Migration cost of alternative layouts exceeds their efficiency gain

What it is

When a person leaves a fluent letter layout for another, the cost of rebuilding motor programs usually exceeds the cruise-speed gain the new map can offer. Switching cost here is not a claim that the new geometry is bad. Speed drops while relearning, the old mapping intrudes on the new actions, and the two programs may interfere for a long time. The efficiency gain is the time saved by shorter finger travel at expert level, often far smaller than the deficit paid during retraining.

Why it happens

Rewriting procedural memory is not an overwrite: two stimulus–response rules coexist. The same letter still triggers the old finger; errors look like pressing a neighbour that used to be the target, especially on the opening of frequent words. Expert time is spent stitching chunks together, not covering single-key aiming distances, so moving a few high-frequency letters under the index fingers saves only a short open-loop fragment for someone already chunked. During relearning the old layout still has to be maintained—mail clients, kiosks, and other people’s machines do not migrate with the user—and switching attention between maps burns control resources. Switching cost is therefore the sum of the speed trough, intrusion errors, and a dual-layout tax; the benefit is at most a few percent at cruise.

Studying it

Transfer studies retrain already-fluent typists onto a target layout and track time to regain former speed, fade-out of intrusion errors (keystrokes aimed at the old location), and negative transfer when returning to the old map. Independent variables include prior fluency, daily practice, and whether the two layouts are interleaved; dependent measures include sessions to criterion, intrusion rate, and switching cost under dual use. Comparing only asymptotic speed hides the tuition; comparing only the first-day drop condemns every new layout. Report the integral of the learning curve (characters not produced during relearning) separately from the asymptotic difference.

Where it stops holding

Novices have no old program to scrap, so switching cost is near zero and geometric differences can appear. Very low daily volume never amortizes the tuition, but not switching is then also cheap. People who type on one dedicated machine and will tolerate weeks of low speed can make the arithmetic work. If a touch layout changes every day, no stable mapping forms and there is nothing to migrate. The gain also depends on language: a layout tuned to one letter-frequency profile may offer nothing on another.

Applying it

• Do not treat “a few percent faster at asymptote” as a reason to switch the default; estimate practice hours needed to regain prior speed.
• When an alternative layout is an advanced option, provide one-step return to the original and do not change the map mid-session.
• Record speed curves and intrusion errors on the same fluent users; do not sell migration with final speeds from people who never used the old layout.

Related

• Same group: C6.01.1 Path dependence of layouts outweighs efficiency optima · C6.01.3 Layouts vary by language and region, and shortcut positions move with them
• Adjacent: C6.03 Shortcut systems · C6.17 Path dependence of keyboard layouts
• Search: switching cost · motor skill transfer · layout migration