Hokkien phonology

Hokkien is a Southern Min language spoken in southern Fujian and Taiwan. It has one of the most diverse phoneme inventories among Sinitic languages.

Along with other Min languages, which are not directly descended from Middle Chinese, Hokkien is of considerable interest to historical linguists for reconstructing Old Chinese.

This article uses Pe̍h-ōe-jī and IPA for transcription.

Syllables
A syllable in Hokkien consists of an initial, a final, and a tone.

In total, Hokkien uses around 800 toneless combinations of initials and finals, and around 2250~2450 total syllables with tones (counting only attested meaningful syllables, not all theoretically possible combinations).

The number of syllables in Hokkien is far greater than in any other Sinitic language. For comparison, Cantonese and Hokchew use around 1800 toned syllables, Beijing Mandarin has 1350 syllables, and Suzhou Wu has only 1100 syllables.

Initials
Hokkien has aspirated, unaspirated as well as voiced consonant initials.

A total of 15 initials (or 14, in dialects with /dz/ merged with /l/) are used in Hokkien. This number does not include the three nasal consonants ([m], [n], [ŋ]), which are usually considered allophones of the non-nasal voiced initials (e.g. is analyzed as /bĩã⊇/, but pronounced as [mĩã²²]). This allophony also leads to a notable feature of the Hokkien accent in other languages, such as Japanese or Mandarin, when the nasal sounds like [m] are denasalized into non-nasal voiced consonants like [b].

Syllables starting with vowels or approximants (//, //) are considered to have the zero initial /∅/ (which can be articulated as a glottal stop [ʔ]).

Affricates and fricatives (,, ~, ) are palatalized before , becoming , , ~,.

The phoneme /l/ in Hokkien has many possible articulations. It ranges between [d], [ɗ], [l] and [ɾ]. Its nasal allophone is uniformly described as [n]. In directly-borrowed loanwords (i.e. those not borrowed orthographically via Mandarin or Japanese), foreign /d/ may be represented with Hokkien /l/:  (from Dutch duit via Malay),  (from English),  (from Malay dukun),.

As a phoneme, /dz/ (also realized as [z]) is found predominantly in southern dialects of Hokkien. In many northern dialects, such as urban Quanzhou and Amoy, it has merged with /l/. This merger is still incomplete in some peripheral northern dialects, such as those of Tong'an or Yongchun, where /dz/ is reported to be present in some localities, in the speech of older speakers, or in a limited set of words (usually the more common ones, such as or ). Some Southern and Western Hokkien dialects merge /dz/ with /g/ (among the Southern Hokkien dialects, such merger is found in the north of Zhangpu, in Taichung, or, under Western Hokkien influence, in Hua'an).

While generally preserving /dz/, Zhangzhou Hokkien may still merge it with /l/ in some words, usually before /-u-/. E.g., characters like jūn 'wet',  joán 'soft',  jǒa 'how much' are now pronounced lūn, loán, lǒa even in Zhangzhou, although older Hokkien dictionaries record them with /dz/.

Finals
A final in Hokkien consists of a nucleus (a vowel, a diphthong, or a syllabic nasal /m̩/ or /ŋ̍/), with an optional medial (/i/ or /u/, some dialects also allow /ɯ/) and coda (/m/, /n/, /ŋ/, /p/, /t/, /k/, /ʔ/).

There are around 80 to 90 finals in Hokkien. The exact number can vary depending on the specific dialect, as well as the inclusion of marginal finals from onomatopoeia and contractions. Western branch of Hokkien (the Lengna dialect) has around 60 finals, due to the loss of the /-ʔ/ coda.

In the tables below, rare rimes used in a small number of words are shaded. Finals used only in the northern or southern dialects of Hokkien are marked with N and S respectively.

Open-vowel finals
Finals with the coda /-ʔ/ are considered to be the checked tone counterparts for the open-vowel finals.

The vowel -a is usually.

The vowel ee is only found in Southern dialects of Hokkien.

