Can Emotional Self-Awareness Help Ameliorate Unhealthy Snacking Habits? The Role of
cue Deliberation on Implementation Intention Effectiveness
Vincent D. J. van Dijk
Universiteit van Amsterdam
Author Note
Vincent D. J. van Dijk, student at Department of Clinical Psychology, Universiteit van Amsterdam. This research was supervised by Aukje Verhoeven and Sanne de Wit. The final version of this paper was written on June 16. It is 7894 words long.
Abstract
Changing your bad habits is easier said than done. In other words, the intention- behaviour gap may be difficult to bridge. Here to help are implementation intentions: specific ‘if-then’ plans that specify a cue and relate it to a certain goal-directed response (e.g. ‘If I am at home and I am feeling bored, then I will eat an apple’). To better understand if and why these plans work, the current study investigated whether implementation intentions outperform goal intentions (plans that specify a desired state, e.g. ‘My goal is to eat healthy’) and if implementation intention effectiveness differs for the type of cues that it specifies. It was examined whether unhealthy snacking behaviour is more successfully changed in response to internal cues (e.g., feeling bored) than to external cues (e.g., being at home) and whether self-knowledge (emotional self-awareness) assists in identifying and recognizing relevant cues, therefore moderating the success of changing ongoing unhealthy snacking habits. Implementation intentions did not outperform goal intentions in the reduction of unhealthy snacking but internal cues did reveal more relevant than external cues. Emotional self-awareness was of aid regarding goal intentions but not implementation intentions, meaning that the explication of relevant cues inherent to implementation intentions may have annulled the benefit of cue explication caused by self-awareness. It was concluded that emotional self-awareness may be beneficial for those attempting to break unwanted habits without access to a professional familiarized with the implementation intention intervention.
Key words: breaking habits; implementation intentions; critical cues; emotional self-awareness
The Effect of Internal Cue Deliberation on Implementation Intention Effectiveness ‘We are what we repeatedly do. Excellence, then, is not an act but a habit’ (Durant, 1926). Pulitzer prize winner Will Durant wrote these words in his endeavour to paraphrase the work of Aristotle, dating back to the 4th century B.C. In the philosophical sense, habitual behaviour has intrigued us for centuries. Psychologists however, did not start examining the construct until the early 1900’s. To this day they attempt to discover how it affects our
behaviour and attempt to find out how we might (to our own advantage) influence our habits ourselves.
This last aspect seems most relevant in a clinical context of promoting healthy behaviour among individuals, which is probably one of the first practical applications of habit research that may come to mind. Unhealthy habits such as smoking, drinking or overeating may lead to serious illness. In the case of overeating, this may lead to overweight or obesity. Obesity is currently gathered among the largest medical and societal issues related to health behaviour. Globally, 37% of men and 38% of women are estimated to be overweight (Ng et al., 2013). This leads to an average of 3-9% of life years lost, due to consequences such as type II diabetes and cardiovascular disease (Ng et al., 2013). That means there is an immense number of people to benefit from prevention strategies, psychological or other interventions and scientific
knowledge of etiological factors. This is why much research on changing habits has been focused on changing eating behaviour. First however, it was necessary to understand how unhealthy habits are formed.
Habits can be seen as routine behaviour following slow associative learning. They form in accordance with Thorndike’s Law of Effect (1911) which states that when a behaviour is
rewarded, this reward strengthens the association between the stimulus (S) and response (R) making the process increasingly automatic and deeming it resilient to change (de Wit & Dickinson, 2009). For instance, one might reward himself by eating a bag of chips when watching television. In the future, the TV (S) will become progressively associated with the eating of chips (R) and possibly make it eventuate more frequently. The isolated S-R relationship may even become so strong that it can persist despite changes in the gratifying value of
outcome or reward (O). Habits are distinguished from goal-directed behaviour, which instead depends on the relationship between the action and its associated outcome (R-O). Two criterions must be fulfilled for behaviour to be regarded as goal-directed (de Wit & Dickinson, 2009). First, there must be knowledge of the causality of the R-O relation. One must believe that performing an action will produce the desired reward, which is why this is also known as the belief criterion. Second, one must be able to conclude whether the reward is currently desired to be able to decide whether or not to perform the behaviour, also known as the desire
criterion. In evaluating a reward and choosing how to act instead of responding subconsciously, goal-directed behaviour is not automatic. At times habits and goal-directed behaviour can co-occur meaning that when a present habit is unhealthy, it can counteract the course of healthy conscious intentions (Cheung & Limayem, 2005).
