Why Most Self-Improvement Fails, and What the Behaviour-Change Science Says Instead

Follows: How Habits Actually Form · How Practice Actually Works

In this deep dive

1. The gap between wanting and doing
2. If-then plans: the most tested trick in the book
3. How long a habit really takes
4. Change the room, not the person
5. Identity: what the evidence actually supports
6. Relapse is the rule, not the exception
7. What self-help books get wrong

It's the second week of January. The running shoes are by the door, the meal-prep containers are stacked, the app is downloaded. You meant every word of the resolution. By February the shoes have migrated to the back of the wardrobe and the app sends notifications you swipe away without reading.

The usual explanation is that you didn't want it enough. The research says something more interesting and much kinder: wanting it was never the problem. Most failed self-improvement fails after the decision, in a stretch of ordinary life that motivational books barely mention.

How Habits Actually Form in this library laid out the habit loop (cue, craving, routine, reward) and the practical levers: obvious cues, small routines, quick rewards, and an identity that makes the habit feel like you. How Practice Actually Works showed that "10,000 hours" is a misreading of Anders Ericsson's research, and that the kind of practice matters far more than the hours. This piece picks up from both and asks harder questions. How strong is the evidence behind each of those levers? Which ones have survived replication, and which ones were never tested at all? And what should you actually do on the Tuesday afternoon when the plan meets real life?

Part 1 — The gap between wanting and doing

Psychologists have a name for the January-to-February slide: the intention–behaviour gap. It is the distance between what people sincerely intend to do and what they actually do.

You might think the solution is simply to want things more. If you raise someone's motivation, surely their behaviour follows? This was tested properly by Thomas Webb and Paschal Sheeran in 2006. Most earlier research had only shown that people with strong intentions tend to act on them, which can't tell you whether the intention caused the action. So Webb and Sheeran collected 47 experiments in which researchers had deliberately raised people's intentions (with persuasive messages, information and the like) and then measured what people did afterwards.

The result is one of the most useful numbers in behaviour-change science. A medium-to-large change in intention (an effect size of d = 0.66) produced only a small-to-medium change in behaviour (d = 0.36). Pushing motivation up does move behaviour, but only about half as much as you'd hope. Something is leaking between the decision and the deed.

Raising motivation moves behaviour, but only about half as far as it moves the motivation. Most self-help works on the half that leaks least.

Exercise shows the leak most clearly. In a 2013 meta-analysis, Ryan Rhodes and Gert-Jan de Bruijn pooled ten studies covering nearly 3,900 people and sorted them by what they intended and what they then did:

Look at the third row. Almost nobody exercises without intending to; the problem is the people who do intend to. Of everyone who meant to meet the activity guidelines, close to half did not. Across physical-activity research more broadly, intentions account for only around a quarter of the variation in what people actually do.

Where the leak happens

Researchers have spent two decades mapping why good intentions fail. Broadly, the failures fall into a few predictable kinds:

• Failing to get started. You forget, or the moment slips past, or you never decided when.
• Getting derailed. A competing goal, a bad mood, a friend's invitation, a tempting alternative takes over at the critical moment.
• Old habits winning. The automatic routine fires before your intention has a chance to.
• Running out of steam. Effort decays as the novelty wears off.

Notice that none of these is "not wanting it enough". They are problems of timing, attention and context. An intention is a mental state you have in a calm moment. Behaviour happens later, in a different moment, when you are tired, distracted, or surrounded by cues pulling the other way.

A 2016 study by Jennifer Inauen and colleagues made this vivid. Young adults reported their intentions and their snacking five times a day for a week. Whether someone was, on average, a person with strong healthy-eating intentions did not predict their snacking. What predicted it was their intention at that moment. When a person's intention rose one point above their own usual level, they ate roughly half as many unhealthy snacks over the next three hours; when it sagged, the snacks crept back. Intentions aren't a fixed trait you possess. They rise and fall through the day, and behaviour follows the dips.

That is why the rest of this piece is mostly about things other than motivation: plans that fire automatically, environments that make the right action the easy one, and ways of recovering from the slips that will happen. Motivation matters. It's just the part of the system that self-help already over-serves.

