This corpus-based study investigates to what extent L1 Turkish speakers of English produce lexical bundles in their academic writing. To this end, a corpus of published research articles in six academic disciplines was collected. The corpus included one-million words in total. The four and five-word lexical bundles in the corpus were identified with the help of a corpus software and analyzed for their frequency, structural and functional features. The analysis yielded a total of 99 four-word and 22 five-word lexical bundles in the corpus. The results showed that the lexical bundles frequently used by Turkish authors in research articles had structural correlates and performed strong functions to construct the discourse of academic writing. Also, the study revealed a new group of bundles called research referential bundles. This finding might indicate that genre plays a significant role in the use of lexical bundles. The discussions given in this article could provide insights for further multi-word studies.