Listeners can rapidly adapt to non-native accented speech, yet the underlying mechanisms remain debated. This study examines whether accent adaptation reflects adjustments in phonetic representations or shifts in decision-making processes. Using a pretest-exposure-posttest paradigm, we examined native English listeners’ perception of the Mandarin-accented /θ/-/s/ contrast across two exposure conditions: exposure to Mandarin-accented sentences (Experiment 1) or to pure tones (Experiment 2). In both experiments, listeners showed increased acceptance of ambiguous /θ/ and /s/ tokens when they formed real words, suggesting that adaptation stems from changes in lexical decision criterion reinforced through task repetition rather than accent exposure alone. Additionally, we observed evidence suggestive of rapid within-test distributional learning from limited trials. Our findings support the notion that listeners lower lexical decision criterion when processing accented speech, while also demonstrating remarkable adaptability to novel accent features even with minimal exposure.