Appendix:Koreanic reconstructions

This appendix discusses trends in Koreanic historical phonology up to the fifteenth century, with the goal of allowing readers to understand the probable Old Korean reflexes of Middle Korean forms.

Note that "Old Korean" here is used interchangeably to refer to Old Korean proper, the attested language of a limited corpus of texts from the sixth to thirteenth centuries, and Proto-Koreanic, a reconstructed language based principally on internal evidence from fifteenth-century Middle Korean. While the two appear to reflect largely the same language at various stages of change, Proto-Koreanic reconstructions (especially ones that involve pitch accent analysis) often go back significantly before the eighth century and are thus much more archaic than the bulk of surviving Old Korean texts, which date from the early second millennium.

All transliterations are given in the Yale Romanization of Korean, which is standard in linguistics.

In the previous century, Lee Ki-Moon proposed a "Korean Great Vowel Shift" in the thirteenth and fourteenth centuries based on alleged inconsistencies in vowel quality in Middle Mongol loanwords into Middle Korean. While still found in some sources, especially ones by Korean scholars, the hypothesis was motivated by an apparent misunderstanding of Middle Mongol phonology^[1] and is largely rejected in modern scholarship.^[2]

In 2007, Itō Chiyuki reconstructed an alternative and more plausible vowel system for Old Korean of the eighth and ninth centuries, based on the Sino-Korean reflexes of Middle Chinese. This involves */ɛ/ > /ə/ and */ə/ > /ʌ/, but no other major change in the vowel system.^[3] Both shifts have been contested, given phonological difficulties in the resulting vowel harmony system.^[4]

Several recent analyses, including more recent work by Itō, also suggest that Middle Korean ㅕ (ye /jə/) may originally have been an eighth vowel, usually reconstructed */e/. This is particularly because the seven-vowel system of reconstructed Old Korean otherwise produces an unusual motivation for Middle Korean vowel harmony.^[4] American linguist Mark Miyake also gives Sino-Korean evidence for OK */e/ > MK /jə/.^[5]

In any case, it appears that there have been at most only two or three changes in the vowel inventory between the eighth and fifteenth centuries. The system in Whitman 2015 is given below.

The prestige Seoul dialect of Middle Korean did not permit the diphthongs ᆢ (yo /jʌ/) and ᆜ (yu /jɨ/), but the fifteenth-century source Hunmin jeong'eum haerye states that these survived in some outlying dialects. yo /jʌ/ is actually still preserved in the divergent and nearly extinct Jeju language, which is closely related to but mutually unintelligible with mainland dialects. Jeju-Korean cognates show that Middle Korean merged yo into ye.^[6]

Thus, Jeju evidence can be used to reconstruct Old Korean */jə/. However, due to heavy superstrate influence, many Jeju forms originally with yo have since been replaced by ye. In these cases, the South Gyeongsang dialect, which merged yo with ya in the first syllable (although this too has often been undone by superstrate influence), can sometimes be used to reconstruct */jə/.^[7]

The nature or ultimate fate of yu /jɨ/ is not clear because modern dialects do not preserve it.^[6] While many dialects do have yu, this is clearly a secondary (and often non-phonemic) development from ye. Meanwhile, there is no evidence of Old Korean or any of its descendants ever having had yi /ji/.^[6]

Between Old and Middle Korean, a number of final *-wo became -woy. This may be due the fusion of the common nominal suffix -i into the stem.

The limited data strongly suggests that Middle Korean initial consonant clusters, such as ps-, st-, pc-, pst-, and so forth, are a secondary development that emerged after the twelfth century, due to the vowels between the consonants dropping out.^[8] These vowels are conventionally reconstructed as u */ɨ/ or o */ʌ/, the two Korean "minimal vowels" particularly vulnerable to deletion.^[9] (The choice of which exact minimal vowel is given is based on Middle Korean vowel harmony, although this probably did not exist in Old Korean,^[10] or at least not in the same form as in Middle Korean.) Thus we may posit the following reconstructed Old Korean forms for Middle Korean consonant clusters, with the caveat that the minimal vowels given are tentative:

