechanism of behavior change. Nature Neuroscience, 17, 625-630. doi:10.1038/nn.3673
Shin, A. C., & Berthoud, H-R. (2011). Food reward functions as affected by obesity and bariatric surgery. International Journal of Obesity, 35, S40-S44.
doi:10.1038/ijo.2011.147
Stice, E., Spoor, S., Bohan, C., Veldhuizen, M. G., & Small, D. M. (2008). Relation of reward from food intake and anticipated food intake to obesity: A functional magnetic
resonance imaging study. Journal of Abnormal Psychology, 117, 924-935. doi:10.1037/a0013600
Stice, E., Spoor, S., Ng, J., & Zald, D .H. (2009). Relation of obesity to consummatory and anticipatory food reward. Psychology & Behavior, 97, 551-560.
doi:10.1016/j.physbeh.2009.03.020
Strack, F., & Deutsch, R. (2004). Reflective and impulsive determinants of social behavior. Personality and Social Psychology Review, 8, 220-247.
doi:10.1207/s15327663jcp1603_2
Stroebe, W. (2008). Dieting, overweight, and obesity: Self-regulation in a food-rich environment. Washintgon: American Psychological Association.
Veling, H., Aarts, H., & Papies, E.K. (2011). Using stop signals to inhibit chronic dieters’ responses toward palatable foods. Behaviour Research and Therapy, 49, 771-780. doi:10.1016/j.brat.2011.08.005
Veling, H., Aarts, H., & Stroebe, W. (2013a). Stop signals decrease choices for palatable foods through decreased food evaluation. Frontiers in Psychology, 4, 875.
doi:10.3389/fpsyg.2013.00875
Veling, H., Aarts, H., & Stroebe, W. (2013b). Using stop signals to reduce impulsive choices for palatable unhealthy foods. British Journal of Health Psychology, 18, 354-368. doi:10.1111/j.2044-8287.2012.02092.x
Veling, H., Chen, Z., Tombrock, M., Verpaalen, I., Schmitz, L., Dijksterhuis, A., & Holland, R. W. (2016). Creating impulsive choices for sustainable and healthy food.
Unpublished manuscript, Radboud University, Nijmegen, NL.
Veling, H., Holland, R. W., & Van Knippenberg, A. (2008). When approach motivation and behavioral inhibition collide: Behavior regulation through stimulus devaluation.
Journal of Experimental Social Psychology, 44, 1013-1019. doi:10.1016/j.jesp.2008.03.004
Veling, H., Van Koningsbruggen, G. M., Aarts, H., & Stroebe, W. (2014). Targeting impulsive processes of eating behavior via the internet. Effects on body weight. Appetite, 78C, 102-109. doi:10.1016/j.appet.2014.03.014
Verbruggen, F., & Logan, G. D. (2008a). Automatic and controlled response inhibition: Associative learning in the go/no-go and stop-signal paradigms. Journal of Experimental Psychology: General, 137, 649–672. doi:10.1037/a0013170
Verbruggen, F., & Logan, G. D. (2008b). Response inhibition in the stop-signal paradigm. Trends in cognitive sciences, 12, 418-424. doi:10.1016/j.tics.2008.07.005
WHO (2005a). Dietary intake of fruit and vegetables and risk of diabetes mellitus and cardiovascular diseases. Geneva: World Health Organization.
WHO (2005b). Dietary intake of fruit and vegetables and management of body weight. Geneva: World Health Organization.
WHO (2015a). Fact sheets - Obesity and overweight. Retrieved January 26, 2016, from http://www.who.int/mediacentre/factsheets/fs311/en/
WHO (2015b). World Health Statistics 2015. Geneva: World Health Organization. Young, D. L., Goodie, A. S., Hall, D. B., & Wu, E. (2012). Decision making under time
pressure, modelled in a prospect theory framework. Organizational Behavior and Human Decision Processes, 118, 179-188. doi:10.1016/j.obhdp.2012.03.005
Ziauddeen, H., Subramaniam, N., Gaillard, R., Burke, L. K., Farooqi, I. S., & Fletcher, P. C. (2012). Food images engage subliminal motivation to seek food. International Journal of Obesity, 36, 1245-1247. doi:10.1038/ijo.2011.239
Table 1.
Participants’ characteristics per experiment.
