| Summary: | 碩士 === 國立中正大學 === 語言學研究所 === 104 === Steriade (2001) and Moreton (2008) argue that two cognitive mechanisms are responsible for phonological typology: analytic bias and channel bias. Analytic bias is cognitive biases, like Universal Grammar, facilitating the learning of some patterns and penalizing others. Channel bias (Ohala 1993, Blevins 2004) is the systematic errors in transmission between speakers and listeners to be the precursors of phonologization. Language distribution of property-sensitive stress also presents the typological asymmetry (Gordon 2006). Property-sensitive stress is a general term to refer to the quantity-sensitive and quality-sensitive stress. Previous studies tested the learning bias of property-sensitive stress in English and Canadian French speakers (Carpenter 2006). English and Canadian French speakers learn natural quality-sensitive stress better than unnatural one, but they did not learn natural quantity-sensitive stress better than unnatural one. The positive results do not match the typology of property-sensitive stress: quality-sensitive stress is learned more easily than quantity-sensitive stress. In addition, they only test the speakers who already speak stress languages (quantity-sensitive stress, as in English vs. quantity-sensitive insensitive stress, as in French). The present study tests analytic bias in learning of property-sensitive stress by speakers who speak tone language (Mandarin).
Three experiments adopted artificial grammar learning (Esper 1925). All stimuli are disyllabic words. Participants in each experiment were asked to try to learn training items in their assigned artificial languages. Then they were presented with a set of pairs of grammatically and ungrammatically stressed items and asked to judge which item was grammatical in their artificial languages. Experiment 1 tested quality-sensitive stress: all items had open syllables contrasting low vowels /ɑ, æ/ and high vowels /i, u/. Stress in the artificial natural and unnatural languages is attracted to low vowels and high vowel respectively. Result shows that learners may be affected by analytic bias at first, but the effect of analytic bias may become weak or disappear after more training. Besides, vowel /æ/ may increase the complexity to make the learning difficult. Experiment 2 tested quantity-sensitive stress: items were open (CV) and heavy syllables (CVC) contrasting in syllable weight. Stress in the artificial natural and unnatural languages was attracted by heavy and light syllables respectively. Result supports analytic bias because of a learning advantage for the language with stress on heavy syllables but not light syllables. Experiment 3 crossed two property-sensitive stress systems within each language, but the results support analytic bias: language with natural grammar in terms of both stress properties systems had the worst performance than language with one natural stress property system.
In sum, the present study shows that quantity-sensitive stress is easier to learn than quality-sensitive stress (comparing Experiment 1 with 2). This gives evidence that quantity-sensitive stress more preferred than quality-sensitive stress due to analytic bias. Moreover, Mandarin speakers can learn quantity-sensitive stress and a language with an overly complex stress system may hurt learning.
|
|---|
