2-2-2018
Implementation Intentions
as a Resistance Strategy
Towards Food Related
Temptations
Master’s Thesis
Mayke Meuwese (11384093)
Supervisor: prof. dr. S.J.H.M van den Putte
UNIVERSITY OF AMSTERDAM GRADUATE SCHOOL OF COMMUNICATION MASTER’S PROGRAMME PERSUASIVE COMMUNICATION Number of words: 6683 Number or words abstract: 267
Table of Contents
Abstract ... 2
Introduction ... 3
The Present Research ... 4
Theoretical Framework ... 6 Methods ... 10 Participants ... 11 Procedure ... 11 Measures ... 13 Automatic evaluations. ... 13 Explicit attitude... 15 Demographics. ... 15 Results ... 15 Discussion ... 19
Strengths and Limitations ... 21
References... 23
Appendix A Characteristics ... 27
Appendix B IAT Pictures ... 28
Appendix C IAT Screenshots ... 29
Abstract
The aim of my study was to replicate and build on the findings of Hofmann et al. (2010), that implementation intentions have a reducing effect on the positive automatic evaluation elicited by a tempting stimulus. Participants could choose between chocolate (N = 62) or chips (N = 58) as their temptation. An experimental design with one experimental (implementation intentions, N = 33), one active control (general goal intentions, N = 36) and one control group (N = 51) was used. Participants were shown their condition specific task and hereafter
performed an implicit association task (IAT) in which their automatic evaluation towards the temptation was measured. Results showed a positive automatic evaluation of the tempting stimulus. However, no differences were found between the experimental (implementation intentions) and the control and active control (general goal intentions) condition for chocolate as the temptation. An explanation for this unexpected absence of an effect could be the small sample size, resulting in low power. For chips as the temptation, there was a significant difference in automatic evaluation between the implementation intentions condition and the control group, however these results were opposite to what was expected. Meaning that participants in the implementation intentions condition had a more positive automatic evaluation compared to the participants in the control group. This could be explained by the discrepancy in whether people truly see chips as a temptation, especially compared to
chocolate. Further research could use this design with a larger sample size, different forms of implementation intentions, different temptations and rating of those temptations to further unravel the best resistance strategy for food related temptations.
Introduction
Cues towards attractive, high-calorie foods are very present in our everyday lives. These cues provide temptations that are hard to resist and ask for a certain amount of self-control in individuals who like to watch their weight (Muraven & Baumeister, 2000). However, the satisfaction of giving in to these temptations is direct, whereas resisting the temptation asks for an activation of their long-term weight watching goal. Moreover, it seems that these temptations are even harder to resist for people who are already overweight in comparison to normal-weight individuals (Ouwehand & Papies, 2010). Especially for these individuals it may be important to have a strategy which they can use to provide resistance towards these temptations or even weaken the power of the temptation at hand in the first place.
A strategy that has promising results, indicating that individuals could use it to resist temptations, are so called ‘implementation intentions’. These implementation intentions are “if…then” statements that make the goal more concrete and thus make the goal more
achievable by linking an intended goal-directed behaviour to an anticipated situational context (Gollwitzer, 1993). Implementation intentions differ from goal intentions (“I intend to reach Z”, p.69) in the sense that they build upon general goal intentions by specifying not only the what that is intended to be achieved but also when, where and how this will be achieved
(Gollwitzer & Sheeran, 2006). Research suggests that because of making these
implementation intentions, mental representations become more accessible and that certain specified situational cues induce direct control of the intended behaviour. This indicates that, although the implementation intention originates in the reflective system, the specified cue induces an automatic goal-directed response (i.e. immediately, without the need for conscious intent; Gollwitzer & Sheeran, 2006).
Not only do implementation intentions have an automatic effect on the cognitive level, research suggests that implementation intentions also have a direct effect on implicit attitudes (Hofmann, Deutsch, Lancaster, & Banaji, 2010; Sheeran, Gollwitzer, & Bargh, 2013; Webb, Sheeran, & Pepper, 2012). In their study, Webb et al. (2012) found that implementation intentions (“e.g., If Muslim names and peace appear together, then I respond especially fast to Muslim words and peace words!”, p. 15) were effective in decreasing the time it took
participants to react in an Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998) with putatively incongruent associations (“e.g., woman-manager, Muslims-peace”, p. 16). This indicates that people gained more control over their implicit attitude responses. Moreover, Hofmann et al. (2010) found that implementation intentions directed at the anticipated temptation of chocolate in an imagined everyday scenario made the automatic evaluation elicited by this tempting stimulus less positive.
The Present Research
These studies show promising results indicating that implementation intentions could possibly be used as a strategy to reduce or resist temptation which could help a lot of people who have trouble resisting temptations in their everyday lives. However, according to Porte (2013) there is a strong need for replication studies in the social sciences. These replication studies are needed to be sure that there were no uncontrolled and/or unknown variables that influenced the found results in earlier studies. To date, no study has replicated the findings of Hofmann et al. (2010), and several variables can be thought of that may have influenced the reducing effect of implementation intentions on the positive automatic evaluation of a
tempting stimulus. Additionally, because Hofmann et al. (2010) only focused on chocolate as tempting stimulus, it is not certain that the found results are generalizable to other food related temptations. Therefore, my aim was to replicate the findings of Hofmann et al. (2010), but with a different sample and some additions.