Part 2 — If-then plans: the most tested trick in the book

In 1999 the psychologist Peter Gollwitzer published a paper with an unusually confident title: "Implementation intentions: Strong effects of simple plans". The idea is almost embarrassingly simple. Instead of only setting a goal ("I will exercise more"), you link a specific situation to a specific response, in an if-then format:

If it is 7 a.m. on Monday, Wednesday or Friday, then I will put on my shoes and walk to the park.

The theory is that the plan hands control to the situation. You do the deciding once, in advance, and then the cue itself triggers the action, the way a habit would. It attacks the first two leaks from Part 1: failing to start, and getting derailed.

Unlike much pop psychology, this idea has been tested a great deal.

The field evidence is the most convincing because it isn't a lab exercise. In a 2011 study led by Katy Milkman, a large company mailed its employees reminders about free on-site flu vaccination clinics. Some mailings also asked people to write down the date they planned to go; others asked for the date and the time. In the control group, 33.1% got vaccinated. Writing down just a date added 1.5 percentage points, which wasn't statistically reliable. Writing down a date and a time added 4.2 points, which was. The more concrete the plan, the more it worked.

A goal says what you want. An if-then plan says what you'll do when a specific moment arrives. Only the second can be triggered by the world.

Where it doesn't work

Honest accounting matters here, because "strong effects of simple plans" can sound like a promise. It isn't one.

• Effects shrink in the real world. A Dutch trial by Emely de Vet and colleagues asked 709 adults to form if-then plans to add two hours of physical activity a week. Across the whole sample, the plans made no measurable difference to activity or weight. Repeated planning helped only the people who already had strong intentions.
• Plans need a goal behind them. The 2024 review found bigger effects in highly motivated people. An if-then plan is an amplifier for a goal you already hold, not a substitute for one.
• Published effects tend to run high. Like most of psychology, this literature grew up in an era when null results were often left in drawers. Expect the effect in your own life to be smaller than the headline numbers.

Even so, the case is strong enough that it's the first tool to reach for. It costs nothing and takes a minute.

Good plans and weak plans

The strong versions name a cue you will actually encounter and an action small enough to start on the spot. The third one shows that the cue can be internal (a feeling) as well as external (a time or place).

A cousin: mental contrasting

Gabriele Oettingen added a twist. In studies of job-seekers, students with a crush, students facing an exam and patients recovering from hip surgery, Oettingen and Doris Mayer found that positive expectations (judging that success is likely) predicted more effort and better outcomes, but positive fantasies (vividly enjoying images of the success) predicted the reverse. Dreaming about the finish line can feel so good that it quietly substitutes for running the race.

Her response was a technique often taught as WOOP: name the Wish, imagine the best Outcome, then the main inner Obstacle, then make a Plan in if-then form for that obstacle. It combines the motivation of a goal with an honest look at what will get in the way. The evidence is promising but more mixed than for plain if-then plans; some recent independent trials have found no effect. Treat it as a useful way to write a better if-then plan, not as a separate miracle.

Part 3 — How long a habit really takes

How Habits Actually Form already retired the "21 days" myth. It's worth knowing where the number came from, and what replaced it, because the real figures change how you should feel in week three.

The 21 days trace back to Maxwell Maltz, a plastic surgeon whose 1960 book Psycho-Cybernetics observed that patients seemed to take "a minimum of about 21 days" to get used to a changed face or a lost limb. He was describing his own clinical impressions of adjusting to a new self-image, not measuring habits, and he said minimum. Over the decades, retellings dropped the word "minimum" and the caveats.

The study that replaced it was led by Phillippa Lally at University College London, published in the European Journal of Social Psychology (online 2009, in print 2010). Ninety-six volunteers each picked one new eating, drinking or exercise behaviour, such as drinking a bottle of water with lunch or running for 15 minutes before dinner, and did it daily in the same context for 12 weeks. Every day they rated how automatic it felt.

