The acquisition of resyllabification in Spanish by English speakers

Phonetics;Phonology;Spanish;Speech-segmentation;Second language acquisition

Scarpace, Daniel L

关键词: Phonetics;    Phonology;    Spanish;    Speech-segmentation;    Second language acquisition;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/97243/SCARPACE-DISSERTATION-2017.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

How do second language learners produce and recognize words in continuous speech when speech segmentation strategies differ between languages? This thesis explores phrases that have been affected by the process of resyllabification of consonants across word boundaries, where syllable and word boundaries are misaligned (e.g. las alas `the wings' as [la.s#a.las] in Spanish). For production, learners must acquire a new gestural coordination pattern for producing these types of sequences across the word boundary, as well as avoid overmarking vowel-initial words with glottalization, which is the norm in some languages, such as English (Bissiri et al., 2011), but not Spanish (Hualde, 2005). For word recognition, models of speech segmentation that emphasize the consonant as a likely segment for the beginning of a word (such as the Possible Word Constraint (Norris et al., 1997) or the Syllable Onset Segmentation Heuristic (Dumay et al., 2001)) would correctly segment salas `rooms' out of a string of sounds [la.s_a.las], but would miss the possibility that alas is also a word. For languages with resyllabification, speakers must consider both segmenting the sequence into both V#CV and VC#V; that is, for /lasa/, both /s/ initial words and /a/ initial words must be activated equally in the lexicon. In English, where vowel-initial words are more likely to be accompanied by glottalization, for the same string /lasa/, /s/ initial words will be more strongly activated than /a/ initial words. Thus, English speakers learning Spanish must suppress their tendency to preclude the activation of lexical entries with different onsets (i.e. vowel initial words when preceded by a consonant).Additionally, English-speaking learners of Spanish must learn an additional language specific cue to word segmentation: the tap and trill contrast at the word boundary in Spanish. While both the tap and trill contrast between vowels word-medially (pero `but' vs. perro `dog'), at the word boundary, the tap is associated with a resyllabified /r/, as in ver ocas `to see geese', whereas the trill is obligatorily a word-initial segment, as in ve rocas `s/he sees rocks'. Several tasks were designed to test both the production and perception of resyllabification. In all cases, native Spanish speakers were compared with learners of Spanish whose first language was English at various proficiency levels. Production data is given from a laboratory read-speech task in both Spanish and English designed to elicit VC#V sequences with three consonants: /n/ and /s/ to test overall resyllabification, and /r/ to test the allophonic distribution of this phoneme at the word boundary. The results show that second language learners struggle overall to resyllabify consonants: they favor marking vowel-initial words with glottal stops, although they do so less than they do in English. They do not produce the tap and the trill in the appropriate positions at the word boundary either, although this is complicated by the difficulty in producing these segments in any position.Perception data is given in three different experiments: a word-monitoring task geared towards investigating /n/ and /s/ at the word boundary, a word-monitoring gating task investigating /r/, and finally a visual-world paradigm eye-tracking experiment investigating whether or not multiple lexical candidates are considered while listening to phrases where /n/, /s/, and /r/ straddle the word boundary. While ultimately perception results show that non-native speakers are able to successfully parse the intended word, even in resyllabification contexts, results show that they still have a bias towards consonant-initial words although this is lessened with increased proficiency. Additionally, native speakers appear to have this bias in perception as well, which is unreported in the literature. This finding calls into question the idea of complete resyllabification (Harris, 1983), wherein perception and production of V#CV and VC#V sequences are expected to be identical. Overall, this work adds descriptive contributions to word-juncture phenomena in Spanish and English and theoretical contributions to the ideas of lexical access and storage, phonological encoding, phonological universals in syllabification, and speech segmentation in a second language, as well as methodological concerns regarding eye tracking.