Frequent switching of input method is itself a cognitive and motor cost; it must not be assumed free

What it is

Each move from pen to keyboard or back, people put down or pick up an effector, wait for a panel or keyboard animation, check that the caret is still there, and sometimes reseat the palm. That time and attention is a switch tax. A design that assumes “both work, so switch at will” hides the tax in every gap between fields, and round trips eat the task.

Why it happens

The tax has layers: motion (pen into a silo or hand swapping for a keyboard), vision (soft keyboard covering context, handwriting pad occupying the screen), state checks (is the last character still there, which mode is this), and sometimes palm rejection and hover resyncing. On short tasks the tax is a large fraction: raising a keyboard for one email address and taking the pen back is dearer than a slightly worse character path in pen mode. Switching also breaks the motor plan for writing, so the next stroke starts slower. Offering two inputs does not mean micro-tasks should round-trip; cheaper is to finish a kind of content in one channel, or give the current channel a good-enough degraded path (a temporary keyboard under the pen, a smear under the keyboard). Animations and focus rebuilds of a few hundred milliseconds inflate the tax.

Studying it

Build a task that must alternate (a label then a number, repeatedly) and one that can be batched (mark all, then fill the form). Measure switch count, wall-clock per switch, and how much tax falls when like items are grouped.

Independent variables: forced alternation vs batching allowed, switch animation duration, degraded path present. Dependent measures: fraction of time spent switching, errors, irritation, voluntary strategies that cut switches.

Where it stops holding

With only one channel there is no tax. Accessibility switches may already be slow; the absolute tax differs, but “not zero” still holds. Demos that pre-raise both inputs undercount real pick-up and put-down.

Applying it

• Group like fields so a channel can finish a run, rather than interleave.
• Give the current channel a good-enough degrade: a short character strip in pen mode, a simple ink line in keyboard mode, so one action does not pay the tax.
• Keep switch animations short and the caret still, to lower per-switch tax.
• How to check: an interleaved form versus a batched one; non-input time should be higher when interleaved. After adding a short character strip, pen users leaving the pen for an email field should drop. Lengthen the switch animation as a control; irritation should rise.

Related

• Same group: C5.16.1 Mixing ink and type needs explicit join rules · C5.16.2 Switching must not drop content or the caret · C5.16.3 Handwriting fits drawing and markup; typing fits long prose
• Adjacent: C5.07 Division of Labor Between Pen and Finger · C5.17 Pen Pairing, Battery, and Loss
• Search: mode-switch cost · input thrashing · channel switch