Relatedly, simply setting a goal (forming a goal intention) often proves to be insufficient in breaking unwanted habits (Orbell 1998; Sheeran 2002). Goal intentions are defined as a broad formulation of the objective a person is trying to achieve (e.g. ‘My goal is that I will eat healthier’). In many cases, these intentions fail, leading Sheeran (2002) to state that the so
called ‘intention-behaviour gap’ may be difficult to bridge. However, a publication from Gollwitzer (1999) brings to light that we can in fact influence many different habits, using a strategy known as an implementation intention. An implementation intention is an if-then plan that specifies a context and an approach related to a certain goal (e.g. ‘If I am at home and I am feeling bored, then I will eat a piece of fruit’ as an implementation intention for snacking healthily). After being formed, implementation intentions are most often visualized, written down or spoken out. The hypothesized mechanism behind their effect is believed to be two-fold (Hagger & Luszczynska, 2014). Because of its specificity, one can get sensitized to the chosen cue, which then becomes more mentally accessible (1). Additionally, the cue is more likely to activate the newly linked (desired) response (2), because of a heightened association between the two. In this manner, performing the coveted behaviour gets delegated to the environment; not to an exercise of prolonged cognitive effort. In an account of this effect, Webb and Sheeran (2008) discovered that both accessibility and association strength mediate the effect of
implementation intentions by using a lexical decision task with relevant and non-relevant primes. Advocating this conclusion; Bayer, Achtziger, Gollwitzer & Moskowitz (2009) showed that in presence of the cue, the desired behaviour can follow in the absence of conscious processes. This was done by using subliminal cues, keeping them outside of awareness.
The current study will attempt to compare goal intentions to implementation intentions, investigate which element of implementation intentions is the most effective and examine whether some individuals might benefit more strongly from this than others. Prospering on its theoretical background, implementation intentions have been found to work better than goal
intentions (concerning snack behaviour) in multiple studies (Adriaanse, Vinkers, de Ridder, Hox & de Wit, 2011) containing a medium effect size for unhealthy snacking, Cohen’s d = .29
(Adriaanse et al., 2011). This corresponds to the first expectation that implementation
intentions are more effective at decreasing unhealthy snacking than goal intentions. The proven value of implementation intentions has led to the investigation of their etiology.
Adriaanse, de Ridder and de Wit (2009) made a dichotomous distinction of implementation intention constituents, namely of internal and external cues. Somewhat expanding on their definition, an external cue will be defined as a location (e.g. at home), company (e.g. being alone) or activity (e.g. studying) in regard to snacking behaviour. An internal cue will be defined as the private internal inducement of snacking, such as a physical/sensory event (e.g. feeling hungry and faint) or an affective state (e.g. to deal with negative emotions). To gain a better understanding of the way implementation intentions function Adriaanse et al. (2009) examined the contribution of each these parts, targeting both healthy and unhealthy snacking. In the first study participants were assigned a pre-formulated implementation intention, whereas in the second study they were allowed to choose one based on their personal situation. Only when implementation intentions were tailored to the
individual (second study) did both healthy and unhealthy snacking change in the differentially aimed directions; the first study only affected healthy snacking. This showed that using
personally relevant cues is recommended, meaning that the cue should be a relevant instigator of the targeted habitual behaviour. They also discovered that using internal cues works better than using external cues. Researchers however, seem divided on this issue. Gollwitzer (1999) for instance, argued the opposite in saying that cues should be readily available in the environment
instead of “internal states” and Hagger and Luszczynska (2014) stated in their review that the type of cue does not make a difference as long as the cue is personally relevant. We intent to contribute to solving this disagreement.
A difference in the favour of internal cues is expected because we hypothesize that in triggering snacking, internal cues might generalize more broadly across external cue situations than external cues do across internal cue situations; illustrated by the following. People might snack in context of a single relevant internal cue (e.g. dealing with negative emotions) in varying external cue situations (e.g. at home, while studying, being alone) but in the context of a single relevant external cue (e.g. at home), only the relevant internal cue (dealing with negative emotions) will induce snacking; so not feeling happy or feeling bored, etc. This leads us to conclude the results from Adriaanse et al. (2009) may have originated from the internal cue being the determining factor. Our second expectation is therefore that unhealthy snacking behaviour will be more effectively reduced in response to internal cues than to external cues concerning implementation intentions. In examining this, the current study will examine a longer period than Adriaanse et al. (2009) so that the effect of time can be examined. This is of interest since all variables might change over time; perhaps the interventions will yield results only at first but not later.
Whether or not it is more effective to focus on external or internal cues, additional variables may play an influential role. It might be argued that one needs a certain proficiency of self-knowledge in order to more effectively recognize or identify personally relevant cues. Attaining self-knowledge indeed possesses many empirically identified obstacles (Wilson & Dunn, 2003). Therefore, our third expectation is that high self-knowledge leads to higher
implementation intention effectiveness for (determinative) internal cues. To operationalize self-knowledge, the Trait Meta Mood Scale (TMMS) will be administered. The TMMS is a test of emotional self-awareness. It assesses stable differences in attending to emotions and states and comprehending them (Salovey et al., 1995). In the implementation intention literature, internal cues are almost always specified as ‘When I feel …’ (bored/social/etc.). Emotional
self-awareness therefore seems to be an adequate measure.
The goal of this paper is to unearth variables associated with higher implementation intention effectiveness. It will attempt to answer if implementation intentions are more
effective than goal intentions (1); if snacking behaviour is reduced more in response to specified internal cues than to external cues (2); and if snacking behaviour is more strongly reduced with internal cues in individuals with high TMMS scores than in those with lower TMMS scores (3).
Method Participants
Participants were female students or recent graduates. They were recruited by handing out flyers at the University of Amsterdam and surrounding supermarkets, putting up posters at student gyms and via social media (Facebook). Distribution across conditions was organized considering BMI, to ensure their equivalence. Participants were not compensated in any way.
