How fast infrequent gestures are forgotten
Aliases: gesture forgetting · skill decay · command fade
What it is
Rarely used gestures bleach out with delay. An Ebbinghaus-like curve is blunt for command strokes: after someone learns “two-finger pull opens the shortcut shade,” weeks without use leave a blank, not a motion. Infrequency is not the odd miss; it is retrieval failure—people hunt for a button, or conclude the function was removed.
Why it happens
Procedural memory consolidates through repetition. Taps and scrolls happen daily and become automatic; a three-finger screenshot or a canvas rotate used monthly never consolidates, so the trace decays with the gap. Decay is uneven: discrete features such as direction and finger count go first, continuous following later. Interference from similar strokes speeds the fade—“two-finger pull” next to “one-finger pull to refresh” is easy to cross after a pause. Recognition scores on a help page overestimate in-place retrieval.
Studying it
After training, retest on a log spacing (1, 7, 30 days) with the task embedded naturally, not preceded by a lesson. Measures include total misses, partial hits (right direction, wrong count), and switches to a visible entry. The real gap between two successful uses in the wild is closer to product life than a lab spacing. Opening a gesture cheat-sheet once is not rehearsal.
Where it stops holding
System gestures met daily across apps (back, home) barely show this curve. User-defined shortcuts that people actually reuse are a different case. Illness and aging can speed sequence loss, but not every older-adult difficulty is forgetting—duration and target size are separate. Forgetting speed describes failed retrieval, not never having learned.
Applying it
- Put commands expected fewer than a few times a month in menus; keep the gesture as an accelerator.
- On the first use after a long gap, offer a skippable weak hint in place, rather than a tutorial after failure.
- From logs, inspect the day-gap distribution between successful gestures; for long median gaps, check that a visible entry exists and is used.