First of all, to see whether the results by Hofmann et al. (2010) also hold for other temptations, a second option as temptation in addition to chocolate was added. What counts as temptation differs between people, some people prefer sweet products (e.g., chocolate) and some people prefer savoury products (e.g., chips; Griffioen-Roose, 2012). A good match between preference and temptation may increase the effectiveness of the implementation intentions, because this match may enhance the automatic evaluation of the tempting stimulus. Therefore, at the start of the study, participants could choose between either chocolate or chips as imagined temptation.
Additionally, a possible moderator was taken up in this study, namely the Body Mass Index (BMI) of the participants. As said before, Ouwehand and Papies (2010) concluded that food temptations are harder to resist for overweight individuals in comparison with normal-weight individuals. It is possible that Hofmann et al. (2010) only had normal-normal-weight
individuals in their study which could have increased the effectiveness of the implementation intentions. To date, no research on implementation intentions and food related temptations has taken up BMI as a possible moderator. To see whether BMI is a true moderator, the
participants in the present research were asked about their length and weight.
Finally, an extra control variable was taken up in this study following the study by Webb et al. (2012). In their second study, Webb et al. (2012) looked if the effects they found were truly ascribable to implementation intentions and not just to a more general goal
intention (e.g., ‘I will not assume that Muslims are terrorists!’). They found that the effects could not be explained by a more general goal intention. Applying this to my study on food temptations in which a possible implementation intention could be; ‘When my friend offers me chips while watching Netflix, then I say no thanks and continue watching Netflix’, it would be good to eliminate a more general goal intention (e.g., ‘I will refrain from eating chips’) as a possible underlying cause of the found results.
In sum, the main research question was: Can implementation intentions be used as a form of resistance towards food related temptations? More specifically, four issues were
studied: (1) whether the evidence of implementation intentions reducing the positive automatic evaluation elicited by a tempting stimulus (chocolate) could be replicated, (2) whether the same pattern of results were obtained for a different tempting stimulus (chips), (3) whether the BMI of the participants moderated the effect of implementation intentions on the automatic evaluation of the tempting stimulus and (4) whether this effect could truly be ascribed to implementation intentions and not to a more general goal intention (see Figure 1 for the conceptual model).
Theoretical Framework
It takes a large amount of self-control to achieve a long-term goal. People often fail, because they experience a lot of self-regulatory problems on the road to achieving their goal (Gollwitzer & Sheeran, 2006). As people are confronted with food related temptations on a daily basis, they need cognitive self-regulatory strategies to control themselves (Hoch & Loewenstein, 1991; Metcalfe & Mischel, 1999), and having a strong intention to reach the goal of losing weight is not enough (Gollwitzer & Sheeran, 2006). When people encounter problems with translating their goals into action, it is often because they fall back into automatic processes (e.g., falling into bad habits, getting distracted; Gollwitzer, 1999). As a matter of fact, as Verplanken and Faes (1999) describe, automatic processing often leads to
Implementation Intentions/ Goal Intentions Automatic Evaluation of Tempting Stimulus BMI of Participants Figure 1. Conceptual Model
habitual behaviour which induces the unwanted behaviour of giving in to cravings that are experienced in the now. Implementation intentions could serve as a solution for this.
Implementation intentions can namely be seen as the next step after goal intentions. Goal intentions specify a certain end state that a person wants to achieve (e.g., “I intend to lose five kilograms”), and by making these goal intentions, people feel committed to reach their goal (Gollwitzer, 1999). Implementation intentions, on the other hand, specify situations in which the set goal could be threatened and actions that need to be taken in order to attain the set goal. By making implementation intentions (e.g., “If I get home after a long workday and crave some chocolate, then I will grab an apple instead”) the person commits him- or herself to react in a certain manner in a specific situation, providing that person with specific tools to resist temptations (Gollwitzer, 1999). In other words, when the specified situation occurs (the feeling of craving chocolate after a long workday), this serves as a cue for the made implementation intention and induces the automatic response (immediately, without conscious intent) of grabbing an apple instead (Verplanken & Faes, 1999). Thus, although the creation of the implementation intentions asks for deliberate processing (with conscious intent), the situational cues induce automatic cue-responses (Verplanken & Faes, 1999). Moreover, by visualizing possible situations in which the goal could be threatened,
accompanied by a behaviour in favour of the intended goal ("Whenever situation x arises, I will initiate the goal-directed response y!", p. 494), the chance of goal attainment is higher (Gollwitzer, 1999).
Furthermore, research suggests that these implementation intentions do not only influence the actual behaviour when a specific situation arises, they also have a direct impact on implicit attitudes, or more precisely the automatic evaluation of a tempting stimulus (Hofmann et al., 2010; Sheeran et al., 2013; Webb et al., 2012). Implicit attitudes are “best characterized as automatic affective reactions resulting from the particular associations that
are activated automatically when one encounters a relevant stimulus” (Gawronski &
Bodenhausen, 2006, p. 693). When such an implicit attitude entails a tempting stimulus (e.g., chocolate) it results in a positive automatic evaluation, because the temptation is associated with positive feelings/memories/thoughts (Sheeran et al., 2013). In the study conducted by Hofmann et al. (2010), they found that making implementation intentions had a decreasing effect on the positive automatic evaluation elicited by chocolate as a tempting stimulus. In other words, chocolate as a temptation became less tempting because the participants imagined situations in which their goal of not eating chocolate could be threatened and
accordingly came up with actions to attain their goal. This had a direct effect on their response time in an IAT in which the control group (who did not make implementation intentions) strongly associated chocolate with positive images. The implementation intentions condition, on the other hand, showed a significant less positive association between chocolate and positive images. To date, no research replicated these findings by Hofmann et al. (2010). Therefore, the first aim of this study was to do so, and accordingly the first hypothesis was as follows:
H1: Implementation intentions have a reducing effect on the positive automatic
evaluation of chocolate as tempting stimulus.