Seven predetermined criteria could lead to exclusion. Participants had to be between 18 and 35 years old and acknowledge that they (at times) display unhealthy snacking behaviour and were motivated to change this, since room for improvement should theoretically be achievable. That is why during the first week of measurement, three or more unhealthy
snacking occasions had to be present. There could be no history of an eating disorder. Participants could not leave on vacation for more than four consecutive days during data collection and none during baseline measurement. The reason for this was that on vacation participants may eat differently, lack access to an internet source, or simply forget to keep their snack diary. BMI had to be between 20 and 35. We included those within the normal range of BMI, since those who are underweight or overweight would be better off consulting a dietician. Lastly, participants had to own a smartphone in being able to keep a snack diary.
A total of 65 participants signed up, five of which were excluded from the program. They either had an eating disorder (n = 3), did not report enough unhealthy snacking occasions during baseline (1) or were unmotivated (1). From the remaining 60 participants, 13 failed to complete the study. Reasons for discontinuation where that the program did not accommodate their expectations (2), a lack of time (2), personal reasons (2), not feeling motivated (1), becoming ill during the study (1), switching to a program that included sports (1) losing track while on vacation (1) and reasons unknown (3). Lastly, 6 participants were excluded from our analysis, since they failed to keep their snack diary. This resulted in the final sample of 41 participants.
A drop-out analysis was conducted as a standardization test to see if there were any differences in pre-existent characteristics between those that were included in our analysis and those that were not. For the continuous variables, an independent t-test was used. Levene’s test showed equal variances for age, BMI, T1 SRHI, T1 motivation to reduce unhealthy snacking and T1 motivation to increase healthy snacking. The t-test revealed no significant differences, meaning that the assumption of equality across inclusion was met (all p’s > .395). For the categorical variables, a Chi-square test was used. This resulted in a non-significant value for
education level (Pearson χ2(4) = 4.821, p = .306) meaning included and excluded participants were similar concerning this variable. For long term goal however, the result was significant (Pearson χ2(5) = 11.380, p = ,044) meaning those that were included had different long term goals from those that were excluded (see Table 1). The fact that those who persisted set different goals from those that did not is not unexpected though. For instance, certain goals could be too easy or difficult leading to prematurely terminated participation.
Table 1.
Long Term Goal for Included and Excluded Participants
Included Excluded
Losing weight 21 6
Staying healthy 13 3
Achieving a goal weight 2 4
A tight body 1 3
Maintaining clothing size 0 1
Different 4 2
Participants in the final sample had a mean age of 24 (SD = 3.80). The average BMI was 24,9 (SD = 3.12) and the diary was filled in for a total of 30 days on average (SD = 2.80).
3) or WO (n = 20). Motivation to reduce unhealthy snacking at T1 was 5.8 (SD = .64), which was 5.76 (SD = .86) for increasing healthy snacking.
Design
A 5 (week) X 2 (conditions) X 2 (types of cues) design was used. Even though both types of cues were specified in one implementation intention, it was nevertheless possible to test the difference between them when they were treated as different levels of a within-subject
outcome variable. In this way, it would become clear in response to which type of cue the number of snacking occasions (compared to baseline) is most pronouncedly reduced: linked to the specified internal cue or linked to the specified external cue.
Procedure
In short, there was a telephone call followed by three appointments and five weeks of keeping a snack diary. All appointments were held with a personal coach who was trained by an experienced instructor familiar with the current procedure.
Participants were informed about the study and asked about exclusion criteria via telephone before the first appointment could be scheduled. In this appointment, further information was provided and an entry questionnaire was filled out. Height, weight and waist size were determined. During this day and the following 6 days, participants received text messages leading to an online diary where they filled out their regular snacking behaviour.
During the second appointment, the intermediate questionnaire was administered. Weight and waist size were measured a second time. Most relevantly, the conditions were created at this point by manipulating the type of plan. Participants in the implementation intention condition (II) formed a specific intention containing both an internal and an external
cue (“If external cue X and internal cue Y, then I will eat a piece of fruit”) and participants in the goal intention condition (GI) formed a more general intention (“My goal is to eat fewer
unhealthy snacks and more fruit and vegetables”). This occurred in consultation with a researcher applying motivational interviewing techniques. For the following 28 days,
participants filled out their post-intervention snack diary, registering the present internal and external cues.
In the final appointment, participants completed an exit questionnaire and had their weight and waist size measured a third time. The TMMS was filled in. The TMMS could only be administered during the final meeting, since 21 participants had already been tested previously (de Vries, 2015). This way, alteration of the design was minimalized. Participants were then thanked and debriefed. Finally, a number of measures that were not used in this study were administered across appointments1. For an outline of the general procedure, see Figure 1.
1
These include: Verplanken and Orbell’s Self-Reported Habit Index (SRHI, 2003) as a measure of unhealthy snack habit strength (T2 and T3); the Slips Of Action Task by de Wit et al. (SOAT, 2013) as a computerized measure of habit propensity (T1 and T3); and a Lexical Decision Task (LDT) as a
Figure 1. General Procedure Check participant criteria Informed consent Instructions snack diary Text message with link to snack diary (twice a day) Identification external and internal cues Formulating implementatio n intention or goal intention
Text message with link to snack diary (each day) Check implementation intention Debriefing Day 1/ T1 Screening Day 1-7 Baseline snack diary Day 7/T2 Manipulation Day 10-34 Post-intervention Snack diary Day 35/T3 End Entry questionnaire Measure: - Height - Weight - Waist Snack diary - Internal cue - External cue Intermediate questionnaire Measure: - Weight - Waist Snack diary - Internal cue - External cue - Number of times specified internal and external cue were encountered Exit questionnaire TMMS Measure: - Weight - Waist Materials First appointment.
