The Role of Omissions in Habit Formation Nora Delvendahl
Student number: 11390433 Supervisor: Dr Aukje Verhoeven
Abstract
We investigated when and how omissions influence habit formation in the context of medication adherence. Participants were instructed to take a pill every morning or evening depending on the condition for three weeks (habit formation phase). Then, participants had to change their behaviour by taking the pill at a different time of day for one week (switch phase). Habit formation was tracked using a medication adherence device that recorded each lid opening and a daily automaticity measure. Automaticity significantly decreased after a single omission. Furthermore, neither early nor late omissions predicted habit formation and omissions overall did not predict participants’ ability to change their behaviour during the switch phase. Unexpectedly, automaticity scores followed a u-shaped curve, but large variations within the data were present. Our findings question the shape of the automaticity curve and raise questions regarding the role of omissions in habit formation, highlighting the need for further research.
Only 50% of prescribed medication doses are actually taken by patients (Nieuwlaat et al., 2014). As a result, many patients do not experience the effects of their treatment, which can pose serious health risks (Nieuwlaat et al., 2014). Commonly reported reasons for missed doses are simple forgetfulness and distracting situations (Stawarz, Cox & Blandford, 2014). Forming a habit can protect the behaviour from being forgotten, because habitual behaviours are performed automatically and without conscious cognitive effort (Gardner, 2015). Indeed, linking medication intake to one’s daily routine is associated with missing doses less often (Stawarz et al., 2014). Therefore, understanding how medication taking habits develop is crucial to improving people’s medication adherence.
While the term habit is widely used by researchers and health professionals, there are many different perspectives of habit. James’ (1890) classic view postulates that habits form through continuous training, emphasizing that each lapse severely disrupts the learning process. With regard to medication adherence, this suggests that an adherence habit develops through repeatedly taking the medication. More recent research, however, has found this definition of habit to be outdated (Lally, Van Jaarsveld & Wardle, 2010; de Wit, 2017). “[W]ithin health psychology ‘habit’ is defined as a phenomenon whereby behaviour is prompted automatically by situational cues, as a result of learned cue-behaviour associations” (Gardner, 2015, p. 277). From this perspective, habit formation occurs through situational cues being associated with the behaviour. Brushing one’s teeth in the bathroom, for example, could trigger the habitual behaviour of taking the pill. In the animal learning literature, the definition of habit
emphasises de-sensitivity to devaluation (de Wit, 2017). As such, once a behaviour has
become habitual, it will be performed in response to cues even if the outcome is devalued. For instance, if a person is used to taking a painkiller every morning, he or she may continue to do so even when the pain has subsided.
One model encompassing all mentioned aspects of the term habit is the dual-process model (de Wit & Dickinson, 2009). The model differentiates between goal-directed and habitual
behaviour. The former refers to consciously planned behaviour that is flexible in order to pursue a valued outcome, while the latter is performed automatically in response to cues, irrespective of whether or not the outcome is currently valued. As such, goal-directed behaviour – unlike habitual behaviour – is considered to be inflexible to change. As a behaviour is repeatedly performed, it becomes more habitual and less goal-directed.
Despite much research on habits, only few studies have tracked real-life habit formation (Lally et al., 2010; Fournier at al., 2017). One of these exceptions is a study looking at participants’ habit formation of a desired eating, drinking or activity behaviour (Lally et al., 2010). Participants reported daily whether or not they performed the behaviour as well as their behavioural automaticity. The findings showed that repeated performance in the same context resulted in an asymptotic automaticity curve with larger increases in automaticity in the beginning than later on in the process. Over time, automaticity cannot further increase, therefore reaching a plateau. Recently, Fournier et al. (2017) reported that automaticity may follow an s-shape rather than an asymptote. In their study, automaticity did not increase much during the first repetitions followed by a steep growth in automaticity until it reached a plateau. These findings highlight the need to examine whether automaticity indeed follows an asymptote or whether a different shape may better represent the data.
