Food temptations measured by using the mobile version of the Approach Avoidance Task.
Elsa van der Meer
Master thesis psychology, specialization Economic and Consumer Psychology Institute of Psychology
Faculty of Social and Behavioral Sciences – Leiden University Date: 15-05-2017
Student number: s1741306
First examiner of the university: Hilmar Zech Second examiner of the university: Lotte van Dillen
Abstract
In the current research, we tested if there was a difference in approach bias towards desirable food and non-desirable food, measured with the mobile version of the Approach Avoidance Task (MAAT). We also made a distinction between the approach bias towards unhealthy and healthy desirable food. Besides testing the approach bias, we tested if body mass index (BMI), the Power of Food Scale (PFS) score, hunger level, and having healthy food habits correlated with an approach bias towards desirable and unhealthy desirable food. During the MAAT, pictures of food and objects were presented to the participants, to which they responded with push or pull movements. The food picture stimuli were divided into four categories: unhealthy desirable, healthy desirable, unhealthy desirable, and healthy non-desirable. The approach bias in the current research is defined as the difference in reaction time between push and pull movements. We expected that people have stronger approach biases towards desirable food than towards non-desirable food. We also expected that participants have stronger approach biases towards unhealthy desirable food than towards healthy desirable food. However, the results of the current research did not support our
expectations. We did not find any differences in approach biases, nor did we find a correlation between the approach biases towards different types of stimuli and BMI, PFS score, hunger level, and food habits. The stimuli set used might account for a large part of the unexpected results. Despite the lack of confirmed results in the current research, the MAAT seems to be a promising tool and, with the use of proper stimulus sets, it could be of great use to determine what type of food is tempting to people. This can be very important when treating health issues like obesity. Additionally, the MAAT can be extremely useful in many different fields such as marketing and psychology. However, future research is necessary to confirm the results found in earlier approach-avoidance research.
Food temptations in the real world
The current research studies food temptations, as this is an important factor of health issues nowadays. According to the World Health Organization (WHO), worldwide obesity has more than doubled over the last thirty years (WHO, 2016). Obesity has become such a problem that now more people die from being overweight than underweight. In modern society, each person makes around 200 food decisions per day (Voedingscentrum, 2017), mainly because people can be exposed to (unhealthy) cues via food advertisements on television and radio at every hour of the day (Havermans, 2013). This growing field of food marketing has likely impacted people to experience many food-related desires, temptations, and cravings everyday. Food cravings can be best defined as intense desires or feelings to consume a specific food without necessarily being physically hungry (Hill, 2007, Weingarten and Elston, 1990). Hofmann, Förster, Baumeister, and Vohs (2012) described a desire as a feeling of wanting or having to do something. When a desire becomes problematic, meaning it conflicts with other goals, Hofmann et al. (2012) call it a temptation. According to
Hofmann et al. (2012), people experience desires about half of the time they are awake. People not only experience many desires every day, but almost half of these desires conflict with other goals, and therefore become temptations. These desires and temptations can be related to, for example, ideas of drinking, eating, socializing, working, or sleeping. Hofmann, Vohs, and Baumeister (2012) showed that eating temptations are experienced throughout the whole day and far more than other temptations such as sleeping, working, sports participation, and alcohol.
Because so many eating desires and temptations are experienced every day, it is a widely studied subject in psychology. Most of these studies have been conducted inside artificial laboratory environments. An important disadvantage of laboratory environments is that they lack external and ecological validity. Moreover, it takes much time and money to test participants in a laboratory (Reips, 2000). For this reason, the current research tested a method that allows research to be done outside of the laboratory. Food desires were investigated with the use of a smartphone application. Using smartphones improves
ecological validity in two ways. Firstly, smartphones are very easily accessible; they play a prominent role in people’s lives. Smartphones therefore allow access to domains of
behavioral data that were previously not accessible or only measurable by using observations and self-reports. Secondly, tests with smartphones can be done without researchers being present (Raento, Oulasvirta and Eagle, 2009). Furthermore, according to Miller (2012), smartphones have huge potential to collect precise and objective data in psychological
research. Smartphones can be used to gather real-life data outside of the laboratory while also having the advantages of automatic data entry and easily replicable lab experiments. Because there are so many advantages of using smartphones in psychological research, Zech (2015) developed a mobile version of the Approach Avoidance Task (AAT) for his thesis in the hopes that it might provide more insights into the fields of temptation and addiction.
The AAT is a task that measures approach and avoidance biases to certain types of stimuli. The AAT requires participants to respond to pictures as quickly as possible by
pushing certain picture types away or pulling certain picture types towards themselves (Rinck and Becker, 2007). According to Solarz (1960), pushing objects or stimuli away from oneself is associated with avoidance, while pulling objects or stimuli towards oneself is associated with approaching. The AAT is often used to test if people have an approach bias towards certain stimuli. An approach bias is detected when a participant has a behavioral tendency to more quickly approach rather than avoid stimuli. If this difference in reaction time is bigger for one particular type of stimuli, a person has an approach bias towards these stimuli. According to Rinck and Becker (2007), the rationale behind the AAT is that it reveals spontaneous behavioral responses. It detects spontaneous avoidance towards unpleasant stimuli and spontaneous approach towards pleasant stimuli. When people have a negative attitude towards one of the stimulus categories, the reaction time towards compatible combinations (avoiding unpleasant stimuli and approaching pleasant stimuli) is shorter than towards incompatible combinations (avoiding pleasant stimuli and approaching unpleasant stimuli). This effect is found in different phenomena such as arachnophobia (Rinck and Becker, 2007), alcohol addiction (Wiers et al., 2014), and food temptations (Dickson, Kavanagh and MacLeod, 2016). In the current research, the AAT was used to measure food desires. The AAT was originally completed with a joystick and a computer but, as indicated above, Zech (2015) developed a mobile version of the AAT. For the current research, we measured food desires with the mobile version of the AAT. This research was designed to see if people have an approach bias towards certain types of food stimuli and to test if the mobile version of the AAT is a feasible technique for measuring food desires and temptations.
More and more people are having health issues because of overeating (WHO, 2016), and the food marketing industry is growing simultaneously (Voedingscentrum, 2017). Because of these changes, it is useful to determine if the AAT also detects spontaneous approach and avoidance biases towards different types of food stimuli. If people indeed show an approach bias towards different types of stimuli, diet programs can be adjusted to be more efficient. For example, dieticians can use the AAT to show what type of stimuli tempt people,
whereupon a specific plan can be made for how to handle those temptations. The AAT can also give more insights into what types of people are more easily tempted. For example, are people with a high or low body mass index (BMI), or people with unhealthy or healthy eating habits more easily tempted by pictures of desirable food? When this information is known, dieticians can offer more specific services to their clients. To know what types of people are more easily tempted and what types of food tempt people is not only helpful for dieticians, but marketers can also use this knowledge to influence their (potential) customers. For example, marketers can test certain stimuli on their target groups, thereby learning what kind of pictures they should use in their marketing campaigns to be successful in attracting
customers to their brand. Those examples show that more knowledge on which stimuli tempt people and how, can be of great use in several professional fields.
