The organization of bilingual semantic has been a hot topic for cognitive researchers. A common view indicated that bilingual speakers access their languages in a shared semantic system, in which each has a independent lexical representation. However, how the lexical representation of a second language accesses its conceptual representation was still controversial. Revised Hierarchical Model (RHM) assumed that there was a connection between lexical representation and semantic representation of the two bilingual languages, and the strength of links was unequal. So far, the Revised Hierarchical Model has got proved by a number of bilingual researches, yet most of the evidences were collected from national bilinguals but not diglossia speakers. Though belong to a same language family, Teochew dialect and Cantonese are featured by two types of phonetic systems with a shared character system using diverse semantic parameters. The present study aimed to explore the accessing process of the two languages of less-proficient Teochew- Cantonese bilinguals with semantic decision task, and test the Revised Hierarchical Model. Two experiments were designed with cross-language priming semantic decision tasks. All of the participants were native Teochew speakers and began to learn Cantonese in teenage. Two experiments were carried out in Experiment 1, in which the participants were required to judge whether the spoken target words were food-word or not. Stimuli included 64 Teochew words and 64 Cantonese words, half of which were translation equivalents. In Experiment 1a, the Cantonese (L2) target words were primed by their Teochew (L1) translations, while Cantonese (L2) worked as priming words in Experiment 1b in contrast. To confirm whether L2 accesses its conceptual representation without the medium of L1, a same task was retained in Experiment 2(2a and 2b) while the translation equivalents were changed to pair words with the relationship of semantic association. Again, the prime word were L1 in experiment 2a while L2 in experiment 2b. Error rates and reaction times for correct responses were analyzed with ANOVA. Significant priming effects were found in Experiment 1a and Experiment 1b. Typically, it was found that the strength of the priming effect from L1 to L2 was significantly larger than the priming effect from L2 to L1. Also, significant cross-language priming effects were found in Experiment 2 when the relation between the Teochew words and the Cantonese words was semantic association. However, no significant difference was found between the strength of the priming effect from L1 to L2 and vise versa. Total experimental results showed that the two languages of less proficient Teochew -Cantonese bilinguals shared a same conceptual representation while each had independent lexical representation. The findings of cross-language priming effects under different priming condition, priming relationships and priming languages, indicated that the less proficient Teochew -Cantonese bilinguals directly accessed a shared conceptual representation during the processing of spoken words. Nevertheless, the link strength between lexical representation and semantic representation of the two languages were still asymmetric.