Interestingly, Lally and colleagues’ (2010) found that a single missed opportunity to perform the behaviour did not affect automaticity. This is in contrast to the notion that omissions in general impair habit formation (Armitage, 2005; James, 1890). It should be noted, however, that performing the behaviour more consistently overall (i.e. having fewer omissions) was linked to a better fit for the asymptotic automaticity curve, indicating that repeated omissions may influence habit formation. In line with this idea, Armitage (2005) found that one week of failing to perform a behaviour negatively affects habit formation. A more recent study found that while the time to form a habit was moderated by performance frequency, up to three omissions a week had no negative influence on the habit formation
process (Kaushal & Rhodes, 2015). Thus, when and how omissions affect habit formation remains unclear.
Seeing as automaticity increases greatly in the beginning of the habit formation process and those increases in automaticity become smaller with repeated performance, the influence of missed opportunities on automaticity may depend on the timing of the omission (Lally et al., 2010). Specifically, omissions early on in the habit formation process may disturb the process more than omissions later on. Therefore, it would be worth examining how the timing of omissions influences habit formation.
The present study looked at habit formation in the context of medication adherence. Participants were instructed to take medication daily. Previous habit-tracking studies have primarily relied on self-report (Lally et al., 2010; Fournier et al., 2017), which may fall victim to response bias or participants not accurately recalling their behaviour. Therefore, we
employed a measure of behaviour along with self-reported automaticity. In order to capture all aspects of habit formation according to the dual-process model (de Wit & Dickinson, 2009), we not only looked at automaticity and behaviour but also at inflexibility to change. Specifically, we evaluated whether this newly formed medication taking habit becomes resistant to change by manipulating the current goal (i.e. changing the pill taking instructions). Including this switch phase as a novel approach enabled us to capture all aspects of habit formation according to the dual-process model (de Wit & Dickinson, 2009).
We aimed to investigate how omissions influence real-life habit formation. Do omissions impair habit formation differently depending on their timing? Firstly, we attempted to replicate Lally and colleagues’ (2010) finding that habits develop following an asymptotic curve and that singular omissions do not affect habit formation. Therefore, we hypothesised that failing to take the pill once will generally have no significant influence on behaviour automaticity (Hypothesis 1). Based on the asymptotic automaticity curve reported in Lally et al. (2010), missed opportunities in the beginning of the habit formation process, where the
increase in automaticity gradient is highest, may impair habit formation more than missed opportunities at lower gradients (i.e. later on). Hence, we expected that early omissions will impair habit formation more than those later on in the process (Hypothesis 2). Lastly, the dual-process model (de Wit & Dickinson, 2009) proposes that stronger habits would lead to more ‘slips of actions’, which refer to habitual behaviour being unintentionally performed despite a changed goal (de Wit, 2017). In line with this idea, we expected that more omissions overall will be linked to fewer slips of action during the switch phase (Hypothesis 3).
Method Participants
This report is part of a larger study which aims to link participants’ brain imaging and habit tracking data. Therefore, participants were recruited only if they had previously participated in the Population Imaging of Psychology (PIoP) 2017 study. An a priori power analysis revealed that for a minimal effect size (f 2=.02; Faul et al., 2013), power = .80 and a = .05, a sample of 395 participants would be required for our analyses1. During the time of writing, the available sample may have been smaller, however, as data collection was still ongoing.
Participants were excluded from our analyses if they had not completed the behavioural measure until June 2017. Furthermore, participants who completed fewer than l2 out of 21 daily questionnaires were also omitted as previous research showed that a frequency of at least four days per week is required to form a habit (Kaushal and Rhodes, 2013).
A total of 67 participants was recruited and reimbursement was either 50 Euros or course credit. Fourteen participants were excluded from the analysis because they failed to complete
1 Minimum sample size based on linear multiple regression with two predictors (R2 increase) as this part of our analysis required the largest sample.
sufficient daily questionnaires and a further 17 participants were omitted due to lacking behavioural data. This resulted in a final sample of 36 participants (14 males). The mean age was 21.71 (SD = 1.46) and 21 participants reported to regularly take medication at baseline.
An exclusion analysis indicated that excluded participants did not differ in age, gender, motivation or baseline medication (all p-values > .17). However, the percentage of
participants in the morning condition was higher for included than for excluded participants, c2(1) = 4.21, p = .04, j = -.27.