The Middle Korean aspirates ㅋ (kh), ㅌ (th), ㅍ (ph), and ㅊ (ch), which are /kʰ/, /tʰ/, /pʰ/, and /t͡sʰ/ respectively, behave as consonant clusters of hC or Ch in Middle Korean, where C is the relevant consonant. They are ultimately believed to arise from Old Korean sequences involving */k/ or */h/.^[11] The Sino-Korean evidence and variation in the orthography of place names suggests that * and * existed as allophones or in free variation with unaspirated equivalents in the eighth and ninth centuries, and that * may or may not have existed. * was the last aspirate to appear, and almost certainly did not exist in the first millennium.^[12] In any case, we may posit the following Old Korean reconstructions for aspirates:

The prestige Seoul dialect of Middle Korean lenited the Old Korean plosives /p/, /s/, /k/, and /t/ in some, but not all, inter-sonorant environments to the fricatives /β/, /z/, /ɣ~ɦ/, and (the last phonemically /l/), respectively.^[13]

/β/ is romanized as <W> and /ɣ~ɦ/ as <G> in Yale.

In the modern Seoul dialect, remains, /β/ has weakened further to /w/ or simply a rounding of the vowel, and /z/ and /ɣ/ have been lost entirely. This process did not occur in the Yukjin and Gyeongsang dialects, which even today regularly preserve /p/ and /s/, and often /k/ as well.^[13]

Because of the irregularity of lenition, some scholars have chosen to see the lenited phonemes as having been different from their non-lenited equivalents even in Old Korean, usually with a phonemic voicing contrast that applied only in inter-sonorant position. However, there is direct phonogramic evidence that Old Korean did not distinguish the two sets of sounds, such as 沙音 (sam-) > 삼〯 (sǎm-) but 沙 (-sa) > ᅀᅡ〮 (-zá).^[14] There are also many compounds where lenition has obviously occurred.^[15] Even if one explains these by positing a series of voiced phonemes that surfaced word-initially as voiceless ones in Middle Korean, there are still serious difficulties, such as why 沙 (MC srae|sraeH) would have been chosen to represent putative */za/ or attestations of lenition in Sino-Korean words where the Middle Chinese form was not voiced. A phonemic split seems therefore the more likely theory.

The reason why lenition occurred in some environments but not others is unknown. Gyeongsang and Yukjin have fully lost whatever contrasts prompted lenition in the first place, so dialectal evidence is not helpful here. The issue is further complicated by the fact that the same Middle Korean morpheme was sometimes lenited and sometimes not lenited. This may suggest that lenition ceased to be productive at some point.

Two major hypotheses on lenition are presented below. However, both fail to satisfactorily explain all cases of lenition or lack thereof.

• Samuel E. Martin proposes that lenition occurred word-internally at the onset of a high-pitch syllable, but only if the subsequent vowel was o or u. Apparent exceptions to lenition are the result of a consonant cluster which was then simplified. Apparently exceptional lenitions are the result of a subsequent dropping of o or u, which, as discussed above, are minimal vowels vulnerable to deletion.^[16]
• Alexander Vovin proposes that lenition occurred intervocally in most environments (except at verbal morpheme boundaries when preceded by a minimal vowel), and that the non-lenited plosives were originally a consonant cluster of NC where N stands for a nasal, or alternatively liquid, consonant of unknown quality. The nasal or liquid was then dropped, producing apparent exceptions to lenition. Apparently exceptional lenitions are the result of a subsequent loss of a vowel, reconstructed as a minimal one. Vovin's key evidence is that many Korean dialects have nC where Middle Korean only has C, although the majority view is that Korean dialects underwent sporadic insertion of /n/ before fricatives; several cases can be clearly demonstrated to be secondary developments.^[17]

Alternative reconstructions are given below.