In the current research, we determine if there is an approach bias towards desirable food stimuli, based on the findings of earlier studies which showed a natural impulsive reaction and approach bias towards palatable foods (Dickson et al., 2016 and Veling, Aarts and Stroebe, 2013). Veenstra and De Jong (2010) found that people have a stronger automatic liking association for high-fat foods than for low-fat foods. This preference for high-fat foods is in line with the results of the study by Raghunathan, Naylor, and Hoyer (2006), as they showed that people have an implicit intuition that tells them unhealthy food is tasty. The above-described studies did not all make a specific distinction between healthy and unhealthy stimuli in the desirable and non-desirable categories. They used healthy and unhealthy
categories but they did not divide the healthy and unhealthy stimuli into desirable and non-desirable categories. However, many non-desirable stimuli can be healthy (e.g., a fruit salad), and many non-desirable stimuli can be unhealthy (e.g., a plain piece of white bread). For this reason we divided our stimuli into four categories (healthy desirable, unhealthy desirable, healthy non-desirable, and unhealthy non-desirable) to be able to detect a potential approach bias. If people are actually tempted by pictures of healthy desirable food as well as by pictures of unhealthy desirable food, a different approach might be possible for solving health issues. Therefore, we aimed to see if the approach bias is visible only towards unhealthy desirable stimuli or whether it also appears with healthy and desirable food stimuli. In the current research, we looked at the approach bias towards desirable and unhealthy desirable food stimuli because earlier studies showed an approach bias towards desirable, palatable, and unhealthy food (Dickson et al, 2016, Veenstra and de Jong and Veling et al., 2013).
In line with earlier studies, the first research question of this study is the following: do people have an approach bias towards desirable food? The approach bias in this research is
defined as the difference in reaction time between push and pull movements. This hypothesis was tested by looking at and comparing the approach bias for desirable food and
non-desirable food as measured with the mobile AAT (MAAT). Because earlier research shows a preference for palatable and desirable food, our first hypothesis is as follows:
H1: The approach bias is larger for desirable stimuli than for non-desirable stimuli.
Multiple studies (Veenstra and De Jong, 2010, Raghunathan et al., 2006 and Veling et al., 2013) showed that people have preferences for desirable and unhealthy food. However, since desirable food can be healthy or unhealthy, we want to know if there is a difference in approach bias between healthy and unhealthy desirable stimuli as well. The research of Veenstra and de Jong (2010) showed an automatic liking association towards unhealthy food, and therefore the second hypothesis is
H2: The approach bias is larger for unhealthy desirable stimuli than for healthy desirable stimuli.
It is useful to know if people in general have an approach bias towards unhealthy or healthy desirable food for practical (health or marketing) reasons, but differences between people can also be influential to understanding why one person is more vulnerable to food temptations than another. The study of Cleobury and Tapper (2013) revealed that people react to internal and external cues when it comes to eating and snacking behavior. Polivy, Herman, and Girz (2011) said that those cues are often situations that have been associated in the past with food intake or internal states like affect, arousal, and stress. Lowe and Butryn (2007) developed the Power of Food Scale (PFS), a test which measures the psychological impact that food plenitude environments haveon individuals. According to Schüz, Schüz, and Ferguson (2015), the score on the PFS moderates how people react towards snacking cues during everyday food choices. People with a higher PFS score are thus more likely to succumb to temptations of desirable food. Because the PFS-score can influence the way people react to temptations, our third hypothesis is as follows:
H3: The approach bias towards desirable food increases with the PFS score. The higher the PFS score is, the larger the approach bias towards desirable food. Not every person has the same eating habits, which might not come as a surprise. But can this also influence how people react to food stimuli? According to Lin, Wood, and Monterosso (2016) the answer to that question is yes, because their study showed that when healthy eating becomes a habit, it protects against temptations. At least in situations where people are not deliberating because people fall back on their habits when the specific behavior
is more available in memory. When certain behavior is performed repeatedly with similar goals and in the same context, the information stored will become more available because it is rooted in memory (Papies, 2016). Because eating is a habitual behavior and therefore
regularly performed, eating behavior is highly available in memory. Additionally, the research of Verhoeven, Adriaanse, Evers, and De Ridder (2012) showed that habit strength (for
unhealthy snacks) is the most important predictor of the daily intake of unhealthy snacks. This means that the stronger the habit to eat unhealthy snacks, the more people succumb to
temptations of unhealthy snacks. According to this and the research of Lin et al. (2016), participants with healthy eating habits should have less trouble resisting unhealthy desirable stimuli because their healthy eating habits protect against the temptation of unhealthy desirable stimuli. Therefore the following hypotheses are formulated:
H4a: The approach bias towards unhealthy desirable stimuli decreases with healthier food habits.
H4b: The approach bias for non-desirable stimuli stays the same regardless the eating habits of a person.
There is a saying which most people have experienced themselves: never go grocery shopping when hungry, because hungry people will buy products they do not really need. According to Mogg, Bradley, Hyare, and Lee (1998), this is because hunger, a non-emotional motivational state, is associated with a selective-attention bias. Mogg et al. (1998) showed in their experiment that hungry participants had a greater attentional bias for food-related words. A non-emotional state such as hunger can bias information processing by selective attention for stimuli that are relevant to the motivational state. In this way, hungry people might react differently than satiated people in the current research with the MAAT. Also, according to Loeber, Grosshans, Hepertz, Kiefer, and Herpertz (2013), hunger can induce an approach bias towards food-associated cues in normal-weight individuals. Furthermore, Berridge, Ho, Richard, and DiFeliceantonio (2010) also stated that when people are hungry, palatable foods become even more attractive. As an approach bias towards palatable food cues, might drive people to overeat and consume food it can have a strong influence on people’s daily lives and their health, therefore the following hypothesis is formulated:
H5: The approach bias towards desirable stimuli increases when participants are hungrier.
Next to hunger, eating habits, and the PFS-score, body mass index (BMI) might correlate with how people react towards food stimuli (Batterink, Yokum and Stice, 2010). According to WHO (2017), BMI is an index of weight-for-height that is often used to classify
adults as underweight, overweight, or obese. The research of Batterink et al. (2010) shows that a higher BMI appears with greater impulsivity towards food cues. Additionally,
according to Havermans, Giesen, Houben, and Jansen (2011), people with a high BMI have more difficulty avoiding unhealthy desirable food stimuli. When both a greater impulsivity towards food cues and more difficulty in avoiding desirable food with an increased BMI also apply to the participants in the current research, the BMI score of participants might influence the approach bias. Therefore, the following hypothesis is presented:
H6: The approach bias towards desirable food stimuli increases with the increase of BMI. The higher the BMI score, the higher the approach bias.
