IPA transcription of Korean

This site uses a custom system for transcribing Standard Korean in the International Phonetic Alphabet (IPA). If we were transcribing English, French, or German, we would be better off adhering to the traditional transcriptions which are familiar to many learners of these languages due to their use in teaching materials and reference works. Such broadly agreed systems do not yet exist for Korean, however.

Furthermore, the system currently used by the English-language Wikipedia is unsatisfactory for non-Korean speakers in that it transcribes word-initial “plain” (also “lax” or “lenis”) consonants with simple voiceless symbols in the IPA. This gives the wrong idea about these sounds for those who are used to seeing the same symbols for sounds that are much closer to the “tense” (or “tenuis”) consonants in Korean. For instance, the Wikipedia system would write 불 bul, which starts with a plain sound, as [pul], and 뿔 ppul, which starts with a tense sound, as [p͈ul], employing the Extended IPA diacritic for strong articulation in the latter case. However, [pul] as pronounced in most languages with voiceless unaspirated stops (e.g. French and Italian, though less so with Spanish where the stops tend to be somewhat weaker) would sound almost exactly like 뿔 ppul in Korean, not 불 bul. Not only is the plain sound much weaker (the duration between the closure and the release is quite shorter and is accompanied by a much smaller build-up of air pressure compared to the tense or aspirated sounds), it is also usually weakly aspirated. Since the aspirated series is written with the aspiration diacritic ʰ, e.g. 풀 pul [pʰul], a voiceless symbol in the same system without the aspiration diacritic such as [p] risks being interpreted as a completely unaspirated sound, conforming to the usual practice in transcribing languages where aspiration is contrastive (e.g. Indo-Aryan languages, Chinese, and Thai, where the sound written [p] is completely unaspirated).

This choice of symbols gives the mistaken impression that the Korean tense consonants are much more strongly articulated than voiceless unaspirated consonants in other languages, when in fact they are of comparable strength and it is the Korean plain consonants that are much more weakly articulated. In our system, we write 불 bul as /bul/ [b̥ul] (more on the phonetic transcription below) and 뿔 ppul as /pul/ [pul], adopting the alternative analysis that the tense series is fundamentally regular voiceless and that the plain series consists of fundamentally voiced consonants that undergo devoicing initially.[1]

We use both a phonemic transcription of Korean that uses one IPA symbol for each phoneme, and a phonetic transcription that indicates what we judge to be the most pertinent ways in which the phonemes are pronounced differently depending on the position. We enclose the phonemic transcription in slashes (/ /) and the phonetic transcription in square brackets ([ ]).

We write the non-tense sibilant ㅅ s as /z̥ʰ/ in order to show that it has characteristics of both the plain series and the aspirated series. We use the ring diacritic that indicates devoicing even in the phonemic transcription because it is voiceless in all positions in normal speech. The tense ㅆ ss is simply written /s/.

1. Phonemic transcription

The following symbols are used for the phonemes of Korean:

vowels: a ɛ ʌ e o ø u y ɯ i
glides: j w ɰ
consonants: m b p pʰ n d t tʰ l z̥ʰ s ʣʲ ʦʲ ʦʲʰ ŋ ɡ k kʰ h

The correspondence with hangul (Korean alphabet) orthography is as follows:
a / ㅐ ɛ / ㅑ ja / ㅒ / ㅓ ʌ / ㅔ e / ㅕ / ㅖ je / ㅗ o / ㅘ wa / ㅙ / ㅚ ø / ㅛ jo / ㅜ u / ㅝ / ㅞ we / ㅟ y / ㅠ ju / ㅡ ɯ / ㅢ ɰi / ㅣ i
ɡ / ㄲ k / ㄴ n / ㄷ d / ㄸ t / ㄹ l / ㅁ m / ㅂ b / ㅃ p / ㅅ z̥ʰ / ㅆ s / ㅇ ŋ / ㅈ ʣʲ / ㅉ ʦʲ / ㅊ ʦʲʰ / ㅋ / ㅌ / ㅍ / ㅎ h

There is also the length mark (ː) to indicate lexical length.

