This paper presents an empirical study on the difficulties in learning Chinese as a second language based on learners’ corpora written by native English speakers and native Japanese speakers at CEFR-based A2 and B1 levels. The first part of this paper will discuss the procedures for how to collect learners’ corpora, proofread, establish an error tag system and annotate errors. Later it will focus on a significant difference in the production of “ 一 + Classifier” among the corpora of native English speakers and native Japanese speakers. The corpus of English native speakers displays an overuse of “ 一 + Classifier”, even in an atelic context like a negative construction or a conditional construction where a “ 一 + Classifier” should not occur. On the other hand, the corpus of Japanese native speakers displays a lack of “ 一 + Classifier”. This striking contrast is due to whether or not a determiner position exists in each language. Since English has a determiner position which accommodates an article, “a/an, the”, “this/that/my/your/~’s”, English-native learners tend to treat the “ 一 + Classifier” as an article although it does not appear in an atelic event structure. On the other hand, Japanese does not have any determiner position before a Noun Phrase, therefore it is assumed that Japanese learners find it difficult to learn the conditions where a “ 一 + Classifier” is necessary.