Entry questionnaire. This measure was used to examine necessary exclusion and to assess motivation. It first asked about education level. Participants were asked to specify their long term goal of participating the in current study, if they had one. Adjacently, two checks were present, pertaining to weight changes and eating disorder history. Eating disorder history was determined by directly asking. If someone had indeed had an eating disorder, additional questions were asked such as: ‘Do you think you are fat, while others find you skinny?’. It also contained the Self-Reported Habit Index (Verplanken & Orbell, 2003) as an indication of
preceding habit strength, containing items such as ‘Eating unhealthy snacks is something I do without thinking’. This was assessed on a 7-point Likert scale (1 = not at all motivated; 7 = very strongly motivated). Cronbach’s alpha was determined at .83. Lastly, motivation to reduce unhealthy snacking and motivation to increase healthy snacking were questioned on the same scale.
Baseline snack diary. The daily diary forms the main dependent measure and was used to ascertain regular snacking behaviour (see Appendix A). Keeping a snack/diet diary has been the method of choice in countless food-related studies for years (De Castro, 1994). It is
considered sufficiently reliable and valid, viewed as the measure of choice in answering these types of questions and the standard up to which new instruments are tested (De Castro, 1994). The diary started with an instruction encouraging participants to maintain their regular snacking behaviour, to fill out each moment separately, to only enter snacks apart from their daily meals and to label quantities adequately. An explanation of internal and external cues was presented. To commence, the button ‘I have read the instructions’ had to be clicked. Healthy and
unhealthy snacks could be selected from a list (or entered freely, if necessary). The amount had to be entered as well. External cues (separate for company, location and activity) and internal cue could be typed in or selected from a number of options, e.g., ‘alone’, ‘visiting someone’, ‘studying’ and ‘to enjoy a special occasion’. It could also be indicated that no snacks had been eaten that day.
Second appointment
Posttest snack diary. This was the same as the baseline diary except for the first instruction, clarifying the additional question at the end which registered the number of encounters with a persons specified internal and external cue. (see Appendix B)
Intermediate questionnaire. A second measure of motivation to reduce unhealthy snacking and motivation to increase healthy snacking.
Third appointment
Exit questionnaire. To investigate several post-hoc variables, an exit questionnaire was used. Two items were again used to measure motivation for a third time.
Trait Meta Mood Scale. In measuring emotional self-awareness, the Dutch version of the TMMS was adopted from an earlier study by Nyklícek and Denollet (2009), containing 34 items. Answers were scored on a 5-point Likert scale, ranging from 1 = strongly disagree to 5 = strongly agree. The first subscale, Attention to Feelings, contains items such as: ‘I don’t think it’s worth paying attention to your emotions or moods’. An example from the second subscale, Clarity of Feelings, is: ‘I am rarely confused about how I feel’. The subscale Mood Repair was not included as this scale concerns repairing unpleasant moods, which has no relevance in the pursued measurement of identifying or recognizing emotions. Advocating its reliability, Cronbach’s alphas for the subscales were .86, .88 and .82, respectively. In addition, evidence supporting its validity has been produced numerously (Gorostiaga, Balluerka, Haranburu & Alonso-Arbiol, 2011; Li, Yan, Yin & Wu, 2002).
Data analyses
The first main analysis, namely the comparison of the two conditions, will consist of a One-way Repeated Measures ANOVA using the five weekly averages of unhealthy snacking occasions per day.
The second main prediction, namely comparing the contribution of the internal cue component to that of the external cue component, will be tested by a Two-way Repeated Measures ANOVA. To acquire a more pure statistical representation of cue-related snacking behaviour, the averages will need to be corrected for the number of times that the specified cue was encountered. For instance, if someone had three snacking occasions in context of their specified external cue in week 1 while it was present 10 times; and also had three snacking occasions in week 2 but with it now being present 20 times, it could be argued that even though the number of snacking occasions remained equal, improvement had in fact occurred since the cue was resisted more often. Correction was not yet implemented in the first analysis because of its contrasting focus: unhealthy snacking in general, regardless of cues. By dividing the number of snacking occasions in presence of the specified cue by the total number of encounters with that same cue, a weekly cue specific (internal/external) fail-score will be calculated. This can then be transformed to a success-score by subtracting it from 1. In the first week, the diary was constructed as such that it did not contain an enquiry into the number of encounters, meaning that only the four post-intervention weeks can be transformed and included.
In the case of not encountering a specified cue in a particular week, calculation of a success-score will not be possible. (Partly) therefore, most participants will have some missing
data. This is problematic, since repeated measures analyses automatically exclude all cases with missing data points, meaning adjustments might be necessary. The most gentle statistical way to deal with missing data is by using the multiple imputation method (Schlomer, Bauman & Card, 2010) but this does not allow for subsequent repeated measures analyses. Therefore, the Expectation Maximization (EM) algorithm will be applied. The same pre-existent variables as in the standardization checks will be used in the estimation process, since Collins, Schafer and Kam (2001) dictated that incorporating auxiliary variables in the imputation process can be beneficial in reducing estimation bias and partially restoring the loss of power caused by missingness. Imputations will be generated separately for the external cue and internal cue variable to maintain present constitutional correlations, relevant for the third and final analysis and thus adhering to the outlined theoretical framework.
