What The Newest Research Says About Forming Habits

Before the studies, the practical version. Each point links to the finding behind it below.

• Expect weeks for something simple and months for something effortful. A glass of water after brushing your teeth is not a gym routine, and the research no longer pretends they run on the same clock.
• Keep a steady cue, but don't make the clock a rule. A consistent time or routine went with doing more, yet paying people only for workouts in a fixed window backfired, and flexible timing went with longer use in one large app dataset.
• Make the start automatic. In a year-long trial, building a habit around getting ready for exercise worked better than building one around the workout itself.
• Choose a version you enjoy. Enjoyment predicted habit strength over weeks; using exercise to relieve stress did not.
• Don't trust your own sense of how habitual something is. Your records are a better guide than your feeling, which is one argument for writing them down.

In HabitKit, a weekly goal is one way to keep the cue without the rigid clock: three distinct days a week, whichever days suit you (how streaks are counted). The grid then shows the pattern you actually kept, which the research suggests is more reliable than how automatic it feels.

The demo's weekly goal requires three distinct days meeting the daily goal, leaving room for days off.

One of the largest studies of habit formation to date did not ask anyone how automatic a behaviour felt. Anastasia Buyalskaya, Colin Camerer, Katy Milkman and colleagues used machine learning on behaviour itself: more than 12 million check-ins from members of a US gym chain, and more than 40 million handwashing opportunities logged by sensors in 30 hospitals. They treated a habit as behaviour that becomes predictable from context, and measured how long that predictability took to level off.

For handwashing, it took about nine to ten shifts, a couple of weeks. For the gym, the published paper first said four to seven months, and press coverage rounded that to about six. A correction in August 2023 replaced it with 68 to 78 days, about two to three months. The six-month figure is still widely repeated; it's wrong.

The gym data also showed what made attendance predictable. The time since the last visit mattered most, lowering the odds of going for 76% of members. Going on the same weekday as recent visits raised them for 69%. For hospital staff, washing on the previous shift predicted washing on the next for everyone.

The spread runs the other way too. In a 2025 study, 194 people tracked a habit they were trying to break, such as snacking or smoking, every day for 91 days. How long it took for the habit's decline to settle ranged from 1 to 65 days, and it didn't depend on which behaviour it was. Our essay on how long it really takes to form a habit covers the older estimates this sits beside.

"Do it at the same time every day" is the most common habit advice there is. Recent work supports the spirit and questions the letter.

A 2026 meta-analysis of 14 studies and 6,069 people found that doing physical activity in a consistent context went with stronger habits, but modestly: a correlation of about 0.20 with habit strength and 0.11 with how much people actually did. Of all the kinds of consistency studied, only a consistent time or a consistent routine went with doing more.

Then the complications. In a field experiment, John Beshears, Katy Milkman and colleagues paid employees to exercise. One group was paid for visits at any time; another was paid only for visits inside a two-hour window they chose themselves. The routine group went less, both during the incentive and after it ended, even among people who had exercised at similar rates while being paid. In data from 14,879 users of a meditation app over their first 180 days, morning sessions went with longer use, but so did less consistent timing in the long run.

And timing seems to interact with who you are. In a preregistered trial with 317 young women, a simple habit, taking a calcium pill, worked best with a morning cue for morning types. Exercise worked best when the planned time matched the person's chronotype: mornings for larks, evenings for owls.

The takeaway isn't that consistency doesn't matter. It's that a rigid clock rule isn't the point: a cue like "after I drop the kids at school" keeps working on a morning that starts late, while a habit that only counts at 8:15 sharp turns one late start into a miss.

The most interesting trial of the last two years tested what part of a behaviour should become the habit. Phillips and colleagues followed 751 new users of an online exercise platform for a year, in a full-factorial design: three strategies, tested alone and in every combination. One built motivation and identity ("I'm someone who exercises"). One built a habit around preparing to exercise. One built a habit around starting the session itself.

Only the preparation habit made a difference. It increased how many classes people took, and it strengthened both the preparation habit and the habit of starting. The identity strategy had no significant effect, and neither did the starting-the-session strategy on its own.

That fits where the theory has landed. A 2024 review by Benjamin Gardner, Amanda Rebar, Sanne de Wit and Phillippa Lally separates the habit, the automatic link between a cue and an action, from the behaviour that follows, and argues that habit is only one influence on what you do at any moment. In their words, "habit formation may be neither necessary nor sufficient to sustain real-world behaviour change". A 2023 meta-analysis of 267 studies found habits predict simple behaviours better than complex ones.

In practice: aim the habit at the first step. Shoes by the door, the yoga mat unrolled, the book on the pillow. Let the habit open the door and expect to walk through it on purpose.

What kind of reward builds a habit? A study during the first UK COVID lockdown followed 580 adults weekly for eight weeks. People who exercised because they enjoyed it built stronger habits over time. People who exercised mainly to relieve stress did not, and that held even for the anxious participants.

A preregistered diary study of 189 people trying to fill half their dinner plate with vegetables found the same pattern between people: those who did it more, liked the taste more and ate in more stable settings had stronger habits. But a good day didn't make the next day's habit stronger, a reminder that habits build slowly, not from one great evening.

