General Chinese

泥 and 娘 are both transcribed ⟨n⟩, as these are not distinct in modern dialects. 喻, a conflation of two older initials, 云 hy~hw and 以 y~w, is transcribed ⟨h⟩ or ∅ according to modern rather than ancient forms; when palatalization is lost, it is transcribed ⟨w⟩. The palatal and retroflex fricatives 照穿牀審禪 fell together early on in the rime tables of Classical Chinese, but are still distinguished in some modern dialects, and so are distinguished here. The convention ⟨q⟩ for nasal 疑, which drops in many dialects, is repeated in the finals, where it represents *[ŋ] with a departing tone.

Although to some extent systematic—the retroflex series are digraphs ending in ⟨r⟩, for example—this is overridden in many cases by the principle of using short transcriptions for common sounds. Thus ⟨z⟩ is used for 精 rather than for the less common 邪, where it might also be expected; ⟨v⟩ is used for frequent 微 rather than for 奉; and ⟨c⟩ and ⟨g⟩, for the high-frequency 見 and 羣, have the additional benefit of being familiar in their palatalized forms (Peking ~ Beijing for example is -⟨cieng⟩) from English words like cello and gem.

The voiced obstruents (the 濁 "muddy" column) are only distinct in Wu dialects. In Min, they are collapsed with the consonants of the tenuis column. Elsewhere they are generally collapsed with the aspirated column in the even tone, and with the tenuis column in other tones. An exception is Cantonese, where in the rising tone they are aspirated in colloquial speech, but tenuis in reading pronunciations. The sonorants do not vary much apart from ⟨v⟩, ⟨r⟩, which in Wu are nasals [m], [ɲ] colloquially but fricatives [v], [z] when read.

Velars ⟨c⟩, ⟨k⟩, ⟨g⟩ are palatalized to affricates before ⟨i⟩, ⟨iu⟩ (the high-front vowels [i], [y]) apart from Min and Yue, where they remain stops before all vowels; ⟨x⟩, ⟨h⟩ also palatalize, but remain fricatives. For instance in Mandarin, they are g, k, h before non-palatalizing vowels and j, q, x before palatalizing vowels, whereas in Cantonese they remain g, k, h everywhere. (Compare the alternate spellings of Beijing and Peking; see Palatalization (sound change) § Mandarin Chinese) The alveolar sibilants ⟨z⟩, ⟨ts⟩, ⟨dz⟩, ⟨s⟩, ⟨sz⟩ (Mandarin z, c, s) are also generally palatalized before ⟨i⟩, ⟨iu⟩ (to Mandarin j, q, x), collapsing with the palatalized velars ⟨c⟩, ⟨k⟩, ⟨g⟩, ⟨x⟩, ⟨h⟩ in dialects which have lost the "round-sharp" distinction so important to Peking opera.

The palatal stops ⟨dy⟩, ⟨ty⟩, ⟨dhy⟩ remain stops only in Min among the Chinese topolects (though also in Japanese and Vietnamese loan words); elsewhere they are conflated with the affricates. The palatal and retroflex sibilants are generally conflated; in Yue and Min, as well as in much of Wu and Mandarin, they are further conflated with the alveolar sibilants. This contrast remains in Beijing, where [sán] 'three' is distinct from [ʂán] 'mountain'; both are [sán] in Sichuanese and Taiwanese Mandarin.

There are numerous more sporadic correlations. For instance, the alveolar affricates ⟨z⟩, ⟨ts⟩, ⟨dz⟩ become stops [t], [tʰ] in Taishan Yue, whereas the alveolar stops are debuccalized to [ʔ], [h], as in Hoisaan for Cantonese Toisaan (Taishan). In Yüchi, Yunnan, it is the velars ⟨c⟩, ⟨k⟩, ⟨g⟩ which are debuccalized, to [ʔ], [ʔʰ]. In the Min dialects, ⟨f⟩, ⟨fv⟩ become [h] or [ʍ]. In Xi'an Mandarin, the fricatives ⟨sh⟩, ⟨sr⟩, ⟨zh⟩, ⟨zr⟩ are rounded to [f] before rounded vowels, as in [fêi] 'water' (Beijing shuǐ [ʂwèi]).

The categories of the Late Middle Chinese rime tables are reduced to the four medials of modern Chinese, plus an intermediate type ⟨e⟩:

⟨i⟩ and ⟨iu⟩ are omitted after labiodental initials.