A number of Middle Korean nouns or verb stems end in /-h/. This reflects a lenition of Old Korean coda /-k/. In many modern dialects—often including Seoul itself, which must have borrowed these forms from non-prestige dialects—the cognate of Middle Korean /-h/ forms involve /-k/, /-ŋ/, or /-j/, testifying to the velar nature of the ancestral form.^[18]

Besides final /-h/, the Gyeongsang dialect often has inter-sonorant /-k-/ where Middle Korean has /-h-/. While the Gyeongsang dialect has sometimes been seen as having innovated /-k-/, Shin Seung-yong presents several arguments for considering the modern Gyeongsang form as more conservative, and the Middle Korean form as the result of a wider trend of inter-sonorant lenition that also led to the phonemes discussed above.^[19] Apparently independently, Vovin proposed similar arguments for seeing Gyeongsang /-k-/ as the conservative form, and /-h-/ as the result of intervocalic lenition.^[20]

It appears that the lenition of */k/ to either /h/ or /ɣ/ was phonologically conditioned. Vovin notes that /ɣ/ is only attested after /l/, /z/, or /i~j/ and before a vowel, whereas /h/ appears freely in all parts of the word. However, /h/ sometimes appears where /ɣ/ could also have appeared. The reason for this is still undetermined.^[20]

A number of Middle Korean nouns and verb stems have an exceptional allomorphy. These take the form CVCV in isolation (for nouns) or before the dictionary citation suffix -ta, but CVCk-, CVCɣ-, or CVCl- before most suffixes:

These forms are believed to originate from an Old Korean CVC(o/u)k or CVC(o/u)l form. The reason they came to undergo this unusual allomorphy is not clear, although it must be connected to the fact that they uniformly have an unusual low-low pitch accent in Middle Korean.^[21]

However, there are words such as 그릏 (kùlùh, “stump”) which do not undergo this allomorphy despite having low-low pitch. It is possible that words such as 그릏 were originally trisyllabic (e.g. *kulukV), with /h/ undergoing intervocalic lenition and the final vowel then dropping, but this is still unclear.

Old Korean had two liquid phonemes, one written with 尸 and part of the irrealis gerund, the other written with 乙 and part of the object marker. Given the reconstructed Chinese readings of these characters, 尸 (OC *hli) and 乙 (MC 'it), it is usually thought that the first was roughly */l/ and the other was a rhotic, here written for convenience's sake as */ɾ/. The two phonemes may have merged as late as after the thirteenth century.^[23] In the case of nouns, the distinction between */l/ and */ɾ/ is largely unrecoverable except in the fortunate cases where direct Old Korean attestation survives or where the word has apparently been loaned to Manchu, which distinguishes the two.

However, the distinction may still survive in verb stems. S. Robert Ramsey identifies two types of final /-l/ in Middle Korean verbs. /l₁/ creates a low pitch on single-mora stems, which is associated with obstruents in Middle Korean. /l₂/ creates a high pitch, which is associated with sonorants. Most verb stems that end in /l/, however, have a bimoraic stem (impliying an earlier bisyllabic form) where these forms cannot be distinguished. Ramsey suggests that even for these verbs, */l₁/ and */l₂/ can be discerned if their causative form is attested, because /l₁/ seems to have taken the /-ɣi/ allomorph of the causative suffix while /l₂/ may have taken the /-i/ form.^[24] In Modern Korean, these allomorphs usually correspond to /-li/ and /-i/ respectively.

Because the Middle Korean irrealis gerund appears to behave as an obstruent, Ramsey suggests that /l₁/ was originally the phoneme represented by 尸, hence /l/, and that /l₂/ was represented by 乙, hence /ɾ/.^[24]

Old Korean has had some form of pitch since at least the eighth century, as Middle Sino-Korean regularly reflects Middle Chinese tone. The reconstructions discussed below reflect an idealized ancestor of the Korean language at a chronologically uncertain point, although many of the phonological processes involved continued to be relevant into the fifteenth century.