Subsequently, in their research, Hofmann et al. (2010) only focused on chocolate as temptation. To see whether their found results can also be generalized to other food related temptations, the present study incorporated another food related temptation. As Griffioen-Roose (2012) describes, people differ in whether they prefer sweet or savoury products. Therefore the present study added chips as possible savoury temptation. This made it possible for the participants to choose the temptation of their liking. Moreover, because the theoretical explanation of why these implementation intentions have a reducing effect on the positive
automatic evaluation of a tempting stimulus should, in theory, also hold for chips as temptation, the second hypothesis was as follows:
H2: Implementation intentions have a reducing effect on the positive automatic
evaluation of chips as tempting stimulus.
Hypothesis 3 builds on the two former hypotheses. As food temptations are harder to resist for individuals who are overweight in comparison to normal weight individuals
(Ouwehand & Papies, 2010), it was expected that the reducing effect of implementation intentions on the automatic evaluation of a tempting stimulus was less prominent for
overweight individuals. This was expected because research by Ouwehand and Papies (2010) showed that cues toward food related temptations increased the wanting of these high-calorie foods for overweight individuals, whereas people with a lower body weight displayed less wanting for high-calorie foods. Moreover, no study to date has incorporated BMI as a possible moderator when it comes down to implementation intentions and temptations. Therefore the third hypothesis read as follows:
H3: Implementation intentions are less effective for people who are overweight in
comparison to normal weight individuals.
Furthermore, as Gollwitzer and Sheeran (2006) describe, having a strong goal
intention (“I intend to reach Z!” or “I intend to achieve outcome Z!”, p.69, p.71), in which the what you want to achieve is specified, only leads to a 20-35% chance of goal achievement.
Implementation intentions, on the other hand, specify the when, where and how one will perform when a certain situation arises (“If situation Y is encountered, then I will initiate goal-directed behaviour X!”, p.69) and thus enlarges the chance of goal achievement. Following this, it was expected that people who would make implementation intentions, would show a bigger effect of a reduced positive automatic evaluation of a tempting stimulus
in comparison to people who only formed general goal intentions. Thus, hypothesis 4 was as follows:
H4: The reducing effect on the positive automatic evaluation of a tempting stimulus is
bigger for implementation intentions than for general goal intentions.
Lastly, in the context of replicating the findings of Hofmann et al. (2010), a last variable was taken up in this study, namely the explicit attitude of the participants towards the temptation. Explicit attitudes differ from implicit attitudes in the sense that implicit attitudes are attitudes of which people cannot control the activation and do not have conscious access to (Rydell & McConnell, 2006). Explicit attitudes, on the other hand, are attitudes that people can report and for which the activation of it can be consciously controlled (Rydell &
McConnell, 2006). Hofmann et al. (2010) found that although there was a significant difference between the experimental and control group for both the automatic evaluation (implicit attitude) and explicit attitude towards chocolate as tempting stimulus, there was a low correlation between them. Indicating that there is a difference between how implicit and explicit attitudes change (Gawronski & Bodenhausen, 2006; Rydell & McConnell, 2006). Because an important aim of this study was to replicate the findings of Hofmann et al. (2010), explicit attitude was taken up as an explorative variable to see whether making
implementation intentions would also have an effect on the explicit attitude of the participants towards the tempting stimulus.
Methods
To be able to answer the research question, an online experiment was conducted. It consisted of three different elements; a condition specific message, an IAT, and a section about demographics, like age, education level, height and weight (for calculation of the BMI). Participants were randomly assigned to three different conditions, namely:
control variable, N = 36) and a control group (N = 51). During the condition specific message, participants in the implementation intentions condition and the general goal intentions
condition were asked to either make implementation intentions or goal intentions. Participants in the control condition were asked to list city names from A-Z. After the condition specific instructions, an IAT tested the automatic evaluation of a tempting stimulus (dependent variable).
Participants
The snowball sampling method was used via the Facebook page of the researcher, meaning that people were approached online to participate and share the participation link in their network. A total of 124 people participated in the study, however 3 people needed to be excluded due to missing values and 1 person was excluded due to excessive speed on the IAT (10% of this persons’ trials had a response time less than 300 ms.). This led to a total sample of 120 participants (mean age = 33.16, SD = 16.11; 70.6% female) who completed the study (see Table 1 in Appendix A for further characteristics). Of these 120 participants, 62 chose chocolate and 58 chose chips as temptation.