The findings, carefully stated:

• Automaticity rose along a curve: quick gains early, then a slow flattening toward a plateau.
• The median time to reach 95% of that plateau was 66 days.
• The range was enormous: 18 to 254 days. The upper figures are extrapolations, since the study only lasted 84 days.
• Only 39 of the 82 people with usable data fitted the curve well. For many people, the behaviour never settled into a clean habit during the study.
• Missing a single day did not materially affect how the habit formed.

A 2024 systematic review by Balpreet Singh and colleagues pulled together 20 studies with about 2,600 participants. Where studies reported a typical time, medians ran 59 to 66 days, means ran 106 to 154 days, and individual times ranged from 4 to 335 days. In one healthy-eating study, only 23% of participants reached the threshold the researchers had set for a formed habit. In another, a daily stretching habit took an average of 106 days to form when done in the morning and 154 days when done in the evening. The reviewers also warned that most of the included studies had a high risk of bias. The honest summary is "about two months for a simple habit, for some people much longer, and for complex habits we don't really know."

Habits don't switch on at day 21 or day 66. They thicken slowly, like a path across a lawn, and a single day off the path barely shows.

Why this changes week three

The 21-day story sets a trap. If you believe a habit should be automatic after three weeks, then day 22 feels like failure: it still takes effort, so you must be doing it wrong. The real curve says day 22 is supposed to feel effortful. You're on the steep early part of a slow climb.

There's a second lesson in what makes a habit a habit. Wendy Wood and colleagues ran diary studies in which people reported, every hour, what they were doing and thinking. During habitual behaviour, people's minds were usually elsewhere; they didn't need to think about the action to do it, and they reported less stress. A habit is not a strongly held intention. It's an action that has been handed off to context, so that the intention is no longer needed.

The most striking demonstration came from a cinema. David Neal, Wood and colleagues gave moviegoers popcorn that was either fresh or stale. People with a strong habit of eating popcorn at the movies ate about as much of the stale popcorn as the fresh. Their hunger and their liking for the food barely mattered. But the habit only ran in the right context. Outside a cinema, habitual eaters were put off by stale popcorn like anyone else, and when researchers had them eat with their non-dominant hand, the habit lost its grip too.

That cuts both ways. A good habit, once formed, will carry you on days when you don't feel like it. A bad one will run even when you no longer enjoy it, which is why "just stop wanting it" so rarely works. Which leads to the most reliable lever of all.

Part 4 — Change the room, not the person

If habits are triggered by context, then changing the context is one of the most direct ways to change the habit. The habits piece listed the moves: remove the cue, add friction to the bad routine, make the good one easier. This part asks how good the evidence for "environment design" really is, because it's more complicated than the books suggest.

The case for

The theory is solid and well supported. Habits are cue-dependent, as the popcorn study showed. Disrupting the cue disrupts the habit. In a field experiment with 800 households, Bas Verplanken and Deborah Roy found that a campaign promoting greener habits worked better on people who had moved house recently, with the window of opportunity lasting about three months after the move. The old cues had gone, and the new routines weren't set yet. And people who seem to have exceptional self-control often turn out to be people who arrange their lives so that they face fewer temptations in the first place (the next piece, Self-Discipline Without Self-Hate, covers that research).

The complication: the nudge debate

Much of the popular enthusiasm for environment design comes from "nudges", the field of designing choices so that the better option is easier or more obvious. In 2022, a large meta-analysis by Stephanie Mertens and colleagues pooled more than 200 studies and over two million participants and found an average effect of d = 0.45, with food choices especially responsive.

Months later, the same journal published a reanalysis by Maximilian Maier and colleagues with a blunt title: "No evidence for nudging after adjusting for publication bias". Correcting for the tendency of positive results to get published and null ones to vanish, they estimated the average effect at close to zero. Their breakdown was more nuanced than the title, though. Evidence was against nudges that merely give information or reminders. For nudges that change the structure of a choice (defaults, the order and placement of options, portion sizes), the evidence was undecided. And the results varied so much between studies that some nudges probably work well even if the average one doesn't.