For the third main analysis, examining the role of self-awareness, a Pearson correlation will be used to examine relations between the TMMS score and either of the success-scores (internal/external) in both conditions. This will require the formation of a success-score total for both types of cues. These will be calculated by averaging the success-scores of its four
constituentweeks. Again, only posttest weeks can be included. Since this procedure is not problematic with missing values, the non-estimated scores will be entered. Finally, to decide whether the correlations within each condition are notably different from each other, Steiger’s Z test will be carried out (Lee & Preacher, 2013).
Results Standardization
In advance, it was of interest to see if drop-out from the main analyses was equal across conditions. The Chi-square test showed that this expectation was not violated (Pearson χ2(1) = .749, p = .387). A manipulation check was then installed to indicate whether there were any differences in pre-existent qualities across conditions for those included. For the continuous variables, an independent t-test was used. Levene’s test revealed equal variances for age, BMI, T1 SRHI, T1 motivation to reduce unhealthy snacking and T1 motivation to increase healthy snacking. The following t-test did not reveal any significant differences meaning that the assumption of equality across conditions had been met. For the categorical variables, a Chi-square test was used. Neither for education level (Pearson χ2(4) = 1.093, p = .895) nor for long term goal (Pearson χ2(5) = 4.704, p = .453) was this test significant, confirming that education level and goal were similar for the two conditions.
To decide on a final form for the self-knowledge variable, the TMMS was tested for its internal consistency. Unfortunately, the value was below what was originally anticipated (α = .65). However, after deletion of the item ‘I constantly think about what I am feeling’, an acceptable value of α= .70 was attained.
First main analysis
Now that all standardization checks were investigated, the first main analysis of
comparing the two conditions could be conducted. A one-way repeated measures ANOVA was intended, but the assumption of normality had been violated in the GI condition in every week (all p’s < .045).Neither log transformation, square root transformation, or reciprocal
transformation could fully overcome this. Although the same variables did not violate the assumption of normality in the II condition (all p’s were .200), application of the non-parametric alternative was still preferable.
To investigate whether the number of unhealthy snacking occasions had decreased over time overall, a Friedman test was conducted. By looking at Figure 2, it seems the median
number of unhealthy snacking occasions declined between baseline and post-intervention.
There was indeed a statistically significant difference in unhealthy snacking over time, χ2 (4) = 19.189, p = .001. To see where this difference occurred, a paired-samples sign test was used instead of the Wilcoxon signed rank test in order to avoid the requirement for symmetrical distributions. Considering the Bonferroni adjustment, the post-hoc sign test showed that a statistically significant change in unhealthy snacking had occurred between baseline and weeks 2, 3, 4 and 5 (see Table 2).
Table 2.
Sign Test Statistic (Z), p-value and Effect Size (r) for the Difference in Unhealthy Snacking Between Baseline and Subsequent Weeks
Weeks Z p r
1 – 2 -4.707 <.001 .52
1 – 3 -3.843 <.001 .42
1 – 4 -3.731 <.001 .42
1 – 5 -4.380 <.001 .49
Note. Significant at the p<.05 level.
To determine if this drop in unhealthy snacking between baseline and post-intervention differed for the two conditions, Mann-Whitney’s U test was used to compare mean ranks. Ranks are calculated after aggregating the two conditions into one column and assigning them in consecutive order. The scores are then split back and the mean of the rank in each group is computed. Viewing Figure 3, they seem to be relatively similar.
Figure 3. Mean Ranks for the II condition and the GI condition per Week
Correspondingly, the test revealed no significant differences in unhealthy snacking between the GI condition and the II condition in week 1 (U = 170, p = .307), week 2 (U = 194, p = .695), week 3 (U = 195, p = .714), week 4 (U = 170, p = .591) or week 5 (U = 160.5, p = .450). It could
therefore be concluded that unhealthy snacking did not decrease more strongly in the II condition than in the GI condition.
Second main analysis
For the second main analysis, the goal was to compare the contribution of the internal cue component to that of the external cue component. Overall, 20 out of 41 participants had
some missing data, amounting to 16.77% of incomplete values. To avoid exclusion, data imputation was desirable. A non-significant Little’s MCAR test revealed that there were no apparent patterns of missing values and the EM algorithm could be applied. Table 3 displays regular and imputed means and standard deviations for both types of cues.
Table 3.
Regular and Imputed Mean and Standard Deviation per Type of Cue
Unfortunately the resulting success-scores did not meet the normal distribution requirement (all p’s < .039), except for week 3 for the external cue. None of the earlier mentioned
transformations were capable of resolving this difficulty. Since no robust test of the factorial repeated-measures design is currently available (Field, 2013), a regular two-way repeated measures ANOVA was used. Regrettably, this strongly limits the comprehensiveness of any subsequent conclusions.
