How are habits formed?
In 2010 Dr Lally and colleagues published the first study to investigate how habits form in everyday life. Participants chose a new behaviour that they wanted to make into a habit. These were mostly eating, drinking or exercise behaviours, for example “doing sit-ups” or “eating a piece of fruit”. They also chose a cue; a context in which they would do the behaviour, for example, one participant chose “after having a shower”, and another chose “with breakfast”. The cues had to be situations that happen everyday but only once a day.
Participants (university students) were asked to try to do this behaviour every time they encountered the cue for 12 weeks. They were also asked to log into the study website each day and report if they had done the behaviour the previous day and complete items from a validated questionnaire (the Self Report Habit Index (Verplanken and Orbell, 2003)). This questionnaire is intended to capture how automatically participants are performing the behaviour. Example items are ‘I do automatically’, ‘I do without thinking’ and participants rate their agreement with each item on a 7 point scale. These are summed to create a scale score. If they missed a day they could report on whether they had done the behaviour for up to three previous days but only reported how automatic it felt on the current day.
Each participant’s data was examined individually, and an asymptotic curve fitted to their automaticity scores over time (each day of the study). The figure below shows what an asymptotic curve looks like. Initially each time you do the behaviour in the relevant situation (when you encounter the cue) it increases your automaticity. After further repetitions each time you perform the behaviour the increase in automaticity is less until you reach your asymptote, at which point each repetition does not change the automaticity with which you initiate the behaviour.
The curve did not fit for all participants in the study. Of the 96 participants, 14 dropped-out and the curve was an adequate fit for 39 participants. Among those for whom the model was a poor fit, a good number were less consistent in performing the behaviour during the study. For each participant, for whom the curve was a good fit, the authors calculated the time it took them to reach 95% of their asymptote (it isn’t mathematically possible to calculate time to reach 100% of the asymptote). The median time it took for these participants to reach their asymptote (for their habit to be as strong as it will ever be) was 66 days with a range from 18 to 254 days. It is important to note that although the study was only 84 days long the time to reach asymptote was calculated from a statistical model and could therefore be longer than 84 days.
The authors compared automaticity scores on days before and after a day when participants had missed an opportunity to perform their planned behaviour. Occasions when a participant had performed the behaviour for 3 days in a row and then missed a day were identified and the difference in automaticity scores on the day before and the day after the missed day was calculated. This was compared to cases where there were three days in a row when the behaviour was performed and the change in score between day 1 and 3 was calculated. When there had been a missed opportunity the difference was 0.55 and when there hadn’t the difference was 0.79. This suggests that one missed opportunity does not de-rail the habit-forming process.
This paper highlights people need to persist with new desired behaviours for long enough for habits form.
Lally, P., van Jaarsveld, C.H.M., Potts, H.W. and Wardle, J. (2010). How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology 40, 998–1009
Verplanken, B., & Orbell, S. (2003). Reflections on past behavior: A self-report index of habit strength. Journal of Applied Social Psychology, 33, 1313–1330.