Effects of Word Segmentation Methods on Tibetan Reading Across Different Reading Conditions

This study investigated the effects of word segmentation cues on Tibetan reading and their moderating factors through four eye-tracking experiments. Experiment 1 examined the interaction between material universality (low vs. high) and word segmentation method (normal sentences vs. dictionary-based segmentation vs. psycholinguistic-based segmentation); Experiment 2 explored the moderating role of reading ability (low vs. high); Experiment 3 tested the influence of material familiarity (low vs. high); Experiment 4 compared reading modes (silent reading vs. reading aloud). 52 (Experiments 1 and 3), 30 (Experiment 2), and 44 (Experiment 4) participants from Tibetan communities in Shanxi Province, China, participated in the experiments, with eye movements recorded using a Tobii X3-120 eye-tracker. Results revealed that segmentation effects in Tibetan reading are context-dependent. Under challenging conditions (low material universality/familiarity, lower reading ability, or reading aloud), segmentation robustly facilitates late lexical processing, saccade target selection, and parafoveal processing and reduces uncertainty; under simpler conditions (high universality/familiarity, higher ability, silent reading), these benefits largely diminish or disappear. In silent reading, dictionary- and psycholinguistic-based segmentation show broadly similar effects, with the latter yielding more robust sentence-level gains; in reading aloud, dictionary-based segmentation is superior across most lexical stages. Material universality and familiarity interact strongly with segmentation, whereas interactions with reading ability and mode are weaker. Results also supported a trade-off whereby high text familiarity can offset segmentation benefits for skilled readers.