Identity is trickier. Across 19 studies and more than 13,000 people, how strongly someone sees themselves as, say, a runner correlates strongly with how habitual their running is. But that's a correlation, and in the trial above, a strategy aimed at identity didn't change behaviour. Seeing yourself as a runner may be what a running habit feels like from the inside, rather than the way in.

Here is the strange part. Animal research has long produced habits in the lab: an overtrained rat keeps pressing a lever for a reward it no longer wants. Humans are a different story.

In 2018, Sanne de Wit and colleagues reported five attempts to create a habit in the lab by overtraining people, two of them replications of the best-known positive study. None worked. A 2023 registered report with 123 people, using that original study's own task with its author on the team, also couldn't show habits growing with training. A 2024 registered report with 220 people found that repeating a choice did make people choose it more often, regardless of reward, but the effect was small, and none of the lab measures lined up with people's own reports of their real-life habits.

People aren't good judges either. Across two studies, Asaf Mazar and Wendy Wood found that habit strength predicted behaviour as well as or better than mood did, yet people explained their behaviour through inner states like mood, even when they were paid to be accurate. In 2025, a study of 6,816 people found that the standard habit questionnaire correlated only about 0.3 with their actual tracked app use; their past usage predicted future use far better.

One more nuance for the "habits are dopamine" story you'll see online. A 2025 study in mice found a dopamine signal that reinforces repeating an action without any reward value attached, separate from the familiar reward signal. Interesting, early, and in mice; it says nothing yet about dopamine detoxes or habit apps for people.

The upshot: your feeling of how automatic something is isn't reliable, and neither is a questionnaire about it. What you actually did is the better record.

The newest trials on breaking habits are sobering in a useful way. In a preregistered randomised trial published in 2026, 313 people trying to snack less were assigned to substitute another behaviour, to inhibit the urge, to make snacks harder to reach, or to a control group, with or without rewards, and rated their habit daily for 13 weeks. Every group's habit weakened, including the control group's. The strategies made it fall faster in the first week, but by the end the difference from the control group was not significant, no strategy beat another, and rewards added nothing.

A 2025 online trial of five strategies for changing a wake-up habit, among adults with sleep problems, found the same shape: almost every group improved, including the comparison group that only got information, with few differences between them. For nail biting, skin picking and hair pulling, a self-help technique called decoupling, which replaces the habitual movement with a different one, held its gains over two years, though only about half of participants were still in the study by then.

Habits also protect behaviour through disruptions. Across a Christmas break, use of a reusable food container system by 17,284 people dropped 16.7% on average, and dropped less for people whose use had been consistent in time and place beforehand.

Put together: the control groups improving suggests that deciding to change and rating the habit every day did much of the work, the first week moves fastest, and a strong habit is what carries you through a holiday. If you do break the chain, our guide to getting back on track after a broken streak covers what the research says about restarting.

• Most of it measures habit by self-report or by predictability, not automaticity directly. The studies that tried to measure habit objectively in the lab mostly couldn't.
• The samples lean young, female and online. Several of the trials above were 77 to 91% women.
• Short follow-ups dominate. A 2023 meta-analysis of habit interventions for physical activity found a small effect, larger at 12 weeks or less than over longer follow-ups.
• Popular rules remain untested. We found no published trial of the "two-minute rule", and the one preregistered test of "habit stacking" we found is a protocol without results yet. They may be good heuristics; they are not findings.

