Non-native phonetic learning is destabilized by exposure to phonological variability before and after training.

Phonological variability is a key factor in many phonetic training studies, but it is unclear whether variability is universally helpful for learners. The current study explored variability and sleep consolidation in non-native phonetic learning. Two groups of participants were trained on a non-native contrast in one vowel context (/u/) and differed in whether they were also tested on an untrained context (/i/). Participants exposed to two vowels during the test were less accurate in perception of trained speech sounds and showed no overnight improvement. These findings suggest that introducing variability even in test phases may destabilize learning and prevent consolidation-based performance improvements.