The final -o͘ is realized as a diphthong -ou  in many rural Zhangzhou dialects (in Pinghe, Zhangpu, Yunxiao, Chawan, etc), similarly to Teochew. Likewise, many of those dialects diphthongize -e into -ei. Some dialects, such as Zhangpu, realize them as triphthongs and. Changtai dialect uniquely pronounces general Hokkien -o͘ as. However, it still uses the vowel, mainly in place of general Hokkien.

-er and -ir  are found in many Northern Hokkien dialects, including Quanzhou and Tong'an. In Amoy, Jinjiang and among some younger urban Quanzhou speakers, -er is merged with /e/. Similarly, the final -ir may be merged with /i/ or /u/ in those dialects. These two finals are not found in Southern dialects of Hokkien.

Chawan dialect in Zhangzhou also has -er and -ir. Thus, Chawan distinguishes the following finals: -ee, -er , and -ei (latter corresponding to urban Zhangzhou -e ). The dialects of Yunxiao and rural Dongshan are notable for having -ee, -e (in place of Chawan -er ) and -ei  (in place of urban Zhangzhou -e ) as distinct finals.

In Northern Hokkien dialects where the final -er is present, it is generally realized as [ɤ̟], and -o  is realized as [o̜]. In dialects where -er is absent, [ɤ̹] is a possible realization of -o.

Finals with codas
The exact realization of and  varies throughout the dialects. For most of them, they are described as [iɪŋ]/[iɪk] or [iəŋ]/[iək]. Many Hokkien dialects in rural Zhangzhou and SEA have them as [eŋ] and [ek]. In urban Quanzhou and Jinjiang, is merged with, but  is preserved.

In dialects with -erng and -ng, the two finals are often confounded. Likewise, -m and -erm  may be used intechangeably. -m is assigned mainly to the syllables with zero initial, e.g. in 毋 m̄ "not", 媒 m̂, hm̂ "matchmaker", 梅 m̂ "plum", 莓 m̂ "berry", etc.

Hui'an dialect merges -im and -iam  into, or -ip  and -iap  into , etc, and thus it has the following rhymes not found in other dialects: , , , ,.

While -er and -ir  are phonemically distinct as standalone finals, they are not distinct in compound finals, and /əm/, /əŋ/, /ən/, /ək/, /ət/ may be described as /ɯm/, /ɯŋ/, /ɯn/, /ɯk/, /ɯt/ as well (and they are usually described as such for Teochew). In Quanzhou operatic pronunciation, this sound is pronounced as [-ɯə-] in compound finals (except for the final /ək/, which is realized as [-ɯak], influenced by the urban Quanzhou -erk/-ek/-iak merger). These rimes also share some phonological constraints with rather than, e.g. they do not cooccur with labial initials (so */pɯ/, */pʰək/~/pʰɯk/ or */mət/~/mɯt/ are not valid syllables in Hokkien, while /pə/ or /pʰə/ are possible).

Rimes with medial /-ɯ-/
Finals with the medial -ir- are mentioned in Lūi-im Biāu-gō͘ (彙音妙悟), an early 19th century Northern Hokkien rimebook, but now they are obsolete in most dialects of Hokkien. They are found only in a few exceptionally conservative dialects, such as Quanzhou operatic, or, per Ang Ui-jin's survey, in the Taiwanese "Old Anxi accent", spoken among older generations in some areas of New Taipei (namely Sanxia, Linkou, Pinglin, Xizhi, Qidu, Pingxi, and Taishan), in Baozhong Township, and in a few villages in Xihu and Puyan. For these dialects, Ang Ui-jin describes this medial as -er- rather than -ir-, except in the final -iriⁿ.

/uɛ/ and related rimes
Some southern dialects (Yunxiao, Chawan, Lengna) have and  as distinct finals. The latter is used in a small number of vernacular readings: in, , , , , , in all of these dialects, and , , , ,  additionally in the Chawan dialect, and , , ,  in the Lengna dialect.

Similarly, those dialects differentiate between and, but the latter is used exclusively in.

Furthermore, in Chawan dialect, is used in, , and , while  is used in , , , , , and. Other rural Zhangzhou dialects (Nanjing, Pinghe, Changtai, Yunxiao, etc) have /uẽ/ in both groups of characters.

