Korean phonology

1.^ ㅈ, ㅊ, ㅉ are pronounced [tɕ, tɕʰ, t͈ɕ] in Seoul, but typically pronounced [ts, tsʰ, t͈s] in Pyongyang. Similarly, /s, s͈/ are palatalized as [ɕ, ɕ͈] before /i, j/ in Seoul. In Pyongyang they remain unchanged.

2.^ /m, n/ tend to be denasalized word-initially. /ŋ/ and /l/ cannot occur word-initially in native words.[5]

3.^ /p, t, tɕ, k/ are voiced [b, d, dʑ, ɡ] between voiced sounds but voiceless elsewhere. Among younger generations, they may be just as aspirated as /pʰ, tʰ, tɕʰ, kʰ/ in initial position; the primary difference is that vowels following the plain consonants carry low tone.[6][7] /pʰ, tʰ, tɕʰ, kʰ/ are strongly aspirated, more so than English voiceless stops. The affricates /tɕ͈, tɕʰ, tɕ/ may be pronounced as alveolar ([ts͈, tsʰ, ts~dz] by some speakers, especially before back vowels.

4.^ The IPA diacritic ⟨◌͈⟩, resembling a subscript double straight quotation mark, shown here with a placeholder circle, is used to denote the tensed consonants /p͈/, /t͈/, /k͈/, /t͈ɕ/, /s͈/.[lower-alpha 1] Its official use in the Extensions to the IPA is for strong articulation, but is used in literature for faucalized voice. The Korean consonants also have elements of stiff voice, but it is not yet known how typical that is of faucalized consonants. They are produced with a partially constricted glottis and additional subglottal pressure in addition to tense vocal tract walls, laryngeal lowering, or other expansion of the larynx.
An alternative analysis[8] proposes that the "tensed" series of sounds are (fundamentally) regular voiceless, unaspirated consonants: the "lax" sounds are voiced consonants that become devoiced initially, and the primary distinguishing feature between word-initial "lax" and "tensed" consonants is that initial lax sounds cause the following vowel to assume a low-to-high pitch contour, a feature reportedly associated with voiced consonants in many Asian languages, whereas tensed (and also aspirated) consonants are associated with a uniformly high pitch.

5.^ The analysis of /s/ as phonologically plain or aspirated has been a source of controversy in the literature.[9] Its characteristics are nearest to those of plain stops, as it generally undergoes intervocalic voicing word-medially.[2] It shows moderate aspiration word-initially (in the same way "plain" stops /p/, /t/, /k/ do), but no aspiration word-medially.[2]

6.^ Between vowels, /h/ may either be voiced to [ɦ] or deleted.

7.^ /l/ is an alveolar flap [ɾ] between vowels or between a vowel and an /h/; it is [l] or [ɭ] at the end of a word, before a consonant other than /h/, or next to another /l/. There is free variation at the beginning of a word, where this phoneme tends to become [n] before most vowels and silent before /i, j/, but it is commonly [ɾ] in English loanwords.

8.^ When pronounced as an alveolar flap [ɾ], ㄹ is sometimes allophonic with [d], which generally does not occur elsewhere.

Korean consonants have three principal positional allophones: initial, medial (voiced), and final (checked). The initial form is found at the beginning of phonological words. The medial form is found in voiced environments, intervocalically and after a voiced consonant such as n or l. The final form is found in checked environments such as at the end of a phonological word or before an obstruent consonant such as t or k. Nasal consonants (m, n, ng) do not have noticeable positional allophones beyond initial denasalization, and ng cannot appear in this position.

The table below is out of alphabetical order to make the relationships between the consonants explicit:

All obstruents (stops, affricates, fricatives) become stops with no audible release at the end of a word: all coronals collapse to [t̚], all labials to [p̚], and all velars to [k̚].[lower-alpha 2] Final ㄹ r is a lateral [l] or [ɭ].

ㅎ h does not occur in final position,[lower-alpha 3] though it does occur at the end of non-final syllables, where it affects the following consonant. (See below.) Intervocalically, it is realized as voiced [ɦ], and after voiced consonants it is either [ɦ] or silent.

ㅇ ng does not occur in initial position, reflected in the way the hangeul jamo ㅇ has a different pronunciation in the initial position to the final position. These were distinguished when hangeul was created, with the jamo ㆁ with the upper dot and the jamo ㅇ without the upper dot; these were then conflated and merged in the standards for both the North Korean and South Korean standards.

In native Korean words, ㄹ r does not occur word initially, unlike in Chinese loans (Sino-Korean vocabulary). In South Korea, it is silent in initial position before /i/ and /j/, pronounced [n] before other vowels, and pronounced [ɾ] only in compound words after a vowel. The prohibition on word-initial r is called the "initial law" or dueum beopchik (두음법칙). Initial r is officially pronounced [ɾ] in North Korea. In both countries, initial r in words of foreign origin other than Chinese is pronounced [ɾ].

• "labour" (勞動) – North Korea: rodong (로동), South Korea: nodong (노동)
• "history" (歷史) – North Korea: ryŏksa (력사), South Korea: yeoksa (역사)

This rule also extends to ㄴ n in many native and all Sino-Korean words, which is also lost before initial /i/ and /j/ in South Korean; again, North Korean preserves the [n] phoneme there.

• "female" (女子) – North Korea: nyŏja (녀자), South Korea: yeoja (여자)

Because they may follow consonants in initial position in a word, which no other consonant can do, and also because of Hangul orthography, which transcribes them as vowels, semivowels such as /j/ and /w/ are sometimes considered to be elements of rising diphthongs rather than separate consonant phonemes.