• Buyalskaya, A., Ho, H., Milkman, K. L., Li, X., Duckworth, A. L., & Camerer, C. (2023). What can machine learning teach us about habit formation? Evidence from exercise and hygiene. Proceedings of the National Academy of Sciences, 120(17), e2216115120. 10.1073/pnas.2216115120
• Correction to Buyalskaya et al. (2023). Proceedings of the National Academy of Sciences, 120(34), e2312763120. 10.1073/pnas.2312763120
• Edgren, R., Baretta, D., & Inauen, J. (2025). The temporal trajectories of habit decay in daily life: An intensive longitudinal study on four health-risk behaviors. Applied Psychology: Health and Well-Being, 17(1), e12612. 10.1111/aphw.12612
• Zhu, L., Tao, Y., Zhou, B., Ma, H., Kaushal, N., et al. (2026). Not all cues are equal: A systematic review and three-level meta-analysis of context consistency, physical activity and habit strength. Applied Psychology: Health and Well-Being, 18(4), e70204. 10.1111/aphw.70204
• Beshears, J., Lee, H. N., Milkman, K. L., Mislavsky, R., & Wisdom, J. (2021). Creating exercise habits using incentives: The trade-off between flexibility and routinization. Management Science, 67(7), 4139-4171. 10.1287/mnsc.2020.3706
• Berardi, V., Fowers, R., Rubin, G., & Stecher, C. (2023). Time of day preferences and daily temporal consistency for predicting the sustained use of a commercial meditation app. Journal of Medical Internet Research, 25, e42482. 10.2196/42482
• Phillips, L. A., More, K. R., Russell, D., & Kim, H. S. (2024). Evaluating the impact of individuals' morningness-eveningness on the effectiveness of a habit-formation intervention for a simple and a complex behavior. Journal of Behavioral Medicine, 47(5), 804-818. 10.1007/s10865-024-00503-4
• Phillips, L. A., More, K. R., Lamoureux, N. R., Dixon, P. M., Meyer, J. D., Ellingson, L. D., Welk, G. J., & Hastings, B. (2025). A full-factorial test of motivational and volitional intervention strategies for promoting exercise habit formation and exercise maintenance among new users of an online exercise class platform. Applied Psychology: Health and Well-Being, 17(1), e12597. 10.1111/aphw.12597
• Gardner, B., Rebar, A. L., de Wit, S., & Lally, P. (2024). What is habit and how can it be used to change real-world behaviour? Narrowing the theory-reality gap. Social and Personality Psychology Compass, 18(6), e12975. 10.1111/spc3.12975
• Hagger, M. S., Hamilton, K., Phipps, D. J., Protogerou, C., Zhang, C.-Q., Girelli, L., et al. (2023). Effects of habit and intention on behavior: Meta-analysis and test of key moderators. Motivation Science, 9(2), 73-94. 10.1037/mot0000294
• Fremling, L., Phillips, L. A., Bottoms, L., Desai, T., & Newby, K. (2025). Comparing positive versus negative intrinsic rewards for predicting physical activity habit strength and frequency during a period of high stress. Applied Psychology: Health and Well-Being, 17(1), e12650. 10.1111/aphw.12650
• Kilb, M., & Labudek, S. (2022). Effects of behavioral performance, intrinsic reward value, and context stability on the formation of a higher-order nutrition habit: An intensive longitudinal diary study. International Journal of Behavioral Nutrition and Physical Activity, 19, 105. 10.1186/s12966-022-01343-8
• Zhu, L., Tao, Y., Guo, Y., et al. (2025). The relationship between habit and identity in health behaviors: A systematic review and three-level meta-analysis. Applied Psychology: Health and Well-Being, 17(2), e70017. 10.1111/aphw.70017
• de Wit, S., Kindt, M., Knot, S. L., Verhoeven, A. A. C., Robbins, T. W., Gasull-Camos, J., Evans, M., Mirza, H., & Gillan, C. M. (2018). Shifting the balance between goals and habits: Five failures in experimental habit induction. Journal of Experimental Psychology: General, 147(7), 1043-1065. 10.1037/xge0000402
• Gera, R., Bar Or, M., Tavor, I., Roll, D., Cockburn, J., Barak, S., Tricomi, E., O'Doherty, J. P., et al. (2023). Characterizing habit learning in the human brain at the individual and group levels: A multi-modal MRI study. NeuroImage, 272, 120002. 10.1016/j.neuroimage.2023.120002
• Nebe, S., Kretzschmar, A., Brandt, M. C., & Tobler, P. N. (2024). Characterizing human habits in the lab. Collabra: Psychology, 10(1), 92949. 10.1525/collabra.92949
• Mazar, A., & Wood, W. (2022). Illusory feelings, elusive habits: People overlook habits in explanations of behavior. Psychological Science, 33(4), 563-578. 10.1177/09567976211045345
• Sharpe, M., Bowen, M., & Lambiotte, R. (2025). Quantifying digital habits. EPJ Data Science, 14, 72. 10.1140/epjds/s13688-025-00581-7
• Greenstreet, F., Vergara, H. M., Johansson, Y., Pati, S., Schwarz, L., Lenzi, S. C., et al. (2025). Dopaminergic action prediction errors serve as a value-free teaching signal. Nature, 643, 1333-1342. 10.1038/s41586-025-09008-9
• Edgren, R., Baretta, D., & Inauen, J. (2026). Habit degradation strategies promote faster early reductions in unhealthy snacking habit strength in intensive longitudinal randomised controlled trial. Communications Psychology, 4, 67. 10.1038/s44271-026-00432-9
• Sarfan, L. D., Milner, A. E., Tiab, S., Tuli, D., & Harvey, A. G. (2025). Let's kick that habit: An experiment of five habit-change strategies on habits and symptoms among adults with sleep problems. Journal of Behavior Therapy and Experimental Psychiatry, 89, 102049. 10.1016/j.jbtep.2025.102049
• Moritz, S., Hoyer, L., & Schmotz, S. (2024). Two-year follow-up of habit reversal training and decoupling in a sample with body-focused repetitive behaviors. Cognitive Therapy and Research, 48(1), 75-81. 10.1007/s10608-023-10434-0
• Ratay, C., Webb, T. L., Wood, W., & Mohnen, A. (2024). Does a holiday break disrupt pro-environmental behaviors? Using field data to test the durability of pro-environmental behaviors and the moderating effect of habit. Resources, Conservation and Recycling, 203, 107440. 10.1016/j.resconrec.2024.107440
• Ma, H., Wang, A., Pei, R., & Piao, M. (2023). Effects of habit formation interventions on physical activity habit strength: Meta-analysis and meta-regression. International Journal of Behavioral Nutrition and Physical Activity, 20, 109. 10.1186/s12966-023-01493-3