How many new habits at once

January likes a list: run, journal, meditate, cook, read, and stop scrolling. That list is a wish to be several people by February. Attention cannot staff several new pairings at once. Wendy Wood and David Neal's habits are context-cued responses. Each new response needs a context that is not already fully occupied, and a stretch of repetitions. Phillippa Lally and colleagues (2010, European Journal of Social Psychology) watched one behaviour at a time. The median to an automaticity plateau was about 66 days, range 18 to 254. Five behaviours is five curves. People pretend it is one montage.

The overhaul fantasy

Overhaul is identity theatre. It is also a reasonable reaction to feeling lost. The repair is still one hinge, not a committee of selves. Thomas Webb and Paschal Sheeran's intention-behaviour gap predicts the committee: many intentions, little behaviour. Charles Duhigg popularised a cue-routine-reward loop that is easy to photocopy. Photocopying the loop five times is not five habits. It is five cartoons.

Ambition is not the scarce resource

Ambition is abundant in week one. Prompts, ability, and a stable if are scarce. BJ Fogg's Tiny Habits method is almost aggressive about this: one tiny behaviour, attached to something you already do, celebrated. You can have more later. Later is allowed. Later is how you still have a practice in March.

Keystone thinking makes the list worse

If you believe one behaviour will cascade, you pick a dramatic one and three satellites. Keystones, as Duhigg used the word, are a publishing device more than a personal protocol. A dramatic first behaviour has a longer Lally curve and a higher shame cost. Shame then takes down the satellites. Pick the boring one that will actually repeat.

Attention, not character

Each new if-then occupies working memory until the context takes over. Peter Gollwitzer's plans work because they are specific. Specificity does not scale to a manifesto. Two implementation intentions can be held. Twelve become wallpaper. Wood's context cannot be redesigned in five rooms on the same Sunday without turning the house into a project that dies on Monday.

Shared cues get congested

If everything is stacked on coffee, coffee becomes a ceremony and then a chore. One extra then after coffee is a stack. Five is a collapse, as the stacking post argued. Use different spines: coffee, commute, teeth, shutting the laptop. Even then, introduce them in series, not in parallel.

Tracking five things is a sixth thing

Self-monitoring changes behaviour until the log is heavier than the life. Five yes/no ticks can be fine. Five quantity targets with notes is a second job. Webb and Sheeran would call the second job another intention.

A queue, not a stack

Put the wanted behaviours in a queue. Staff the first until the cue retrieves it on a bad day, not until it is photogenic. Then admit the second. A bad-day test is more honest than a day-count. Lally's 21-day myth cousin, the "give each habit three weeks then add," is still a calendar pretending to be a mechanism. The mechanism is automaticity in context. If the first behaviour still needs a speech, you do not have spare attention for a second speech.

Count only the behaviours that still need a speech

Brushing your teeth is not a new habit. Neither is locking the door. People pad January lists with already-automatic actions so the list looks like a life. That padding hides the real load: the items that still require Gollwitzer's sentence every day. Count those. One is plenty. Two is a stretch if they live in different rooms. Five is a committee.

Wood's context design also takes attention. Rearranging one kitchen is work. Rearranging a kitchen, a desk, and a bedtime in the same weekend is a renovation, and renovations slip. Do the furniture for habit one, run it until the cue is boring, then touch the next room. Duhigg's loop photocopies easily on paper. Rooms do not.

Policy	What it assumes	What usually happens
Five new habits on 1 January	Attention is free; identity can be bulk-edited	By week two, none of them have a stable if
One tiny habit, then a queue	Attention is the bottleneck	March still has a practice
One dramatic keystone plus satellites	Cascade will pay the bill	The dramatic one dies; satellites follow
Many if-thens written in one sitting	Specificity scales	Wallpaper

Exceptions exist. If the extras are almost zero-ability (put a glass out, hang the keys) and they share a cue you already never miss, you can sneak them in. They are not a personality overhaul. They are Fogg-sized beads. Do not use the exception as a loophole for a marathon, a novel, and a new language in the same fortnight.

Willpower cannot staff the parallel project. Automaticity might staff it later, when several cues are already doing the retrieving. You do not get to start at later. You start at one sentence.

A useful review, once a month, is: which behaviour still needs a speech, which is now mostly the room, and which was never a behaviour, only a slogan? Queue the first. Protect the second. Delete the third. Lally's 66-day median is not a permission slip to add a new item every weekend. It is a warning that the first item may still be in the messy middle while you are already shopping for a fifth.

Habit AI's one-sentence parse, plus a list you can keep short, is a queue in software: add the next habit when the first one is boring, not when a New Year page is still blank.

Read next

Habit discontinuity: moving house, new job, new baby is when you should shrink the list, not expand it. Habit stacking that does not collapse is what happens if you hang the whole queue on one kettle. the iPhone tracker is built for a short list.