Almost everybody trying to change something starts with motivation and finds it insufficient. The research suggests this is the expected result rather than a personal failing.
What a habit actually is
In behavioural psychology, a habit is a learned association between a context and a response, acquired through repetition, that operates with minimal deliberation.
Which means the defining feature is automaticity, not frequency.
Something done daily with effort and decision each time is not a habit. Something done automatically in response to a cue is.
The context dependency
The most practically useful finding.
Habits are triggered by context — location, time, preceding action, emotional state, presence of particular people.
Which means the same person behaves differently in different settings, and behaviour that seems like a stable trait is frequently a context effect.
Studies of people moving house or changing jobs find that established habits are disrupted, and that this window is when behaviour change succeeds most readily.
Which suggests that changing the environment is more effective than trying harder in the same one.
Why motivation fails
Motivation fluctuates with mood, sleep, stress and countless other things.
A system depending on it works on good days and fails on bad ones, which are exactly the days when the behaviour matters most.
Habits work because they do not require motivation once established, which is the entire point of building them.
The difficulty is that establishing them requires motivation, during a period when the behaviour is still effortful.
The timeline
The widely repeated figure of twenty-one days has no research basis and comes from a surgeon's observation about patients adjusting to changed appearance.
The actual research found a median of around sixty-six days to reach maximum automaticity, with an enormous range across individuals and behaviours.
Simple behaviours automated faster than complex ones, and the variation between people was larger than any average conveys.
Which means the useful takeaway is that it takes considerably longer than people expect, and that a missed day does not reset anything.
Implementation intentions
The intervention with the most consistent evidence.
Specifying when, where and how a behaviour will occur — in a concrete if-then form — substantially improves follow-through compared to a general intention.
The proposed mechanism is that it creates the cue-response link directly rather than relying on remembering to decide.
The effect has been replicated across many domains and it is close to free to apply.
Friction
Small changes in effort have effects out of proportion to their size.
Removing a few steps from a desired behaviour, or adding a few to an unwanted one, changes rates measurably.
Which is why the practical advice — laying things out the night before, removing an application, keeping something out of the house — works better than it should.
The effect is not about the effort itself. It is that friction interrupts automaticity and forces a decision, and decisions frequently go the other way.
Identity
Some research suggests behaviours framed as part of who someone is are maintained better than those framed as things they do.
The evidence is more mixed than popular accounts suggest, and it is plausible enough to be worth trying since it costs nothing.
What to do with all this
Pick one thing rather than several, since the effortful establishment period consumes limited capacity.
Attach it to an existing routine so the cue already exists.
Make it small enough to do on a bad day, since consistency matters more than magnitude during establishment.
Change the environment rather than relying on resolve.
And expect it to take months rather than weeks, which prevents the abandonment that occurs when three weeks pass and it still feels difficult.
Breaking a habit
Structurally harder than building one, for a specific reason.
The learned association does not disappear. Research on extinction suggests the original association remains and is overlaid by new learning.
Which is why old behaviours return under stress, in the original context, or after a period away — the new learning is context-dependent and the old association resurfaces.
Substitution generally works better than suppression, since the cue still occurs and something will follow it.
Tracking
Monitoring a behaviour changes it, which is well established across domains.
The effect appears to come from the attention itself rather than from the record, and it diminishes when tracking stops.
Which makes it useful during establishment and unnecessary afterwards, since a habit that requires tracking has not become automatic.
Streaks
Motivating and fragile.
The evidence on missing a day is reassuring — a single lapse does not measurably affect eventual automaticity in the studies that examined it.
What does damage progress is abandonment following a lapse, which is a response to the streak framing rather than to the lapse.
Rewards
The evidence here is more complicated than habit advice suggests.
External rewards can undermine intrinsic motivation for behaviours people already enjoy, which is a well documented effect.
For behaviours nobody enjoys initially, rewards during establishment appear more useful, and the risk is dependence on them.
Which suggests using them briefly and removing them, rather than building a permanent system around them.