If a mid-air move had to be learned, it must be learnable again from the task, not from last month's setup wizard

What it is

Any mid-air action that cannot be guessed from a visible control, and had to be learned, must still have its learning door at the moment of use. A first-boot tutorial whose entrance then hides in deep settings assumes memory never decays. Mid-air has no keycap reminder. A few unused days and direction, return, and which hand, go. On-demand relearning means: from the current task, open this item's demonstration, practice once, go back to work. Discoverability is knowing the vocabulary exists. Relearning is not having to find the first tutorial again after forgetting.

Why it happens

Motor skill forgets, especially low-frequency skill with no physical trace. Retrieval cues for mid-air moves are weak; over a gap, direction and reset go first. First-run tutorials happen before people understand the product's tasks, so encoding is poor. The effective relearning window is the moment the command is actually needed. If the entrance is three layers off the task, people switch to a non-gesture path or drop the function, and logs show “low use” that is really “high forgetting + high retrieval cost.” Relearning materials must also match the live firmware vocabulary. An old clip teaching a moved item creates new neighbour errors.

Studying it

Teach, then return after a day, a week, a month. Measure unprompted success, then time to find the relearning entrance and succeed. Factor: entrance in-task vs in settings vs absent. Control: auto-pop the item's demo on failure. Log how many people switch modality after forgetting. Do not measure retention only on the teaching day. Experts returning after a season are a more realistic forgetting window.

Where it stops holding

Fully guessable items with one-to-one form and function need little relearning, but should still be tappable in a legend. Games that teach again on death put the entrance in the level design. A one-shot kiosk user has no “later”; relearning collapses to on-the-spot teaching and should be shorter. If the vocabulary changes often, relearning materials must version with it, or on-demand relearning teaches the wrong move. If relearning means running the entire tutorial again, cost exceeds forgetting and people refuse to open it.

Applying it

• Every taught gesture can open its own demo on failure and from help. Ban a single bundled tutorial as the only door.
• The entrance is within two steps of the current task. A “gesture guide” buried in settings is backup only.
• Accept on a week-later return: after an unprompted miss, time to find the demo and succeed. If they cannot find it, or it takes too long, the entrance fails.

Related

• Same group: C4.14.1 The gesture set must be bounded and discoverable · C4.14.2 Similar shapes must not carry very different consequences · C4.14.3 Gesture meaning is validated with target users, not designer intuition
• Adjacent: C3.32 When to teach gestures · C4.07 Mid-air swipe
• Search: relearning · gesture legend · skill decay