Why Typing Cadence Is Treated as a Signal

Typing rhythm is one of the oldest behavioural signals in use, predating the web by decades. Its persistence comes from a simple property: the timing between keystrokes is produced by processes the typist does not control.

Two intervals carry the information

Systems measure how long each key is held down and how long elapses between releasing one key and pressing the next. These are usually called dwell and flight times.

Dwell times are fairly consistent for a given person and vary little with what is being typed. Flight times vary much more, because they depend on which fingers and which hands are involved.

The pair together is more informative than either alone, since the relationship between them reflects technique rather than merely speed.

Familiar sequences are faster than unfamiliar ones

Common letter pairs in a language are typed as learned motor units, so the interval between them is much shorter than between letters that rarely occur together.

This produces a characteristic pattern tied to what someone types often. People who write frequently in a particular language show distinctive fast transitions for that language's common pairs.

It also means passwords and usernames, typed thousands of times, have unusually consistent timing signatures compared to novel text the same person enters.

Hand alternation shapes the rhythm

Sequences that alternate hands are faster than sequences requiring the same finger twice, because the second hand can begin moving before the first has finished.

The resulting rhythm depends on keyboard layout and on the physical arrangement of the typist's hands. Someone typing on a different layout produces measurably different timing for the same text.

Touch typists and hunt-and-peck typists differ so sharply here that the two groups are distinguishable from timing data alone.

Automation reproduces the mean but not the shape

Scripted input can be given realistic average delays easily. Reproducing the correct distribution around that average, with the right correlations between adjacent intervals, is considerably harder.

Real typing contains bursts and stalls: a fluent run followed by a pause to think, then a correction. Randomised delays produce none of that structure.

The absence of structure is what detection keys on. It looks for input that is too evenly distributed rather than for input that is too fast.

The limits are practical rather than theoretical

Short fields provide too few intervals to say anything, so cadence contributes little on a login form with a brief username.

Pasted input produces no cadence at all, and pasting a password from a manager is a security-positive behaviour that must not be penalised. Systems that treat missing cadence as suspicious punish exactly the users behaving best.