There's also a cautionary tale. Many of the most quoted examples of "mindless eating", like soup bowls that secretly refill themselves or the claim that bigger plates make you eat more, came from the Cornell food lab of Brian Wansink. In 2018 Cornell found that he had committed academic misconduct, including misreporting research data and using problematic statistical techniques. He was removed from teaching and research and resigned, and more than a dozen of his papers were retracted. Not every finding he reported is necessarily false, but none of it should be repeated as settled fact.

Temptation bundling, and what replication did to it

One clever idea survived its follow-up, at a smaller size. Katy Milkman, Julia Minson and Kevin Volpp tested temptation bundling: allowing yourself something you crave only while doing something you should. In their study, participants could listen to gripping audiobooks only at the gym. At first those participants went 51% more often than controls (29% more for a group merely encouraged to save audiobooks for the gym). The effect faded over the weeks, especially after the Thanksgiving holiday, but 61% of participants said they would pay to keep the arrangement.

A much larger follow-up with 6,792 people found a real but smaller effect: giving people audiobooks and encouraging them to bundle raised the chance they worked out in a given week by around 10 to 14%, and the effect lasted up to 17 weeks. That's the typical arc of a good behavioural finding: real, modest, and smaller than the first study.

The humbling megastudy

In 2021, Milkman and dozens of colleagues ran what they called a megastudy: 54 different four-week programmes to encourage exercise, designed by 30 scientists from 15 universities, all tested on the same 61,293 members of a US gym chain. Forty-five per cent of the programmes increased weekly gym visits during the four weeks, by 9% to 27%. But only 8% produced an effect that was still measurable after the programme ended. And independent judges, asked to predict which programmes would work best, mostly couldn't.

Even the experts can't reliably predict which behaviour-change trick will work. That's an argument for trying things on yourself and keeping what works.

The best-performing programme, incidentally, gave small rewards for returning to the gym after a missed workout. Hold on to that. It comes back in Part 6.

Part 5 — Identity: what the evidence actually supports

The most inspiring idea in modern habit books is that lasting change starts with who you believe you are. James Clear's Atomic Habits describes three layers of change: outcomes (what you get), processes (what you do), and identity (what you believe about yourself), and calls identity the deepest. His suggested starting question is "Who is the type of person that could get the outcome I want?" The habits piece in this library put it as "I'm trying to run" being fragile and "I'm a runner" being durable.

It's an appealing idea and it fits common experience. So what does the research say?

Where the evidence is solid

Psychologists have measured "self-identity" for decades: how much a person sees a behaviour as part of who they are ("I think of myself as someone who recycles", "being healthy is an important part of me"). A 2010 meta-analysis by Jostein Rise, Paschal Sheeran and Silje Hukkelberg pooled 40 studies of more than 11,000 people. Self-identity was strongly linked to intentions (a correlation of about .47), and it added real predictive power beyond the usual factors like attitudes and social pressure: about 6% more of the variation in intentions, and 9% once past behaviour was also accounted for.

So identity matters. People who see a behaviour as "me" are more likely to intend it, and through those intentions, more likely to do it.

Where the evidence wobbles

The harder claim is that you can switch on a new identity with a change of words. In 2011, Christopher Bryan, Gregory Walton, Todd Rogers and Carol Dweck reported that asking people how important it was to "be a voter" rather than to "vote" increased actual turnout, checked against official records, by over ten percentage points in their experiments. It was a celebrated result: a noun, not a verb, and people's behaviour changed.

In 2016, a team led by Alan Gerber ran a much larger field replication. They found no statistically significant difference between the noun and verb wordings; the estimated difference was approximately zero. Both wordings were less effective than a short, ordinary get-out-the-vote message. The original authors and the replicators disagree about why the results differed, and that debate isn't settled. But it is a warning against the idea that a label, by itself, rewires behaviour.

The pattern is a familiar one. The original paper reported three randomised experiments, two of them in statewide elections; the follow-up was a single, much larger field experiment. A striking effect in smaller studies that shrinks toward nothing in a big one is the usual shape of a replication failure, and it's worth remembering whenever a book promises that one reframe will change everything.

