| Summary: | 碩士 === 國立中正大學 === 語言所 === 98 === According to Steriade (2001), there are two general mechanisms responsible for phonological typology: analytic bias, systematic cognitive predispositions like Universal Grammar (UG) helping people to learn some patterns than others, and channel bias (Ohala 1993), the diachronic phonologization of phonetically systematic errors in speech transmission, without their being represented explicitly in grammar. To test for analytic bias, the present study examines nasal spreading by testing the learnability of artificial grammars.
In Johore Malay (Onn 1976), nasality spreads rightward from a nasal consonant to and past vowels, glides, and glottals (e.g., mãɁ̃ãp, ‘pardon’), while being blocked by all other consonants (e.g., [pəŋãw̃ãsan], ‘supervision’).
McCarthy (2009) notes that standard Optimality Theory (Prince & Smolensky 2004) has a surprising problem with this kind of pattern. Namely, standard OT predicts that rather than spreading until a blocker is reached, nasality will not spread at all if there is a blocker anywhere within a word (e.g. [pəŋawasan]); [mãɁ̃ãp]) (the so-called “sour-grapes” pattern).
McCarthy (2009) applies Harmonic Serialism (HS), where word forms are built step by step. Thus nasality in Johore Malay can spread from one segment to another, step by step, until a blocker is reached. This correctly predicts that the blocking type of nasal spreading is attested, while the sour-grapes type of nasal spreading is not.
Since HS is claimed to be part of UG, the HS analysis implies that this cross-linguistic pattern is due to analytic bias. However, phonetic coarticulation also applies locally within words, and diachronic phonologization applies in a step-like fashion. Thus the sour-grapes pattern may be unattested merely because it cannot arise via channel bias. McCarthy presumably predicts that a grammar with a blocking type pattern should be easier to learn than a grammar with a sour-grapes pattern.
Exps. 1-3 tested the blocking and sour-grapes artificial grammars in adults and children and found that the blocking grammar was learned better than the sour-grapes grammar in Exp. 1 (adults). However, Exp. 2 (children) and Exp. 3 (adults) showed no significant learning difference between the two grammars, and no interaction between grammars and age.
Exps. 4-1 and 4-2 were memory and grammaticality judgment task respectively. The results showed that the sour-grapes grammar was learned better than the blocking grammar in both task, though this learning asymmetry was more evident for the grammaticality judgment task than for the memory task.
Exp. 5 was a concept formation task to examine how participants learned both grammars over time. The overall accuracy of the sour-grapes grammar was always higher than that of the blocking grammar, and the burst of learning for the sour-grapes grammar was also earlier than the burst of learning for the blocking grammar, suggesting that the learnability of the sour-grapes grammar was easier than that of the blocking grammar.
In sum, the present study shows that the sour-grapes grammar was learned better than the blocking grammar (the result of Exp. 1 was due to the design flaws), contrary to the analytic bias explanation. The present study gives evidence that instead of analytic bias, this cross-linguistic favoring of blocking over sour-grapes grammars is due to channel bias.
|
|---|