Method
The current research was part of a broader study that consisted of lab and field research with the use of the MAAT. This paper mainly regards the results found in the lab study. However, in this section of the paper, the procedure and materials for the field study are also discussed. The current research studied whether an approach bias existed towards desirable and unhealthy desirable food in comparison with non-desirable and healthy
desirable food in a laboratory setting. The research looked at the difference in reaction times when participants approached and avoided different types of food stimuli. The stimuli were categorized as desirable or non-desirable stimuli and healthy or unhealthy stimuli. Possible moderators were investigated, such as the level of hunger, food habits, BMI, and the PFS score. The lab study took place to test the application, instructions, and stimulus sets and to calculate an effect size on which the number of participants for the field study could be based. The results of the lab study can be used for comparison with the results of the field study.
Research design
The current observational research was cross-sectional and had a within-subject design. Each participant completed two blocks of the MAAT. During one block, participants approached food/avoided objects and during the other block they avoided food/approached objects. Each participant also completed several questionnaires. The ethics committee of Leiden University approved the procedure.
The independent variables were ‘picture type desirability’ (desirable and non-desirable), ‘picture type healthiness’ (unhealthy desirable and healthy desirable) and
‘response direction’ (pull or push). The dependent variables were the reaction times for each category and the approach bias (the difference between push and pull reaction times).Possible moderators were the participants’ score on the PFS, their food habits, BMI, and hunger level.
Participants
The participants were recruited from Leiden University via flyers and posters, the online page SONA, and via different Leiden University social media pages in return for 1 credit or €3.50 monetary reward. The participants were between the ages of 18 and 29 years. In total, 58 people participated in the lab study. In the field study, only people with an Android smartphone can participate because the application is only available for Android operating systems.
Materials and instruments
Participants received a link via email to download the mobile application. Several questions about hunger and when they had last eaten were asked. Participants also answered several demographic questions about age, weight, height, and gender. When participants were doing the MAAT, they moved the smartphone away from their body in order to signify
avoiding stimuli, and they moved the smartphone towards their body to represent approaching stimuli. In the field study, the proper movements were shown via an instructional video.
The stimuli of desirable food (healthy and unhealthy), non-desirable food (healthy and unhealthy), and objects that were shown in the MAAT were from the food-pics database of Blechert, Meule, Buschan, and Ohla (2014). Two blocks of 60 pictures each were created from this database. Each block began with 15 practice trials in which the participants received feedback in the form of a red or green screen. Subsequently, the real stimuli were shown. After 30 stimuli, participants could have a short break. Each block consisted of 10 pictures of healthy desirable food, 10 pictures of unhealthy desirable food, 10 pictures of healthy non-desirable food, 10 pictures of unhealthy non-non-desirable food, and 20 pictures of objects. All of the stimuli were shown for a maximum of 2 seconds. The congruent/incongruent order was randomized within each participant. The stimuli were selected on the craving and
recognizability scores of the food-pics database. The cutoff point chosen was one standard deviation above (for desirable stimuli) and one standard deviation below (for non-desirable stimuli) the mean. Subsequently, the stimuli were divided into healthy and unhealthy categories. Figure 1 shows examples of two different types of stimuli.
Figure 1. Examples of healthy desirable and unhealthy non-desirable food stimuli. Reprinted from “Food-pics: an image database for experimental research on eating and appetite,” by Blechert, J., Meule, A., Busch, N. A., and Ohla, K., 2014, Frontiers in Psychology, 5.
The participants also completed a picture-rating task. In this task, all of the stimuli were showed again and the participants answered on a 5-point scale how desirable the
pictures were to them. The participants completed the 21 questions of the Power of Food Scale (Lowe and Butryn, 2007) and several questions about their eating habits. Four additional questions about self-control from the impulsive-control scale of Neubach and Schmidt (2007) and three questions about health motives from The Eating Motivation Survey (Renner et al., 2012) will be asked in the field study.
Procedure lab study
After providing informed consent, the participants opened the application on their smartphones (in the case that a participant had an iPhone on which the MAAT did not function, another smartphone was ready to use in the laboratory) and started the procedure. Dutch participants completed the study in Dutch, whereas participants with other nationalities completed the study in English. The participants first answered the questions about hunger and the last time they had eaten, and then they completed the demographic questions. The participants then continued with the MAAT. After the participants finished the second block of the MAAT, they continued with the picture-rating task. Next, the participants completed the PFS, and finally, they answered questions about their eating habits. Afterwards the participants were debriefed and received their reward (1 credit or €3.50). The total procedure took approximately 20 minutes.
Procedure field study
In most ways, the procedure for the field study will be similar to the procedure for the lab study and therefore only the differences in procedure are discussed. After giving informed consent, participants will receive instructions on what the study expects from them and how many times they should participate. Participants are supposed to do the MAAT twice at different time points, once before a meal and once after a meal. An instructional video on how to do the right push and pull movements will be shown. For both rounds, the MAAT will be preceded by the questions about hunger, self-control, and the last time the participants have eaten. A push messages will be send to the participants when they need to do the MAAT for the second time. When the last part of the study is completed, the participants will receive the debriefing and explanations on how to receive the reward (2 credits or €6.50). The
participants can also join a lottery to win a €30 gift card for a restaurant in Leiden. To join the lottery, they need to provide their email addresses. The total procedure takes approximately 1 hour.
Results
In this part of this paper, we discuss only the results of the lab study. In total, 58 people participated in the lab study, but because of technical failures, the data of only 50 participants could be used. Of these 50 participants, 2 were excluded because they had an error rate higher than 25%. The total number of participants included in analysis was
therefore 48 (39 females, 9 males). The characteristics of the participants are shown in Table 1. Participants ranged in age from 18 to 29 years old with an average of 22.42 years (SD = 2.78). Participants BMI ranged from 16.90 to 27.18 with an average of 21.76 (SD = 2.68). Four participants (8.3%) indicated they were on a diet and 10 participants were vegetarians (20.8%). Out of the 48 participants, 27 (56.3%) of them completed the study in Dutch; the other 21 (43.8%) participants used the English application.
Table 1. Participant characteristics
Characteristics Total group Male Female
N (%) 48 (100%) 9 (18.8%) 39 (81.3%) Mean age (SD) 22.42 (2.78) 23.11 (3.30) 22.26 (2.67) On a diet - yes 4 1 3 Mean BMI (SD) 21.76 (2.68) 23.71 (2.21) 21.31 (2.60) Vegetarian (%) 10 (20.8) 2 8 Language Dutch (%) 27 (56.3) 4 23 Language English (%) 21 (43.7) 5 16
The reaction times on the stimuli were aggregated with the median by participant, picture type, and response type. All alpha levels for the hypothesis tests were set at .05. The sample used was not random as it consisted of mostly students from Leiden University. The design of the study took care to preserve the independence of observations. People did not see each other when participating, and they were asked not to spread information about the hypothesis to other persons who were still going to participate. However, because we used a within-subject design, the observations are always partially dependent. This problem is solved by using repeated measures ANOVAs when analyzing the data. The assumptions for the repeated measures ANOVAs were not violated unless stated otherwise. There were no significant deviations from normality according to the Kolmogorov-Smirnov tests. As there were only two conditions in our analyses, we did not need to look for sphericity. According to
Cook’s D values, the few outliers that were detected did not have any influence, therefore we did not remove them from the dataset. Preliminary analyses were also conducted for all regression analyses to ensure no violation of the assumptions of normality, linearity, and homoscedasticity.
