The Japanese Sentence Corpus: Eye-movement data on sentence reading with word predictability norms

We introduce the Japanese Sentence Corpus (JSC), a sentence corpus containing eye-movement data from young adults reading horizontally and vertically presented sentences, along with large-scale predictability norms for each word. Eye-movement corpora have expanded rapidly, with resources available for alphabetic scripts such as English, German, and French, as well as for logographic Chinese. Japanese stands out for its unique linguistic features, combining a multiscriptal writing system of logographic Kanji and phonographic Hiragana and Katakana with one of the few cases of flexible text direction. This makes Japanese, alongside Chinese, an excellent test case for investigating how readers adapt their visual processing and oculomotor activity to different text orientations. Using predictability norms, we analyzed eye-movement data from 67 young native Japanese readers, examining how properties of preceding, current, and upcoming words influence fixation durations. Together with existing corpora, the JSC adds a typologically distinct language, strengthening cross-linguistic comparisons and identifying language-universal and language-specific effects. Ultimately, JSC provides a benchmark dataset for computational models of Japanese sentence processing, bridging psycholinguistics and large language models.