Two bol.com vouchers worth €10,- were raffled among the 64 participants who left their email address at the end of the survey. This was done by combining the email address with a number, and hereafter a random number generator was used twice to select the winners. No other form of compensation was awarded for partaking in this study. Procedure
The post on the researchers’ Facebook page contained two links, one to the survey with chocolate as temptation and one with chips as temptation. Participants had to choose which one they found more tempting and click on that specific link. This was done because it was not possible to incorporate two separate IAT’s into one Qualtrics project, thus two projects had to be made resulting in two different links. After clicking on one of the two links
participants were randomly assigned to one of three conditions; the implementation intentions, general goal intentions or control group.
Hereafter, the participants were shown the condition specific message. In the
implementation intentions condition, participants were asked to imagine that they had the goal
of not eating chocolate/chips and therefore had the motivation to resist any such temptation in every situation that their temptation was present. They were asked to envision specific
situations in which the temptation could be present and for all these situations write down in the text box provided in Qualtrics an “if.., then I will” statement. For example: “If my friend offers me chocolate/chips during the film, I will say ‘no thanks’ and concentrate on the film” and to repeat these intentions silently to themselves until the time run out. Participants were told they had 2 minutes to write down as many situations and corresponding intentions to refuse as they possibly could, and that they would automatically be redirected to the next section when the two minutes were over. In the general goal intentions condition participants were asked to envision that their goal was to refrain from eating chocolate/chips and therefore were motivated to resist chocolate/chips. The participants were asked to write down in the provided text box their intentions to resist eating chocolate/chips, for example “I will refrain from eating chocolate/chips to lose weight” or “I will refrain from eating chocolate/chips to snack less” and to repeat these intentions silently to themselves until the time run out. Participants were told they had 2 minutes to perform this task and that they would
automatically be redirected to the next section when the two minutes were over. In the control condition participants were asked to write down city names from A to Z (e.g., Almere, Breda,
Castricum, etc.; Hofmann et al., 2010). In all conditions a timer was set at two minutes to carry out the task and the participants were then redirected to the IAT. All participants were given instructions to the IAT and were asked to do this test.
Subsequently, the participants indicated their explicit attitude towards their temptation (chocolate/chips) on a seven point rating scale. Finally, the demographics of the participants were asked, as well as their weight and height to be able to calculate their BMI.
Measures
Automatic evaluations. The automatic evaluations of the participants towards the temptation was measured as in Hofmann et al. (2010), using an image IAT. An image IAT is a psychological research tool that uses pictures as stimuli (e.g., a picture of chocolate, see Figure 2 in Appendix B) to examine the automatic (implicit) association between a target (e.g., chocolate or furniture) and a category (e.g., positive or negative), that may exist below awareness (Carpenter et al., 2017; Greenwald et al., 1998). Within such an IAT participants are asked to rapidly ‘sort’ shown stimuli to the corresponding word that is shown either in the upper left or right corner of the screen, by pressing either the ‘E’ or ‘I’ key on the keyboard with their left or right hand, correspondingly. For example, as can be seen in Figure 8 in Appendix C, the participant had to press with their left hand for all furniture +
negative stimuli and with their right hand for all chocolate + positive stimuli. The idea behind
an IAT is that the task is easier to complete, and thus that person will also react faster, when the sorting manner is in accordance with the person’s associations. For instance, when a person is asked to sort all chocolate + positive stimuli with one hand and furniture + negative with the other hand, it is expected that they react faster than when chocolate is combined with negative. This is, because chocolate is (often) seen as temptation and therefore combining chocolate with positive is more consistent with people’s implicit associations. Alternatively, when people have to sort chocolate combined with negative, people have to overcome these implicit associations and it thus makes them slower at performing the task.
In the present study’s IAT, the primary target was either chocolate or chips (depending on the participants choice) and the neutral target was furniture (e.g., Blanton, Jaccard,
Gonzales, & Christie, 2006; Hofmann et al., 2010). Furniture served as the neutral target, because people (most likely) do not have a strong positive or negative association with it. The colour of the furniture was adjusted to the specific condition so the colour matched with the primary target (dark brown for chocolate and light brown for chips, see Figure 4 and 5 in Appendix B). Finally, pictures of positive and negative valence served as category stimuli (see Figure 6 and 7 in Appendix B). In total 24 pictures (4 chocolate, 4 chips, 4 dark brown furniture, 4 light brown furniture, 4 positive, and 4 negative) were used to build the IAT (see Figure 2-7 in Appendix B). The pictures of the chocolate, furniture, positive valence, and negative valence were obtained via email contact with dr. Hofmann. He provided the
researcher with the needed pictures to enhance the possibility of replicating his findings. The chips pictures were obtained through the internet by the researcher.
The IAT was build using iatgen (an online IAT building tool; Carpenter et al., 2017) and consisted of seven blocks (sets of trials). Five of these blocks were training blocks and two blocks were the critical blocks. In these critical blocks the primary target was combined with the word ‘positive’ or ‘negative’ and the participant was ought to ‘sort’ the shown
picture as fast as possible by pressing the ‘E’ or ‘I’ key on the keyboard (see Figure 8 and 9 in Appendix C). The key assignment (position of the temptation either upper left or right corner) and compatibility order (whether the temptation was combined with the word ‘positive’ or ‘negative’ in the first block) was randomized over all participants.
