Lexical bundle studies have attracted much interest recently. Many studies have examined how nonnative speakers employ bundles in their language production but few studies have explored discipline-specific bundles in the field of applied linguistics. This study thus aims to investigate how non-native speakers in the field of applied linguistics employ discipline-specific bundles. The data includes three corpora: a 1-million-word expert writer corpus, a 1.8-million-word nonnative-speaker (NNS) writer corpus, and a 1.8-million-word native-speaker (NS) writer corpus. Bundles identified in the corpora were categorized based on functions. Findings include: (1) Expert writers used 118 types of bundles, whereas NNS and NS writers used 197 and 115 types respectively. Yet, only 41 types were shared by all. (2) The distributions of bundle usage patterns were similar for the expert writers and the NS writers. (3) Although the NNS group used more bundle types, 121 types were not found in either the expert or NS groups. (4) The three corpora had similar functional distributions, with research-oriented bundles being the most common, followed by text-oriented and participant-oriented bundles. (5) NNS writers tended to rely on certain bundles to specify research topics and results, but they didn’t seem to have a good command of participant-oriented bundles.