Speech perception is flexible in that perceptual processes of individuals can adapt as they gain experience in a new sound system of a non-native language. In second language learning, High Variability Phonetic Training (HVPT) in particular has proven effective, as it exposes learners to the variability (e.g., different speakers or phonetic contexts) needed to develop robust sound categories. However, speech perception often takes place in adverse listening conditions (e.g., noise, multitasking), and learning to recognize sounds in these conditions requires adapting to the insufficient cues available, as well as modulating attention to stay focused on the target speech. Our lab explores how learners can cope with these difficulties, and how training methods can be optimized to support speech learning in real-world conditions.