Keeping and Losing Contrasts

In this paper, I will first show that contrasts aren\u27t licensed in particular contexts because they are necessarily realized with more salient cues in those contexts than in contexts where they aren\u27t licensed. Next, I will argue that the releases of consonants, particularly non-continuants, are not packages of particularly salient acoustic information about the consonant\u27s identity but instead only one of many kinds of acoustic information about how the utterance is segmented prosodically. Prosodic segmentation helps the listener find words in the stream of speech. Finally, I will show that segments are usually perceived to be different from their neighbors, i.e. to contrast with them, except when the target sound is C1 in a CVC1C2V string, which is instead often perceived to be the same as C2, i.e. to assimilate to C2. Both the general and specific effects of neighboring sounds phonetically explain where contrasts are kept or lost quite differently than the licensing by cue account