Materials2 and Procedure
Baseline meeting. Participants were informed about the importance and harmlessness of vitamin B1. In order to prevent participants from acting in line with our expectations of habit formation, we invented a cover story. The apparent study aim was to investigate the effects of vitamin B1 on concentration. Participants were told that they will receive vitamin B1 or placebo pills.
In the following questionnaire, participants reported their age, medication intake and substance use, and completed a two-item motivation measure (e.g. ‘I have the intention to take the pills in the upcoming period’) rated on a 7-point Likert scale from 1 = completely disagree to 7 = completely agree, and additional measures not discussed here (see ‘Additional measures’ below).
Then, participants obtained their MEMS caps, an adherence measure which records each opening of the lid, along with the instruction to store the device in a specific location (i.e. participants could only take the pill when they were at home) and to take a pill once daily: either within 4 hours after waking or within 4 hours before sleeping, depending on the condition. The time of day condition was matched to participants’ previously reported
medication intake. The timing of the daily questionnaire was set individually to differ from the pill taking time (i.e. participants in the morning condition received the questionnaire in the evening and vice versa).
To maintain the cover story, participants subsequently completed a digit span task as a concentration measure.
Habit formation phase (3 weeks). For three weeks, participants’ adherence was tracked by the MEMS caps. Using the software-program Lotus, participants were sent a daily questionnaire via email and/or text message asking whether or not they took the pill. After a ‘yes’ response, participants were presented with the four-item Self-Reported Behavioural Automaticity Index (SRBAI, Gardner et al., 2012; Cronbach’s a = .81 on day 1), rated on a visual analogue scale (VAS; 0-100), such as ‘When I take the pill, I do it without thinking’.
Whilst ‘no’ response participants were asked why they did not take the pill. The answer options were: ‘I was not at home’, ‘I was at home but I forgot’, or ‘other’.
For plausibility of the cover story, participants also reported how concentrated they felt during the last 24 hours.
Lastly, all participants were asked about their wake and sleep times on the previous day. Second meeting. The second meeting was held via ‘Gruveo’ (https://www.gruveo.com), a web-based video call service. During this meeting, participants completed the motivational measure and additional measures (see ‘Additional measures’ below). Then, participants were instructed to change the time of their pill intake. Participants in the morning condition were instructed to take the pill in the evening and the reverse instructions were given to participants in the evening condition. Half of the participants were asked to form an implementation intention and the other half formed a goal intention. For the purpose of this report, however, we collapsed our analyses across both groups.
Switch phase (1 week). The one-week switch phase measures were the same as described in the habit formation phase apart from the SRBAI items being reversed for the pill that was not being taken anymore (automaticity of not taking the pill).
Final meeting. Participants completed the motivation measure and additional measures (see ‘Additional measures’ below). For the cover story, another digit span task was employed. Finally, participants were informed that they will receive more information upon project completion.
At the end of the study, participants will be sent an email debriefing them about the aim of the study.
Additional measures. The following measures were also employed as part of the project, but will not be discussed in the present report.
Baseline meeting. Personal Need for Structure scale (PNS; Neuberg & Newsom, 1993), Short Mindful Attention Awareness Scale (MAAS; Schroevers, Nykliček & Topman, 2008) and questions regarding participants’ level of stress, quality of sleep and whether they moved recently.
Second meeting. Same measure as the baseline questionnaire, but omitting the PNS, the MAAS, and the question regarding moving, and adding the Self-Report Habit Index (SRHI; Verplanken & Orbell, 2003).
Final meeting. Identical to the previous questionnaire, but also asking questions about stressful events and holidays during the study period.
Data analyses
All analyses described were conducted using IBM SPSS Version 24. Prior to our main analyses, the mean of each participant’s SRBAI item scores for each repetition was calculated to derive a single daily automaticity score. For Hypothesis 1 and 2, the analyses focused on the habit formation phase. For Hypothesis 3, we included the switch phase.
Hypothesis 1. The predictor variable was the adherence behaviour, measured through the opening of the pillbox. Within the adherence behaviour, an omission was defined as a missed opportunity to take the pill. The dependent variable was self-reported automaticity.
Based on Lally et al. (2010), a single omission was defined as a missed opportunity followed by three consistent behaviour performances. A missed opportunity was only said to have occurred when participants did not take the pill, but indicated that they were at home (i.e. they had the opportunity to take the pill).
