The current study compares and contrasts the structure of recurrent multi-word units, lexical bundles, in Korean and American university students` writing. Two sets of corpora were prepared from the students` argumentative writings. A series of chi-square tests with standardized residuals was conducted to identify the idiosyncratic characteristics of the lexical bundle structures in the two corpora. The study found that the structure and the distribution of the lexical bundle types and tokens in the Korean university students` writing were different, i.e., the Korean students used lexical bundle types and tokens more frequently than the American university students. The major contributors of the difference were found to be the overuse of noun phrase-based bundles, preposition phrase-based bundles, and personal pronoun with certain lexical bundles. The results also confirmed that the Korean students used more lexical bundles that are usually found in conversation register. Pedagogical implications of the findings and future research suggestions are also introduced and discussed.