←  All research Advances in Experimental Social Psychology · 2006

Intending to do something is not enough. If-then plans are.

Holding a strong intention does not produce the behaviour. Deciding in advance exactly which situation triggers which action does. Across 94 tests the effect was d = .65 — large, by the standards of this literature.

Design: meta-analysis of experimental evidence on implementation intentions. Included: 94 independent tests. Overall effect: d = .65. Getting started: d = .61. Avoiding derailment: d = .77.
d = .65
Effect of if-then planning on follow-through, across 94 independent tests
SMALLd = 0.2MEDIUMd = 0.5LARGEd = 0.8GETTING STARTED · d = 0.61AVOIDING DERAILMENT · d = 0.77OVERALLd = .6594 independent tests of implementation intentions across goal domains.
Placed against the conventional benchmarks for small, medium and large effects. The gap between intending and pre-committing is one of the larger reliable effects in behavioural research.

The gap this closes

Almost everyone intends the right things. Intentions are a famously poor predictor of what actually happens — the gap between the two is one of the most studied problems in behavioural science.

An implementation intention is a narrower commitment. Not “I intend to train more” but “if it is Tuesday at 7pm, then I go straight from work to the gym.” A specific cue, a specific response, decided before the moment arrives.

“I INTEND TO…”GOAL INTENTIONWEAK · UNRELIABLEBEHAVIOURIF X → THEN YIMPLEMENTATION INTENTIONd = .65BEHAVIOUR
Same goal, different wiring. Naming the cue and the response in advance is what carries the effect.

The evidence

Gollwitzer and Sheeran pooled 94 independent tests and found an overall effect of d = .65. Broken down by the problem being solved: getting started, d = .61; avoiding derailment once underway, d = .77.

That second number is the interesting one. If-then plans help most exactly where willpower is usually invoked — the moment something interrupts you.

94 INDEPENDENT TESTS00.20.40.650.81.0POOLED d = .65
Every dot is an independent test. The pooled effect lands at d = .65 — large for this literature.
WHERE THE EFFECT IS BIGGESTGETTING STARTEDd = 0.61STICKING WITH ITd = 0.65RE-ENTRY AFTER DISRUPTIONd = 0.77
The advantage is largest exactly where systems die: coming back after the day went sideways (d = .77).

The mechanism

Pre-committing appears to hand control of the behaviour to the situation instead of to deliberation. When the cue arrives, the response is already decided, so there is no decision left to lose.

Read alongside the self-control study on this site, the picture is consistent: you win by not having the argument, not by winning it. An if-then plan is the mechanism for not having the argument.

CUETIME · PLACE · TRIGGERRESPONSEDECIDED IN ADVANCEGOThe decision happens once — before the moment. The moment only executes.
The mechanism is delegation: control passes from willpower in the moment to a cue set in advance.
What to build from this

Turning the finding into a mechanic

  • Write plans as cue and response, not as aspirations. If X, then Y.
  • Bind the cue to a time and a place. Vague triggers do not fire.
  • Plan the derailment too — that is where the larger effect was found.
  • Decide once, in advance. Every decision left for the moment is a decision you might lose.
Find your missing mechanism See the systems
IN PRACTICE
In LEVEL UP SYSTEM every quest is an if-then with a consequence attached: a cue, a response, and a damage number. The mechanism from these 94 tests, made playable.
See LEVEL UP SYSTEM →
Gollwitzer PM & Sheeran P (2006). Implementation Intentions and Goal Achievement: A Meta-Analysis of Effects and Processes. Advances in Experimental Social Psychology 38, 69–119.
Read the original paper →