Nasalization and Tapping in Yonbyon Korean

전체 글

(1)N a s aliz ation an d T appin g in Y on b y on K ore an H y u n s o ok K an g (H any an g U niv ers ity ) * 1) Je on g - Im H an ( K onkuk U niv ers ity ) K an g , H y un s o ok & H an Je ong - Im . 200 2 . N a s aliz at ion and T apping in Y on by on K ore an . T he L ing uistic A ss ociation of K orea J ournal, 10 (4), 1- 21. T his study fir st present s acou stic data on the ph onetic realization of an underly ing liquid in Yonby on Korean w hich is n ot n asalized after a surface n asal, but r ather appears as a t ap . It also presents the ph onetic realization of a pre - liquid obstruent . It then presents an optim ality - theoretic acc ount of the different surface form s of an underlying liquid (McCarthy and Prince 1993, 1995, etc .) and th ose of a pre - liquid obstruent in Yonby on Korean . Ke y W ord s : Yonby on dialect , liquid, later al, t ap, pre - liquid nas alization. 1 . In tro du c tion T his study has tw o major goals . One is to report an interesting cas e of nas alization in the Yonby on dialect of Korean which is spoken in China (henceforth YB), using acoustic data. T he other is t o present an optim ality - theoretic account of allophones of liquid and the tr ansformed nasal fr om an obstruent before an underlyin g liquid in Yonby on Korean . T raditionally , the nasalization of the / obstruent +liquid/ clusters in South Korean (/ ky kly / - > [ky ny ] ' encouragement ' ) is accounted for by Later al Nasalization follow ed by Nasal Assimilation as in (1) (Kim - Ren aud 1974; Cho 1997 among others ).. *. T his w ork w as supported by Hany ang Univ ersity in the prog ram year of. 2001.. (2) 2. Hy un s ook Kang & Jeong - Im Han. (1) Nasalization in (south ) underlyin g form Lateral Nasalization Nasal Assimilation surface form. Korean / ky kly / ' encouragement ' ky kny ky ny [ky ny ]. How ev er , nasalization not triggered by a nasal is attest ed in YB. (2) Nasalization in Yonbyon Korean underlyin g form / ky kly / surface form [ky y ]. ' encouragement '. As (2) indicates , the underlying / l/ in YB does n ot appear as a n asal on the surface, ev en though the preceding obstruent / k/ is realized as a homor ganic nasal. Here w e con sider the phonetic output from / l/ as a tap. A tap is argued to be a segment derived fr om a trill - more generally rhotic - , or a stop, inv olving contact betw een the tongue tip/ blade and the region on/ near the alveolar ridge (Ladefoged 1993; Kahn 1976; Catford 1988; Bann er - Inouy e 1995; Ladefoged and Maddieson 1990; de Jong 1998). Usually the tap has a v ery brief contact , and thus its closure duration appears to be significantly shorter than that for a stop or a trill. As for the American En glish tap, its mean dur ation is 26 ms , as against 75 ms for [d] an d 129 m s for [t] (Steriade 2000). How ev er , ex cept for this alternation relationship betw een a trill/ stop and a t ap, and extra short duration of the tap, the phonetic characteristics the tap show s are v arious . F or ex ample, Ladefoged (1993) point s out that the tap allophone of a trill has different articulation from that of a stop. T hey ar e ar gued to be regarded as comprising a single class of "t ap" only because they are audit orily similar . F or this reas on , the tap realized as an alv eolar rhotic on the surface can be described with the feature combin ation [+s onor ant , +continuant ], distin guishable from the corresponding stops . On the other hand, the tap which is r ealized as a voiced alv eolar st op. (3) Nas alization and T apping in Yon by on Korean. 