The purpose of this research is to investigate how syllable structure and phonotactic sequencing constraints affect speech perception in Nepali. Word initial and word final consonant clusters are not allowed in Nepali. As a result, when borrowing words from other languages, Nepali speakers often break up clusters by inserting vowels. For example, they pronounce the English word 'slip' as 'silip'. Earlier research has found that Nepali speakers find it difficult to distinguish between consonant clusters at word boundaries and the same clusters broken up by vowels. In other words, 'slip' and 'silip' tend to be indistinguishable to Nepali speakers. This suggests that these language-specific phonotactic constraints cause Nepali listeners to perceive nonexistent vowels at word boundaries.
This research investigates whether similar phenomenon can be observed word internally. A perceptual experiment will be conducted to test the hypothesis that monolingual Nepali speakers also find it difficult to distinguish between consonant clusters word internally and the same clusters broken up by vowels. If the results of the experiment support this hypothesis, then such perceptual confusion would indicate that consonant clusters at word-internal syllable boundaries, like the ones at word boundaries, affect the speech perception of Nepali speakers.
Since Nepali is an understudied language, this research promises to shed new light on the role of syllable structure and phonotactic sequencing constraints on speech perception. Additionally, this research can inform phonotactic learning models by identifying a distinct role for syllable structure. Broader impacts of the proposed research include introducing underprivileged Nepali high school students to linguistic and psychological sciences.
Understanding how humans perceive speech is an important goal for understanding human cognition, as well as for machine recognition of speech. Much earlier research has focused on how speakers of well-studied languages, such as English or Japanese, perceive speech. In this project, different theories of speech perception were studied with respect to Nepali, an understudied Indo-Aryan language spoken in Nepal in south Asia. Specifically, the project sheds light on the role syllables play in judgements of word well-formedness. We conducted perceptual experiments with monolingual Nepali speakers that test perception of illusory vowels at syllable edges. Nepali disallows word initial and word final consonant clusters. As a result, Nepali speakers find it hard to distinguish between consonant clusters at word boundaries and the same clusters broken up by a vowel. For instance, [skul] and [iskul] are indistinguishable as distinct words to Nepali speakers. Hence the [i] vowel is called illusory because they perceive [skul] as [iskul]. This research investigated if similar phenomenon can be observed word internally at positions hypothesize to correspond to syllable boundaries. The results show that illicit consonant clusters at these word-internal positions had the same effects on speech perception. It was difficult for the subjects to discriminate [sibilant][voiceless stop] clusters from [sibilant][voiceless stop][vowel] sequences at hypothesized syllable boundaries. For instance, the degree of confusability was high for non word pairs like [nask • po ] and [naskë • po]. In this example, the hypothesized syllable boundary is represented using a ‘•â€™ and the vowel schwa is represented with [ë]. Also, it was found that such degree of confusability was higher with the vowel [ë] than with [i]. This result is in accord with the results for word boundary clusters in which the subjects perceive illusory [ë] vowel at word boundaries. These observed perceptual confusion is consistent with hypotheses positing syllable structure in Nepali. First, as Nepali allows word medial clusters, the perceptual confusion that resulted in the hearing of an illusory vowel cannot be merely due to word-internal illicit sequences. Second, the illicit word-internal sequences had the same effects on the word-internal speech perceptions as they had on the word-boundary speech perceptions. It was not only difficult for the subjects to discriminate [sibilant][voiceless stop] clusters from [sibilant][voiceless stop][vowel] sequences at positions word-internally that correspond to hypothesized syllable boundaries, but this confusability was higher with [ë] than [i], just as with word boundaries . These experiments help us better understand that syllable structure plays an important role in the nature of speech perception in Nepali. In particular, the perceptual phenomenon previously observed at word edges can be unified with the results in this experiment which looked at word-medial clusters under the lingustic concept of the syllable. These results suggest that computational models of human's perception of language need to appropriately incoroporate the language's syllable structure.