The Middle Korean "rising pitch", which is really a bimoraic sequence of a low and high pitch in a single syllable, implies an earlier bisyllabic form. The bisyllabic form is not predictable from the Middle Korean reflex alone, however, and reconstruction is difficult without a direct attestation of some sort. The inverse is not true; not all Old Korean bisyllabic forms that contracted to a single Middle Korean syllable have rising pitch.^[25]

Note also that Sino-Korean words have rising pitch corresponding to both the departing and rising tone of Middle Chinese, and hence some words with rising pitch may actually be nativised Sino-Korean words instead of originating in Old Korean bisyllabic forms. This has been suggested for 죠ᇰ〯 (cywǒng, “servant, slave”), which may be from 從 (MC dzjowngH, “follower”).^[26]

Middle Korean has six accentual classes for monosyllabic verbal stems. S. Roberts Ramsey was among the first to identify these classes and propose the likely pre-Middle Korean forms of three of them: Class 1, Class 2, and Class 6.^[27]

Class 1 verb stems consist of a single, always low-pitch syllable which ends in a non-lenited obstruent, /l₁/, or /j/. Class 6 verb stems consist of a low-pitch syllable before vowel-initial suffixes but takes the rising pitch before a consonant-initial suffix. They end in a lenited obstruent, a nasal consonant, /l₁/, /l₂/, or /j/. Ramsey takes only the obstruent-final stems into consideration in his reconstruction.^[28]

Class 2 verb stems mostly consist of two dissimilar types of stems which share a conjugational paradigm in which they always take high pitch. The first type consists of one-syllable stems that end in sonorant consonants: /m/, /l₂/, or /j/. The second type consists of one-syllable stems whose onset is always a consonant cluster or an aspirate consonant, whose vowel is always the minimal vowel ó or ú, and which have no coda consonant. Ramsey also noted that unlike Class 2 verb stems, very few Class 1 or 6 verb stems had a minimal vowel.^[29]

Ramsey initially concluded that at least for the classes he discussed, Old Korean verb stems could only end in sonorants (/m/, /l₂/, /j/, or a vowel) and that the final syllable of the stem always carried high pitch, while the first syllable carried low pitch. Thus:

• Sonorant-final Class 2 verb stems were unchanged from Old Korean.
• Vowel-final Class 2 verb stems were formed from an Old Korean bisyllabic stem with minimal vowels in both syllables. The first minimal vowel was lost to produce the initial consonant cluster or aspirates, and the pitch of the second, surviving syllable was retained.
• Obstruent-final Class 1 and Class 6 verb stems were both formed from an Old Korean bisyllabic stem with a minimal vowel in the second syllable. The minimal vowel was lost and the pitch of the first, surviving syllable was retained. Class 1 verb stems did not undergo lenition, while Class 6 verb stems did; the latter also retained the originally bisyllabic form as a rising pitch. Ramsey believed that a phonemic voicing distinction, existing even prior to lenition, was responsible for this.
• Sonorant-final Class 6 verb stems were also bisyllabic with a minimal vowel in the second syllable, explaining the paradigmatic match.

Ramsey later revised his position to one more akin to Martin's, accepting that Middle Korean Class 1 verbs were actually unchanged from Old Korean, but Vovin continues to follow a reconstruction more akin to Ramsey's original position, positing bisyllabic ancestors for Class 1 verbs.

Almost all Middle Korean verb stems of (C)V shape, such as 오 (wo-, “to come”), ᄒᆞ (ho-, “to do”), and 주 (cwu-, “to give”), belong to Class 3/4. These stems had a highly irregular pitch pattern, apparently morphologically conditioned, in which they took low pitch before most word-final verbal suffixes (and a small number of word-internal suffixes apparently derived from word-final ones), but took high pitch before most word-internal suffixes as well as in all compounds. The reason for this is unknown, although at least one of the Class 3/4 verbs was actually CVC at an earlier stage: *ho-, which retains its original form *hoy- in the infinitive ᄒᆞ〮야〮 (hóy-á). Vovin suggests that most Class 3/4 verbs originally ended with a nasal consonant, but evidence for this is not conclusive.^[31]

As for the other two classes, a class of irregular low-pitch stems (Class 7) consist mostly of compounds. Ramsey had no particular explanation for a small class of fixed rising-pitch stems (Class 5).