Eventually a D-score, which serves as an index of automatic evaluation, was
calculated in the same program as where the IAT was build (Carpenter et al., 2017). This D-score was calculated according to the guidelines of Greenwald, Nosek and Banji (2003) and is a number between -2 and 2 that stands for the difference in reaction time between the two critical blocks. Increasing positive values indicates a faster reaction time in the block where chocolate or chips was combined with positive stimuli and had the same response key as
compared to the block where chocolate or chips was combined with negative stimuli and had the same response key. This thus indicates a positive automatic evaluation of the tempting stimulus. Increasing negative values indicates the reverse effect and thus a more negative automatic evaluation of the tempting stimulus. Internal consistency of the D-score was good at α =.86 for both the chips and chocolate IAT.
Explicit attitude. To measure the explicit attitude of the participants towards the temptation (chocolate or chips) two seven-point semantic differentials were used (Hofmann et al., 2010). The question that participants saw was: “On a scale from 1-7, what is your opinion towards chocolate(chips)?”. The poles ranged from ‘not at all tasty’ to ‘very tasty’ and ‘I do not like it at all’ to ‘I like it a lot’. These poles were exactly the same as used in the study by Hofmann et al. (2010). Afterwards the two poles were combined into one single explicit attitude index with a higher score indicating a more positive attitude. Internal consistency of the one single explicit attitude index was exceptional for the chips survey (α = .97) and good for the chocolate survey (α = .85).
Demographics. In the demographics section participants were asked about seven characteristics, namely: gender, age, height, weight, highest level of completed education, marital status, ethnic background and device on which the participant took part in the experiment. Afterwards the BMI variable was calculated with the height and weight of the participants, using the formula weight(in kg)/height*height(in m).
Results
All assumptions for an ANOVA were checked and all were met with the exception of the assumption of normality. To compensate for the violation of this assumption,
bootstrapping was used during the analysis of all main hypotheses (Field, 2013). Multiple AN(C)OVA’s were performed to see whether the hypotheses could be supported by the data or not. The first hypothesis was an exact replication of study two in
Hofmann et al. (2010) and therefore it was expected that their findings would be reproduced. More specifically, it was expected that implementation intentions would have a reducing effect on the positive automatic evaluation for chocolate as tempting stimulus. However, an ANCOVA with the D-score as the dependent variable, condition (implementation intentions versus control) as independent between subject variable, and compatibility order, and key assignment as covariates yielded no main effect of condition, F(1, 39) = .45, p = .51, ηp2 = .01. Indicating that there was no significant difference in automatic evaluation between the
implementation intentions group and the control group for the participants who chose chocolate as temptation.
The second hypothesis expected the same as hypothesis 1, however now the
temptation was chips instead of chocolate. Therefore, the same ANCOVA as for hypothesis 1, but with chips as temptation, was performed, and this yielded a significant effect of condition, F(1, 37) = 4.75, p = .04, ηp2 = .11 , indicating that there was a significant difference between the implementation intentions group and the control group for the participants who chose chips as temptation. However, this difference was in the opposite direction than what was hypothesized, thus the participants in the implementation intentions condition had a more positive automatic evaluation on average than the control condition.
Additionally, an one-way ANOVA with explicit attitudes as dependent variable and condition as factor yielded no main effect for both chips, F(1, 39) = 2.12, p = .15, ηp2 = .05, and chocolate, F(1, 41) = .57, p = .46, ηp2= .01. Meaning that there was no significant difference in explicit attitudes between the implementation intentions and control group for both temptations.
Following the study by Hofmann et al. (2010), key assignment and compatibility order were taken up in the analysis for the first two hypotheses as possible confounding variables. These analyses indeed showed a significant effect of compatibility order on condition. More
specifically, a one-way ANOVA with D-score as dependent variable and compatibility order as independent between subjects variable indicated that participants over all conditions who performed the compatible block first (chocolate/chips-positive) rather than second had on average a more positive automatic evaluation for both the chocolate, F(1, 60) = 6.03, p = .02, η2= .09 and chips, F(1, 56) = 47.22, p <.001, η2= .46 IAT. This is in line with previous research (Greenwald et al., 1998; Lane, Banaji, Nosek, & Greenwald, 2007; Hofmann et al., 2010).
To facilitate a sufficient power for the analysis of hypothesis 3 and 4 it was decided to combine both temptations (chocolate and chips) into one variable (temptation). However, to see if the D-scores for both these variables were comparable enough to combine into one variable, a non-parametric test was performed in which the mean D-scores over all conditions for chocolate and chips were compared. This test indicated a significant difference between the mean D-scores for chocolate and chips, U = 1243, z = -2.92, p <.001, r = -.27 (see Figure 10 and Table 2 in Appendix D). Indicating that chocolate was automatically evaluated as more positive than chips. Because of this significant difference, but in order to also keep sufficient power, it was decided to add temptation as a possible additional moderator in both the analysis for hypothesis 3 and 4.
As for the third hypothesis, it was expected that implementation intentions would be less effective in reducing the positive automatic evaluation of the tempting stimulus for people who were overweight in comparison to normal weight individuals. To see if the
participants’ BMI influenced the effectiveness of the implementation intentions, an ANCOVA was performed with D-score as the dependent variable, condition (implementation intentions versus control) as the independent between subject variable, compatibility order and key assignment as covariates, and BMI and temptation as possible moderators. This analysis yielded no main effect on condition, F(2, 105) = 1.00, p = .73, ηp2 = .86, no effect for the
interaction between condition*BMI, F(2, 105) = .39, p = .72, ηp2 = .28, and also no effect for the three-way interaction between condition*BMI*temptation, F(2, 105) = 1.85, p = .16, ηp2 = .03. Thus, participants’ BMI, contrary to what was expected, did not influence the
effectiveness of implementation intentions versus the control condition. Moreover, the temptation type did not moderate the effect.