3. on the phonetic r epresentation can be described with the feature [+v oice] to be distinguished from [t] and [- sonorant], from [r] (Banner - In ouye 1995). Despite thes e v arious phonetic characteristics that the tap show s , the tap is uncontr ov ersially argued t o be derived from the corr esponding stop or trill in terms of lenition , w eakening of sounds . T he tap is considered to have a reduced oral strictur e as compared to the corresponding st op or trill. American English T appin g , for ex ample, is argued to result from the reduction of an alveolar stop to a tap (Kahn 1976). T hus , together w ith all other lenited segm ents , the tap occurs more readily in prosodically w eak position . According to the typological surv ey of the tap by Banner - In ouye (1995), the tap is found either in final or interv ocalic position which is typical lenition environment . In this study w e will observe a type of tap w hich is unique in that even though it show s general properties of the tap such as short duration and an alv eolar contact , it appears in syllable- initial position , as an obstruent , not a sonorant . T he organization of this paper is as follow s : in section 2, w e w ill conduct an acoustic study on the phonetic realization of an underlying liquid / l/ in Yonby on dialect . Based on the spectrographic data, the phonetic realization of each liquid w ill be ex amined. In Section 3, w e will motiv ate constraints t o account for the v arious surface form s of an underlyin g liquid / l/ and the transformed nasal from an underlyin g obstruent in YB. S ection 4 concludes the paper .. 2 . P h on e tic E x p e rim e n t s on a Liqu id in Y on b y on K ore an 2 .1 . Ex perim ent T his study has grown out of a YB nativ e speaker ' s observation that an underlyin g / obstruent + liquid/ sequence does not surface as tw o nasal segments in her dialect as it does in South Korean . She claimed. (4) 4. Hy un s ook Kang & Jeong - Im Han. that "no underlying laterals in YB surface as n asals , but appear as [l] or [ ]." T his study uses acoustic data to test such nativ e ' s intuition or impressionistic obs erv ation of laterals . 2 .2 . M eth ods 2 .2 .1 . S tim uli T he test token s w ere constructed by considering v arious phonological an d morphological contexts as show n below . (3) test w ords phonological context a. V__V. #__ __# V__ __V b.. [- - obstruent + __- - ]w o r d. obstruent]w o r d +__ c.. test w ords yulæ ' origin ' kolip ' isolation ' suly ek hy draulic pow er ' muloto ' no w ork ' loto 'labor ' tol ' stone ' tallyek calendar ' silly e ' rude ' ky kly ' encouragement ' hy ply k 'cooperation hapli ' reasonablenes s y kkoklo 'Yokkok street ' tapsipli ' T apsip area. [- - nasal + __- - ]w o r d. k nlo 'w ork ' c nly k ' electric pow er ' lanli ' disturbance '. nasal]w o r d +__. sinmunlo ' Sinmun street ' hyangsinlyo ' spice '. T he ex amples in (3a ) ar e provided to check w hether YB has / l/ variant s similar t o SK liquids in v arious phonological positions other. (5) Nas alization and T apping in Yon by on Korean. 5. than the post - consonantal one. More specifically , un derlying / l/ s in interv ocalic, w ord- initial, and w ord- final positions are considered. Later al geminat es in interv ocalic position are als o considered for this purpos e. T he ex amples of post - con sonantal lat erals in underlying representation are presented in (3b ) and (3c). In (3b ), the three ex amples in the second column hav e the sequence of an underlying obstruent plus a later al within a w or d, while in the tw o other ex amples in the third column a w or d boundary is posited betw een the obstruent and the lateral. T he ex amples in (3c) are those w here an un derlying nasal is immediately follow ed by a lateral. Again in the first three ex amples of the second column , the s equence of an underlying nasal and a lateral is included within a w ord, and in the other tw o ex amples the nasal an d the later al are located across the w ord boundary . Each test w ord w as recorded both in an isolated form and in a fr ame sentence. T he frame sentence used in this experim ent is pr esented in (4). (4) frame sentence ik si - - - - - - ipnita.. 