External cue Internal cue
Week Regular Mean (SD) Imputation Mean (SD) Regular Mean (SD) Imputation Mean (SD)
2 .74 (.29) .75 (.27) .79 (.29) .80 (.28)
3 .67 (.29) .67 (.26) .82 (.25) .83 (.23)
4 .48 (.41) .48 (.38) .78 (.31) .78 (.29)
Mauchly’s test of sphericity indicated that the assumption of sphericity had been
violated for time (χ2(5) = 16.994, p = .005) and for the interaction between time and type of cue (χ2(5) = 17.919, p = .003). Since the estimated epsilon was greater than 0.75 in both cases (ε = 83 and ε =.77 respectively), the Huyn-Feldt correction was utilized. Figure 4 and Figure 5 display the course of the success-scores of the separate conditions.
Figure 5. Posttest Success-scores per Type of cue in the GI Condition
Comparing the two figures, it seems as though they are reasonably alike. The difference
between conditions was indeed found not significant, F(1, 39) = .077, p =.783. The course of the separate lines does seem to be disparate however. The interaction between time and type of cue revealed significant (F(2.54, 90.07) = 4.608, p = 0.007), meaning the number of snacking occasions per week differed for each level of the type of cue. In interpreting this interaction, it was investigated for simple main effects.
First, this was done for type of cue. The paired-samples sign test showed a significant simple main effect of type of cue in weeks 3 and 4 and a marginally significant effect in week 5. In both cases, the success-score was higher for the internal cue than for the external cue. The associated statistics are illustrated in Table 4.
Table 4.
External and Internal cue Success-score Median (EC Mdn; IC Mdn), Sign Test Statistic (Z), p-value and Effect Size (r)
Week Mdn EC Mdn IC Z p r
2 .81 .89 -1.644 .100 .18
3 .73 .92 -3.601 <.001 .40
4 .56 .87 -2.959 .003 .33
5 .83 .89 -1.833 .067 .20
Note. Significant at the p<.05 level.
Second, the simple main effect of time was investigated by the use of a Friedman test. For the external cue success-score, there was a statistically significant difference between weeks (χ2(3) = 22.554, p < .001). The post-hoc sign test showed a significant drop from week 2 (Mdn = .81) to week 4 (Mdn = .56), Z = -4.056, p < .001 and a significant rise between week 4 (Mdn = .56) and week 5 (Mdn = .83), Z = -3.523, p < .001. For the internal cue success-score, the Friedman test was not significant (χ2(3) = 2.889, p = .409), indicating the effects were stable over time. Third main analysis
For the third main analysis, the goal was to discover whether self-knowledge plays a part in internal or external cue effectiveness. The included sample was smaller (n = 20), since some participants had been tested prior to the current outline as mentioned in the method section. As shown in Figure 6 and Figure 7, TMMS score was significantly correlated with both the
external cue success-score (rs (10) = .699, p = .024) and the internal cue success-score (rs (10) = .640, p = .046) in the GI condition.
Figure 7. Relationship Between TMMS Score and the Internal cue Success-score
In the II condition, TMMS score was not significantly correlated with either the external cue success-score (rs (8) = .282, p = .498) or the internal cue success-score (rs (8) = -.099, p = .815). Steiger’s Z test revealed no significant difference in either the II condition (Z = 0.696, p = .486) or the GI condition (Z = .230, p = .818).
Discussion
This study compared goal intentions to implementation intentions in reducing unhealthy snacking. A comparison was made in boosting the effectiveness by an internal and external cue component and additionally, the role of self-knowledge was examined. No effect of forming implementation intentions was found, although unhealthy snacking did decrease over time. At
first, both types of cues seemed equally important, but the contributing value of the external cue component deteriorated over time; only to recover somewhat in the final week. Self-knowledge was found to play an assisting role, but only in the goal intention condition and concerning both types of cues.
It was unexpected that forming an implementation intention was no more effective than forming a goal intention. Nonetheless, this does not necessarily mean that the two are equal, only that equivalence of effects could not be rejected. The inability to confirm a difference in favor of implementation intentions might have originated in the large comparability of the intervention in both groups in the current study. In being able to compare improvement for each type of cue, they had to be specified and tracked in both the II condition and the goal intention condition. This could possibly have lead participants in the goal intention condition to focus on their relevant cues as well, thus incorporating the active element of the experimental condition. Unfortunately, this drawback could not be overcome since the mentioned
measurements were required to test the two following predictions. A future study might evade this problem however, simply by investigating only the first prediction. As a less persuasive argument, the stated explanation could be explored by inquiry in the exit questionnaire of a similar study. Although the current finding contradicts what was found in many unhealthy snacking related studies (e.g. Adriaanse, de Ridder & de Wit, 2009; Kroese, Adriaanse & Evers, 2011), the meta-analytical evidence of implementation intention effectiveness (Adriaanse, Vinkers, de Ridder, Hox & de Wit, 2011) persuades to state that their usefulness should seemingly not be doubted at this point.