How to use identity honestly

Putting the evidence together suggests a more careful version of the identity idea:

1. Identity follows behaviour at least as much as it leads it. Clear himself frames small wins as the way you prove a new identity to yourself. You don't become a runner by deciding you are one. You become one by running often enough that the label stops feeling like a lie.
2. Use identity as a direction, not a verdict. "What would a healthy person do here?" is a useful question at a decision point. "I am a healthy person" is a claim that the first missed workout can puncture.
3. Watch out for identity as a trap. If a habit becomes too central to who you are, a lapse can feel like a threat to your whole self, which makes it more likely to turn into a full collapse. That's the subject of the next part.

Identity is earned in repetitions, not announced in affirmations. The label works best as a question you ask at the moment of choice.

Part 6 — Relapse is the rule, not the exception

Here is the most important fact in this piece: almost everyone who successfully changes a behaviour slips along the way. Lapses aren't a sign the change failed. They're part of what change looks like.

The best long-term data come, of all things, from New Year's resolutions. In the late 1980s, John Norcross and Dominic Vangarelli followed 200 resolvers for two years:

That second number looks grim, but look at who succeeded. Among the successful group, 53% had slipped at least once, and the average number of slips over the two years was 14. The successful resolvers were not people who never lapsed. They were people who lapsed and kept going. Their slips, the study found, were usually triggered by stress, negative emotions and situations where they felt they'd lost control.

Smoking shows the same pattern at a larger scale. Michael Chaiton and colleagues followed 1,277 smokers in Ontario who tried to quit. Depending on the statistical assumptions, the estimated average number of attempts before quitting for good ranged from about 6 to 142; their preferred, least biased method gave about 30. The authors' conclusion was that for many smokers it may take 30 or more attempts, and that clinicians and smokers should expect this.

The lapse that becomes a relapse

Addiction researcher Alan Marlatt, who developed the relapse-prevention approach to treatment, described a specific trap he called the abstinence violation effect. A single lapse (one cigarette, one drink, one skipped week) triggers guilt and a sense of failure: I've blown it, I'm obviously someone who can't do this. That reaction, not the lapse itself, raises the odds of sliding into full relapse.

Self-help culture accidentally feeds this. Streak counters, "never break the chain" and all-or-nothing challenges make the first miss feel catastrophic. Lally's study offers the correction: one missed day did not materially affect habit formation. What matters is the next day.

That's why the winning programme in the gym megastudy is so telling. Out of 54 programmes, the best one rewarded people specifically for coming back after a missed workout. It treated the return, not the streak, as the thing to reinforce. The habits piece put it as a rule of thumb: never miss twice.

The people who change aren't the ones who never slip. They're the ones who've practised getting back up so often that it's become the habit.

Making the decision still helps

None of this means resolutions are pointless. In a 2002 follow-up, Norcross and colleagues compared 159 people who made New Year's resolutions with 123 who wanted to change the same kinds of problems but hadn't made one. At six months, 46% of the resolvers were continuously succeeding, compared with 4% of the non-resolvers. Confidence, skills and readiness measured before January predicted who would succeed.

A 2020 Swedish experiment with 1,066 people added two practical details. People who framed their goal as something to approach ("eat more vegetables") succeeded more often than those who framed it as something to avoid ("stop eating sweets"): 58.9% versus 47.1%. And the group given a little support (some information and exercises about goal-setting) did better than both the control group and a group given extended support. More help was not better help. (The study relied on people's own reports, and its authors receive royalties from a book partly based on it, which they disclosed.)

What about the "stages of change"?

You may have met the Transtheoretical Model, which says people move through stages: precontemplation, contemplation, preparation, action and maintenance. It's useful vocabulary, and it helped make relapse a normal part of the conversation. But it's contested. In 2005, the addiction researcher Robert West published an editorial in Addiction titled "Time for a change: putting the Transtheoretical (Stages of Change) Model to rest", arguing that the model had outlived its usefulness. The journal ran several replies alongside it, some defending the model and some agreeing with West, and the argument has never fully settled. The fair conclusion is to use the stages as a description if they help, not as a rule that you must be "ready" before you try.