Around five obstruent-final stems—all but one with initial consonant clusters, and all but one with minimal vowels—belong to Class 2. Ramsey apparently did not offer an explanation for them either:

• ᄠᅳᆮ〮/ᄩᅳᆮ〮 (pt(h)út-, “to pluck”)
• ᄧᅳᆽ〮 (pcúc-, “to tear”)
• ᄎᆞᆽ〮 (chóc-, “to look for”)
• 슺〮 (súc-, “to clean”)
• 치ᇦ〮 (chíW-, “to be cold”)

Most Middle Korean nouns have high pitch on their final, or only, syllable. It is sometimes speculated that Proto-Koreanic did not have phonemic pitch in native words, with the final syllable being automatically emphasized.^[32] (The earliest phonogramic evidence is ambiguous on the existence of pitch, although the existence of pitch in Sino-Korean shows that it must have formed by the eighth century at the latest.) Itō Chiyuki gives suggestive evidence that Middle Korean low-pitch monosyllabic nouns should be reconstructed as having originally been multisyllabic, with a final minimal vowel having dropped out as with Class 1 verb stems:^[33]

• Low-pitch monosyllabic nouns tend to feature final consonant clusters or aspirates
• Low-pitch nouns tend to take the ᄋᆡ〮 (-óy) or 의〮 (-úy) allomorph of the locative marker rather than the conventional 애〮 (-áy) or 에〮 (-éy), which suggests a historical final -ó/ú which dropped out
• Low-pitch monosyllabic nouns often end in /-ŋ/, which does not occur word-initially and does not have an associated Old Korean phonogram, and may be a secondary development from */-nk(o/u)/^[34]

Note that a number of low-pitch nouns appear to be nativised Sino-Korean forms or other loans from various sources, and thus the same caveat applies as with rising-pitch forms.

This raises the question of why certain disyllabic nouns became low-pitch and others became rising-pitch. Both John Whitman and Itō suggested that the consonantal leniting factor, which they identified as per Ramsey's original analysis of the verb stems, determined the pitch. CV̀CV́ resulted in CV̌C, just as Ramsey originally suggested for Class 6 verbs, while CV̀CV́ became CV̀C. But given evidence against a distinction in Old Korean phonograms, it is difficult to justify this hypothesis. Martin's vowel-conditioned leniting environment is incompatible with Whitman and Itō's hypothesis, while Vovin's suggestion of nasals as the lenition-blocking factor, if combined with the disyllabic hypothesis for low pitch, would require positing the recent loss of an immense number of nasals which left little trace in Middle Korean or the modern dialects.

All multisyllabic nouns ending in the high pitch always retained the high pitch. But Middle Korean high-pitch monosyllabic nouns were also divided into two classes: a class which always retained its high pitch, and a class which became low before the locative suffix.

Itō argues that the permanently fixed class of high-pitch monosyllables was also originally bisyllabic or bimoraic (similar to Class 2 verb stems). Thus, the pitch behavior of forms such as ᄲᅧ〮 (spyé) can be easily explained through historical *sùpyé. For forms without clusters or aspirates, where a bisyllabic form is difficult to justify, Itō suggests two possibilities, both speculative. The first is that an initial minimal vowel dropped out so that a V̀CV́C form became CV́C, given that only one Middle Korean word, 읏듬 (ùstùm, “leading, foremost”), begins with a minimal vowel, potentially suggesting a process of initial minimal-vowel dropping. Itō also notes that Middle Korean had a restriction against rising pitch on a minimal vowel, and hence that forms such as 믈〮 (múl) with a fixed high pitch may have formed from processes that would otherwise have produced a rising pitch.^[33] Possible justification for this includes ᄃᆞᆯ〮 (tól), which is attested in Old Korean with bisyllabic 月羅理 (*TOlal-i) (where the -i is believed to be a nominative marker).

However, analyses along the lines of Itō's are not yet generally accepted.

