9/2/2009 Introduction Korean af rean affricat fricates and es and We propose an analysis of consonant-tone consonant-to tone i inte teraction interaction for Korean affricates based on Lee (2008) Seunghun Julio Lee (Central Connecticut State University) [continuant] specifications are unordered in the Jeremy Perkins (Rutgers University) phonological representation of affricates (Lombardi 1990) The 6 th Workshop on Altaic Formal Linguistics Nagoya University May 22-24, 2009 2 Laryngeal contrasts Our proposal in Korean obstruents (1) affricates stops The distribution of Korean affricates supports a. [ ʦ am] ‘sleep’ cf. [tal] ‘moon’ L lax Lombardi’s (1990) proposal that the continuant b. [ ʦ * am] ‘a moment’ cf. [t*al] ‘daughter’ H tense specifications for affricates do not have to be p c. [ ʦ h am] ‘very’ cf. [t h al] ‘mask’ H aspirated ordered as [-continuant] preceding [+continuant]. (2) affricates fricatives a. [ ʦ am] ‘sleep’ L Cf. Sagey (1986) lax N/A b. [ ʦ * am] ‘a moment’ cf. [s*ada] ‘to be cheap’ H tense [-continuant] precedes [+continuant] c. [ ʦ h am] ‘very’ cf. [sada] ‘to buy’ H aspirated The place of articulation is represented following Kim (1999). 3 4 Analysis - Overview Phonological representation We will derive the consonant-tone interaction regarding Phonologically, affricates can be represented affricates via constraint interaction. either as unordered [-continuant] and [+continuant] (Lombardi 1990) or as ordered so that [-continuant] precedes [+continuant] Adopting Lee (2008), tones can directly associate with (Sagey 1986) (Sagey 1986) consonants consonants. Lombardi 1990: 381 Sagey 1986: 96 Such a direct association of tone is only restricted by markedness constraints. [-continuant] Root | There are no faithfulness constraints that preserve tones directly Root associated to consonants. | [-continuant] [+continuant] As a result, tones cannot be contrastive on consonants, but [+continuant] tones on adjacent vowels can be affected by consonants. Cf. Affricates as strident stops (Jakobson, Fant & Halle 1951: 24) 5 6 1
9/2/2009 Ordered [continuant] specifications If the ordering of the [continuant] specifications were to play a crucial role, we expect that the [+continuant] feature Methods & Results would affect the pitch of the vowel following affricates since it is closer to the vowel. → Then, affricates would pattern on par with fricatives (see Then, affricates would pattern on par with fricatives (see [s] and [s*] in (2)), contrary to fact. [tsa] L | [ts] [a] [-continuant] [+continuant] 7 Emerging tone in Standard Korean Methods Silva (2006: 305) Native speakers of Seoul Korean List of words beginning with obstruents varying in terms of “… standard Korean is coming into place of articulation and manner of articulation alignment with other varieties of the alignment with other varieties of the Recorded in a carrier phrase at the Phonetics Lab of Rutgers language (most of which employ either University lexical or phrasal pitch accent), as well as with other East Asian languages, which use 이건 ___ 이라고 하지요 . F0 in phonemically relevant ways.” “This is called _____.” 9 10 Effects of affricates on pitch Phonetics of affricates Lax affricates have lower pitch on following vowels. Table (3) shows that Korean affricates pattern as stops. Tense and aspirated affricates have higher pitch on following vowels. Affricates and stops share the feature [-continuant]. (3) Initial F0 st.err. (cf. affricates and fricatives share the feature [+continuant]) a. lax [tsa] 193.2 Hz 9.0 Phonetically, affricates consist of a stop closure followed by b. tense [ts*a] 297.3 Hz 15.4 fricative release. [ts h a] c. aspirated 337.1 Hz 14.2 This stop closure works in concert with laryngeal activity, and so segments with [-continuant] are better anchors for tone (Cf. Steriade 1994). This is the same pattern seen in stops (Silva 2006, Wright 2007). 11 12 2
