Merged production of /ɪ/ and /ɛ/ before nasal consonants is well documented in Southern US English. Perception studies of this merger are more limited (cf. Austen 2020). One possible source of pre-nasal merger is anticipatory vowel nasalization. A 2AFC perception task asked US listeners from inside or outside the South, to categorize stimuli on continua from bid to bed and bin to Ben. To test the effects of consonant and vowel nasality separately, we cross-spliced stimuli in a 2 by 2 design. We ask (1) whether Southern speakers are merged in perception, as is generally assumed, and (2) whether it is vowel nasality, consonant nasality, or both that gives rise to the merger in perception. We fit a logistic regression model on the probability of /ɛ/ responses as a function of continuum step, vowel nasality, consonant nasality, and subject region, with all interactions. We found that Southern listeners had a flatter categorization function with lower accuracy at the continuum ends than non-Southern listeners whenever a nasal coda was present, regardless of vowel nasality, confirming that (1) Southern speakers are, to some degree, merged in perception, and (2) that the presence of the nasal coda, and not vowel nasality, conditions merger in perception.
Smith et al. (Tue,) studied this question.