The present study investigates speech segmentation of English by semi-bilingual Japanese speakers of English. Two experiments were conducted with forty subjects. The first experiment used a syllable-monitoring task and the results did not show a moraic segmentation strategy, which is a languagespecific segmentation strategy for native speakers of Japanese. This suggests that they employed a general or segment-bysegment sub-lexical segmentation strategy rather than a language-specific moraic segmentation strategy for accessing the lexicon. The second experiment used a phonememonitoring task and here the results supported a languagespecific moraic segmentation. Taken together, experiments suggest that semi-bilingual Japanese speakers have two segmentation strategies for spoken language recognition, supporting a model proposed by Cutler and colleagues. These results also suggest that these semi-bilinguals only partly suppress their native language-specific segmentation, which makes them comparable to the perfectly balanced French- English bilinguals in a similar study by Cutler and colleagues.