Atomic Habits, Rebuilt on the Research

Railway Technical Research Institute, Japan, 1994 — 'pointing and calling' (shisa kanko)

In a controlled task, workers made 2.38 errors per 100 actions with no intervention, dropping to 0.38 per 100 when they physically pointed at the object and said the check aloud — a reduction of roughly 85%. Verified: this is the source of the widely quoted figure, and the figure is right.

What it does and doesn't show

This is a *vigilance and error-checking* result on a simple, repetitive task — not a demonstration that habit-scorecard-style self-monitoring changes long-run behaviour. It is strong evidence for the narrow claim (saying it out loud catches mistakes) and no evidence at all for the broad one.

Gollwitzer & Sheeran, 2006 — meta-analysis of 94 studies

"Implementation intentions" — concrete if-then plans specifying when, where and how — produce a medium-to-large increase in goal attainment (d ≈ 0.65) over goal intentions alone. This is a large, well-replicated literature and it is the engine under this chapter.

Habit stacking ('after [current habit], I will [new habit]')

A sensible extension of the same principle — using an existing routine as the cue rather than a clock time. Clear's specific formulation has not itself been trialled, but it is a faithful application of a well-supported mechanism.

Robins et al. — US servicemen returning from Vietnam

Around 34% of enlisted men used heroin in Vietnam and roughly 20% showed signs of dependence. On returning home — a total change of context — only about 12% relapsed into addiction at any point over the following three years. Roughly nine in ten simply stopped. The variable that changed was not willpower or treatment. It was the environment.

Wood & Neal — context-dependence of habit

Habits are bound to the settings in which they were learned; move the person, and the cue structure that sustained the behaviour disappears with it.

Galla & Duckworth, 2015 (JPSP) — 6 studies, N = 2,274

People high in trait self-control succeed not by out-muscling temptation but because they have more beneficial *habits*: routines that run on autopilot, not moment-to-moment resistance, are what link self-control to good life outcomes.

Hagger et al., 2016 — multi-lab registered replication (Perspectives on Psychological Science)

The "ego depletion" effect — willpower as a limited daily fuel tank that gets used up — was tested across roughly two dozen labs under a pre-registered protocol. The best estimate of the effect came out at approximately zero. The muscle metaphor that dominated a decade of pop psychology did not survive replication.

Asch, 1951 — conformity experiments

Faced with a group unanimously giving an obviously wrong answer about the length of a line, about 75% of participants went along with the group at least once, and conformity ran at roughly a third of critical trials — against an error rate under 1% when people answered alone.

Christakis & Fowler — obesity and behaviour in social networks

⚠️ Real finding, contested interpretation. Behaviours like smoking and obesity do cluster in social networks. But the *causal* reading — that behaviour spreads contagiously through friendships — drew serious statistical criticism (Cohen-Cole & Fletcher; Lyons; Shalizi & Thomas), whose core point is that homophily (people befriending others already like them) can produce the identical pattern with no contagion at all. In fairness: later sensitivity analyses suggest the obesity and smoking results are reasonably robust to that confound, while the happiness and loneliness results are less so.

Johnson & Goldstein, 2003 (Science) — 'Do Defaults Save Lives?'

Countries where organ donation is opt-out show effective consent rates frequently above 90%; comparable countries where it is opt-in cluster far lower, often under a quarter. Same populations, same values — the entire difference is which box is pre-ticked. In the authors' own experiment, switching the default roughly doubled donation rates (about 42% to about 82%).

The general principle

Tiny changes in the effort required to do something produce changes in behaviour far larger than most people's stated intentions or motivation ever do.

Thaler & Benartzi, 2004 (Journal of Political Economy) — 'Save More Tomorrow'

Employees who pre-committed to raise their savings rate out of *future* pay rises lifted their average savings from 3.5% to 13.6% over 40 months, with about 80% of enrolees staying in the plan through their fourth raise. Pre-commitment nearly quadrupled saving where exhortation had failed.

Commitment contracts generally

Binding your future self — automatic transfers, cancelled subscriptions, deleted apps — consistently outperforms relying on your future self's judgement, precisely because it removes the judgement.

Delay discounting (Ainslie and successors)

Humans steeply devalue rewards that arrive later — which is exactly why habits with delayed payoffs (study, exercise, saving) lose to habits with instant ones. Robust and heavily replicated.

Watts, Duncan & Quan, 2018 (Psychological Science) — the marshmallow replication

⚠️ Important correction to a story this chapter's neighbourhood is built on. The famous finding that a preschooler's ability to delay gratification predicts life success largely dissolved on replication: the association was about half the original size, and shrank by roughly two-thirds once family background, early cognitive ability and home environment were controlled. Delay of gratification looks far more like a *proxy for a stable, resourced childhood* than a personal superpower.

Giné, Karlan & Zinman, 2010 (American Economic Journal: Applied Economics) — the CARES smoking study, Philippines

Smokers were offered a savings account they would forfeit to charity if they failed a nicotine test six months later. Only 11% took it up — but those randomly offered it were **3 percentage points** more likely to pass the test, and the effect persisted in surprise tests at 12 months. Note the honest size: a real, durable effect, and a modest one. Commitment devices work; they are not magic.

Public commitment and social accountability

Telling a specific person a specific commitment raises follow-through — the immediate social cost of admitting you skipped substitutes for the missing instant penalty.

Vukasović & Bratko, 2015 — meta-analysis of personality heritability

Around 40% of the person-to-person differences in the Big Five personality traits are linked to genes (roughly: openness .41, extraversion .36, emotional stability .37, agreeableness .35, conscientiousness .31). Substantial — and equally, the majority of the difference is *not* genetic.

Polderman et al., 2015 (Nature Genetics) — fifty years of twin studies

Across human traits generally, the share of differences linked to genes averages around 49%. Genes are a real constraint and never a complete explanation.

Ericsson — deliberate practice

Once a skill becomes automatic, performance plateaus. Continued improvement requires effortful, feedback-rich practice deliberately aimed at the edge of your competence — the opposite of the comfortable, unconscious repetition habits produce.

The reflection point

Habits deliver consistency, not excellence. Without periodic review, an automated routine will happily carry you in a direction that stopped being the right one years ago.