Approach bias towards desirable and non-desirable stimuli
To test the hypothesis that the approach bias is larger for desirable stimuli than for non-desirable stimuli, the reaction times were analyzed with a 2x2 (picture type desirability: desirable versus non-desirable stimuli, and response type: push versus pull) repeated measures ANOVA.
The results from the repeated measures ANOVA showed a significant main effect of picture type desirability F(1,47) = 34.13, p < .001, ηpartial2 = .421, with a longer mean reaction
time for non-desirable stimuli (M = 467.30, SD = 9.13) than for desirable stimuli (M = 443.66, SD = 9.07). There was also a significant main effect of response type F(1,47) = 14.47, p < .001, ηpartial
2 = .235, with a longer mean reaction time for push movements (M = 468.78, SD =
9.10) than for pull movements (M = 442.18, SD = 9.96) (see Figure 2). Subsequently, to be able to find a difference in approach bias between the two types of stimuli, we expected to find a significant interaction effect between picture type desirability and response type. However, there was no significant interaction effect between picture type desirability and response type F(1,47) = .07, p = .789, ηpartial2 = .002. Figure 3 also shows that there was no
difference between the approach bias towards desirable and non-desirable stimuli.
Figure 2. Reaction times and standard errors on the MAAT.
360 380 400 420 440 460 480 500 Non-desirable Desirable R eac ti on ti me s (ms ) Picture type Pull Push
Figure 3. Approach biases (difference between push and pull reaction times) and standard errors on two different types of stimuli.
Approach bias towards unhealthy desirable stimuli and healthy desirable stimuli To test the hypothesis that the approach bias is larger for unhealthy desirable stimuli than for healthy desirable stimuli, the reaction times were analyzed with a 2x2 (picture type healthiness: unhealthy desirable stimuli, versus healthy desirable stimuli and response type: push versus pull) repeated measures ANOVA.
The repeated measures ANOVA showed there was a significant main effect of picture type healthiness F(1,47) = 21.09, p < .001, ηpartial2 = .310, with a longer mean reaction time for
unhealthy desirable stimuli (M = 457.03, SD = 9.58) than for healthy desirable stimuli (M = 435.13, SD = 9.005). There was also a significant main effect of response type F(1,47) = 14.54, p < .001, ηpartial2 = .236, with a longer mean reaction time for push movements (M =
460.27, SD = 9.85) than for pull movements (M = 431.89, SD = 9.60) (see Figure 4).
Additionally, to be able to find a difference in approach bias between the two types of stimuli, we expected a significant interaction effect between picture type healthiness and response type. However, there was no significant interaction effect between picture type healthiness and response type F(1,47) =1.50, p = .227, ηpartial2 = .031. This result contradicts
our hypothesis that the approach bias would be larger for unhealthy desirable stimuli than for healthy desirable stimuli.
10,0 15,0 20,0 25,0 30,0 35,0 40,0 Desirable Non-desirable A p p roac h b ias (ms ) Picture type Desirable Non-desirable
Figure 4. Reaction times and standard errors on the MAAT.
Also Figure 5 shows that there is no difference between the approach bias towards unhealthy desirable and healthy desirable stimuli. Even if the effect had been significant, the result was in the opposite direction as expected – a larger approach bias for healthy desirable stimuli then for unhealthy desirable stimuli.
Figure 5. Approach biases (difference between push and pull reaction times) and standard errors for two types of stimuli.
The PFS score and approach bias
To test the hypothesis that the approach bias towards desirable food would increase with the increase of the PFS score, a regression analysis was conducted with the PFS score as the independent variable and the approach bias towards desirable food as the dependent variable. Results of the regression analysis showed that the PFS score was not a significant predictor for the approach bias towards desirable food F(1,46) = .142, p = .708, R2 = .003.
360 380 400 420 440 460 480 500
Unhealthy desirable Healthy desirable
R eac ti on ti me (ms ) Picture type Pull Push 0 5 10 15 20 25 30 35 40 45
Unhealthy desirable Healthy desirable
A p p roac h b ias (ms ) Picture type Unhealthy desirable Healthy desirable
Figure 6 summarizes these findings. The regression line in Figure 6 shows that even if the results had been significant, it would have been a negative relation instead of the expected positive relation.
Figure 6. Scatterplot with the approach bias towards desirable stimuli and the PFS score. Healthy eating habits and approach bias
To test the hypothesis that the approach bias towards unhealthy desirable stimuli would decrease with having healthier food habits, a regression analysis was conducted with healthy food habits as the independent variable and the approach bias towards unhealthy desirable food as the dependent variable. Results of the regression analysis showed that having healthy food habits was not a significant predictor for the approach bias towards unhealthy desirable food F(1,46) = 3.662, p = .062, R2 = .074. Figure 7 shows this result.
To test the second part of this hypothesis, another regression analysis was conducted to test if the approach bias indeed remained the same for non-desirable stimuli independently of having healthy eating habits. This appeared to be the true as the regression analysis showed that having healthy food habits was not a significant predictor for the approach bias towards non-desirable food F(1,46) =.169, p = .683, R2 = .004. Figure 8 summarizes these results.
Figure 7. Scatterplot with the approach bias towards unhealthy desirable stimuli and the healthy eating habit score.
Figure 8. Scatterplot with the approach bias towards non-desirable stimuli and the healthy eating habit score.
Hunger level and approach bias
To test the hypothesis that the approach bias towards desirable stimuli would increase when participants are hungrier, a regression analysis was conducted with the level of hunger as the independent variable and the approach bias towards desirable food as the dependent variable. Results of the regression analysis showed that the hunger level was not a significant predictor for the approach bias towards desirable food F(1,46) = 1.001, p = .322, R2 = .021. Figure 9 summarizes these findings.
Figure 9. Scatterplot with the approach bias towards desirable stimuli and hunger level. BMI and approach bias
To test the hypothesis that the approach bias towards desirable food stimuli increases with the increase of BMI, a regression analysis was conducted with BMI as the independent variable and the approach bias towards desirable food as the dependent variable. Results of the regression analysis showed that BMI was not a significant predictor for the approach bias towards desirable food F(1,46) = 1.536, p = .221, R2 = .032. Figure 10 summarizes these findings.
Figure 10. Scatterplot with the approach bias towards desirable stimuli and BMI.