It was shown that following manipulation, both cue components (internal and external) were initially effective. The fact that this effect remained constant for the internal cue but not for the external cue could mean that internal cues generated a more durable response, showing that internal cues are indeed more relevant for snacking behaviour. Favorably, this result seems to be in line with the resembling study by Adriaanse et al (2009), somewhat strengthening the confidence that may be held in their results. The recovery of external cue effectiveness in the final week was unanticipated, but does not necessarily dismiss the indicated interpretation. For instance, reaching the end of the program and approaching face-to-face contact with a coach after a four week individual period could have caused a boost in motivation, therefore
temporarily restoring the external cue effect. Also, the difference was still marginally significant in favor of the internal cue. Keep in mind that the baseline week could not be included in forming this depiction, so even though the effect of the internal cue seems more durable, there is (unfortunately) no information on how this would relate to the state of affairs previous to manipulation. This drawback might be overcome in the future by installing a second
pre-intervention week. In that case, relevant cues could be decided after the first week, serving this single purpose. The second week would then allow for registration of the number of cue
encounters and be able to function as the week of baseline measurement.
It should be noted that seeing there was no non-parametric alternative for this analysis, the results should only be interpreted with caution. Additionally, it could also be argued that the necessary estimation of missing data points further diluted the comprehensiveness of following conclusions. Schlomer et al. (2010) did note that EM might be more appropriate for exploratory
testing than for inferential analysis because it does not provide standard errors and confidence intervals. However, Graham (2009) reminds us in his comprehensive work that many important analyses do not use these anyway and that excluding cases with missing data could in fact be considered worse than using modern ways of replacing them. He states that substituting missing data should serve as ‘our basic platform’ and that it would ‘raise the quality’ of all contemporary research.
The reliability of the acquired data might still have been affected by another cause however, because participants were never reminded of their specified cues which may have lead to forgetting them. This should have no effect on snacking reduction since implementation intentions can function subconsciously (Bayer et al, 2009). When forgotten however, answering the number of times that the cues were encountered becomes unreliable, ultimately influencing success-scores. Future studies could overcome this drawback by incorporating the specified cues as a customized header in the diary.
Interestingly, emotional self-awareness was shown to be related to habit decline in the goal intention condition, but not in the implementation intention condition. Although this contradicts what was initially predicted, the proposed mechanism might still apply. Since the relevant cues are made explicit in the implementation intention condition, self-knowledge may no longer be required and consequently, lose its aiding value. In the goal intention condition however, focusing on relevant cues is not encouraged; which might be why those who possess self-knowledge, exercising reflection or self-observation, end up battling their unhealthy habits
more successfully. Contrary to what was expected, this was equal for both types of cues suggesting a more generic effect.
Unfortunately, no definite statement about causality can be made on account of the analysis being correlational. Nevertheless, degree of self-knowledge is arguably a pre-existent characteristic, driving presumptions in the stated direction. Also, the associations emerged despite a strikingly low sample size, strengthening the confidence that might be held in regard to their accuracy. Lastly, it could be noted that self-knowledge is not necessarily correct (Veenman, 2015, Nesbit & Wilson, 1977). This is not necessarily problematic in the light of the current study however. Correct or not correct, those who engage in this type of reflection might benefit. The current study has added onto the existing literature by being the first to examine this effect.
Future research will have to indicate whether the current findings are replicable, ideally under improved methodological circumstances such as having a larger sample size (hopefully resulting in the enabling of parametric testing). The external validity will also have to be addressed, investigating different populations from the current selective (student) sample. It would also be interesting to see whether the same results apply to healthy snacking or to
alternative types of complex habitual health behaviour such as smoking or alcohol consumption. In those cases, similar results might be expected.
The current study has tentatively added on the work by Adriaanse et al. (2009) in making an argument for the relevance of internal cues in breaking habitual snacking behaviour. It was concluded with caution that self-knowledge may be beneficial for those attempting to break
unwanted habits without access to a mental health professional capable of instructing the implementation intention formulating process.
References
Adriaanse, M. A., de Ridder, D. T.D., & 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(1), 60-71.
Adriaanse, M. A., Vinkers. C. D. W., De Ridder, D. T. D., Hox, J. J., & De Wit, J. B. F. (2011). Do implementation intentions help to eat a healthy diet? A systematic review and meta-analysis of the empirical evidence. Appetite, 56, 183-193.
Bayer, U. C., Achtziger, A., Gollwitzer, P. M., & Moskowitz, G. B. (2009). Responding to subliminal cues: do if-then plans facilitate action preparation and initiation without conscious intent?. Social Cognition, 27(2), 183-201..
Cheung, C., & Limayem, M. (2005). The role of habit in information systems continuance: examining the evolving relationship between intention and usage. ICIS 2005 Proceedings, 39.
Churchill, S., & Jessop, D. C. (2011). Too impulsive for implementation intentions? Evidence that impulsivity moderates the effectiveness of an implementation intention
intervention. Psychology and Health, 26(5), 517-530.
Collins LM, Schafer JL, Kam CM. 2001. A comparison of inclusive and restrictive strategies in modern missing data procedures. Psychol. Methods 6:330–51
assessment of diet: a comparison of a 7-day diary with a food frequency questionnaire using urinary markers of nitrogen, potassium and sodium. International Journal of Epidemiology, 30(2), 309-317.
De Castro, J. M. (1994). Methodology, correlational analysis, and interpretation of diet diary records of the food and fluid intake of free-living humans. Appetite,23(2), 179-192. Durant, W. (1926). The story of philosophy: The lives and opinions of the world's greatest
philosophers. Time Inc.
Field, A. (2013). Discovering statistics using IBM SPSS statistics (pp. 555). Sage.
