The aim of this project was to identify factors contributing to cross-language semantic preview benefits. In Experiment 1, Russian-English bilinguals read English sentences with Russian words presented as parafoveal previews. The gaze-contingent boundary paradigm was used to present sentences. Critical previews were cognate translations of the target word (CTAPT - START), noncognate translations (CPOK - TERM), or interlingual homograph translations (MOPE - SEA). A semantic preview benefit (i.e., shorter fixation durations) was observed for cognate and interlingual homograph translations, but not for noncognate translations. In Experiment 2, English-French bilinguals read English sentences with French words used as parafoveal previews. Critical previews were interlingual homograph translations of the target word (PAIN - BREAD) or interlingual homograph translations with a diacritic added (PÁIN - BREAD). A robust semantic preview benefit was found only for interlingual homographs without diacritics, although both preview types produced a semantic preview benefit in the total fixation duration. Our findings suggest that semantically-related previews need to have substantial orthographic overlap with words in the target language to produce cross-language semantic preview benefits in early eye fixation measures. In terms of the Bilingual Interactive Activation + model, the preview word may need to activate the language node for the target language before its meaning is integrated with that of the target word.