The above results show that none of the hypotheses can be confirmed. There were no differences in approach bias towards desirable or non-desirable stimuli nor unhealthy
desirable or healthy desirable. Subsequently, the predicted possible moderators, hunger level, BMI, PFS score, and food habits, did not influence the approach biases. Explanations for these findings are given in the discussion section of this paper.
Correlation participants’ desirability ratings and stimuli set
To see if the participants rated the data in the same way as they were categorized beforehand (desirable or non-desirable), we checked if the participants desirability rating scores correlated with the existing ratings of the food-pics database of Blechert et al. (2014). The relationship between participant ratings and the craving and desirability scores was investigated using the Pearson product-moment correlation coefficient. There was a medium positive correlation between participant ratings and the craving and desirability scores, r = .450, n = 4000, p < .001 and r = .435, n = 4000, p < .001, with high participant ratings associated with high craving and desirability scores.
Approach bias towards food and objects
In the more exploratory part of this research, we checked if the approach bias towards food in general was larger than the approach bias towards objects. A 2x2 (picture type food: food versus objects, and response type: push versus pull) repeated measures ANOVA was conducted. The repeated measures ANOVA showed there was a significant main effect of picture type food F(1,47) = 75.55, p < .001, ηpartial2 = .613, with a longer mean reaction time
for objects (M = 484.98, SD = 8.28) than for food stimuli (M = 453.21, SD = 8.57). There was also a significant main effect of response type F(1,47) = 7.14, p = .010, ηpartial2 = .132, with a
longer mean reaction time for push movements (M = 473.52, SD = 8.16) than for pull movements (M = 464.68, SD = 8.61) (see Figure 11).
The interaction effect between picture type and response type was also significant F(1,47) = 15.26, p < .001, ηpartial2 = .245. The participants had a longer reaction time when
pulling objects than pushing objects, and they had a shorter reaction time when pulling food than when pushing food. As shown in Figure 12, we can conclude from these results that the approach bias towards food is bigger than the approach bias towards objects.
Figure 11. Reaction times and standard errors on the MAAT.
380 400 420 440 460 480 500 520 Food Objects R ea ct io n ti m e (m s) Picture type Pull Push
Figure 12. Approach biases (difference between push and pull reaction times) and standard errors on two different types of stimuli.
-30 -20 -10 0 10 20 30 40 50 Food Objects A p p roac h Bi as (ms ) Picture type Food Objects
Discussion
The current research studied if the approach bias towards desirable stimuli was larger than the approach bias towards non-desirable stimuli. We also studied the difference between the approach bias towards unhealthy desirable and healthy desirable stimuli, all measured with the MAAT. Different possible moderators such as the PFS score, food habits, hunger level, and BMI were taken into consideration as well. Surprisingly, no differences were found between the approach biases towards desirable and non-desirable stimuli or the approach biases towards unhealthy desirable and healthy desirable stimuli. Furthermore, the results also showed that the PFS score, food habits, hunger levels, or BMI did not correlate with the approach biases towards the different types of stimuli.
The results found in earlier research on approach biases towards food did not
correspond to the results found in the current research. Possible explanations for not finding the expected results are discussed below.
Limitations
For the current research, choices were made regarding the design, participants, materials, and procedure. Those choices had several advantages but unfortunately also some disadvantages, therefore the limitations of the current research are discussed in this section. Firstly, the stimuli set used was not extensively tested beforehand. Although we found that the participant ratings correlated with the cravings and desirability scores of the food-pics
database by Blechert et al. (2014), the data could not tell us anything about the healthy or unhealthy categories within the desirable and non-desirable categories because the participant rating score was only based on a single desirability question. For this reason, it might be possible that people did not perceive all stimuli in the same way as we categorized them. There are many different opinions of what is healthy or not, and that might be a reason for not finding a difference between the approach bias towards unhealthy desirable and healthy desirable stimuli.
Further, because our participants were both international and Dutch students, differences in food cultures and preferences may have influenced our results. People can judge particular foods as desirable, but when they are not used to seeing and eating them regularly, it might not lead to the same reaction time and thus the same approach bias as for people who are used to the particular food types. Furthermore, the stimuli set consisted of pictures of processed and unprocessed food and cooked and uncooked food. This might have also influenced how people perceived the stimuli, because it might take longer to recognize
uncooked and unprocessed food than to recognize cooked and processed food. Therefore the reaction time for uncooked and unprocessed food might be longer, this would influence the approach bias. However, it does not necessarily mean that people were less tempted by the stimuli, because the longer reaction time may be due to the longer recognition time. Altogether, as earlier research with the MAAT showed an expected approach bias (Zech, 2015), it is more likely to question the stimulus sets used than to question the functioning of the MAAT application. However, because the MAAT is a newly developed application and used on Android devices for the first time, more research with the MAAT should be done to make sure the MAAT is the right method to test approach biases.
Secondly, the set up of the experiment was in such a way that people needed to look at pictures of food and objects, and their task was to either pull food and push objects or the other way around. This design of the experiment was chosen to make sure participants were blinded to our hypothesis. However, that might have led to not directly measuring the approach avoidance bias we were interested in, as the participants were reacting to food and objects and not to different types of food. This might be a problem because afterwards, we split the food stimuli into different categories but the participants did not respond to these categories, only to food in general. Participants might not have paid attention to what type of food they had seen and merely responded to food stimuli, as that was what they were asked to do. Therefore, their reaction time might just consist of the time they need to recognize food, instead of recognition time plus the time needed to solve the conflict in their mind when pushing away desirable food or pulling non-desirable food. If the design of the study indeed influenced the reaction time, this also influenced our approach bias, because we calculated the approach bias towards different stimuli types as the difference in reaction time between two response types. All in all, the design of the study might be an important factor of not finding the expected results.
Thirdly, the results of this research cannot be generalized to different populations or outside the laboratory. This is because we only used students from Leiden University, thus the current research cannot determine if the lack of approach bias is also evident in different populations. Besides the lack of generalization to different populations, the research took place in a laboratory environment to test the application and the stimulus sets, and therefore we cannot say anything about results outside of the lab. Further, the number of participants might be a point of concern. In the current study, we tested only 48 participants whereas most approach-avoidance studies test approximately 100-150 participants.
Fourthly, the measures of food habits, BMI, hunger, and the PFS score were all based on self-reports. This might be a disadvantage because the following studies of Sherry,
Jefferds, and Grummer-Strawn (2007) and Gorber, Tremblay, Moher, and Gorber (2007) show that self-report measures for height and weight are not always accurate in that people overestimate their height and underestimate their weight and BMI. Although we informed the participants that their results would be processed anonymously, it might be the case that participants reported their food-habits or hunger levels different from reality because they knew the results would be examined. The research of Adams, Soumerai, Lomas, and Ross-Degnan (1999) also showed that self reported research had a substantial overestimation bias. One of the reasons for this might be social-desirability bias. Social norms can influence behavior and therefore can influence how people respond on self-report measures. As a lot of young people use social media quite extensively (Centraal Bureau van de Statistiek [CBS], 2016), the so-called ‘fitgirl/boy hype’ – a trend of showing healthy meals and fit bodies – might lead to a social norm of having healthy eating habits, which might influence people to respond to a question about their eating habits in a social desirable way. Although we emphasized that the results would be saved anonymously, we cannot rule out the possibility that the answers of our participants were influenced by the fact that participants knew their answers were studied.