Next, the fourth hypothesis stated that the reducing effect of implementation intentions on the positive automatic evaluation of a tempting stimulus would be bigger for these
implementation intentions in comparison to general goal intentions. An ANCOVA with D-score as the dependent variable, condition (implementation intentions versus general goal intentions) as the independent between subject variable, compatibility order and key
assignment as covariates, and temptation as possible moderator, was performed. This analysis yielded no main effect of condition, F(1, 63) = 1.29, p = .46, ηp2 = .56, and no significant interaction effect of condition*temptation, F(1, 63) = 1.28, p = .26, ηp2 = .02. Indicating that there was no significant difference in automatic evaluation between the implementation intentions group and the general goal intentions group, and that the type of temptation did not moderate the effect.
Finally, the implicit-explicit correlations across conditions were inspected for both chips and chocolate as temptation. A significant correlation was found for the general goal intentions condition with chips as temptation, r = .49, p = .05, but all the other conditions did not show a significant relationship, rimplementation/chocolate = .02, p = .95, rimplementation/chips = .09, p
= .74, rcontrol/chocolate = -.09, p = .68, rcontrol/chips = .14, p = .50, and rgoal/chocolate = .42, p = .08.
This indicates that for most conditions the explicit attitude did not correspond with the automatic evaluation. Thus, it could not be said that if the automatic evaluation was high, the explicit attitude would be high as well, and vice versa.
Discussion
The aim of my study was to answer the research question: Can implementation intentions be used as a form of resistance towards food related temptations? and to replicate
and build on the findings of Hofmann et al. (2010). An experimental design was used with one experimental (implementation intentions), one control and one active control group
(general goal intentions). The results showed that both chips and chocolate were automatically evaluated as positive. However, in contrast to the hypothesis for chocolate as temptation, there were no differences in automatic evaluation between the implementation intentions group and the control group. For chips as temptation the results did show a significant difference in automatic evaluation between the implementation intentions group and the control group, however this was in the opposite direction than was expected. Meaning that after participants performed the condition specific task, the participants in the implementation intentions group automatically evaluated chips as more positive than the control group.
Additionally, no significant difference in explicit attitude was found between the experimental and control group for both chocolate and chips as temptation. Moreover, both the BMI of the participants as well as the temptation type did not moderate the effect. When comparing the experimental with the active control group (general goal intentions), the result showed that also between these conditions there were no significant differences. Thus, it could be concluded that all hypotheses could be rejected.
A possible explanation for not finding a difference in automatic evaluation for chocolate as temptation between the experimental group and the control group could be that the number of participants of the study (small sample size) was probably too low for sufficient power. As Field (2013) describes, the higher the power of a study, the higher the probability that an effect is found, assuming that the effect exists in the population. Hofmann et al. (2010) had a total of 506 participants, for only chocolate as temptation. My study only had a total of
120 participants who were divided over two temptation types and three conditions. Because of this, the power in my study varied between .10 and .57, indicating a poor to moderate power. Even though the power was moderate to poor for chips as temptation a significant effect was found, and although this effect was opposite to what was expected, the effect size of, ηp2 = .11 was medium to large (Cohen, 1969, p. 278-280). Indicating that a larger sample size would have probably lead to the same results.
Moreover, the results for chocolate as temptation did show that the difference between the implementation intentions and control group was in the same direction as found by
Hofmann et al. (2010). More specifically, the automatic evaluation of the participants in the implementation intentions condition was indeed less positive than of the participants in the control condition (see Table 2 in Appendix D). However, the effect size for chocolate as temptation was so poor, η2 = .01 that even with a larger sample size, a significant effect would
probably not have been found (Brown, 2007).
These results are not in line, even contrary for chips, with former research by
Hofmann et al. (2010) about the reducing effect of implementation intentions on the positive automatic evaluation elicited a tempting stimulus.
A possible explanation for chips as a temptation having an opposite effect on the automatic evaluation between the experimental and the control group could be that a framing effect occurred. Chong and Druckman (2007) describe that framing effects occur “when (often small) changes in the presentation of an issue or an event produce (sometimes large) changes of opinion” (p. 104). This could have happened in the present study for the simple reason that the participants who had to make implementation intentions were confronted with the message that chips is a temptation, whereas the participants in the control condition were not. This in itself could have activated the participants in the implementation intentions
condition to truly see chips as a temptation, even if they did not see it like that beforehand. And this could have led them to implicitly evaluate chips as something more positive.
Furthermore, no significant difference was found between the implementation
intentions condition and the general goal intentions condition. A possible explanation for why no significant difference was found, could be that the examples that were provided for people in the general goal intentions condition also contained a reason for why the specific goal was set (e.g., I will refrain from eating chocolate/chips to lose weight). However, research indicates that the addition of the reason why could have affected the results (Adriaanse, De Ridder, & De Wit, 2009). More specifically, Adriaanse et al. (2009) found that
implementation intentions that specified the ‘why’ were more important in changing
unhealthy snacking behaviour than implementation intentions that only specified the ‘where and when’. Although no significant difference was found in the present study, the results did show that the automatic evaluation of the participants in the general goal intentions condition (where participants were asked to add a ‘why’) was less positive, especially for chips as the temptation, than the automatic evaluation of the participants in the implementation intentions condition (where participants were asked to specify the ‘where and when’; see Table 2 in Appendix D). This does not provide significant support for the findings of Adriaanse et al. (2009), but it may be interesting for future research to use this design to see whether this difference in importance between the ‘why’ and ‘where and when’ can also be supported on the implicit (automatic) level.