• Eom, Ik-sang (1994) “Aspiration and Voicing in Old Sino-Korean Obstruents”, in Young-Key Kim-Renaud, editors, Theoretical Issues in Korean Linguistics, Center for the Study of Language, Stanford University Press, →ISBN, pages 405—418
• 伊藤智ゆき (Itō Chiyuki) (2007) 朝鮮漢字音研究 [Sino-Korean Phonology], Kyūkoshoin, →ISBN
• Itō, Chiyuki (2013) “Korean accent: Internal reconstruction and historical development”, in Korean Linguistics‎, volume 15, number 2, →DOI, pages 125—194
• Ko, Seongyeon (2013) “The end of the Korean vowel shift controversy”, in Korean Linguistics‎, volume 15, number 2, →DOI, pages 1—30
• 곽충구 (Kwak Chung-gu) (2012) “육진방언의 음성과 음운사 [yukjinbang'eonui eumseonggwa eumunsa, The phonology and phonological history of the Yukjin dialect]”, in Bang'eonhak, volume 16, pages 121-154
• Lee, Iksop, Ramsey, S. Robert (2000) The Korean Language, State University of New York Press, →ISBN
• Lee, Ki-Moon, Ramsey, S. Robert (2011) A History of the Korean Language, Cambridge University Press, →ISBN
• Martin, Samuel E. (1996) Consonant Lenition in Korean and the Macro-Altaic Question, University of Hawaii Press, →ISBN
• Miyake, Mark (2017) “Fishy Rhymes: Sino-Korean Evidence for Earlier Korean /*e/”, in Alexander Vovin, William McClure, editors, Studies in Japanese and Korean Historical and Theoretical Linguistics and Beyond, BRILL, →ISBN, pages 405—418
• 백두현 (Paek Du-hyeon) (1994) “이중모음 'ᆢ'의 통시적 변화와 한국어의 방언 분화 [Diachronic change of the diphthong yo and Korean dialectal divergence]”, in Eomun ronchong, volume 28, pages 59—94
• 박석문 (Park Seok-mun), 서장국 (Seo Jang-guk) (2015) “국어 어말음 ‘ㄱ’의 교체와 변화에 대하여 [gugeo eomareum ‘g’ui gyochewa byeonhwa'e daehayeo, On the alternation and change of the Korean coda consonant *-k]”, in Ban'gyo Eomun Yeon'gu, volume 41, pages 315–334
• Ramsey, S. Robert (1991) “Proto-Korean and the Origin of Korean Accent”, in William G. Boltz, Michael C. Shapiro, editors, Studies in the Historical Phonology of Asian Languages, John Benjamins, →ISBN, pages 215—239
• Ramsey, S. Robert (2004) “Accent, Liquids, and the Search for a Common Origin for Korean and Japanese”, in Japanese Language and Literature, volume 38, number 2, pages 339–350
• 신승용 (Shin Seung-yong) (2003) “/k/ > /h/ 變化에 對한 考察 [Study on the /k/ > /h/ shift]”, in Gugeohak, volume 41, pages 93—122
• Vovin, Alexander (2003) “Once Again on Lenition in Middle Korean”, in Korean Studies‎, volume 27, pages 85—107
• Vovin, Alexander (2010) Koreo-Japonica: A Re-evaluation of a Genetic Origin, University of Hawaii Press, →ISBN
• Vovin, Alexander (2013) “Mongolian names for 'Korea' and 'Korean' and their significance for the history of the Korean language”, in Sung-Ock Sohn, Sungdai Cho, Seok-Hoon You, editors, Studies in Korean Linguistics and Language Pedagogy: Festschrift for Ho-min Sohn‎, Korea University Press, →ISBN, pages 200—205
• 魏国峰 (Wei Guofeng) (2017) 고대 한국어 음운 체계 연구: 전승 한자음을 대상으로 [godae han'gugeo eumun chegye yeon'gu: jeonseung hanja'eumeul daesang'euro, Old Korean Phonology: Focusing on Inherited Sino-Korean], Taehaksa, →ISBN
• Whitman, John (2012) “The relationship between Japanese and Korean”, in Nicolas Tranter, editors, The Languages of Japan and Korea‎, Routledge, →ISBN, pages 24—38
• Whitman, John (2015) “Old Korean”, in Lucien Brown, Jaehoon Yeon, editors, The Handbook of Korean Linguistics‎, Wiley, →ISBN, pages 421—483