Korean dictionaries often indicate the pronunciation of words by giving a phonetic respelling that uses the length mark (ː) when necessary. For instance, 대관령 Daegwallyeong (we use the Revised Romanization of Korean in general except for names and words widely known under a different romanization) may be given the pronunciation respelling [대ː괄-], that is, [대ː괄령] with 령 replaced with a hyphen to conserve space since it doesn’t change from the original spelling. Using the above correspondence list, we can convert [대ː괄령] to /dɛː.ɡwal.ljʌŋ/ (the phonetic transcription is [d̥ɛː.ɡwal.ljʌŋ] as we will see below). Our phonemic transcription is therefore based on the surface pronunciation, but when we want to transcribe the underlying morphophonemic representation that is closer to the Korean spelling, we can enclose them in pipes (| |) and write |dɛː.ɡwan-ljʌŋ| (with the hyphen here representing the morpheme boundary).

Following the phonetic respelling used in Korean dictionaries, we represent stops in syllable codas, where phonation contrasts are neutralized, as plain stops /b, d, ɡ/. This is supported by the normal enchaînement rules in Korean across word boundaries—when stops in syllable codas are followed by a vowel-initial independent word, they surface as plain stops.

In standard pronunciation, the glide /j/ in /je/ may be dropped if it is immediately preceded by a consonant other than /l/ in the underlying syllable. The glide /ɰ/ in /ɰi/ may also be dropped when it doesn’t begin a word. We italicize such glides in order to indicate the optional omission.

시계[-계/-게] sigye /z̥ʰi.ɡje/ [ɕʰi.ɡje]
주의[-의/-이] juui /ʣʲu.ɰi/ [ʣ̥ʲu.ɰi]

When the ㅅ s inserted in writing between the elements of compound words (called 사이시옷 saisiot in Korean) is optionally pronounced as /d/, we write it as a subscript.

시냇가[시ː내까/시ː낻까] sinaetga /z̥ʰiː.nɛd.ka/ [ɕʰiː.nɛk.ka, -nɛt-]
이삿짐[-사찜/-삳찜] isatjim /i.z̥ʰad.ʦʲim/ [i.z̥ʰat.ʦʲim]
촛대[초때/촏때] chotdae /ʦʲʰod.tɛ/ [ʦʲʰot.tɛ]

2. Phonetic transcription

The plain stops /b, d, ɡ/ and the plain affricate /ʣʲ/ are devoiced unless they are surrounded by voiced sounds. So word-initial /b, d, ɡ, ʣʲ/ are devoiced, but these do not merge into either /p, t, k, ʦʲ/ or /pʰ, tʰ, kʰ, ʦʲʰ/ as they are produced with much less force and the following vowel usually begins with a lower pitch. They are somewhat like whispered versions of /b, d, ɡ, ʣʲ/ and can be described as lenis voiceless consonants. The IPA does not provide a specific way of notating lenis voiceless consonants, but we can follow the convention often used for some Germanic languages (e.g. Swiss German dialects) and write these as [b̥, d̥, ɡ̊, ʣ̥ʲ] (however, the similar notation does not imply that the pronunciations of lenis consonants are the same in the different languages, because the exact phonetic nature of the contrast with fortis or tense consonants depends on the language). The ring diacritic which we already saw for /z̥ʰ/ is added below the symbol for the consonant (or above if the symbol has a descender) to indicate that it is voiceless. It is important to note that our use of this diacritic is different from the use sometimes seen in transcribing partially voiced sounds in other languages, such as the voiced consonants in English at the beginning or ending of a word—Korean lenis consonants are voiceless word-initially, not partially voiced.

바다 bada /ba.da/ [b̥a.da]
도끼[도ː-] dokki /doː.ki/ [d̥oːk.ki]
겨울 gyeoul /ɡjʌ.ul/ [ɡ̊jʌ.ul]
재미 jaemi /ʣʲɛ.mi/ [ʣ̥ʲɛ.mi]

The plain stops /b, d, ɡ/ in syllable codas are simply written [p, t, k] in phonetic transcription since they are voiceless and do not need to contrast with other stops in this position. In order to make it explicit that they always have no audible release, we could write [p̚, t̚, k̚] in narrower transcriptions.