Strengths and Limitations
A strength of this study was that people could choose between chocolate and chips according to what they saw as temptation. However, the results showed that participants evaluated chocolate as a temptation significantly more positive than chips. Additionally, despite the fact that people could choose their own temptation, the automatic evaluation of the
participants in the control condition and who chose chips as temptation was very low (D -score = .20), especially when compared to the automatic evaluation of the participants in the control condition who chose chocolate as temptation (D-score = .57). This could indicate that
although people like chips as well, it is not as big of a temptation as chocolate. Furthermore, a low correlation was found between the explicit attitude and automatic evaluation (implicit attitude) for all conditions, except for the general goal intentions condition with chips as temptation. Therefore it could be possible that people say and/or think that they have a preference for chips, but in reality they have an implicit preference for chocolate. Future research could shed more light on this topic by asking participants about their explicit attitude towards both chocolate and chips as temptation.
Another strength is that the participants’ BMI was taken up in this study as a possible moderator, however the results showed that the BMI did not have a moderating effect on the automatic evaluation of both tempting stimuli. A possible explanation for not finding BMI as moderator could be that the participants were asked to envision that they had the goal of not eating chocolate/chips, whereas they maybe, in reality, did not have this goal at all. Moreover, according to research done by Papies, Stroebe and Aarts (2008), for people who truly have the goal of losing weight, or eat more healthily, in brief who watch what they eat, temptation cues can activate the relevant goal intention and thus facilitate self-regulation. Therefore it is possible, if future research gathers a sample size that only consists of participants who truly have the intention to watch/control what they eat, that BMI is found to be a true moderator.
A limitation of the present study was the accessibility of this survey. The survey was spread via the researchers’ Facebook page and she experienced that a lot of people wanted to participate when they saw the message, however they saw it on their mobile phone and for the IAT a keyboard was needed to be able to participate. This could have resulted in those people dropping out. Moreover, as most of the researchers’ friends on Facebook are Dutch, it could
be possible that their understanding of the English language was not adequate to truly understand the given tasks.
Next to these possible explanations of why the hypothesised effects were not found and the limitations, there were also some aspects of this study that were positive. First of all, because this is partly a replication study, the usage of mostly the same stimulus material as the original article enhanced the internal validity (Porte, 2013). Moreover, to be sure that the pictures of the furniture matched the temptation, the colours of the furniture were adjusted. This enhanced the construct validity. Lastly, the survey only took 8 -10 minutes to fill in and this made it more attractive for people to participate.
Although this study does not provide support for the usage of implementation
intentions as a resistance strategy towards food related temptations, it does give insights into the classification of what people experience as being a food related temptation. Moreover, although only few significant effects were found, specifying the reason ‘why’ seems to have a bigger effect on reducing the automatic evaluation than specifying ‘where and when’ the temptation occurs. Further research could use this design with a larger sample size, different forms of implementation intentions, different temptations and rating of said temptations to further unravel the best resistance strategy for food related temptations.
References
Adriaanse, M. A., De Ridder, D. T., & De Wit, J. B. (2009). Finding the critical cue:
Implementation intentions to change one's diet work best when tailored to personally relevant reasons for unhealthy eating. Personality and Social Psychology Bulletin, 35, 60-71.
Blanton, H., Jaccard, J., Gonzales, P. M., & Christie, C. (2006). Decoding the implicit association test: Implications for criterion prediction. Journal of Experimental Social Psychology, 42, 192–212.
Brown, J. D. (2007). Statistics Corner. Questions and answers about language testing
statistics: Sample size and power. Shiken: JALT Testing & Evaluation SIG Newsletter, 11, 31-35.
Carpenter, T. P., Pogacar, R., Pullig, C. P., Kouril, M., LaBouff, J. P., Aguilar, … Chakroff, A. (2017, August 11). Building and Analyzing Implicit Association Tests for Online Surveys: A Tutorial and Open-Source Tool. Retrieved from
https://psyarxiv.com/6xdyj/
Chong, D., & Druckman, J. N. (2007). Framing theory. Annual Review of Political Science, 10, 103-126.
Cohen, J. (1969). Statistical power analysis for the behavioural sciences. New York: Academic Press.
Field, A. (2013). Discovering statistics using IBM SPSS statistics. London: SAGE.
Gawronski, B., & Bodenhausen, G. V. (2006). Associative and propositional processes in evaluation: An integrative review of implicit and explicit attitude change.
Psychological Bulletin, 132, 692–731.
Gollwitzer, P. M. (1993). Goal achievement: The role of intentions. European Review of Social Psychology, 4, 141-185.
Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54, 493-503.
Gollwitzer, P. M., & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta‐analysis of effects and processes. Advances in Experimental Social
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual
differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology, 74, 1464–1480.