Beyond the method of gathering data about food habits, the fact that the score for having healthy food habits was based on only one question might have played a role in not finding the expected results. Not everyone has the same knowledge about eating healthy, and thus people with different eating patterns could evaluate their different eating pattern with the same score. This might explain why we did not find the expected correlation between having healthy food habits and unhealthy desirable food stimuli.
Furthermore, our data showed that there is an approach bias towards food in general in comparison with objects. This approach bias towards food in general might also explain why we did not find a correlation between the PFS score and the approach bias towards desirable food stimuli. According to Lowe and Butryn (2007), the PFS measures the psychological impact that food plenitude environments have on individuals. When the PFS score is high, it means that a food plenitude environment has a strong impact on a person. This could mean that people with a high PFS score are reacting faster and are more attracted to every type of food, and not only to desirable food or only when approaching food, therefore their reaction time towards food stimuli decreases for as well push as pull movements. This means there is a smaller difference in reaction time, and thus a smaller approach bias in comparison with
people with a low PFS score. People with a low PFS score are less influenced by a food plenitude environment and therefore probably only react faster when approaching desirable stimuli and not when avoiding desirable stimuli. This leads to a bigger difference in reaction time between push and pull movements, and thus a bigger approach bias towards desirable stimuli. This might explain why we did not find the expected correlation between PFS score and the approach bias towards desirable food.
The approach bias towards food in general might also play a role when looking at the correlation between hunger and the approach bias towards desirable food stimuli. According to Mogg et al. (1998), hungry people have a selective-attention bias towards food. Therefore, it might be that when people are hungrier, they do not only react faster when approaching desirable food, but also when avoiding desirable food. This leads to a smaller approach bias than expected, because we expected that people were faster when approaching desirable food and had a longer reaction time when avoiding desirable food. If the approach bias towards food in general indeed caused a shorter reaction time when avoiding and approaching desirable food, it might be a reason for not finding the expected correlation between hunger level and the approach bias towards desirable food.
Future research
As the MAAT is a newly developed way of doing approach-avoidance research, it might be useful to first try to replicate the results of the study of Zech (2015) with subjects that are known for having an approach bias. When this is done and the results are confirmed, research can focus on different ways of using the MAAT. As said before, the stimulus set in the current research was not tested extensively beforehand. Future researchers that will study food temptations and desires with the MAAT should test their stimuli set beforehand. This means that they should use pictures from a successfully proven database where the pictures are already separated in the right categories. Alternatively, they can develop a new stimuli set and test this set extensively before using it in the MAAT. It might also be useful to create different stimulus sets for different cultures, as a lot of food habits are culturally dependent (Silver and Archer, 2000). Extending the picture-rating task might be a good way of ensuring that people perceive the stimuli as they should. This can be done for example by showing the four categories and ask the participants in which category they would fit the picture. This could give better insights to if the stimuli set used was appropriate. Because existing research of Dickson et al. (2016) and Veling et al. (2013) showed that people have approach
tendencies or biases towards desirable food, a proper stimuli set might show this approach bias with the use of the MAAT as well.
Further, the design of the study can be improved. With the use of a between-subject design, learning effects and carryover effects can be avoided. This will, however, decrease the power of the study. Furthermore, it is important to think about how the direct approach bias can be measured without clearly showing the hypothesis. In the current research, participants distinguished the stimuli on objects and food but they did not need to consider different categories within the food pictures. Future research should determine how to directly distinguish two types of stimuli without necessarily giving away the hypothesis. One option for this is to create an introduction story in which it seems logical that the participants need to distinguish two types of food, but without letting them know that the research is about food temptations.
Increasing the generalizability is another important factor, so it is important to not only use students from Leiden University, but also people from different populations. The research will also be more valuable when the number of participants increases. Furthermore, one of the main advantages of using smartphones in research is that is possible to gather data in the natural environment of participants (Miller, 2012), thus future research should focus on gathering data outside of the laboratory. This means that the participants should download the application and do the experiments at home. This will be done in the field study, as described in the method section of this paper. Another important point is that currently the application is only available for Android operating system smartphones. However, according to the research of Shaw, Ellis, Kendrick, Zielger, and Wiseman (2016) people with an Android smartphone differ from people with an iPhone. In general, iPhone owners tend to be younger, females, and more concerned about status. Besides differences in types of owners, they also found that iPhone owners have higher levels of emotionality. It is therefore important to develop the application even further so that it can also run on operating systems other than Android. This would increase the generalizability of future research.
When doing field research, using self-report questionnaires is almost inescapable. However, it is not always certain that the information gathered is reliable. To make sure people respond honestly, it is important to make people feel comfortable about participating by ensuring them even more that the data is anonymous and will only be used for research purposes. A good strategy to influence participants to be honest is to use the consistency heuristic (Cialdini, 2007). The consistency heuristic uses the principle that people want to be consistent in words and deeds. So, when the application at different moments in the
experiment explicitly asks the participants to agree with the question to answer honestly, people should be more likely to act in the way they just agreed on doing. This might help future researchers to overcome the self-report bias. Another way of improving the data about healthy eating habits might be by making a questionnaire about food habits. This ensures that the data about eating healthy is unambiguous and not influenced by differences in peoples’ perceptions of what is healthy.
Future research could also focus on testing the approach bias at different times of the day. Comparing the results of testing in the morning, afternoon, and evening might give insights into when people are most likely to give in to temptations. Knowing when people are most vulnerable for temptations could assist in creating efficient plans for dieticians to help their clients to overcome the power of temptation.
Theoretical implications
Our findings that there are no differences in approach biases towards desirable versus non-desirable stimuli are not in agreement with previous research. Previous research found a natural impulsive reaction and approach bias towards palatable food (Dickson et al., 2016 and Veling et al., 2013), therefore we expected to find a larger approach bias towards desirable food than towards non-desirable food. Furthermore, because the research of Raghunathan et al. (2006) showed that people have an association with unhealthy food being tasty, we expected that respondents would have a larger approach bias towards unhealthy desirable food than towards healthy desirable food. However, the results of the current research did not confirm our expectations. Nevertheless, we found a main effect of response type and picture type desirability. This means that the participants were faster at pulling pictures than pushing pictures, and that the reaction time towards desirable pictures was generally shorter than towards non-desirable pictures. Because of the shorter reaction time towards desirable food, it seems that people recognize and process images of desirable food faster than non-desirable food.
