Adult Japanese speakers often experience difficulty learning English /r/-/l/. Previous research has reported the effectiveness of using perception-based high-variability identification training with natural speech stimuli on improving Japanese speakers' perception and production of /r/-/l/. Recent studies have also shown the effectiveness of using production-based multipronged pronunciation training with explicit instruction on articulatory movements and visualized spectrograms showing F3 frequencies of English /r/-/l/. The proposed study will compare the effects of identification training and pronunciation training on Japanese speakers' English /r/-/l/ learning, as well as the generalization of learning gains to novel phonetic environments. Further, the proposed study will contribute to speech perception and production research by exploring the relationship between perceptual learning and production learning.