Target Asymmetries in Consonant Harmony
자음조화에서 관찰되는 타겟 비대칭성
|dc.description||학위논문 (석사)-- 서울대학교 대학원 : 언어학과 언어학전공, 2016. 2. 전종호.||-|
|dc.description.abstract||Consonant harmony refers to non-local assimilation between consonants: for instance, in Kera, /kə-dà:rə̀/ friend becomes [gə-dà:rə̀], in which a nominalizing prefix /k-/ takes on the voicing of the following stem-initial voiced stop /d/. As seen in Kera, there are triggers that cause other sounds to change, and the targets that undergo the change, under the influence of triggers. The present study investigates restrictions on the type of target and trigger consonants. Suppose that A and B are potential target consonants. If A undergoes consonant harmony (A…B → B…B) whereas B doesnt (B... A → B…A, not * A … A), A would be considered a preferred target (or B would be a preferred trigger). This kind of target asymmetry is observed in consonant harmony whereby only certain types of consonants are preferred as targets. This study concerns a question of whether and why target asymmetries exist in consonant harmony typology.
In order to address this question, I conducted a cross-linguistic survey. From the survey results, I found target asymmetries in the following three types of consonant harmony: sibilant, retroflex, and nasal consonant harmony. In sibilant consonant harmony whereby place assimilation occurs between alveolar sibilants ([s, z, ts, tz]) and palatal sibilants ([∫, ʒ, t∫, dʒ]), it is observed that alveolar sibilants are more likely to be targets, whereas palatal ones are more likely to be triggers. Of 25 languages in my survey, 15 languages reveal the target asymmetry in which only alveolar sibilants are targets, to the exclusion of palatal sibilants (e.g. Sarcee, /sì-tʃógò/ → [ʃìtʃógò] my flank, but /sí-tʃíz-àʔ/ → [ʃítʃídzàʔ], *[sítzídzàʔ] my duck (Hansson 2001)). On the other hand, the rest 10 languages show symmetric harmony in which both alveolar and palatal sibilants are equally targeted (e.g. Navajo, /si-dʒé:ʔ/ → [ʃidʒé:ʔ] they lie (slender stiff objects), and /j-iʃ-mas/→ [jismas] Im rolling along (Hansson 2001)). Significantly, however, the opposite case of asymmetric situation whereby only palatal sibilants are targeted is not attested in any language in the survey. Similar kind of target asymmetry is also observed in retroflex and nasal consonant harmony. In retroflex consonant harmony, non-retroflex consonants such as dental or alveolar stops and nasals (/d, t, n/ and /d̪, t̪, n̪/) are more likely to be targets, whereas their retroflex counterparts (/ɖ, ʈ, ɳ/) are more likely to be triggers. In nasal consonant harmony, non-nasal consonants such as plain stops (/b, d, g, p, t, k/) and liquid consonants (/l, r/) are more likely to be targets, while nasal consonants (/m, n, ŋ/) are more likely to be triggers. Based on these survey results, I conclude that the observed target asymmetries should be expressed as the following implicational statements: i) If palatal sibilants are targets of consonant harmony, so are alveolar sibilants. ii) If retroflex consonants are targets of consonant harmony, so are non-retroflex consonants. iii) If nasal consonants are targets of consonant harmony, so are non-nasal consonants.
Inspired by phonetically-based Optimality Theory (Hayes, Kirchner and Steriade 2004), I investigate perceptibility variation in contexts where consonant harmony typically occurs. Based on this investigation, I argue that the target asymmetries are perceptually motivated and can be well understood under P-map hypothesis (Steriade 2001, 2009). The upshot of P-map is that perceptually prominent phonological change is avoided. In Optimality-Theoretic terms, faithfulness constraints preventing more prominent perceptual change invariably outrank those prohibiting less perceptual change. In line with P-map, I claim that consonant harmony is a process preferring less perceptual modification. Various phonetic research show that some phonological features have prolonged phonetic cue spanning over multi-segmental domains, among which the features relevant to palatal, retroflex, and nasal consonants are included. I assert that consonants with prolonged phonetic cue (i.e. more likely triggers of consonant harmony) may weaken perceptibility of the relevant features in nearby consonants and make them less perceptible. This means that phonetic cue of relevant phonological features is weaker before the consonants with long cue (i.e. palatal sibilants, retroflex consonants, and nasal consonants) than before the consonants without long cue.
Reflecting this contextual perceptibility variation, constraints for the corresponding contexts are projected and universally ranked by P-map. The faithfulness constraints prohibiting phonological change before alveolar sibilants, non-retroflex consonants, and non-nasal consonants universally outrank those prohibiting change before palatal sibilants, retroflex consonants, and nasal consonants, respectively. To take an example of sibilant consonant harmony, ID-IO (anterior/__s) is universally ranked above ID-IO (anterior/__∫), explaining a cross-linguistic tendency that faithfulness for anteriority is weaker before palatal sibilants than before alveolar sibilants. Moreover, language-specific consonant harmony patterns are also explained by interaction of these constraints with IDENT-CC, which induces consonant harmony. When IDENT-CC dominates the two ID-IO faithfulness constraints, consonant harmony occurs all the time, regardless of the type of triggers and targets. When it is ranked between the two ID-IO faithfulness constraints, only the alveolar sibilants are targeted, revealing target asymmetry patterns. Finally, when IDENT-CC is dominated by the two ID-IO faithfulness constraints, consonant harmony do not occur at all. The target asymmetries in retroflex and nasal consonant harmony are similarly analyzed: ID-IO (anterior/_d) ≫ ID-IO (anterior/_ɖ), and their interaction with IDENT-CC in retroflex consonant harmony, and ID-IO (nasal/_d) ≫ ID-IO (nasal/_n), and their interaction with IDENT-CC in nasal consonant harmony. Note that this analysis predicts the absence of the pattern in which palatal sibilants, retroflex consonants, and nasal consonants are exclusively targeted in consonant harmony.
The analysis accounts for all and only the attested patterns of consonant harmony typology. Both universal and language-specific patterns are explained in terms of proposed constraint interaction.
|dc.description.tableofcontents||1. Introduction 1
2. Typology 4
2.1. Overview of the survey 4
2.2. Patterns with target asymmetries 4
2.2.1. Sibilant consonant harmony 6
2.2.2. Retroflex consonant harmony 10
2.2.3. Nasal consonant harmony 14
2.2.4. Summary 18
2.3. Patterns with no target asymmetries 19
2.3.1. Non-sibilant coronal consonant harmony 20
2.3.2. Dorsal consonant harmony 22
2.3.3. Secondary articulation consonant harmony 24
2.3.4. Liquid consonant harmony 27
2.3.5. Stricture consonant harmony 32
2.3.6. Laryngeal consonant harmony 37
2.3.7. Summary 39
3. Phonetic and Perceptual Basis 40
3.1. Phonetic grounding: contextual perceptibility variation 41
3.2. Asymmetry in perceptibility of phonological features 44
3.3. Summary 46
4. Analysis 48
4.1. Preliminaries 48
4.1.1. Correspondence-based approach 49
4.1.2. P-map 54
4.2. Proposal 57
4.3. Summary 65
5. Conclusion 66
|dc.title||Target Asymmetries in Consonant Harmony||-|
|dc.title.alternative||자음조화에서 관찰되는 타겟 비대칭성||-|