앞[압] ap /ab/ [ap]
꽃밭[꼳빧] kkotbat /kod.pat/ [kot.pat]
옷차림[옫-] otcharim /od.ʦʲʰa.lim/ [ot.ʦʲʰa.ɾim]
생각 saenggak /z̥ʰɛŋ.ɡaɡ/ [z̥ʰɛŋ.ɡak]

This also applies when homorganic stops (those that are produced at the same place of articulation) are repeated, in which case they result in geminates.

악기[-끼] akgi /aɡ.ki/ [ak.ki]
같다[갇따] gatda /ɡad.ta/ [ɡ̊at.ta]
합법[-뻡] hapbeop /hab.pʌb/ [hap.pʌp]

Tense consonants between vowels are similar in pronunciation to geminates, which we indicate by inserting an italicized copy of the symbol (or for /ʦʲ/, an italicized copy of [t]) before the syllable break.

아끼다 akkida /a.ki.da/ [ak.ki.da]
그때 geuttae /ɡɯ.tɛ/ [ɡ̊ɯt.tɛ]
오빠 oppa /o.pa/ [op.pa]
아가씨 agassi /a.ɡa.si/ [a.ɡaɕ͈.ɕ͈i]
어찌 eojji /ʌ.ʦʲi/ [ʌt.ʦʲi]

The sibiliants /z̥ʰ, s/ are written [ɕʰ, ɕ͈] in front of /i, j/ and [ʃʰ, ʃ͈] in front of /y/ [y, ɥi]. The use of the previously mentioned Extended IPA diacritic for strong articulation is restricted to the allophones of tense /s/ in our system. It may be tempting to write [ʑ̥ʰ, ɕ] and [ʒ̊ʰ, ʃ] respectively to maintain the pattern of using voiceless symbols for the tense consonants. But the normal [ɕ, ʃ] in other languages tend to be weak enough to be identified with allophones of the non-tenseㅅ s /z̥ʰ/ in Korean, unlike [s] in most languages that tend to be identified with the tense ㅆ ss /s/. This probably has to do with the fact that among sibilants, the “hiss” consonant [s] is the most high-pitched and most prominent while “hush” consonants like [ɕ, ʃ] tend to be lower-pitched and less prominent.

마셔 masyeo /ma.z̥ʰjʌ/ [ma.ɕʰjʌ]
날씨 nalssi /nal.si/ [nal.ɕ͈i]
쉽다[쉽ː따] swipda /z̥ʰyːb.ta/ [ʃʰyːp.ta, ʃʰɥiːp-]

If /d/ comes before /s/ (usually underlying |z̥ʰ|), we treat it as if it were completely assimilated to the following /s/ in the phonetic transcription, producing a geminate sequence [s.s], [ɕ͈.ɕ͈], or [ʃ͈.ʃ͈] (see below for the variable treatment of /s/ depending on the following vowel).

젖소[젇쏘] jeotso /ʣʲʌd.so/ [ʣ̥ʲʌs.so]
헛심[헏씸] heotsim /hʌd.sim/ [hʌɕ͈.ɕ͈im]
뱃사공[배싸-/밷싸-] baetsagong /bɛd.sa.ɡoŋ/ [b̥ɛs.sa.ɡoŋ]

The liquid phoneme /l/ is written as a tap [ɾ] between vowels and after a vowel and before /h/ (which becomes [ɦ] in this position). It is written [l] otherwise. Phrase-initial /l/, which occurs only in loanwords in the South Korean standard, can be either [ɾ] or [l]. We indicate both, with the order depending on which is more suitable for the source language (although this is somewhat arbitrary since [ɾ] and [l] are essentially in free variation in such words).

사람[사ː-] saram /z̥ʰaː.lam/ [z̥ʰaː.ɾam]
일흔 ilheun /il.hɯn/ [iɾ.ɦɯn]
곤란[골ː-] gollan /ɡoːl.lan/ [ɡ̊oːl.lan]
라디오 radio /la.di.o/ [ɾa.di.o, la-]
레몬 remon /le.mon/ [le.mon, ɾe-]

In voiced environments, /h/ is written as [ɦ] in order to describe its voicing. This is more phonetic detail than we would normally go into, but we indicate it here to help learners of Korean because the resulting [ɦ] is often very weak and sometimes dropped completely. This contrasts with phrase-initial [h], which is quite prominent and can even have strong fricative realizations.