Zhangpu dialect uses, and  consistently in place of ,  and.

Marginal finals
Some marginal finals (not mentioned in the above charts) may occur in specific contexts, such as contractions. For example, in Dongshan dialect there is a final -iohⁿ, used in 即樣 chiohⁿ "like this" and 迄樣 hiohⁿ "like that". In Tong'an dialect, there is a final -iai, used in contractions (遐兮 hiâ--ê > hiâi "those") or in words with the final -ia suffixed with á (e.g.  chhia-á > chhiai-á).

Tones
Traditionally, four Middle Chinese tones are called "level" 平 piâⁿ, "rising" 上 chiǔⁿ, "departing" 去 khìr and "entering" 入 ji̍p. These names are mnemonics illustrating the corresponding tone, e.g. the word "level" 平 piâⁿ has the level tone, the word "to enter" 入 ji̍p has the entering tone, etc. In modern languages, these four tones are further divided into two categories: the "dark" (陰 im) or "upper" (上 chiǔⁿ) tones and the "light" (陽 iông) or "lower" (下 ě) tones, giving a total of 8 tones in traditional system.

The tones can be counted in two patterns: the "dark—light" order (the checked tones are 7 and 8) is more common in works published in China, and the "level—rising—departing—entering" order (the checked tones are 4 and 8) is more popular in Taiwan. This article follows the latter numbering system.

In most dialects of Hokkien, there are only 7 distinct citation tones, as some of the 8 traditional tones merge into a single tone. Certain dialects (the Lengna dialect or the Changkeng-Yidu dialect in Anxi and Yongchun) distinguish all eight tones, and some (Jinjiang) have only 6 citation tones due to additional mergers. Many Northern dialects merge certain tones in the citation form, but not in the sandhi form. Tone contours vary across the Hokkien dialects. Some individual dialects, especially those of the Western branch of Hokkien, have more complex tone sandhi systems, with the post-sandhi tone dependent on the following tone.


 * "Dark level" tone ①陰平
 * High level 44 ˦ ~ 55 ˥ in most dialects.
 * May be slightly lower in Quanzhou dialects (33 ˧ ~ 44 ˦).
 * In urban Zhangzhou dialect it shifts towards high rising 34 ˧˦.


 * "Dark rising" tone ②陰上
 * High falling 53 ˥˧ ~ 51 ˥˩ in most dialects.
 * Coastal Quanzhou dialects (urban Quanzhou, Nan'an, Jinjiang, etc) have it as high level with a small drop at the end (55 ˥ ~ 554 ˥˥˦).


 * "Dark departing" tone ③陰去
 * Low falling 31 ˧˩ in most dialects.
 * May have higher onset (41 ˦˩) in Northern Hokkien and lower onset (21 ˨˩ or even 11 ˩) in Southern dialects.


 * "Dark entering" tone ④陰入
 * Mid-falling 32 ˧˨ in Southern dialects, as well as in Amoy, Yongchun, Tong'an, etc.
 * High falling 54 ˥˦ in Quanzhou dialects.


 * "Light level" tone ⑤陽平
 * Mid or high dipping tone 214 ˨˩˦ in Northern Hokkien, including Amoy.
 * Lower dipping 212 ˨˩˨ ~ 213 ˨˩˧ in Southern Hokkien, although sometimes it may become more level 22 ˨ or lose its rising part (in this case, however, it does not merge with the low-falling tone, but has a longer low segment with an overall contour 211 ˨˩˩).
 * Since the initial falling part is natural for rising tones in tonal languages, many works ignore it and describe this tone as 13 ˩˧ for Southern dialects of Hokkien or 24 ˨˦ for Northern dialects.


 * "Light rising" tone ⑥陽上
 * Mid-level with a slight drop 22 ˨ ~ 221 ˨˨˩ in some Northern dialects (urban Quanzhou, Nan'an, Hui'an, etc).
 * Merged with tone ⑦陽去 in Southern dialects and some peripheral Northern dialects (Amoy, Tong'an, Yongchun, etc).