We performed a Wilcoxon signed-rank test comparing automaticity scores before and after a single omission to test whether they differed.
Hypothesis 2. For each participant, automaticity scores were plotted over time and nonlinear regressions were performed to fit an asymptote to each participant’s data. To test whether the asymptotic model accurately described the data, we compared it to linear, quadratic, s-shaped and exponential functions, employing R2-values.
If the asymptotic model presented the best fit for most participants, we planned to define early and late omissions per participant as having occurred during the first 25% of automaticity and the last 25% of automaticity before 95% of asymptote is reached. In case the asymptotic curve did not present the best fit for the majority of the sample, early and late omissions would be defined as having occurred during the first and the last week of the habit formation phase, respectively.
We conducted a two-step logistic regression with the dependent variable being whether or not participants formed a habit. In line with Lally et al. (2010), forming a habit was defined as an SRBAI score of at least 50% for three consecutive days. In the first step, we examined whether early omissions predicted whether or not a habit was formed and in the second step, we tested whether late omissions predict habit formation over and above early omissions:
1. ℎ𝑎𝑏𝑖𝑡 = 𝛽) + 𝛽+∙ 𝑒𝑎𝑟𝑙𝑦 𝑜𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑠 + 𝑒5
Next, we conducted a two-step linear regression in which the dependent variable was the number of days it took participants to form a habit. Therefore, only participants that formed a habit according to our definition were included in this part of the analysis. Seeing as habit formation was defined as three consecutive days of an SRBAI score of at least 50%, we counted the number of days until the first of those three days. This was our measure of how long it took to form a habit.
First, we examined whether early omissions predict the number of days it took to form a habit and subsequently whether late omissions predict the number of days until a habit was formed over and above early omissions:
3. ℎ𝑎𝑏𝑖𝑡𝑑𝑎𝑦𝑠 = 𝛽)+ 𝛽+∙ 𝑒𝑎𝑟𝑙𝑦 𝑜𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑠 + 𝑒5
4. ℎ𝑎𝑏𝑖𝑡𝑑𝑎𝑦𝑠 = 𝛽)+ 𝛽+∙ 𝑒𝑎𝑟𝑙𝑦 𝑜𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑠 + 𝛽6∙ 𝑙𝑎𝑡𝑒 𝑜𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑠 + 𝑒5
Hypothesis 3. A slip of action was defined as a recorded pillbox opening at the time of day that was previously correct during the habit formation phase. In case we failed to find slips of action in our behavioural data, we intended to use the automaticity score of not taking the pill during the beginning (i.e. the first three days) of the switch phase as an approximation of a slip of action, because it may be a more sensitive measure of participants’ inflexibility to change (i.e. habit strength).
In a regression model, we examined whether the total number of omissions during the habit formation phase predicted the mean automaticity of not taking the pill in the beginning of the switch phase.
Results
Descriptive statistics and randomisation check. Overall, only four participants did not have any omissions during the habit formation phase and the average number of omissions was 2.47 (SD = 1.72). At baseline, there were no significant differences in age, gender and
motivation between participants in the morning and participants in the evening condition (all p-values > .26).
Hypothesis 1. Based on our definition, we identified 41 single omissions3 across 31 participants. From the day before an omission to the day after that omission, the average drop in automaticity was 8.95 points. Counter to our expectations, we found this decrease in automaticity to be significant (Median difference score = 11), Z = -2.43, p = .02, r = -.029, indicating that a single omission can hinder automaticity (see Figure 1).
Figure 1. Median automaticity scores before and after a single omission. Error bars correspond to +/-
1 standard error. * p < .05.
Hypothesis 2. The results of the curve analysis indicated that for only three participants, the asymptotic model provided the best fit. For the majority of participants (n = 23), the quadratic model fitted best. Figure 2 shows the plotted automaticity data of one participant for whom the quadratic curve best explained the data. The quadratic curve is characterised by
3 A post-hoc power analysis revealed that for the Wilcoxon-signed rank test with a = .05 and
d = .37, we achieved power = .61. 0 5 10 15 20 25 30 35 40 45 Pre-omission Post-omission SR BA I s co re
relatively high automaticity scores in the beginning that decrease over the first few days and then increase until the end of the habit formation phase.