'T his is _________ .'. 2 .2 .2 . S ubj ec t s an d R ec ordin g T w o n ativ e speakers (on e male, one female) of YB participated in the recording . All speakers w ere born and raised in Yonby on , China and had spent less than tw o years in South Korea. Neither had any report ed history of either speakin g or hearin g disorders . T hey w ere recorded using a CSL (Comput erized Speech Lab ) (model 4300B) for P C at 16kHz samplin g rate. Before recording , each speaker w as asked to read a few ran domly chosen test tokens to familiarize themselves with the material. T hen they w ere asked to read each test utt erance thr ee times in random order , at a natural, comfortable speed. T he test s entences w ere present ed in Korean orthography in random order . T he total number of tokens w ere 228 (19 w ords x 2 speaker s x 2 recor ding styles x 3 repetition s ).. (6) 6. Hy un s ook Kang & Jeong - Im Han. 2 .2 .3 . A n aly s is T he recorded data w ere stored as files and then w av eforms , and wide- band spectrogr ams for each token w ere generated using the MULT I- SPEECH (model 3700) softw are package. T he judgem ents among v arious phonetic forms of / l/ w ere bas ed on the w av eforms , spectrograms an d recorded s ounds . Both authors of this study ex amined the test tokens for the portion of one repetition by a m ale speaker and then the secon d author did the rest of the phonetic analysis based on their judgement s of these. 2 .3 . Re s ult s an d D is c u s s ion T he acoustic data obtained from the experiment conduct ed in this study w ere compatible with the impressionistic judgements . Before reporting how underlying / l/ s are phonetically realized in the post - cons onant al position , other kinds of lateral v ariants are ex amined. Fir st , F igure 1 show s that the un derlying / l/ s in interv ocalic position are realized as "sonorant taps " with rather clear formant patt erns , ev en though they are w eaker than those of a v ow el. T he n ature of son orancy for a t ap is also attested in many other languages , including the Am erican English tap (St eriade 2000; de Jon g 1998). Figur e 1.. k. o. / kolip/ - - > [ko ip]. ' isolation '. i. p. (7) Nas alization and T apping in Yon by on Korean. 7. W or d- final laterals in YB are also similar t o those of SK in that they show clear formant patterns as in Figur e 2, but some tokens are heard to be more v elarized as in the case of En glish "dark [l]s ". It s eems that these lateral variant s are pr oduced w ith great er t ongue dor sum retraction an d low erin g than those in SK laterals . How ev er , the argument for the "darkness " of YB w or d- final laterals r equir es m ore syst ematic res earch . Figure 2. / t ol/ - - > [tol]. t. o. ' stone '. l. In w ord- initial position , un derlying laterals are realized as taps . In spectrograms , they are show n as having a closur e with a strong r elease as in F igure 3 (/ lot o / - > [ odo ]). In som e cas es , a short dur ation of frication noise follow s the release. In the same phon ological position , SK show s n asals , not t aps (/ loto / - > [nodo ]).. (8) 8. Hy un s ook Kang & Jeong - Im Han. F igure 3. / loto / - > [ odo ]. o. d. 'labour '. o. T hese tw o differ ent kinds of phonetic r ealizations of a tap, n amely , that in int erv ocalic position an d that in w ord- initial position , are also found in an ex periment al study by Lee (1999), w here the phonetic realization s of / l/ s in Korean loanw ords w ere ex amined, focusing on their pros odic status . Accordin g to her study , int erv ocalic / l/ s ar e shown to be realized as a sonorant segm ent w ith w eak formant pattern s , w hile initial / l/ s show realization as a short stop or a fricativ e.1) Based on these results , she claim s that sonorants are observed mor e often in the prosodically w eaker position such as the interv ocalic position , while obstruent s ar e more readily found in the prosodically strong position such as the initial position . Now let us turn to the phon etic r ealization of underlying laterals aft er obstruent s .2) First , w e consider the v arious surface realizations of the underlyin g lat erals recor ded in an is olat ed form , w hich are shown in T able 1.3). 