 * "Light departing" tone ⑦陽去
 * Mid-level 22 ˨ ~ 33 ˧ in Southern dialects, as well as Amoy Hokkien.
 * Merged with tone ③陰去 in many Quanzhou dialects (but still distinguished in sandhi).


 * "Light entering" tone ⑧陽入
 * In greater Quanzhou and Zhangzhou Hokkien, its contour is similar to that of tone ⑤陽平 (mid- or low-rising).
 * In Amoy and Taiwanese Hokkien, it is a high level tone 4 ˦.

Tone sandhi
A phrase in Hokkien is divided into "tone groups", where each syllable except the last one undergoes the tone sandhi.

The suffix 仔 -á
The suffix 仔 -á is related to some special phonetic changes.

Syllables before 仔 -á may induce its change due to assimilation.

Some assimilations are dialect-specific. E.g. in Tong'an dialect, a syllable ending in -a changes it to -ai before 仔 -á: chhia-á > chhiai-á.

The tone sandhi before 仔 -á is different from general Hokkien tone sandhi.

Historical phonology and internal differences
The earliest sources on the Hokkien phonology are the rhyme dictionaries  and. The former describes the Quanzhou Hokkien, while the latter describes the Zhangzhou Hokkien.

Current Hokkien-speaking area mostly coincides with the 10th century Chheng-goân Circuit, a de facto independent polity that emerged after the fall of the Min Empire. The polity was divided into two prefectures, Quanzhou (which also included modern Putian, Changtai and most of Xiamen) and Zhangzhou (which also included the Hokkien-speaking areas of modern Longyan), and the border between these medieval prefectures roughly coincides with certain modern Hokkien isoglosses. The Chiang-bú Circuit (彰武軍), which was under the rule of Wuyue, covers the Hokchew-speaking area, and Tiong-gī Circuit (忠義軍) ruled by Southern Tang lies in Inland Min- and Hakka-speaking area in Fujian.

Changtai dialect contains features of both Northern (Quanzhou) and Southern (Zhangzhou) dialect areas, atop of having some of its own unique characteristics. Changtai was a part of Quanzhou prefecture in 6—10 centuries, until being transferred under Zhangzhou's jurisdiction in 980.

Chawan dialect is a distinct variety of Hokkien. It may have received some influence from Teochew, but its amount is contestable.

The Eastern Namoa dialect shows some traits of Zhangzhou Hokkien, as this half of the Namoa island was previously included in the Zhangzhou prefecture, yet in most aspects it still clusters more with Teochew.

Hai Lok Hong dialect has even more features typical for southern dialects of Hokkien, and may be classified as a distinct dialect of either Teochew or Hokkien, or a variety of Southern Min separate from both of them. The charts below follow the classification of The Language Atlas of China, where Hai Lok Hong is included in Teochew.

The Lengna and Zhangping dialects are very different from mainstream Hokkien. At the same time, they form a continuum with Zhangzhou dialects. They are usually classified as the Western branch of Hokkien.

Datian Min is usually included in Southern Min as a distinct variety, apart from Hokkien and Teochew. It is divided into two dialects, Qianlu (the 'Frontlect') and Houlu, the former lying closer to Hokkien, and the latter having more Central Min influence. An undescribed variety of Southern Min in the north of Dehua is reported to be quite different from other Hokkien dialects and may belong to the same Hokkien—Central Min transitionary area as Datian Min.

Hinghwa is a language closely affiliated with Hokkien, yet it has received heavy Hokchew influence and is not usually considered a part of Hokkien itself.

Denasalization
One large difference between Hokkien and Teochew is the degree of denasalization. Teochew /n/, /m/ and /ŋ/ are usually considered phonemes rather than allophones of the voiced plosives /d/, /b/, /g/.

In Teochew, most syllables with codas preserve the nasal initial, with a few exceptions: denasalization frequently occurs in some specific syllables, like buang, bak (, , but mak: ), leng , long , lang (, , , but nang: ), lung , bung (, , , , but mung: , , ). It may also sporadically occur in some individual characters:, , , , etc. In Hokkien on the other hand, syllables with codas (excluding -h) can never have nasal initials.