The medial ⟨e⟩ is used for syllables which have a palatalizing medial [i] in Mandarin, but no medial in Yue. That is, in Mandarin ⟨e⟩ should be read as ⟨i⟩, with the same effect on consonants as ⟨i⟩ has, whereas in Cantonese it is silent. In Shanghainese both situations occur: ⟨e⟩ is equivalent to ⟨i⟩ in reading pronunciations, as [i] or [y], but is not found in colloquial speech.

In Cantonese, medial ⟨i⟩ can be ignored after sibilants, as palatalization has been lost in these series. That is, siao, shao are read the same.

Chao uses the following rimes. They do not always correspond to the Middle Chinese rimes.

Rimes consist of a nucleus (the main vowel) and optionally a coda. They need to be considered as a unit because of a strong historical interaction between vowel and coda in Chinese dialects. The following combinations occur (note that the vowel ⟨iu⟩ is treated as medial ⟨i⟩ plus nucleus ⟨u⟩):

The most salient dialectal difference in rimes is perhaps the lack of the obstruent codas ⟨p⟩, ⟨t⟩, ⟨c⟩ in most dialects of Mandarin and independently in the Wencheng dialect of Oujiang, though this has traditionally been seen as a loss of tone (see below). In Wu, Min (generally), New Xiang (Hunanese), Jin, and in the Lower Yangtze and Minjiang dialects of Mandarin, these codas conflate to glottal stop /ʔ/. In others, such as Gan, they are reduced to [t], [k], while Yue dialects, Hakka, and Old Xiang maintain the original [p], [t], [k] system.

Nasal codas are also reduced in many dialects. Mandarin and Wu do not distinguish between ⟨m⟩ and ⟨n⟩, with them being reduced to [n] or nasal vowels, or in some cases dropped altogether. In Shanghainese many instances of ⟨nɡ⟩ have conflated as well, or been dropped, but a phonemic distinction is maintained.

In Mandarin, an additional coda is found, -er [ɻ], from GC ⟨ri⟩.[12]

In Cantonese, the simple vowels i u iu o a e are [iː uː yː ɔː aː ɛː], apart from ⟨i⟩ and ⟨iu⟩ after velars, which open to diphthongs, as in ci [kei] and ciu [kɵy]. Diphthongs may vary markedly depending on initial and medial, as in cau [kou], ceau [kaːu], ciau [kiːu], though both ceu ~ cieu are [kɐu], following the general pattern of ⟨e⟩ before a coda (cf. cen [kɐn] vs can [kaːn]). Cantonese does not have medials, apart from gw, kw, though sometimes it is the nuclear vowel which drops: giung [kʰoŋ], xiong [hoŋ], but giuan [kʰyːn].

The following combinations of orthographic medials and rimes occur, taking -iu to be medial i + rime u :

* In entering tone, ang (eang, iang, uang) changes to oc (eoc, ioc, uoc), and ong (iong) changes to ouc (iouc)

^‡ -eaeng is generally shortened to -aeng

^§ In entering tone, iuan changes to iuet

Double cells show discrepancies between analysis and orthography. For instance, Chao analyzes ieng, iueng as part of the aeng series rather than the eng series, and ien as part of the an series.[14] Though not apparent from the chart, eng-ing-ueng-iuing, ung-ong-iung-iong, and en-in-un-iun are similar series. The discrepancies are due to an effort to keep frequent syllables short: en-in-un-iun rather than *en-ien-uen-iuen, for example; as well as a reflection of some of the more widespread phonological changes in the rimes.

The Classical correspondences, with many archaic distinctions lost, are as follows:

These all occur in the velar-initial series, but not all in the others.

In Cantonese, after coronal stops and sibilants, rounded finals such as -on and -uan produce front rounded vowels, as in don [tyːn], and after velars, iung and iong lose their ⟨i⟩. Min dialects are similar, but in certain tones ⟨i⟩ and ⟨iu⟩ become diphthongs rather that their usual [i], [y]. For example, in Fuzhou, even-tone 星 sieng is [siŋ] but departing-tone 性 sieq is [seiŋ].

In Yunnan Mandarin, ⟨iu⟩ is pronounced as ⟨i⟩, so that the name of the province, yunnom, is [inna] rather than [ynnan] as in Beijing.

