The reason substitution beats subtraction

Someone quits the mid-afternoon scroll and simply stops. For a week the phone stays face down. Then a dull Tuesday at 3.40 arrives and the hand moves anyway — not towards the deleted app, but towards a different one that was installed for something else entirely.

What everyone sees

Quitting is pictured as removal, as though the behaviour were an item taken out of a set and the set were now smaller. What is actually removed is one response. The condition that summoned that response — a time of day, a mood, a room, a lull between tasks — is untouched and continues to arrive on its old timetable.

What is actually happening

Bouton’s work on extinction shows that ceasing a learned response does not erase the original association; it adds a competing inhibitory one that is context-dependent, which is why the response returns when the context changes or time passes. Wood and Neal’s account of the habit-goal interface places the cue rather than the intention at the controlling position, so an unaddressed cue remains a live instruction. Substitution works because it assigns the instruction a destination instead of leaving it open, and Verplanken and Wood’s review of habit interventions finds that changes which supply a replacement response in the same context outperform those relying on suppression alone. The effect sizes here are modest and setting-dependent, and the literature does not support any single protocol.

Why it stays hidden

The asymmetry hides because the first week of subtraction succeeds. Suppression is strong while attention is high, which produces early evidence that removal is working. The failure arrives later, in a moment of low attention, and is then attributed to weakness rather than to the structural fact that the cue was never given anything to point at.