Three experiments using a priming lexical decision paradigm were conducted to examine whether cross-language activation occurs via decomposition during the processing of derived words in Korean-English bilingual readers. In Experiment 1, when participants were given a real derived word and an interpretable derived pseudoword (i.e., illegal combination of a stem and a suffix) in Korean as a prime, response times for the corresponding English-translated stem were significantly faster than when they had received an unrelated word. In Experiment 2, non-morphological ending pseudowords (i.e., illegal combination of a stem and an orthographic ending) were included, and this did not show a priming effect. In Experiment 3, non-interpretable derived pseudowords also yielded a significant priming effect just as the interpretable ones. These results together suggest that cross-language activation of morphologically complex words occurs independently of lexicality and interpretability.