9/2/2009 Affricates and Pitch Similar behaviors of stops and affricates LAX TENSE ASPIRATED I nitial F0 for Stops & Affricates - Speaker 2 0 1 400 a. [tsa] (mean = 193.2 Hz) b. [ts*a] (mean = 297.3 Hz) c. [ts h a] (mean = 337.1 Hz) 350 300 250 250 F0 ( Hz) 200 150 100 50 0 Lax Tense Aspirated Laryngeal Manner Stops Affricates 13 14 Unordered [continuant] specifications Unordered [continuant] specifications As shown in (3), affricates’ influence on the pitch of the L tone and [-continuant] following vowel comes from the laryngeal feature associated Stops: [ta] Fricatives: [sa] with stops [-continuant] L [+continuant] L → Thus, the continuant specifications should not be ordered. | | | | [t] [a] [s] [a] [tsa] [ ] [-continuant] L Stops are [-continuant], and therefore occur with L tone. | | Fricatives are [+continuant], and therefore do NOT occur with L tone. [ts] [a] | [+continuant] 15 16 Unordered [continuant] specifications An OT Analysis The consonant-tone interaction data favors Lombardi’s (1990) proposal that the [continuant] specifications of th [ ti t] ifi ti f affricates are not ordered in the phonological representation. 17 3
9/2/2009 Consonants on tone Tone is not contrastive on consonants Lee (2008) Lee (2008) A direct association of tone is only restricted by markedness There are no faithfulness constraints that preserve tone directly constraints. associated to consonants. M ORA → T → This means that no language has tones that are contrastive on g g consonants. moras not linked to a tone are marked. R OOT N ODE → T Unattested language root nodes (including non-moraic consonants) not linked to a H L tone are marked. [b a] [b a] *[+S PREAD G LOTTIS ]/L [+s.g] segments associated with L tone are marked. 19 20 Descriptive OT Generalizations Analysis – optimal output This mapping is optimal. Lax affricates do not occur with H tone (4a). This requirement is accommodated by changing H to L tone. optimal input Tense and aspirated affricates do not occur with L tone (4b, c). output This requirement is accommodated by changing L to H tone. L H | | /ts h a/ [ts h a] (4) affricates [ ʦ am] a. ‘sleep’ L lax [ ʦ * am] b. ‘a moment’ H This does not violate the constraints, *[+S PREAD G LOTTIS ]/L tense and R OOT N ODE → T [ ʦ h am] c. ‘very’ H aspirated It does violate I DENT -T; however, this constraint is ranked below the others 21 22 Analysis (cont’d) Analysis (cont’d) This mapping is not optimal because the aspirated affricate This mapping is not optimal because the aspirated affricate [ts h ] is associated with a L tone. [ ts h ] is not associated with any tone unintended unintended input input output output L L L L L L L L | | | /ts h a/ [ts h a] /ts h a/ [ts h a] This violates the constraint, *[+S PREAD G LOTTIS ]/L This violates the constraint, R OOT N ODE → T 23 24 4
9/2/2009 Theoretical consequences Conclusion Markedness constraints on consonant-tone interaction outranks We have proposed an OT analysis of consonant-tone the tonal faithfulness constraint (I DENT -T). Under this ranking, the interaction for Korean affricates based on Lee (2008) change of tone is allowed to satisfy the markedness requirements. [continuant] specifications are unordered in the phonological In this analysis, consonant-tone interaction results from constraint representation of affricates (Lombardi 1990). representation of affricates (Lombardi 1990). interaction and not from representational requirements (i e interaction, and not from representational requirements (i.e. Korean affricates behave like stops (rather than fricatives). Bradshaw 1999 among others). In addition to *[+S PREAD G LOTTIS ]/L, the constraint *[+ VOICE ]/H, *[+C ONSTRICTED G LOTTIS ]/L are also active in Standard Korean. → Thus, this analysis also accounts for the presence of L tone on lax affricates and H tone on tense affricates. 25 26 Appendix 1a: Appendix 1b: Tableau (lax affricates) Tableau (tense affricates) L *[+C ONSTRICTED | R OOT N ODE → T I DENT -T L G LOTTIS ]/L /ts* a/ | R OOT N ODE → T *[+V OICE ]/H I DENT -T /ts a/ H a. 1 * L [ts* a] [ * ] a. 1 * [ts a] L b. | W * L H [ts* a] b. | W * L [ts a] L c. 1 W * L H [ts* a] c. 1 W * L [ts a] 27 28 Appendix 1c: Acknowledgments Tableau (aspirated affricates) We thank Shigeto Kawahara for comments. All errors are L *[+S PREAD | R OOT N ODE → T I DENT -T responsible of the authors. G LOTTIS ]/L /ts h a/ H The data is collected under the protocol approved by Central a. 1 * Connecticut State University Human Studies Council #Su08011 y [ts h a] [ h ] L This project is partially funded by the CCSU Center for b. | W * L [ts h a] International Education/Hyundai Funding and the CCSU-AAUP travel fund awarded to Seunghun Lee. L c. 1 W * L [ts h a] 29 30 5
Recommend
More recommend