In Hinghwa, Hokkien voiced consonants /b/, /dz/, /g/, including cases when they are derived from nasal initials, are further devoiced into /p/, /ts/, /k/.

Final correspondences
The charts below illustrate the common correspondences in rimes between various dialects of Hokkien, as well as related Southern Min languages. Middle Chinese finals are transcribed using Baxter's transcription, and Proto-Southern-Min reconstructions are per Kwok Bit-chee.

In the example characters, literary and vernacular readings are marked by different types of underlines. Note that the examples are given primarily for Hokkien correspondences, and other languages may lack corresponding readings for some of the example characters.

Open-vowel finals
In the tables below, characters after the double line ‖ have nasalization at least in some dialects of Hokkien.

Finals with -n/-t
Teochew has mostly merged -n/-t with -ŋ/-k, except for some peripheral dialects. The dialect of Fenghuang County in Chaozhou preserves the most -n/-t finals (a total of six: -un, -in, -uan, -ien, -an, -ɯn). The Eastern Namoa dialect preserves only -in and -un. In Hai Lok Hong, while some dialects also preserve -in and -un, most Western Hai Lok Hong dialects only preserve -un, and most Eastern Hai Lok Hong dialects merge all -n/-t finals with -ŋ/-k, like in Teochew.

The choose of -ian/-iat or -uan/-uat for a given character derived from MC rhymes 仙三合 -jw(i)en and 先四合 -wen is not consistent among different languages. For 仙三合 -jw(i)en, the generally used reflex is -uan/-uat for most Southern Min languages, except Hinghwa and Lengna, where it is -ian/-iat. However, there is a tendency in Hokkien to have -ian/-iat here when the MC initial was 以 y-, either as the only reading or a non-standard popular variant. For 先四合 -wen, the general reflex is -ian/-iat.

The shape of a character may influence the choose of -ian/-iat or -uan/-uat. Characters with 肙 as the phonetic element (涓, 罥, 鵑, 鞙) tend to have -uan in Hokkien, but -ian in other languages. Characters derived from 矞, 血 and 穴 tend to have -iat in Hokkien, but -uat in Teochew. Characters derived from 夬 and 癸 tend to have -uat in mainstream Hokkien and Teochew, but -iat in Hinghwa, Lengna, Hai Lok Hong, etc.

Finals with -ʔ
Finals with the coda -ʔ are all used in vernacular readings. Their literary counterparts almost always have -p, -t, -k as a coda in Hokkien.

Nasalized finals
Nasalized finals in Hokkien have two principal etymological sources.

First category includes the nasalized finals that are cognate to finals with a full nasal coda. They are used only in vernacular readings.

Another type of nasalized finals is used in syllables with nasal initials that did not undergo denasalization. Such syllables may be alternatively analyzed as having a plain, non-nasalized final and a nasal initial. Although this analysis is not typical for Hokkien, it is more common in the descriptions of Teochew (e.g. the Peng'im romanization would spell 迷 as mi5, and 棉 as min5, even though both are actually /mĩ⁵⁵/, or mî in Pe̍h-ūe-jī). This type of nasal finals occurs in both literary and vernacular readings.

While finals like /ĩ/, /ɛ̃/, /ã/, /iũ/, /uã/ may be induced by both contexts, finals /ãi/, /ãu/, /iãu/, /õ~ɔ̃/ are used exclusively in syllables that did not have an etymological nasal coda (see the section on the open-vowel finals for examples). Additionally, depending on the dialect, words like 'congee' and  'sister' may have finals /uẽ/ (in some Hokkien dialects: Changtai, Chawan, Southern Taiwan, as well as in Teochew) or /uãi/ (in urban Zhangzhou). Teochew preserves some other combinations of nasal initials and finals not found in Hokkien, such as ngiá 'beautiful' and  ngú 'language' (the latter only in the Teoyeo dialect of Teochew).

Occasionally, nasal finals occur in characters that never had a nasal coda or a nasal initial, e.g. táⁿ,  phīⁿ,  phàⁿ.

Other correspondences
The following correspondences are less regular and common, and as such, they are illustrated by specific characters in which they occur.