Part 7 — What self-help books get wrong

Self-help isn't useless. Several of the most useful ideas in this piece, like if-then plans and temptation bundling, reached most people through popular books. The problem is a set of habits the genre has: compressing messy findings into rules, preferring the inspiring version to the accurate one, and rarely going back when the evidence changes.

The deeper errors

Beyond individual myths, three patterns explain why so much self-improvement fails.

1. It sells motivation, which is the part that leaks. Most books are designed to make you feel ready to change. Part 1 showed why that's the weakest link: a big rise in motivation buys a modest change in behaviour. Readers finish the book fired up, and the fire meets Tuesday afternoon.

2. It runs on survivors' stories. The author who woke at 5 a.m. and built an empire is real. So, invisibly, are the many who woke at 5 a.m. and got nothing but tired. Single stories can't tell you whether a method works; only comparisons can. When the megastudy compared 54 ideas head to head, most did little that lasted, and experts couldn't predict which would win.

3. It blames character for problems of design. If change is about willpower and wanting it enough, then every lapse is a verdict on you. The research points the other way: the people who change most are often the ones who rely least on willpower. They plan for specific moments, change their surroundings, accept the slow climb, and treat lapses as data.

There's one more, quieter error: one-size-fits-all certainty. Habits took between 18 and 254 days in Lally's study. Most of the megastudy's programmes worked for some people and not others. Any book that promises the method is overselling. The honest version of behaviour-change science is a set of tools with modest, real effects, and the job is to find out which ones work for you.

How to read a self-help book

None of this means you should stop reading them. It means reading them the way you'd read a sales pitch from someone you like: with interest and with questions.

• Does it cite studies, or only stories? Stories show what's possible, not what's likely.
• Does it say how big the effects are? "Works" can mean a tiny average change.
• Does it ever mention a study that failed? A field with no failures is a field nobody checked.
• Does it tell you what to do when the method doesn't work for you? Good advice expects variation.
• Is the book an advert for a course, app or programme? That isn't disqualifying, but it's worth knowing.

The science doesn't hand you a programme. It hands you a set of experiments, and the permission to run them on yourself without treating each failure as a verdict.

The takeaway

Most self-improvement fails in the gap between intending and doing. Raising motivation moves behaviour only about half as far as it moves the motivation, and close to half of people who intend to exercise don't. The tools with the best evidence work on the gap itself: concrete if-then plans that hand control to a cue; patient repetition in a stable context (around two months for a simple habit, often much longer); changing the structure of your environment rather than relying on reminders; using identity as a question rather than a label; and, above all, planning for lapses, which nearly everyone who succeeds goes through. Expect modest effects, test them on yourself, and keep what works.

Try this: pick one change you've tried before. Write one if-then plan with a specific time or place and an action small enough to start on the spot. Then write a second plan for the day you miss: "If I skip a day, then the next day I will do the smallest possible version." The second plan is the one most people never write.