1) Lee ' s dat a are m uch m ore com plex th an summ arized here, depending on the pr osodic position in w hich each seg ment occurs . See Lee (1999) for m ore detailed analy sis . 2) Ob struent s in fr ont of underlying later als surface as n asals in this experiment . 3) S om e t okens sh ow both w eak form ant pattern and closure w ith a strong release . T his partially n asalized pattern is classified as "n asalized tap".. (9) Nas alization and T apping in Yon by on Korean. 9. T able 1. Various phonetic realizations of / l/ after obstruents in is olat ed forms (number of tokens ) pros odic context [...C__...]w o r d C]w o r d _ _ _ _ _ _ _ sum. nasal 0 1 1. lateral 0 1 1. tap 26 13 39. nasalized tap 1 3 4. sum 27 18 45. (C=conson ant ) T able 1 show s that most underlying lat erals are realized as a tap or a nas alized tap (43 t okens out of 45), which appear to be similar to stops w ith a very short closure and a strong release. Some surface taps also show a short period of frication noise. T he phonological environment is not show n to affect the phonetic realization of / l/ s : whether the sequence of / obstruent +later al/ in underlyin g represent ation is included w ithin a w ord or separated across the w ord boundary , laterals surface as a tap in most cases . As w e see in T able 1, underlyin g post - con sonantal laterals surface as taps both in w or d- internal position and across the w ord boundary .4) Figure 4 and Figur e 5 show the pattern w here an underlyin g / l/ is realized as a t ap in w ord- internal position (/ hapli/ ) and that across the w ord boundary (/ tapsip#li/ ), respectiv ely . Note that obstruents preceding the un derlying laterals alw ays surface as nasals .. 4) T w o w ord t okens out of 45 surface as nas als, but even in these cases w e can hear and see the str ong release at the beginning of the n asal, w hich can indicate the onset of the second nas al clearly .. (10) 10. Hy un sook Kang & Jeong - Im Han. F igure 4.. h. s. / hapli/ - - > [ham i]. a. m. 'r easonableness '. i. F igure 5.. / tapsip#li/ - - > [tapsim i]. m. i. i. T he surface realization of / l/ s in post - consonantal position is clearly distinguished from that of intervocalic / l/ s or that of w ord- initial / l/ s , even though they are all called as "tap". In Figur e 1, w e already show ed that the interv ocalic t ap has a clear form ant pattern , representing sonorancy . How ev er , the spectrographic data of the post - cons onant al taps as giv en in Figures 4 and 5 rev eal that they are not sonorant , but rather a v ery short v oiced stop, which is thus more similar to the w ord- initial tap. T he proper characterization of thes e tw o or three kinds of v ariants of a tap seems t o require a finer distinction , depending on their phon ological position (and their prosodic position , as suggested in Lee 1999).. (11) Nasalization and T apping in Yonby on Korean. 11. T he surface realization of lat erals following obstruent s recorded in a frame sentence is show n in T able 2. T able 2. Various phonetic r ealizations of / l/ after an obstruent in a sentence (number of t okens ) prosodic context. lateral. [...C__...]w o r d. 3. 0. 19. 5. 27. C]w o r d. 1. 1. 14. 2. 18. 4. 1. 33. 7. 