지혜[-혜/-헤] jihye /ʣ̥ʲi.hje/ [ʣ̥ʲi.ɦje]
은행 eunhaeng /ɯn.hɛŋ/ [ɯn.ɦɛŋ]
공항 gonghang /ɡoŋ.haŋ/ [ɡ̊oŋ.ɦaŋ]

Front rounded vowels /y/ and /ø/ may also be pronounced [ɥi] and [we] respectively. In fact, in contemporary Standard Korean, these glide-vowel pronunciations are predominant, but we put [y] and [ø] first because the pronunciation we describe is of a more conservative variety that preserves lexical length and the distinction between /ɛ/ and /e/.

위치 wichi /y.ʦʲʰi/ [y.ʦʲʰi, ɥi-]
회의[회ː의/훼ː이] hoeui /høː.ɰi/ [høː.ɰi, hweː-]

In the conservative pronunciation which preserves lexical length, /ʌː/ is more central than /ʌ/, so we write it as [əː].

별빛[별ː삗] byeolbit /bjʌːl.pid/ [b̥jəːl.pit]
경주[경ː-] Gyeongju /ɡjʌːŋ.ʣʲu/ [ɡ̊jəːŋ.ʣʲu]

3. Contemporary popular pronunciation

It suffices in most cases to provide the transcription of the conservative pronunciation, because the contemporary popular pronunciation of Standard Korean can be regularly predicted from it. Here are the main differences:

  • Lexical length distinction is all but lost. To indicate this, simply eliminate the length marks (ː) from the transcriptions based on the conservative pronunciation, while remembering that [əː] becomes [ʌ], e.g. 별빛[별ː삗] byeolbit /bjʌːl.pid/ [b̥jəːl.pit] becomes [b̥jʌl.pit].
  • The distinction between the mid vowels /ɛ/ and /e/ is all but lost. To indicate this, simply write [e] for all cases of /ɛ/, e.g. 재미 jaemi /ʣʲɛ.mi/ [ʣ̥ʲɛ.mi] becomes [ʣ̥ʲe.mi].
  • Traditional front rounded vowels /y/ and /ø/ are exclusively pronounced as glide-vowel sequences [ɥi] and [we], e.g. 위치 wichi /y.ʦʲʰi/ [y.ʦʲʰi, ɥi-] becomes [ɥi.ʦʲʰi] and 회의[회ː의/훼ː이] hoeui /høː.ɰi/ [høː.ɰi, hweː-] becomes [hwe.ɰi].

However, some differences are not regular. Contemporary popular pronunciation may also replace standard forms with dialectal forms, e.g. 네가 ne-ga /ne.ɡa/ is frequently replaced by dialectal 니가 ni-ga /ni.ɡa/ in order to distinguish it from 내가 nae-ga /nɛ.ɡa/ which becomes [ne.ɡa] through the /ɛ//e/ merger. In addition, tensification is more widespread than is recognized in standard pronunciation, e.g. standard 김밥[김ː-] gimbap /ɡiːm.bab/ [ɡ̊iːm.bap] frequently becomes /ɡiːm.pab/ [ɡ̊iːm.pap], or [ɡ̊im.pap] for the majority of Korean speakers who have lost lexical length. In this site, the contemporary population pronunciation of Korean will only be given if the pronunciations used are widespread and present irregular differences from the standard pronunciations.

1. Kim, Mi-Ryoung and Duanmu, San, “‘Tense’ and ‘Lax’ Stops in Korean,” Journal of East Asian Linguistics 13 (2004): 59–104. http://www.jstor.org/stable/20100861 

One thought on “IPA transcription of Korean

  1. Detlef 14 March 2017 / 11:32 pm

    It is called “unaspirated voiceless sound”, not “voiceless unaspirated sound”.


Leave a Reply / 댓글을 달아주세요

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s