Figure 2. Example of changes in automaticity scores over the 21 days of the habit formation phase and
the modelled quadratic curve.
With regards to early omissions, participants missed an average of 0.72 (SD = 1.08) opportunities during the first week and no participant reported more than four omissions during that time. The average number of late omissions was somewhat smaller with 0.09 (SD = 0.34) with the maximum number of late omissions being 2.
The majority of participants (n = 21) did not form a habit during the habit formation phase and for those that did form a habit (n = 15), the average time it took was 8.20 days (SD = 4.57). We predicted whether or not a habit was formed from the number of early omissions and found that they were not a significant predictor. Moreover, including late omissions in the second step did not significantly improve the model fit (see Table 1). In summary, neither early nor late omissions significantly predicted whether or not a habit was formed.
0 5 10 15 20 25 30 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Au to m at ic ity sc or e Day
With regards to the number of days to form a habit, the results from our regression analysis failed to support our hypothesis by indicating that neither early nor late omissions
significantly predicted the number of days to form a habit (see Table 2).
Table 1
Summary of the logistic regression analysis for variables predicting whether or not a habit was formed.
Variable B (SE) Wald p
R2 Cox-Snell R2 Nagelkerk e ∆R2 Cox-Snell Step 1 Early omissions -0.17 (0.32) 0.27 .60 .01 .01 - Step 2 Early omissions Late omissions -0.38 (0.37) 1.33 (0.10) 1.07 1.77 .30 .18 .07 .09 .06 Table 2
Summary of the regression analysis for variables predicting the number of days to form a habit.
Variable B (SE) t p R R2 ∆R2 Step 1 Early omissions -0.32 (1.01) -0.32 .32 .09 .01 - Step 2 Early omissions Late omissions 0.10 (1.01) -3.06 (2.14) 0.10 -0.40 .92 .18 .39 .15 .15
Hypothesis 3. Only one participant took the pill at the time of day that was previously correct during the habit formation phase (i.e. a slip of action). Therefore, we used the automaticity score of not taking the pill during beginning of the switch phase as an
approximation of a slip of action. Counter to our predictions, omissions did not significantly predict the mean automaticity of not taking the pill during the first three days of the switch phase, b = 2.70, SE = 2.53, t(34) = 1.07, p = .29, and consequently, the model could only account for 3.2% of the variance in mean automaticity, F(1,34) = 1.14, p = .29, R2 = .03.
Discussion
This report investigated whether and how omissions influence habit formation. By tracking habit formation with a behavioural measure in combination with self-reported automaticity and a measure of de-sensitivity for devaluation, we incorporated all aspects of habit formation as proposed by the dual-process model (de Wit & Dickinson, 2009). The results indicated that self-reported automaticity significantly decreases after a single omission, thus failing to support the assumption that single omissions do not impair habit development (Hypothesis 1). This suggests that a single omission might already have a negative effect on habit formation. In contrast to our prediction, that early omissions impair habit formation more than late omissions (Hypothesis 2), neither early nor late omissions significantly predicted habit formation. With regard to slips of action, omissions were not predictive of how automatic it was for participants to not take the pill, thus not supporting Hypothesis 3.
Although Lally et al (2010) reported that single omissions do not influence automaticity, the present report found no evidence for this claim. The significant drop in automaticity immediately after an omission suggests that omissions, in general, might impair habit formation (Armitage, 2005; James, 1890). While it is possible that each lapse severely
disrupts the process, another explanation for our findings could be that the influence of single omissions might depend on the type of behaviour. The behaviours predominantly studied in previous research refer to exercise and diet (Lally et al., 2010; Armitage, 2005; Kaushal & Rhodes, 2015). It may, therefore, be possible that medication adherence requires more
continuous performance to become habitual than physical activity or dietary behaviours. This would indicate that interventions promoting adherence habits should focus on preventing missed opportunities until a habit has formed. Further research could investigate this by
directly comparing how single omissions influence habit formation for different types of behaviours.