45. _______. sum. t ap. nasalized tap. n asal. sum. T able 2 show s that ev en in a frame sentence, underlying later als are realized as taps or nas alized taps in 92% of tokens . T hus recording style does not affect the realization of un derlying / l/ s as taps . Now w e consider the realization of the underlying / l/ s after an underlyin g alv eolar nasal, which is provided in T able 3 an d T able 4. T able 3 is for lat erals recorded in isolated forms and T able 4 for those in a frame sentence. T able 3. Various phon etic realizations of / l/ after nasals in isolated forms (number of tokens ) pros odic context. nasal. lateral. tap. nasalized t ap. sum. [...N__...]w o r d. 0. 10. 16. 1. 27. N]w o r d. 0. 0. 13. 5. 18. 0. 10. 29. 6. 45. _______. sum (N=nasal). In SK, the sequence of an underlying alv eolar nasal and a lateral within a w ord is r ealized as a lateral geminate. How ev er , even in this case, YB produces an interesting form on the surface: a tap pr eceded. (12) 12 Hy un sook Kang & Jeong - Im Han. by a n asal [n]. T able 3 show s that more than half of the token s are show n t o be taps , nam ely , short v oiced stops . It might be noticed that still a considerable number of underlying alv eolar nasal- later al s equences surface as later al geminates , not as an alv eolar nasal plus a tap sequence. W e suspect that it has to do with av oidance of the consecutiv e coronal gestures . A tap is defined as having a certain part of the tongue strike against the roof of the mouth and then return to its rest position . T he mov ement of the ton gue tip tow ard the alv eolar ridge is crucial for the successful tap gesture. How ev er , if coronal alv eolar contact which is for the alv eolar nas al, / n/ , is immediately before the tap, the tongue will tend to econ omically maint ain the alv eolar contact it began during the / n/ , rather than a nasal and a tap. T his blocks the occurrence of tap in this environment ; instead lateral geminates occur . Similar av oidance of the consecutiv e coronal gestures is also found in English . Banner - Inouye (1995) notes that in the w ord ' int ernational ' , for ex ample, / t/ in the / nt/ s equence is not realized as tap because the tongue tends to economically maintain the alv eolar contact it began during the / n/ , rather than strike again st the r oof of the mouth and then return to its rest position . As for the recording style effect , the frame sentence affects the pattern of the lateral realizations as in T able 4. F ew er taps are show n to occur on the surface of the frame sent ence. T able 4. Various phonetic realization s of / l/ after nasals in a frame sentence (number of t okens ) prosodic contex t. n asal. lateral. tap. nasalized t ap. sum. [...N__...]w o r d. 1. 19. 6. 1. 27. N]w o r d. 0. 2. 15. 1. 18. 1. 21. 21. 2. 45. _______. sum. In w ord- intern al position , sev en tokens out of 27 are realized as taps .. (13) Nasalization and T apping in Yonby on Korean. 13. 19 token s appear as lateral geminates on the surface. When the w ord boundary is posited betw een the nas al and the lat eral, most tokens are realized as taps (16 out of 18). W e conclude that / l/ does not necessarily become [ ] in / n +l/ s equence due to the pr ecedin g alveolar sound. T he surface realizations of the laterals after an alveolar nasal [n] as taps w ord- intern ally and "across the w ord boundary " are shown in Figur e 6 and 7 respectively . Figur e 6.. k. n F igure 7.. s. / k nlo/ - - > [k n o]. i. 'w ork '. o. / hya sin#ly o/ - - > [hy a sin y o]. n. y. ' spice '. o. F rom the acoustic data pres ented in this study , w e can conclude that in YB, 1) the obstruent preceding the underlying lateral surfaces as a. (14) 14. Hy un sook Kang & Jeong - Im Han. nasal; 2) the underlyin g / l/ surfaces as a t ap, not a n asal. T his surface tap show s n o sonorancy , but rather appears as an obstruent . It is also notew orthy that the obstruent tap occur s in phonologically str ong positions , nam ely , w ord- initial an d post - cons onant al. In the following section , w e con sider w hat motiv ates / obstruent +lateral/ to surface as [nas al+t ap], n ot [nasal+nasal].. 