French, S. J. (1999). The effects of specific nutrients on the regulation of feeding behaviour in human subjects. Proceedings of the Nutrition Society, 58(3), 533-540.
Gawrilow, C., Gollwitzer, P. M., & Oettingen, G. (2011). If-then plans benefit executive functions in children with ADHD. Journal of Social and Clinical Psychology, 30(6), 616-646.
Gollwitzer, P.M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54, 493–503.
Gorostiaga, A., Balluerka, N., Aritzeta, A., Haranburu, M., & Alonso-Arbiol, I. (2011). Measuring perceived emotional intelligence in adolescent population: Validation of the Short Trait Meta-Mood Scale (TMMS-23). International Journal of Psychology and Psychological Therapy, 11(3), 523-537.
Graham, J. W. (2003). Adding missing-data-relevant variables to FIML-based structural equation models. Structural Equation Modeling, 10, 80–100.
Graham, J. W. (2009). Missing data analysis: Making it work in the real world. Annual Review of Psychology, 60, 549-576.
Heckhausen, H., & Gollwitzer, P. M. (1987). Thought contents and cognitive functioning in motivational versus volitional states of mind. Motivation and emotion, 11(2), 101-120. Hagger, M. S., & Luszczynska, A. (2014). Implementation intention and action planning
interventions in health contexts: State of the research and proposals for the way forward. Applied Psychology: Health and Well‐Being, 6(1), 1-47.
Kroese, F. M., Adriaanse, M. A., Evers, C., & De Ridder, D. T. (2011). “Instant success”: Turning temptations into cues for goal-directed behavior. Personality and Social Psychology Bulletin, 0146167211410889.
Lee, I. A., & Preacher, K. J. (2013, September). Calculation for the test of the difference between two dependent correlations with one variable in common [Computer software].
Available from http://quantpsy.org.
Li, C., Yan, J., Yin, X., and Wu, Z. (2002). A primary study of the application of the Trait
MetaMood Scale in military medical students. Chinese Journal of Clinical Psychology, 10, 202- 203.
Mayer, J. D., Mamberg, M. H., & Volanth, A. J. (1988). Cognitive domains of the mood system. Journal of Personality, 56(3), 453-486.
Mayer, J. D., & Salovey, P. (1993). The intelligence of emotional intelligence. Intelligence, 17(4), 433-442.
Ng, M., Fleming, T., Robinson, M., Thomson, B., Graetz, N., Margono, C., ... & Gupta, R. (2014). Global, regional, and national prevalence of overweight and obesity in children and adults during 1980–2013: a systematic analysis for the Global Burden of Disease Study 2013. The Lancet, 384(9945), 766-781.
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: verbal reports on mental processes. Psychological review, 84(3), 231.
Nyklíček, I., & Denollet, J. (2009). Development and evaluation of the Balanced Index of Psychological Mindedness (BIPM). Psychological Assessment, 21(1), 32.
Orbell, S., Hodgkins, S., & Sheeran, P. (1997). Implementation intentions and the theory of planned behavior. Personality and Social Psychological Bulletin, 23, 945-954.
Orbell, S., & Sheeran, P. (1998). ‘Inclined abstainers’: A problem for predicting health‐related behaviour. British Journal of Social Psychology,37(2), 151-165.
Pallant, J. (2007). SPSS survival manual: A step-by-step guide to data analysis using SPSS version 15. Nova Iorque: McGraw Hill.
Roth, P. L. (1994). Missing data: A conceptual review for applied psychologists. Personnel Psychology, 47, 537–570.
Scheier, M. F., & Carver C. S. (1985). The Self-Consciousness Scale: A revised version for use with general populations. Journal of Applied Social Psychology, 15(8), 687-699.
Schlomer, G. L., Bauman, S., & Card, N. A. (2010). Best practices for missing data management in counseling psychology. Journal of Counseling Psychology, 57(1), 1-10.
Sheeran, P. (2002). Intention—behavior relations: A conceptual and empirical review. European review of social psychology, 12(1), 1-36.
Steiger, J. H. (1980). Tests for comparing elements of a correlation matrix. Psychological Bulletin, 87, 245-251.
Thorndike, E. L. (1911). Animal intelligence: experimental studies. New York: Macmillan. Veenman, M. V. J. (2015). Metacognition: know thyself. De Psycholoog, 10-21.
Verplanken, B., & Orbell, S. (2003). Reflections on Past Behavior: A Self‐Report Index of Habit Strength. Journal of Applied Social Psychology, 33(6), 1313-1330.
Webb, T. L., & Sheeran, P. (2008). Mechanisms of implementation intention effects: the role of goal intentions, self‐efficacy, and accessibility of plan components. British Journal of Social Psychology, 47(3), 373-395.
Wieber, F., Odenthal, G., & Gollwitzer, P. (2010). Self-efficacy feelings moderate implementation intention effects. Self and Identity, 9(2), 177-194.
Wilson, T. D., & Dunn, E. W. (2004). Self-knowledge: Its limits, value, and potential for improvement. Psychology, 55.
de Wit, S., & Dickinson, A. (2009). Associative theories of goal-directed behaviour: a case for animal–human translational models. Psychological Research PRPF, 73(4), 463-476.
Appendix A
Appendix B
