Instance theory predicts information theory: Episodic uncertainty as a determinant of keystroke dynamics.

How does prior experience shape skilled performance in structured environments? We use skilled typing of natural text to evaluate correspondence between performance (keystroke timing) and structure in the environment (letter uncertainty). We had similar to 350 typists copy-type English text. We reproduced Ostry's (1983) analysis of interkeystroke interval as a function of letter position and word length, that showed prominent first-letter and midword slowing effects. We propose a novel account that letter position and word length effects on keystroke dynamics reflect informational uncertainty about letters in those locations, rather than resource limited planning/buffering processes. We computed positional uncertainty for letters in all positions of words from length one to nine using Google's n-gram database. We show that variance in interkeystroke interval by letter position and word length tracks natural variation in letter uncertainty. Finally, we provide a model showing how a general learning and memory process could acquire sensitivity to patterns of letter uncertainty in natural English. In doing so, we draw an equivalence between Logan's (1988) instance theory of automatization and Shannon's measure of entropy (H) from information theory. Instance theory's predictions for automatization as a function of experience follow exactly the uncertainty in the choice set being automatized. As a result, instance theory stands as a general process model explaining how context-specific experiences in a structured environment tune skilled performance.