3 . A n Optim ality - th e ore tic ac c ou n t of Y on by on / l/ In this section , w e w ill account for the surface realization of / obstruent +lateral/ in YB within the framew ork of Optimality T heory and Correspondence T heory (McCarthy & Prince 1993, 1995 etc.). 3 .1 . Liqui ds in Y onby on D i alec t Various altern ation forms of underlying / l/ are rew ritten in (5) for ease of reference. (5) a. b. c. d. e.. yulæ t ally k tol ky kly loto. [yu æ ] [tally k] [tol] [ky y ] [ odo ]. ' origin ' ' calen dar ' ' ston e ' ' encour agement ' ' labor '. As w e see in (5a), a lateral appears as sonorant tap [ ] in interv ocalic position (cf. F igure 1). As a part of a geminate or in coda position , it appears as [l] as in (5b,c). If it appears after a consonant as in (5d) or at the initial position as in (5e), it appears as (obstruent ) tap [ ]. Int erestin gly , if an obstruent pr ecedes a lateral in the input , it appears as a nasal as w e see in (5d): / k/ chan ges into / / before a surface tap [ ]. W e suggest that an un derlying later al in (5d) triggers the nasalization of a pr ecedin g obstruent , though it appear s as t ap [ ] in the output . We suggest the followin g constraint s (cf. Cho 1997, Davis &. (15) Nasalization and T apping in Yonby on Korean. 15. Shin 1999). (6) a. * : A tap [ ] may n ot appear . b . *s y l [l : A singleton lateral is not licensed as an on set . c. Max [+son ]: A [+sonorant] feature in the input should appear in the output . d. Ident - Onset [sonorant]: T he [±son orant ] feature of an output ons et is identical to the [±son orant ] feature of the corresponding input segm ent . e. Ident - On set [n asal]: T he [nasal] feature of an output onset is identical to the [nas al] featur e of the correspondin g input segm ent . A few w ords are in order for the constraints . First , w e assume that tap [ ] is a member of rhotics (7) (cf. Ladefoged & Maddieson 1990) and that the ex act phonetic nature of tap [ ] in Kor ean depends on phonological environments . F or ex ample, in a lenition environment , it is usually realized as sonorant tap w hereas after a consonant or in the initial position , it is usually realized as tap with a short stop dur ation . (7) Rhotics (r - sounds ) are "segment s in which there is a single or r epeated brief cont act betw een the t ongue and a point on the upper surface of the v ocal tract , i.e. principally apical trills , taps and flaps ." (Ladefoged & Maddieson 1990: 182) Now , let us begin to inv estigate how each of the surface forms of an input / l/ is s elected. F irst , consider laterals in int erv ocalic position as in (8).. (16) 16. Hy un sook Kang & Jeong - Im Han. (8) / yu +læ / - - > [yu æ ] ' origin ' *s y l [l. / yu +læ/ a. yu .læ. Ident - Onset [nas ]. Max [lat]. *!. ☞ b . yu . æ. *. c. yu .næ. *!. *. As w e see in (8), the constraint ranking in (8) w ill select the correct output / yu . æ / . Candidate (a) violat es the fatal *s y l [l and candidate (c) violates fatal Ident - Onset [nas]. On the other hand, candidate (b ) violates only Max [lat], the low est - ranking amon g the r elev ant constraints . T herefore, / yu . æ/ is selected as optimal. If an / l/ - initial morpheme is attached to an obstruent - endin g morpheme, the follow ing con straint tableau will occur . (9) / ky k+ly / - - > [ky / ky k +ly / a . ky kly b . ky. ny. ☞ c . ky. y. d. ky k y. y ]. ' encouragement '. *s y l [l. Ident Ons et [nas]. Max [son]. *. F aith (coda). *! *!. * * *!. *. *. T he most faithful candidate, (a), violat es an undomin ated constraint *s y l [l and thus is eliminated. Candidate (b ) violat es an other fatal constraint Ident - Onset [nas al]. Candidat es (c) and (d) both respect Ident - On set [n asal] but incur a violation of * . Betw een candidat es (c) and (d), candidate (d) respects F aith (Coda)5) but violates constraint Max [s on] since an input / l/ is realized as a short obstruent [ ] and its underlyin g [sonorant] feature is nev er realized either on the segm ent itself or on the adjacent segment [k]. On the other hand, candidate (c ). 