We found no support for the idea that the early stages of habit formation are especially susceptible to omissions. One reason for our finding could be that the asymptotic curve reported in Lally and colleagues (2010) only provided the best fit for three participants. We based our hypothesis on the assumption that automaticity follows an asymptote, however, the quadratic shape provided the best fit for most participants’ data. It may, thus, still be possible that there are certain sensitive periods in habit formation during which omissions impact the process especially, but the time frame might depend on the shape of the automaticity curve. For the quadratic shape, the early stages of the habit formation process might not be a sensitive period but rather the time after which automaticity is at its minimum and increases again. Future research could, therefore, examine whether certain time periods, such as when automaticity increase is greatest, are more susceptible to omissions than others.
Our results indicated that the automaticity curve is quadratic meaning that automaticity followed a u-shape over the course of the habit formation phase. While unexpected, this shape could be explained through participants initially mistaking the saliency of the experimenter’s instructions at the beginning with automaticity. This would explain the relatively high scores at the beginning of the habit formation phase. After some time, the instructions may have become less salient and consequently, automaticity scores decreased until the behaviour actually started to become automatic, and the scores increased. It is critical to mention, however, that although the quadratic curve could account for most participants’ data, there was considerable variation between participants. This finding is in line with Lally et al. (2010), who also found many differences between participants with the asymptotic curve only providing a good fit for less than half of the sample. In another habit tracking study,
automaticity predominantly appeared to follow an s-shape (Fournier et al., 2017). Taken together, the variations in findings highlight that the shape of the automaticity curve might not
be as clear as previously assumed and individual differences may play a major role in habit formation.
With regard to inflexibility to change after the habit formation phase, only one slip of action was present in our data. One reason for this could be that a slip of action would have only been recorded if the lid of the pillbox was unscrewed. This leaves the possibility that participants might have touched the pillbox or picked it up, but this behaviour would have not been recorded. Therefore, a more delicate measure, such as a touch-sensitive pill box, may be needed in future studies to detect these subtle slips of action.
Another possibility that only a limited number of slips was observed could be that the habit formation phase was too short. Previous studies reported average time frames ranging from six to over fourteen weeks for successful habit formation (Kaushal & Rhodes, 2013; Lally et al., 2010; Fournier et al., 2017). While it may be plausible to assume that the time it takes to form a habit might depend on the type of behaviour and that less complex behaviours would become habitual faster (Kaushal & Rhodes, 2013), the three weeks in our study might have not been long enough to form strong adherence habits. This would have consequently influenced participants’ behaviour during the switch phase, as slips of actions are only likely for habitual behaviours (de Wit & Dickinson, 2009). Thus, future studies including a longer habit formation phase could investigate whether this longer time frame would lead to stronger adherence habits and consequently, more slips of action.
One of the limitations of our report was the lack of power in our analyses. It should be noted, however, that data collection was still ongoing at the time of writing. Nevertheless, given the small sample size in this report, it may be possible that we were unable to detect small effects in our data (Field, 2009). This is especially relevant for our third hypothesis as participants might not have formed strong habits and any effects of omissions on how automatic it was for participants to not take the pill may have hence been small. It would, thus, be worth re-conducting the analyses once data collection is completed.
Another point of criticism could be that people may not have been consciously aware of how automatic their behaviour was (Hagger, Rebar, Mullan, Lipp & Chatzisarantis, 2015). This could partially explain the quadratic shape in the automaticity data as participants may have mistaken the ease of remembering the instructions for automaticity. While self-report habit measures are widely employed in the field, whether or not they can accurately capture automatic processes has been widely debated (Gardner & Tang, 2014; Hagger et al., 2015; Orbell & Verplanken, 2015). In order to reduce potential problems associated with self-report measures, we included the switch phase as a more objective measure of habit formation. Inflexibility to change once the outcome is devalued is a key component of habitual behaviour and an indicator of habit strength (de Wit & Dickinson, 2009). In addition to this, future studies could combine the SRBAI with a weekly implicit association test that pairs contextual cues with behavioural responses (Hagger et al., 2015) to assess whether participants’ self-reported automaticity is in line with their implicit associations regarding the behaviour.
A final issue may have been that participants were instructed to take the pill within four hours after waking or within four hours before sleeping. Given the irregular sleep and wake rhythm of many of our participants, this given time frame may have been too vague. If a participant woke up at 1 pm, for example, taking the pill at 3.55 pm the same day would still be coded as having taken it in the morning. Habit formation occurs through associating situational cues with a behaviour (Gardner, 2015). Therefore, people with more structured lives may encounter these situational cues more consistently and thus, may find it easier to from a habit. Seeing as our sample primarily consisted of university students, it may be interesting to see whether people with more regular lives, such as full-time employees or pensioners, would form the habit faster than people with less regular lives.
