The automation of multi-word sequences (MWSs) reduces the cognitive load required for fluent speech (Wood, 2010). However, this poses challenges for L2 learners because considerable phonological reduction occurs in high-frequency MWSs (Bybee, 2002). This study looks at whether attending to reductions during focused listening and speaking practice of the highest-frequency MWSs could help improve students’ spoken fluency and listening perception (accuracy). Over the course of 10 weeks, both a control group (N=17) and an experimental group (N=16) of Japanese L1 English language users studied 10 common bigram collocations each week, while the experimental group additionally studied sets of 10 phonologically reduced high-frequency trigrams. Both groups also described three picture stories that contained the target collocations in class and completed focused listening homework each week. Participants were tested for 1) gains in spoken fluency through both a free speaking and storytelling task and 2) ability to accurately complete a dictation listening task containing target MWSs. This presentation will discuss the study and results, which found a large advantage for the experimental group in spoken storytelling fluency but no statistically significant difference in free speaking fluency and listening perception.