5) F aith (Coda ) say s th at a coda seg ment should be faithful to the corresponding input seg m ent .. (17) Nasalization and T apping in Yonby on Korean. 17. violates F aith (Coda) but respects Max [son ] since [+son] feature of an input [l] is realized on the preceding segment . Since the optimal output is candidat e (c), [ky y ], Max [son] should outrank F aith (Coda ). How ev er , the constraint r anking in (9) cannot eliminat e a possible candidate, [ky ty ], as an optimal output . Consider (10). (10) / ky k+ly / a. ky. y. b. ky. ty. SyllCon *s y l [l. Ident On set [n as]. Max [son]. * *. Note that / ky ty / (b ) does not violate any of the constraints giv en in (10) an d thus it should be s elected as optim al. T he difference betw een tw o candidates in (10) is that the input / l/ appears as a tap in (a ) whereas it appear s as an obstruent [t] in (b). How ev er , this difference cannot be captured with any of the con straints in (6). T hen what constraint (s ) dictates candidate (a) t o be an optimal one when compar ed w ith candidate (b )? T hough ' encouragement ' in (10b) is represented as / ky ty / , [t] in [ky ty ] is v ery likely to be pronounced as [d] by voicing assimilation in Korean an d thus , [v oice] cannot be a feature which distinguishes [t] fr om [ ]. T herefore, feature- wise, tap [ ] seems t o be as different from an input [l] as [t] is . Note that phonetically [t] is an obstruent st op an d [ ] is an obstruent stop as w ell in this position . T he difference is that the duration of closure in tap [ ] is shorter than that of the obstruent [t]. T hus , faithfulness constraints betw een input [l] and outputs [ ] and [t] do not seem to help us select the correct output (a). It is possible, of course, that by ranking some markedness constraint s like *t > > * , w e can select the correct output as is shown in (11).. (18) 18. Hy un sook Kang & Jeong - Im Han. (11) / ky k +ly / ☞. * ]s y l *s y l [l. a . ky. y. b . ky. ty. Ident - Onset [nas]. *t. * *. *!. Since *t is more highly ranked than * , / ky ty / w hich violates *t will be elimin ated and / ky y / in (11b ) will be s elected as optimal. How ev er , it has been argued that for South Korean dialect [t] is the most unmarked conson ant (cf. S ohn 1987, etc.) and as far as w e know , there is no argument against the unmarked nature of [t] in the Yonbyon dialect , either . T herefor e, ranking * the low est contradict s this assumption , n amely that [t ] is the m ost unmarked cons onant in Korean . We conclude that neither faithfulness constraints nor markedn ess constraint s help us select the correct output in (11). 3 .2 . S om e A rg um ent f or a f e ature [liquid] In this subs ection , w e will offer the suggestion why tap [ ], not [t], is pr eferred as an allophone of / l/ in the postcon sonantal position . W e propos e that a tap [ ] appears in YB since [ ] shares some phon etic features with [l]. W e suggest that the shared feature betw een these tw o categories is [liquid] in (12). With the feature [liquid], the close relationship betw een [l] and [ ] can be easily defined. Consider the constraint tableau in (13). (12) [liquid]: Certain phonetic and phonological similarities w hich laterals an d taps (mor e generally , rhotics ) share (Ladefoged & Maddieson 1990).. (19) Nasalization and T apping in Yonby on Korean. (13) / ky k +ly / - - > [ky / ky k +ly /. y ] ' encour agement '. *s y l [l. ☞ a. ky. y. b. ky. ty. 19. Max [liquid]. Max [son]. *. F aith (coda) *. * *!. Candidate (b ) violates Max [liquid] and candidate (a) violates * . Since * is r anked low