In Nanking, ⟨ien⟩ metathesizes to [ein] after alveolars, as in 天 [tʰein] for Beijing tian [tʰiɛn].

The basic spelling is used for the even 平 tone(s). For the rising 上 tone(s), the nucleus is doubled (with the vowel ⟨iu⟩ → ⟨iuu⟩, as that is treated as medial ⟨i⟩ + nucleus ⟨u⟩), or the coda is changed to a 'lighter' letter. For the departing 去 tone(s), the coda is made 'heavier'; if there is no coda, add ⟨h⟩. For the entering 入 tone(s), a stop coda is used.

'Lighter' means that a vowel coda is made more open (⟨i⟩ → ⟨e⟩, ⟨u⟩ → ⟨o⟩); 'heavier' means that a vowel coda is made more close (⟨i⟩ → ⟨y⟩, ⟨u⟩ → ⟨w⟩) and a nasal coda (⟨n⟩, ⟨m⟩) is doubled. The nasal ⟨ng⟩ is 'lightened' to ⟨g⟩ (as in many Polynesian languages) and made heavier as ⟨q⟩ (as in the GC initial):

One consequence of this is that the rimes -e and -ei in the even tone conflate to ⟨ee⟩ in the rising tone. However, since there are no such syllables which begin with the same consonant and medial, no syllables are actually conflated.

The difference between yin and yang tones is indicated by the voicing of the consonant. A zero consonant is treated as voiceless (it is sometimes reconstructed as a glottal stop), so i, iem, uon, iuan are ping yin (Mandarin yī, yān, wān, yuān), whereas yi, yem, won, yuan are ping yang (Mandarin yí, yán, wán, yuán). In a few cases, the effect that voiced ⟨m⟩, ⟨n⟩, ⟨l⟩, ⟨r⟩ have on tone needs to be negated to achieve a ping yin tone. This is accomplished by spelling them ⟨mh⟩, ⟨nh⟩, ⟨lh⟩, ⟨rh⟩.

To mark the toneless Mandarin syllable ma, a centered dot is used: ⟨·ma⟩. The dot is omitted for toneless ⟨me, de, te, ne, le⟩, as tonic me, de, te, ne, le do not exist.

The realization of the tones in the various varieties of Chinese is generally predictable; see Four tones for details. In Beijing Mandarin, for example, even tone is split according to voicing, with muddy consonants becoming aspirates: ba, pa, ma, bha → bā, pā, má, pá (and mha → mā). Departing tone is not split, and muddy consonants become tenuis: bah, pah, mah, bhah → bà, pà, mà, bà. Rising tone splits, not along voicing, but with muddy-consonant syllables conflating with departing tone: baa, paa, maa, bhaa → bǎ, pǎ, mǎ, bà. That is, bhaa and bhah are homonyms in Beijing, as indeed they are in all of Mandarin, in Wu apart from Wenzhounese, in Hakka, and in reading pronunciations of Cantonese. Entering tone is likewise split in Beijing: mat, bhat → mà, bá.[16]

However, the realization of entering tones in Beijing dialect, and thus in Standard Chinese, is not predictable when a syllable has a voiceless initial such as bat or pat. In such cases even syllables with the same GC spelling may have different tones in Beijing, though they remain homonyms in other Mandarin dialects, such as Xi'an and Sichuanese.[17] This is due to historical dialect-mixing in the Chinese capital that resulted in unavoidably idiosyncratic correspondences.

In Yue, there is a straightforward split according to consonant voicing, with a postvocalic ⟨h⟩ in Yale romanization for the latter. Muddy onsets become aspirates in even and rising tones, but tenuis in departing and entering tones: ba, pa, ma, bha → bā, pā, māh, pāh; baa, paa, maa, bhaa → bá, pá, máh, páh; bah, pah, mah, bhah → ba, pa, mah, bah; bat, pat, mat, bhat → baat, paat, maht, baht. In addition, there is a split in entering tone according to vowel length, with Cantonese mid-entering tone for short vowels like bāt, pāt.[18] In reading pronunciations, however, rising tone syllables with muddy onsets are treated as departing tone: bhaa → bah rather than → páh. There is also an unpredictable split in the even ping yin tone which indicates diminutives or a change in part of speech, but this is not written in all Cantonese romanizations (it is written in Yale, but not in Jyutping).