Experiment N Excluded N Sex (F/M) Age BMI RS Hours not eaten Memory recall
N Percentage M (SD) Min. Max. M (SD) M (SD) M (SD) Min. Max. M (SEM)
1 31 9 20/2 90.9/9.1 22.1 (2.60) 18 26 21.9 (1.88) 11.3 (3.88) 5.6 (3.32) 3 13 69.7% (2.4%)
2 31 5 19/7 73.1/26.9 21.5 (2.18) 18 26 22.4 (2.91) 11.2 (5.02) 6.7 (4.37) 3 16 67.6% (2.9%)
Table 2.
Mean WTP during the auctions for high and low value go and no-go items.
Auction High value Low value
Go No-go Go No-go
M SD M SD M SD M SD
1 (pre-training) 1.09 .29 1.09 .30 .27 .27 .27 .27
Table 3.
Participants’ performance during GNG training.
Experiment Go accuracy No-go accuracy Go RT
M SEM M SEM M SEM
1 99.3% 0.3% 94.4% 1.4% 438.0 ms 11.1 ms
2 99.1% 0.3% 95.4% 0.8% 437.3 ms 10.6 ms
Table 4.
Reaction times(ms) during the choice task in.
Experiment All trials Go items No-go items
M SEM M SEM M SEM
1 1326 19 1327 25 1325 28
1. Auction #1
2. GNG
3. Food Choice Task
High go vs High no-go (experimental trials) Low go vs Low no-go (experimental trials) High go vs Low go (filler trials)
High no-go vs Low no-go (filler trials)
4. Auction #2
5. Memory Recall Task
Figure 2. Mean percentage of the choices for go items for high and low value items within the deliberative choice experiment. Error bars represent S.E. mean.
0 10 20 30 40 50 60
0 10 20 30 40 50 60 70
High value go Low value go
Figure 3. Mean percentage of the choices for go items for high and low value items withtin the impulsive choice experiment. Error bars represent S.E. mean.
10 20 30 40 50 60 70 80 90 16 18 20 22 24 26 28 30 P er ce nt ag e of c hoos ing g o ove r no -g o itms BMI
10 20 30 40 50 60 70 80 90 0 5 10 15 20 25 P er ce nt ag e of c hoos ing g o ove r no -g o ite ms RS
0 20 40 60 80 100 120 0 5 10 15 20 25 P er ce nt ag e of c hoos ing g o ove r no -g o ite ms RS
Figure B3. Scatterplot of the RS and the percentage of choosing low value go over low value no-to items.
Appendix A
Exploratory analyses within the deliberative choice experiment
It was explored whether there was a difference in the effect of GNG training in overweight or obese individuals. There was no significant correlation between BMI and the percentage of choosing go items over no-go items, r = -.25, p (two-tailed) = .256. The correlation between BMI and the percentage of choosing go items over no-go items was also separately tested for high and low value items. There was no significant correlation between BMI and the
percentage of choosing go items over no-go items for neither high value items (r = -.11, p (two-tailed) = .638), and low value items (r = -.21, p (two-tailed) = .341).
Furthermore, it was explored whether there was a difference in effect of GNG training in restraint eaters. There was no significant correlation between the RS and the percentage of choosing go items over no-go items, r = -.04, p (two-tailed) = .846. The correlation between the RS and the percentage of choosing go items over no-go items was also separately tested for high and low value items. There was no significant correlation between the RS and the percentage of choosing go items over no-go items for neither high value items (r = -.29, p (two-tailed) = .191), and low value items (r =.15, p (two-tailed) = .498).
Appendix B
Exploratory analyses within the impulsive choice experiment
Just as in Experiment 1, it was explored whether there was a difference in the effect of GNG training in overweight or obese individuals. There was a marginally significant correlation between BMI and the percentage of choosing go items over no-go items, r = .39, p (two-tailed) = .050 (see Figure B1 for the scatterplot). Although the effect is marginally significant, these findings are in line with previous findings (Veling et al., 2014). In Figure 4 the
scatterplot is shown. The correlation between BMI and the percentage of choosing go items over no-go items was also separately tested for high and low value items. There was no significant correlation between BMI and the percentage of choosing go items over no-go items for neither high value items (r = -.24, p (two-tailed) = .246), and low value items (r =.33, p (two-tailed) = .105).
There was no significant correlation between the RS and the percentage of choosing go items over no-go items, r = .35, p (two-tailed) = .083 (See Figure B2 for the scatterplot). The correlation between the RS and the percentage of choosing go items over no-go items was also separately tested for high and low value items. There was no significant correlation between the RS and the percentage of choosing go items over no-go items for neither high value items (r = .12, p (two-tailed) = .555). However, there was a marginally significant correlation between the RS and the percentage of choosing low value go items over low value no-go, r = -.36, p (two-tailed) = .069. In Figure B3 the scatterplot is shown.