From a methodological perspective, this is the first study to implement a switch phase into the habit tracking paradigm. While many studies have focused on habits in general (de Wit, 2017; Gardner, 2015), habit formation remains an understudied area of research. We strongly
believe that habit tracking studies, especially including more objective measures, are a valuable way of gaining insight into the processes underlying real-life habit formation. Understanding when and how omissions influence habit formation is relevant for developing interventions designed to promote habit formation, which is particularly important for improving medication adherence rates.
References
Armitage, C. J. (2005). Can the theory of planned behavior predict the maintenance of physical activity?. Health psychology, 24(3), 235.
de Wit, S (2017). Control of Behaviour by Competing Learning Systems. In T. Egner (Ed.), The Wiley Handbook of Cognitive Control (pp. 190-206). Wiley-Blackwell
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. Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2013). G*Power Version 3.1.7 [computer
software]. Uiversität Kiel, Germany. Retrieved from http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/download-and-register
Field, A. (2009). Discovering statistics using SPSS. Sage publications.
Fournier, M., d’Arripe-Longueville, F., Rovere, C., Easthope, C. S., Schwabe, L., El Methni, J., & Radel, R. (2017). Effects of Circadian Cortisol on the Development of a Health Habit. Health psychology: official journal of the Division of Health Psychology, American Psychological Association.
Gardner, B. (2015). A review and analysis of the use of ‘habit’ in understanding, predicting and influencing health-related behaviour. Health Psychology Review, 9(3), 277-295. Gardner, B., Abraham, C., Lally, P., & de Bruijn, G. J. (2012). Towards parsimony in habit
measurement: Testing the convergent and predictive validity of an automaticity subscale of the Self-Report Habit Index. International Journal of Behavioral Nutrition and Physical Activity, 9(1), 102.
Gardner, B., & Tang, V. (2014). Reflecting on non-reflective action: An exploratory think-aloud study of self-report habit measures. British Journal of Health Psychology, 19(2), 258-273.
Hagger, M. S., Rebar, A. L., Mullan, B., Lipp, O. V., & Chatzisarantis, N. L. (2015). The subjective experience of habit captured by self-report indexes may lead to inaccuracies in the measurement of habitual action. Health Psychology Review, 9(3), 296-302.
James, W. (1890). The principles of psychology. New York: Dover Publications.
Judah, G., Gardner, B., & Aunger, R. (2013). Forming a flossing habit: an exploratory study of the psychological determinants of habit formation. British journal of health
psychology, 18(2), 338-353.
Kaushal, N., & Rhodes, R. E. (2015). Exercise habit formation in new gym members: a longitudinal study. Journal of behavioral medicine, 38(4), 652-663.
Lally, P., Van Jaarsveld, C. H., Potts, H. W., & Wardle, J. (2010). How are habits formed: Modelling habit formation in the real world. European journal of social psychology, 40(6), 998-1009.
Neuberg, S. L., & Newsom, J. T. (1993). Personal need for structure: Individual differences in the desire for simpler structure. Journal of Personality and Social Psychology, 65(1), 113. Nieuwlaat, R., Wilczynski, N., Navarro, T., Hobson, N., Jeffery, R., Keepanasseril, A., ... &
Sivaramalingam, B. (2014). Interventions for enhancing medication adherence. The Cochrane Library.
Orbell, S., & Verplanken, B. (2015). The strength of habit. Health psychology review, 9(3), 311-317.
Schroevers, M., Nykliček, I., & Topman, R. (2008). Validatie van de nederlandstalige versie van de Mindful Attention Awareness Scale (MAAS). Gedragstherapie.
Stawarz, K., Cox, A. L., & Blandford, A. (2014). Don't forget your pill!: designing effective medication reminder apps that support users' daily routines. In Proceedings of the 32nd annual ACM conference on Human factors in computing systems (pp. 2269-2278). ACM. Verplanken, B., & Orbell, S. (2003). Reflections on Past Behavior: A Self-Report Index of