er than Max [liquid], / ky y / is select ed as optim al. T he r anking betw een Max [liquid] an d Max [son ] can be determined if w e ex amine ex amples w ith w ord- initial / l/ . Consider (14). (14) / lo+do / a . lodo b . nodo ☞ c . odo. *s y l [l. Max [liquid]. Max [son]. *. *. *. F aith (coda). *! *!. Candidate (a) violates an un dominated constraint *s y l [l. Candidate (b) incurs a violation of Max [liquid] since an underlyingly distinctiv e segment is neutr alized as [n] on the surface. Can didate (c ), how ev er , violates Max [son] since an underlying [s onor ant] feature is not r ealized either on the segment itself or on a nearby segment in the output . Since candidate [ odo ] is optimal, Max [liquid] should be ranked higher than Max [son] in YB. By adoptin g a feature [liquid], w e are able to explain the surface alternation forms of / l/ in YB.. 4 . Con c lu s ion In this paper , w e hav e considered an optimality - theoretic account for the surface forms of liquid in Yonbyon Korean . In specific, w e hav e show n that a liquid appears as tap [ ] in postconson antal and w or d- initial positions in YB due to the con straint ranking in (15). W e. (20) 20. Hy un sook Kang & Jeong - Im Han. hav e also show n that preliquid obstruent surfaces as nas al due t o the underlyin g [sonorant] feature of an input liquid. (15) YB : *s y l [l > > Max [liquid] > > Ident - On set [s on] > > *C , *## , *. R ef erenc e s Banner - Inouy e, S . (1995). T rills, tap s and s top s in con tras t and variation. Unpublished doctoral dis sertation . UCLA . Los Angeles , CA . Catford, J . (1988). A P ractical I n trod uction to P honetics. Clarendon Pres s : Oxfor d. Cho, Y. (1997). Liquid specification in Korean as geminate alt erability . H arvard S tudies in K orean L ing uis tics VII , 78- 92. Davis , S . & S .- H. Shin . (1999). T he syllable contact constr aint in Korean : An optimality - theoretic an aly sis . J ournal of E as t A s ian L ing uis tics 8:4 , 285- 312. de Jon g , K. (1998). Stress - related variation in the articulation of coda alv eolar stops : flappin g revisited. J ournal of P honetics 26, 283- 310. Jun , J. (2000). Preliquid nas alization . K orean J ournal of L ing uis tics 25:2, 191- 208. Kahn , D. (1976). Sy llable- bas ed g en eraliz ations in E ng lish p honology . Unpublished doctoral dissertation . MIT . Boston , MA . Kim - Renaud, Y- K. (1974). K orean cons onantal p honology , Unpublished doct oral dissert ation . Univ ersity of Haw aii. Manoa , Haw aii. Ladefoged, P . (1993). A Cours e in P honetics. New York : Harcourt , Brace, Jov anovich . Ladefoged, P . & Maddies on . A (1990). T he S ounds of the W orld 's Lang uag es. Blackw ell, Ox ford.. (21) Nasalization and T apping in Yonby on Korean. 21. Lee, S .- H. (1999). On the relationship betw een the phon etic r ealizations of the allophon es of the Kor ean liquid / l/ and their pr osodic status . T he J ournal of the A cous tical S ociety of K orea 18:7, 85- 91. McCarthy , J . & Prince, A . (1993). P rosod ic M orp hology I : Cons train t I nteraction and Satisfaction. Unpublished manuscript . McCarthy , J . & Prince, A (1995). F aithfulness and reduplicativ e identity . Univers ity of M assachus etts Occas ional p ap ers in L ing uis tics: Pap ers in Op tim ality T heory , 249- 384. Amherst : GLSA . Steriade, D. (2000). P aradigm uniformity and the phonetics - phon ological boundary . Pap ers in Laboratory P honology V, 300- 313.. Hyun sook Kang English Language and Culture Division Hany ang Univ ersity Sa- dong , Ansan - si Kyungki- do 425- 791, Korea Phone: 82- 31- 400- 5348 E- m ail: hskang @hanyang .ac.kr Jeong - Im Han Department of English , Konkuk Univ ersity Hw ayang - dong 1, Gw angjin - gu Seoul 143- 701, Korea Phone: 82- 2- 450- 3339 E- m ail: jhan @konkuk .ac .kr Received in September , 2002 Accepted in Nov ember , 2002. (22)