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the Implicit Association Test I: An improved scoring algorithm. Journal of Personality and Social Psychology, 85, 197–216.
Griffioen-Roose, S. (2012). The role of sweet and savoury taste in food intake and food preferences (Unpublished doctoral dissertation). Wageningen University,
Wageningen.
Hoch, S. J., & Loewenstein, G. F. (1991). Time-inconsistent preferences and consumer self-control. Journal of Consumer Research, 17, 492-507.
Hofmann, W., Deutsch, R., Lancaster, K., & Banaji, M. R. (2010). Cooling the heat of temptation: Mental self‐control and the automatic evaluation of tempting stimuli. European Journal of Social Psychology, 40, 17-25.
Lane, K. A., Banaji, M. R., Nosek, B. A., & Greenwald, A. G. (2007). Understanding and using the Implicit Association Test: IV: Procedures and validity. In B. Wittenbrink, & N. Schwarz (Eds.), Implicit measures of attitudes: Procedures and controversies (pp. 59–102). New York: Guilford
Metcalfe, J., & Mischel, W. (1999). A hot/cool-system analysis of delay of gratification: dynamics of willpower. Psychological Review, 106, 3-19.
Muraven, M., & Baumeister, R. F. (2000). Self-regulation and depletion of limited resources: Does self-control resemble a muscle? Psychological Bulletin, 126, 247-259.
Ouwehand, C., & Papies, E. K. (2010). Eat it or beat it. The differential effects of food temptations on overweight and normal-weight restrained eaters. Appetite, 55, 56-60.
Papies, E. K., Stroebe, W., & Aarts, H. (2008). Healthy cognition: Processes of self-regulatory success in restrained eating. Personality and Social Psychology Bulletin, 34, 1290-1300.
Porte, G. (2013). Who needs replication research? CALICO Journal, 30, 10–15.
Rydell, R. J., & McConnell, A. R. (2006). Understanding implicit and explicit attitude change: A systems of reasoning analysis. Journal of Personality and Social Psychology, 91, 995-1008.
Sheeran, P., Gollwitzer, P. M., & Bargh, J. A. (2013). Nonconscious processes and health. Health Psychology, 32, 460-473.
Verplanken, B., & Faes, S. (1999). Good intentions, bad habits, and effects of forming implementation intentions on healthy eating. European Journal of Social Psychology, 29, 591-604.
Webb, T. L., Sheeran, P., & Pepper, J. (2012). Gaining control over responses to implicit attitude tests: Implementation intentions engender fast responses on attitude incongruent trials. British Journal of Social Psychology, 51, 13-32.
Appendix A Characteristics Table 1
Characteristics of Participants in all Groups
Variable Total (N =119) Males (N = 33) Females (N = 84) Other (N = 2)
M age in years (SD) 33.2 (16.1) 44.7 (21.7) 28.6 (10.6) 32.5 (10.6) Education High School/GED 7.6% 12.1% 6.0% . 2-year college 4.2% . 6.0% . 4-year college 5.9% 18.2 1.2% . Bachelor's degree 54.6% 30.3% 64.3% 50% Master's degree 23.5% 30.3% 20.2% 50% Doctorate degree 4.2% 9.1% 2.4% . Marital status
Single, never married 25.2% 12.1% 29.8% 50%
In relationship (not living together) 22.7% 15.2% 26.2% .
In relationship (living together) 27.7% 30.3% 26.2% 50%
Married or domestic partnership 22.7% 39.4% 16.7% .
Divorced 1.7% 3.0% 1.2% . Ethnic background White 93.3% 93.9% 92.9% 100% Asian 3.4% 3.0% 3.6% . Other 3.4% 3.0% 3.6% . BMI Underweight 6.7% . 9.5% . Normal weight 69.7% 60.6% 73.8% 50% Overweight 21.0% 39.4% 14.3% . Obesity 2.5% . 2.4% 50% Device Desktop computer 23.5% 33.3% 20.2% . Laptop 76.5% 66.7% 79.8% 100%
Appendix B IAT Pictures
Figure 2. Picture chocolate Figure 3. Picture chips
Figure 4. Picture furniture in chocolate IAT
Figure 5. Picture furniture in chips IAT
Appendix C IAT Screenshots
Figure 9. Example of training block in the chocolate IAT, with the red X indicating that a mistake was made by the participant. Figure 8. Example of a critical block in the chocolate IAT. Participants now had to press the “I” key.
Appendix D Mean D-scores
Table 2
Mean D-score and Attitude per Condition and Temptation
Condition/Temptation Chocolate Chips
Implementation Intentions D-score(SD) N = 18 .47 (.57) N = 15 .38 (.46) Attitude(SD) 6.64 (.51) 5.67 (.79) Control D-score(SD) N = 25 .57 (.37) N = 26 .20 (.59) Attitude(SD) 6.48 (.78) 6.10 (.97) Goal Intentions D-score(SD) N = 19 .46 (.49) N = 17 .10 (.68) Attitude(SD) 6.53 (.49) 5.35 (1.56)
Figure 10. Mean D-scores sorted by temptation and together across all conditions. Standard Deviation for chocolate .47, chips .59 and total .55. Significant difference between chocolate and chips at p <.05.
0 0,1 0,2 0,3 0,4 0,5 0,6
Chocolate Chips Total
A u to m at ic E va lu at io n
