The status of the word-final domain and linguistically-constrained language learning cues

Past experimental and theoretical research have argued that phonological devoicing as a word-final phenomenon only has inherent phonetic motivation to exist utterance-finally and that its presence at lower domain levels is due to analogy (Hock 1991, 1999; Hualde and Eager 2016). This evinces the idea that the word is not a domain tied to phonology inherently. Juxtaposingly, final devoicing (FD) is found phonetically at various prosodic domain levels. Final lengthening (FL) is also found at various prosodic levels, though is explained as only having a push to exist utterance-finally through a process similar to the “slowing down of a machine” or “tempo change...local to the final gesture in the phrase” (Klatt 1976 and Edwards et al. 1991 respectively). The present artificial language learning (ALL) experiment gives evidence that these cues may have an inherent reason for existing at lower domain levels. Participants were exposed to 7 minutes of AL stimulus in one of three conditions: transitional probabilities (TP) or the likelihood that syllables will occur in sequence together, final lengthening (FL) or greater duration of final segments, and final devoicing (FD) or a trailing off of voicing in word-final segments. They were then tested on their ability to segment the AL words out of the speech stream. The results show there is a perceptual effect to having phonetic FL at the right edge of words for listeners, in that it increases novel word segmentation ability above the level of only TPs. The same effect was not found for FD. The FL results indicate that phenomena in the word-final area may be used for perceptual reasons, providing evidence speakers use phonetic effects at the word-level