For the two desirable categories (unhealthy and healthy), we also found a main effect of picture type healthiness and response type. However, this effect for picture type was in a different direction than expected; participants had a shorter reaction time towards healthy desirable pictures than towards unhealthy desirable pictures. A possible explanation for this might be that people experience less inner conflict while reacting to healthy desirable pictures, as it is socially accepted and not bad for your health to eat healthy food, and
our hypotheses were not confirmed but the results showed that people’s responses to pull movements were faster than to push movements, it is alluring to see if this is an effect towards the specific stimuli or just a physiological difference in how people respond. Future research should investigate how these push and pull movements have been established.
Furthermore, based on the findings of Schüz et al. (2015), Lin et al. (2016), Loeber et al. (2013), and Batterink et al. (2010), we expected that food habits, BMI, hunger levels, and PFS scores would correlate with the approach bias towards desirable or unhealthy desirable food. However, we did not find any evidence that the above concepts correlate with the approach bias towards desirable and unhealthy desirable food. Previous research on the influence of hunger on approach biases mostly used food stimuli in general and did not specify different categories. Because our results confirmed that the participants had a larger approach bias towards food in general rather than towards objects, it is possible that people have an approach bias towards food in general but not specifically towards one type of food in comparison with another type of food.
Practical implications
People carry around smartphones everyday as part of their social lives. Therefore, psychological research with smartphones can be of great added value because it allows access to domains of behavioral data that were previously not accessible (Raento et al., 2009). Because of the popularity of the smartphone, gathering data in the field is now more
accessible than ever before. This is a productive means of enhancing research because people can be tested in their everyday environment. Because the MAAT can be easily accessible for a wide variety of people, research can focus on aspects that were not always possible to study. For example, testing approach biases at different times of the day or testing if different
populations have different approach biases. Previously, these types of research were very time consuming and asked a lot of effort from the participants and researchers. With developing the MAAT, Zech (2015) made a significant progress in the possibilities of conducting large-scale research without being time consuming and high in costs.
Although we did not find the results we expected, we found an approach bias towards food stimuli in general in comparison with objects and a shorter reaction time towards desirable stimuli in comparison with non-desirable stimuli. With this knowledge, and with future research to show exactly to what extend these preferences or biases are present, dieticians can create specified plans for their clients and help them to control their weight. Earlier research showed that the AAT is not only capable of detecting approach biases, but the
AAT can also assist in training avoidance biases (Kakoschke, Kemps and Tiggemann, 2017 and Dickson et al., 2016). This can be extremely useful in helping people resist temptations. Of course, this is not only valuable in the food domain but other addictive domains can also profit from avoidance training. When doing avoidance training via an application at home, it can help much more people in a shorter time frame in comparison with helping people in a clinic or office.
The MAAT can also be a useful tool for marketers, as the results of the MAAT can give valuable information. When marketers know the exact type of products towards which people have an approach bias, they can make advertising more efficient and effective. Before this knowledge can be of practical use, it is important that more research, with proper
stimulus sets, will be conducted.
Conclusion
In conclusion, the results of the current study did not confirm our expectations. The MAAT did not detect a difference in approach bias towards desirable and non-desirable stimuli or a difference in approach bias towards unhealthy desirable stimuli and healthy desirable stimuli. The results also showed that the PFS score, BMI, hunger level, and food habits did not correlate with the approach biases. However, it is quite likely that this is
because of the stimuli set used, and because of the design of the study. The MAAT is a newly developed tool with a lot of promising practical applications for dieticians, psychologists, and marketers. Before the MAAT can be used on a wider scale, it is important that more research is conducted to show the application can detect approach biases. Therefore, it is important that strong and extensively tested stimulus sets are used. If future research proves that the MAAT is a reliable tool for detecting approach biases, it can be used on a large scale and in a wide variety of research fields.
References
Adams, A. S., Soumerai, S. B., Lomas, J., & Ross-Degnan, D. (1999). Evidence of self-report bias in assessing adherence to guidelines. International journal for quality in health care, 11(3), 187-192.
Batterink, L., Yokum, S., & Stice, E. (2010). Body mass correlates inversely with inhibitory control in response to food among adolescent girls: an fMRI study. Neuroimage, 52(4), 1696-1703. Doi: 10.1016/j.neuroimage.2010.05.059
Berridge, K. C., Ho, C.Y, Richard, J. M., & DiFeliceantonio, A. G. (2010). The tempted brain eats: Pleasure and desire circuits in obesity and eating disorders. Brain Research, 1350, 43-64. Doi: 10.1016/j.brainres.2010.04.003
Blechert, J., Meule, A., Busch, N. A., & Ohla, K. (2014). Food-pics: an image database for experimental research on eating and appetite. Frontiers in Psychology, 5, 1-10. Doi: 10.3389/fpsyg.2014.00617
Centraal Bureau van de Statistiek [CBS]. Internet; toegang, gebruik en faciliteiten (2016). Retrieved April, 2017 from:
http://statline.cbs.nl/Statweb/publication/?DM=SLNL&PA=83429ned&D1=25-31&D2=0%2c3-6&D3=0&D4=a&VW=T
Cialdini, R. B. (2007). Influence: The psychology of persuasion. New York, NY: Collins. Cleobury, L., & Tapper, K. (2013). Reasons for eating ‘unhealthy’ snacks in overweight and
obese males and females. Journal of Human Nutrition and Dietetics, 27(4), 333-341. Doi: 10.1111/jhn.12169
Dickson, H., Kavanagh, D. J., & MacLeod, C. (2016). The pulling power of chocolate: Effects of approach-avoidance training on approach bias and consumption. Appetite, 99, 46-51. Doi: http://dx.doi.org/10.1016/j.appet.2015.12.026
Gorber, S. C., Tremblay, M., Moher, D., & Gorber, B. (2007). A comparison of direct vs. self-report measures for assessing height, weight and body mass index: a systematic review. Obesity reviews, 8 (4), 307-326. Doi: 10.1111/j.1467-789X.2007.00347.x Havermans, R. C., Giesen, J. C. A. H., Houben, K., & Jansen, A. (2011). Weight, gender, and
snack appeal. Eating behaviors, 12(2), 126-130. Doi: http://dx.doi.org/10.1016/j.eatbeh.2011.01.010
Havermans, R. C. (2013). Pavlovian Craving and Overeating: A conditioned Incentive Model. Current Obesity Reports, 2(2), 165-170. Doi: 10.1007/s13679-013-0053-z
Hill, A. J. (2007). Symposium on ‘Molecular mechanisms and psychology of food intake’, The psychology of food craving. Proceedings of the nutrition society, 66, 277-285. Doi: 10.1017/S0029665107005502
Hofmann, W., Förster, G., Baumeister, R. F., & Vohs, K. D. (2012). Everyday temptations: An experience sampling study of desire, conflict, and self-control. Journal of Personality and Social Psychology, 102(6), 1318-1335. Doi: 10.1037/a0026545 Hofmann, W., Vohs, K. D., & Baumeister, R. F. (2012). What people desire, feel conflicted
about, and try to resist in everyday life. Psychological Science, 23(6), 582-588. Doi: 10.1177/095679761237426
Kakoschke, N., Kemps, E., Tiggemann, M. (2017). The effect of combined avoidance and control training on implicit food evaluation and choice. Journal of Behavior Therapy and Experimental Psychiatry, 55, 99-105. Doi:
http://doi.org/10.1016/j.jbtep.2017.01.002
Lin, P. Y., Wood, W., & Monterosso, J. (2016). Healthy eating habits protect against temptations. Appetite, 103, 432-440. Doi:
http://dx.doi.org/10.1016/j.appet.2015.11.011
Loeber, S., Grosshans, M., Herpertz, S., Kiefer, F., & Herpertz, S. C. (2013). Hunger
modulates behavioral disinhibition and attention allocation to food-associated cues in normal-weight controls. Appetite, 71, 32-39. Doi:
http://dx.doi.org/10.1016/j.appet.2013.07.008
Lowe, M. R., & Butryn, M. L. (2007). Hedonic hunger: A new dimension of appetite? Physiology & Behavior, 91, 432-439. Doi: 10.1016/j.physbeh.2007.04.006 Miller, G. (2012). The smartphone psychology manifesto. Perspectives on Psychological
Science, 7(3), 221–237. Doi: 10.1177/1745691612441215
Mogg, K., Bradley, B. P., Hyare, H., & Lee, S. (1998). Selective attention to food-related stimuli in hunger: are attentional biases specific to emotional and psychopathological states, or are they also found in normal drive states? Behavior research and therapy, 36, 227-237.
