Abstract
As children learn to read, two skills develop in parallel: the complexity of words they can decode and the speed of word recognition. Reading assessments capture these processes with timed and untimed tasks, yet it remains unclear whether word processing speed and word difficulty are separable dimensions of reading ability or two manifestations of the same skill. The distinction has practical stakes: beginning readers decode too few words to be measured precisely, producing floor effects that limit early dyslexia screening. Here, we independently manipulated word difficulty and presentation time in a lexical decision task to test whether they constitute distinct axes of reading development. Through a research--practice partnership, we collected large-scale developmental data from 3,147 students spanning Grades 1--12 in the United States (English) and Brazil (Portuguese). Items presented with infinite presentation time were systematically less difficult but required longer response times than items under fast conditions, for students across development in both language contexts. Importantly, word processing speed and word difficulty were characterized by a unidimensional framework, indicating that presentation time extends a single difficulty continuum rather than measuring a second construct. We leveraged this result to design and validate a multistage adaptive-timing screener for dyslexia (N = 14,393 calibration; N = 841 validation). A K--2 validation study showed that adaptive timing reduced floor effects and improved screening precision and efficiency for beginning readers. By integrating the experimental precision of psychophysics with scalable psychometric frameworks, this work shows how basic science can be translated into universal school-based dyslexia screening.