The present thesis focuses on how native English listeners process phonological rule misapplications in non-native-accented speech. In Experiment 1, we examined whether listeners use information about a speaker’s native language to help them understand that speaker’s accented English. The test case for this scenario was word-final obstruent devoicing in German and German-accented speech. Results showed that participants did not generalize their knowledge cross-linguistically. In Experiment 2, we used a categorization task and an eye-tracking visual world paradigm to investigate listeners’ use of a position-sensitive allophonic alternation, the velarization of /l/, as a word segmentation cue in native English. Participants were able to use velarization as a cue during word segmentation, even though they also showed a later, post-perceptual bias to segment /l/ as word initial. Follow-up experiments will build upon these conclusions using German-accented speech as stimuli, which will have reduced or absent velarization of /l/ in word-final position. In sum, these experiments inform us about the limits of phonological knowledge about foreign-accented speech.