Neubach, B., & Schmidt, K. H. (2007). Entwicklung und Validierung von Skalen zur Erfassung verschiedener Selbstkontrollanforderungen bei der Arbeit [Scale development and validation for assess various self-control demands at work]. Zeitschrift fur Arbeitswissenschaft, 61, 35–45.
Papies, E. K. (2016). Health goal priming as a situated intervention tool: how to benefit from nonconscious motivational routes to health behavior. Health psychology review, 10(4), 408-424. Doi: 10.1080/17437199.2016.1183506
Polivy, J., Herman, C. P., & Girz, L. (2011). Learning to eat: The influence of food cues on what, when, and how much we eat. In Schachtman, T. R., & Reilly, S. S. (Red.), Associative learning and conditioning theory: human and non-human applications. Doi: 10.1093/acprof:oso/9780199735969.001.0001
Raento, M., Oulasvirta, A., & Eagle, N. (2009). Smartphones an emerging tool for social scientists. Sociological Methods & Research, 37(3), 426-454.
Raghunathan, R., Walker Naylor, R., & Hoyer, W. (2006). The unhealthy = tasty intuition and its effects on taste inferences, enjoyment, and choice of food products. Journal of Marketing, 70(4), 170-184.
Reips, U. D. (2000). The web experiment method: advantages, disadvantages, and solutions. In M. H. Birnbaum (Red.), Psychological Experiments on the Internet (pp. 89-114). San Diego: Academic Press.
Renner, B., Sproesser, G., Strohbach, S., & Schupp, H. T. (2012). Why we eat what we eat. The Eating Motivation Survey (TEMS). Appetite, 59(1), 117–128. Doi:
http://dx.doi.org/10.1016/j.appet.2012.04.004
Rinck, M., & Becker, E. S. (2007). Approach and avoidance in fear of spiders. Journal of Behavior Therapy and Experimental Psychiatry, 38(2), 105-120. Doi:
http://dx.doi.org.ezproxy.leidenuniv.nl:2048/10.1016/j.jbtep.2006.10.001
Schüz, B., Schüz, N., & Ferguson, S. G. (2015). It’s the power of food: individual differences in food cue responsiveness and snacking in everyday life. International Journal of Behavioral Nutrition and Psychical Activity, 12:149. Doi: 10.1186/s12966-015-0312-3
Shaw, H., Ellis, D. A., Kendrick, L. R., Ziegler, F., & Wiseman, R. (2016). Predicting smartphone operating system from personality and individual differences. Cyberpsychology, Behavior, and Social Networking, 19(12), 727-732. Doi: 10.1089/cyber.2016.0324.
Sherry, B., Jefferds, M. E., & Grummer-Strawn, L. M. (2007). Accuracy of adolescent self-report of height and weight in assessing over-weight status. Archives of Pediatrics and Adolescent Medicine, 161(12), 1154-1161. Doi: 10.1001/archpedi.161.12.1154
Silver, J. (director), & Archer, D. (producer). (2000). A world of food: tastes & taboos in difference cultures [Video Tape]. Berkeley Media. Retrieved April 12, 2017, from Food Studies Online.
Solarz, A. K. (1960). Latency of instrumental responses as a function of compatibility with the meaning of eliciting verbal signs. Journal of Experimental Psychology, 59(4), 239-245.
Veenstra, E. M., & De Jong, P. J. (2010). Restrained eaters show enhanced automatic approach tendencies towards food. Appetite, 55(1), 30-36. Doi:
http://dx.doi.org/10.1016/j.appet.2010.03.007
Veling, H., Aarts, H., & Stroebe, W. (2013). Using stop signals to reduce impulsive choices for palatable unhealthy foods. British Journal of Health Psychology, 18(2), 354-368. Doi: 10.1111/j.2044-8287.2012.02092.x.
Verhoeven, A. A. C., Adriaanse, M. A., Evers, C., & De Ridder, D. T. D. (2012). The power of habits: Unhealthy snacking behavior is primarily predicted by habit strength. British Journal of Health Psychology, 17(4), 758-770.
Voedingscentrum. Invloed omgeving op eetgedrag (2017). Retrieved February 25th, 2017 from: http://www.voedingscentrum.nl/encyclopedie/invloed-omgeving-op-eetgedrag.aspx
Weingarten, H. P., & Elston, D. (1990). The phenomenology of food cravings. Appetite, 15, 231-246.
Wiers, C. E., Stelzel, C., Park, S. Q., Gawron, C. K., Ludwig, V. U., Gutwinski, S., Heinz, A., Lindenmeyer, J., Wiers, R. W., Walter, H., & Bermpohl, F. (2014). Neural correlates of alcohol-approach bias in alcohol addiction: The spirit is willing but the flesh is weak for spirits. Neuropsychopharmacology, 39(3), 688-697. Doi:
10.1038/npp.2013.252
World Health Organization [WHO]. Obesity and overweight (2016). Retrieved November 30, 2016 from: http://www.who.int/mediacentre/factsheets/fs311/en/
World Health Organization [WHO], BMI Classification (2017). Retrieved March 6, 2017 from: http://apps.who.int/bmi/index.jsp?introPage=intro_3.html
Zech, H. G. (2015). The Mobile Approach-Avoidance Task (master’s thesis). Retrieved from http://hdl.handle.net/1887/34773.