In current pronunciation, /ɰi/ merges into /i/ after a consonant. Some analyses treat /ɯ/ as a central vowel and thus the marginal sequence /ɰi/ as having a central-vowel onset, which would be more accurately transcribed [ȷ̈i] or [ɨ̯i].[11]^:12

Modern Korean has no falling diphthongs, with sequences like /a.i/ being considered as two separate vowels in hiatus. Middle Korean had a full set of diphthongs ending in /j/, which monophthongized into the front vowels in Early Modern Korean (/aj/ > /ɛ/, /əj/ [ej] > /e/, /oj/ > /ø/, /uj/ > /y/, /ɯj/ > /ɰi ~ i/).[11]^:12 This is the reason why the hangul letters ㅐ, ㅔ, ㅚ and so on are represented as back vowels plus i.

The vowel that most affects consonants is /i/, which, along with its semivowel homologue /j/, palatalizes /s/ and /s͈/ to alveolo-palatal [ɕ] and [ɕ͈] for most speakers (but see differences in the language between North Korea and South Korea). As noted above, initial |l| is silent in this palatalizing environment, at least in South Korea. Similarly, an underlying |t| or |tʰ| at the end of a morpheme becomes a phonemically palatalized affricate /tɕʰ/ when followed by a word or suffix beginning with /i/ or /j/ (it becomes indistinguishable from an underlying |tɕʰ|), but that does not happen within native Korean words such as /ʌti/ [ʌdi] "where?".

/kʰ/ is more affected by vowels, often becoming an affricate when followed by /i/ or /ɯ/: [cçi], [kxɯ]. The most variable consonant is /h/, which becomes a palatal [ç] before /i/ or /j/, a velar [x] before /ɯ/, and a bilabial [ɸʷ] before /o/, /u/ and /w/.[16]

In many morphological processes, a vowel /i/ before another vowel may become the semivowel /j/. Likewise, /u/ and /o/, before another vowel, may reduce to /w/. In some dialects and speech registers, the semivowel /w/ assimilates into a following /e/ or /i/ and produces the front rounded vowels [ø] and [y].

As noted above, tenuis stops and /h/ are voiced after the voiced consonants /m, n, ŋ, l/, and the resulting voiced [ɦ] tends to be elided. Tenuis stops become fortis after obstruents (which, as noted above, are reduced to [k̚, t̚, p̚]); that is, /kt/ is pronounced [k̚t͈]. Fortis and nasal stops are unaffected by either environment, though /n/ assimilates to /l/ after an /l/. After /h/, tenuis stops become aspirated, /s/ becomes fortis, and /n/ is unaffected.[lower-alpha 4] /l/ is highly affected: it becomes [n] after all consonants but /n/ (which assimilates to the /l/ instead) or another /l/. For example, underlying |tɕoŋlo| is pronounced /tɕoŋno/.[17]

These are all progressive assimilation. Korean also has regressive (anticipatory) assimilation: a consonant tends to assimilate in manner but not in place of articulation: Obstruents become nasal stops before nasal stops (which, as just noted, includes underlying |l|), but do not change their position in the mouth. Velar stops (that is, all consonants pronounced [k̚] in final position) become [ŋ]; coronals ([t̚]) become [n], and labials ([p̚]) become [m]. For example, |hankukmal| is pronounced /hankuŋmal/ (phonetically [hanɡuŋmal]).[17]

Before the fricatives /s, s͈/, coronal obstruents assimilate to a fricative, resulting in a geminate. That is, |tʰs| is pronounced /ss͈/ ([s͈ː]). A final /h/ assimilates in both place and manner, so that |hC| is pronounced as a geminate (and, as noted above, aspirated if C is a stop). The two coronal sonorants, /n/ and /l/, in whichever order, assimilate to /l/, so that both |nl| and |ln| are pronounced [lː].[17]

There are lexical exceptions to these generalizations. For example, voiced consonants occasionally cause a following consonant to become fortis rather than voiced; this is especially common with |ls| and |ltɕ| as [ls͈] and [lt͈ɕ], but is also occasionally seen with other sequences, such as |kjʌ.ulpaŋhak| ([kjʌulp͈aŋak̚]), |tɕʰamtoŋan| ([tɕʰamt͈oŋan]) and |wejaŋkanɯlo| ([wejaŋk͈anɯɾo]).[17]

1. Velar obstruents found in final position: ㄱ g, ㄲ kk, ㅋ k
2. Final coronal obstruents: ㄷ d, ㅌ t, ㅅ s, ㅆ ss, ㅈ j, ㅊ ch
3. Final labial obstruents: ㅂ b, ㅍ p

The resulting geminate obstruents, such as [k̚k͈], [ss͈], [p̚pʰ], and [t̚tɕʰ] (that is, [k͈ː], [s͈ː], [pʰː], and [tːɕʰ]), tend to reduce ([k͈], [s͈], [pʰ], [tɕʰ]) in rapid conversation. Heterorganic obstruent sequences such as [k̚p͈] and [t̚kʰ] may, less frequently, assimilate to geminates ([p͈ː], [kːʰ]) and also reduce ([p͈], [kʰ]).

These sequences assimilate with following vowels the way single consonants do, so that for example |ts| and |hs| palatalize to [ɕɕ͈] (that is, [ɕ͈ː]) before /i/ and /j/; |hk| and |lkʰ| affricate to [kx] and [lkx] before /ɯ/; |ht|, |s͈h|, and |th| palatalize to [t̚tɕʰ] and [tɕʰ] across morpheme boundaries, and so on.

Hangul orthography does not generally reflect these assimilatory processes, but rather maintains the underlying morphology in most cases.