Sources & Further Reading

Sources
- Webb, T.L. & Sheeran, P. (2006). Does changing behavioral intentions engender behavior change? Psychological Bulletin, 132(2), 249–268.
- Rhodes, R.E. & de Bruijn, G.-J. (2013). How big is the physical activity intention–behaviour gap? British Journal of Health Psychology, 18(2), 296–309.
- Inauen, J. et al. (2016). Mind the gap? An intensive longitudinal study of between-person and within-person intention–behavior relations. Annals of Behavioral Medicine.
- Gollwitzer, P.M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54(7). Gollwitzer, P.M. & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis. Advances in Experimental Social Psychology, 38. Sheeran, P., Listrom, O. & Gollwitzer, P.M. (2024). The when and how of planning: meta-analysis of implementation intentions in 642 tests. European Review of Social Psychology.
- Milkman, K.L. et al. (2011). Using implementation intentions prompts to enhance influenza vaccination rates. PNAS, 108(26). De Vet, E. et al. (2009). International Journal of Behavioral Nutrition and Physical Activity, 6:11.
- Oettingen, G. & Mayer, D. (2002). The motivating function of thinking about the future: expectations versus fantasies. Journal of Personality and Social Psychology, 83(5).
- Lally, P. et al. (2010). How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology, 40(6), 998–1009. Singh, B. et al. (2024). Time to form a habit: a systematic review and meta-analysis. Healthcare, 12(23), 2488.
- Wood, W., Quinn, J.M. & Kashy, D.A. (2002). Habits in everyday life. JPSP, 83(6). Neal, D.T., Wood, W. et al. (2011). The pull of the past. Personality and Social Psychology Bulletin, 37(11). Verplanken, B. & Roy, D. (2016). Empowering interventions to promote sustainable lifestyles: testing the habit discontinuity hypothesis. Journal of Environmental Psychology, 45.
- Mertens, S. et al. (2022). The effectiveness of nudging. PNAS, 119(1). Maier, M. et al. (2022). No evidence for nudging after adjusting for publication bias. PNAS, 119(31). Cornell University provost's statement on Brian Wansink, 20 September 2018.
- Milkman, K.L., Minson, J.A. & Volpp, K.G. (2014). Holding the Hunger Games hostage at the gym. Management Science. Kirgios, E.L. et al. (2020). Teaching temptation bundling to boost exercise. Organizational Behavior and Human Decision Processes. Milkman, K.L. et al. (2021). Megastudies improve the impact of applied behavioural science. Nature, 600, 478–483.
- Rise, J., Sheeran, P. & Hukkelberg, S. (2010). The role of self-identity in the theory of planned behavior: a meta-analysis. Journal of Applied Social Psychology, 40(5). Bryan, C.J. et al. (2011). Motivating voter turnout by invoking the self. PNAS. Gerber, A.S. et al. (2016). A field experiment shows that subtle linguistic cues might not affect voter behavior. PNAS.
- Norcross, J.C. & Vangarelli, D.J. (1988–89). The resolution solution. Journal of Substance Abuse, 1(2). Norcross, J.C. et al. (2002). Auld lang syne. Journal of Clinical Psychology, 58(4). Oscarsson, M. et al. (2020). A large-scale experiment on New Year's resolutions. PLOS ONE, 15(12).
- Chaiton, M. et al. (2016). Estimating the number of quit attempts it takes to quit smoking successfully. BMJ Open, 6(6). Larimer, M.E., Palmer, R.S. & Marlatt, G.A. (1999). Relapse prevention: an overview of Marlatt's cognitive-behavioral model. Alcohol Research & Health, 23(2). West, R. (2005). Time for a change. Addiction, 100(8).
- Carney, D. (2016). "My position on 'Power Poses'", statement posted on her Berkeley faculty page. Cuddy, A.J.C., Schultz, S.J. & Fosse, N.E. (2018). Psychological Science, 29(4).
- Maltz's 21-day remark (Psycho-Cybernetics, 1960) as quoted in James Clear, "How long does it actually take to form a new habit?"

Go deeper
- 📘 How to Change — Katy Milkman (2021). The best single book on this research, by the scientist behind the flu-planning study, temptation bundling and the megastudy.
- 📘 Good Habits, Bad Habits — Wendy Wood (2019). The habit researcher's own account of why context beats willpower.
- 📘 Rethinking Positive Thinking — Gabriele Oettingen (2014). Why dreaming about success can backfire, and the planning method she built from that research.
- 📘 Atomic Habits — James Clear. The most practical popular guide and the source of identity-based habits; read it alongside this piece's caveats.
- 🌐 Singh et al. (2024), "Time to Form a Habit" — free full text. Short, readable, and honest about how thin the evidence on habit timing still is.
- 🌐 Maier et al. (2022), "No evidence for nudging after adjusting for publication bias" — free full text. A two-page lesson in how a field's headline numbers can shrink.
- 🌐 See also, in this library: How Habits Actually Form and How Practice Actually Works, and next in this series, Self-Discipline Without Self-Hate.

Sensitive-topic note: this piece mentions quitting smoking and relapse in addiction. Lapses are normal there too, but nicotine, alcohol and drug dependence are medical conditions, and quitting alone is much harder than quitting with help. A doctor, a quitline or a local addiction service can make a real difference; in the US, the SAMHSA National Helpline (1-800-662-4357) is free and confidential.