RESUMEN
Previous research has indicated that, in viewing a visual word, the activated phonological representation in turn activates its homophone, causing semantic interference. Using this mechanism of phonological mediation, this study investigated native-language phonological interference in visual recognition of Chinese two-character compounds by early Hakka-Mandarin bilinguals. A visual semantic-relatedness decision task in Chinese was given to native Mandarin speakers and early Hakka-Mandarin bilinguals. Both participant groups made more false positive errors and responded more slowly to the pair of two-character compounds containing a homophone; but only Hakka-Mandarin bilinguals made more false positive errors and responded more slowly to the pair containing a near-homophone. We concluded that phonology is needed in both native and nonnative speakers' meaning access of Chinese two-character compounds and that native-language phonological interference is universal in L2 visual word recognition, not language type dependent; phonological and orthographic information are "interactive-compensatory" in helping Hakka readers' resolve the interference.