Abstract The study aims to discuss how English article system is acquired by Mandarin-Chinese speakers at two domains, semantic domain and sentence level, by analyzing oral-story-telling data produced by forty 19-to-20-year old college students in Taiwan (20 males and 20 females), divided into low proficiency and high proficiency levels based on their results of Michigan Listening Comprehension Test. The production data was classified into four semantic types marked by a combination of two universal semantic concepts, specificity and definiteness for the purpose of examining Fluctuation Hypothesis (FH) proposed by Ionin (2004), who argued that L2 access to Universal Grammar by predicting L2 learners without article system will fluctuate between two parameter settings of specificity and definiteness. It is found that overuse of the did occur in [+specificity, -definiteness] contexts where target use is a, particularly for low-level learners. Besides, to closely probe into how L1 Mandarin-Chinese speakers use articles in L2 grammar within Ionin¡¦s framework, a model for linguistic properties marking specificity and definiteness in Chinese was proposed so as to compare the differences between English article system and Chinese classifier system, arguing that the interference of L1 may take place at semantic domain by L1 Mandarin-Chinese speakers. The evidence that the substitution of nage for definite article the in [+specificity, +definiteness] contexts and that of numeral one yige for indefinite article a in only [+specificity, -definiteness] and [-specificity, -definiteness] contexts sheds the light on the possibility of the occurrence of L1 transfer at the semantic domain. As for article use in sentential positions, due to definiteness effect and subject indefinite effect in Chinese, it is predicted that L1 Mandarin-Chinese speakers would drop articles more often in preverbal positions than in postverbal positions. The result showed that low-level learners did drop more articles in preverbal positions than in postverbal positions, but advanced learners showed the contrast, which implies that the beginners are easily governed by the definiteness effect, that is, L1 is at play at the initial state of L2 grammar. Overall, the advanced learners used articles more accurately than the low-level learners did, suggesting that the advanced Mandarin-Chinese L2 English learners may gradually reset the parameter of L2 grammars in acquiring English article system. Furthermore, different error types produced by the participants were classified in the study and provided with theoretical discussion. A surprising finding is that the low-level learners highly misused the Cinderella for Cinderella in the data. Such error production may show the evidence of L2 access to UG since the Cinderella cannot be used in English and there is no determiner the projecting in Chinese proper name. The overuse of the further illustrates the existence of projecting D for L1 Chinese learners. The acquisition rate of article use was measured by SOC (Suppliances in Obligatory Contexts) and TLU (Target-Like-Use). The results showed that the most difficult article use for both proficiency levels is zero article Ø. The advanced learners can use the more accurately than the learners at proficiency level due to the high occurrences of overgeneralization of the by the low-proficiency levels. In general, the result of the current study bears on the issue of accessibility of UG and the possibility of parameter (re-) setting. It is also shown that L1 plays a significant role in L2 article use not only at the semantic domain but also at the sentential level by L1 Mandarin-Chinese speakers, especially for those at the initial state of L2 grammar.