Abstract
<p>Cross-linguistic studies of reading have documented broad similarities across languages, yet the extent to which first-language (L1) reading processes differ once structural confounds are minimized remains unclear. We conducted self-paced reading experiments in English and Czech using tightly aligned translation-equivalent sentences, enabling a direct comparison of L1 reading under parallel input conditions. Matched groups of English and Czech L1 speakers (N = 176 per language) showed near-identical comprehension accuracy but robust differences in processing speed: reaction times (RTs) were reliably faster in English than in Czech, a pattern that persisted after controlling for word length and surprisal. At the same time, the effects of classical predictors such as length and surprisal were directionally similar across languages, consistent with universal reading mechanisms. Our analyses suggest two complementary explanations for the observed cross‑linguistic differences. First, Czech exhibits higher information density than English, which—under conditions that prevent skipping and regressions—manifests as slower per‑word reading. Second, surprisal was distributed differently across the two languages: in Czech, surprisal was markedly higher at sentence onset and decreased more steeply over word position, amplifying early‑sentence processing demands. These findings indicate that while core reading processes are universal, language‑specific typological properties shape how information is packaged and distributed across the sentence, yielding systematic cross‑linguistic differences in processing speed. The results parallel cross-linguistic findings on speech rate and support an emerging framework in which reading adapts to the informational structure of the language being processed.</p>