Cantonese: the Key to the Sinosphere

Let’s say you want to learn some East Asian languages, especially to a high level. Not just one, but several. In most cases, that will be Mandarin, Japanese, and Korean, so I’ll talk about those, but the secrets I’m going to tell you apply to any Chinese language and even Vietnamese. You can learn all of the above languages much more easily if you first invest some time into Cantonese.

A common complaint is that Cantonese is less useful than Mandarin. That may be so, but Cantonese is still a huge language that comes with a distinct and interesting culture, most notably spoken in Hong Kong and southern China, but also in chinatowns all over the world. But the real reason I’m writing about it is its almost magical ability to act as a key to any other language you may want to learn in what I like to call the Sinosphere. The usefulness of knowing Cantonese is bigger for some languages than for others, but it helps a lot in all cases. Let’s go through the benefits of knowing Cantonese.


As you may or may not know, all Chinese languages descended from a common language, and all but a few of them only split after what we call the Middle Chinese period. Well, these transitions are actually impossible to talk about in absolute terms, I’d argue, but that’s another discussion. Early Middle Chinese is described as having four tonal categories: 平 ‘level’, 上 ‘rising’, 去 ‘departing’, and 入 ‘entering’ (though 入 didn’t contrast directly with the others, as we shall see). We can’t really know for sure what the pronunciation of these tones were, but we assume that all tones were lower in pitch if their syllable started with a voiced initial, meaning that the vocal cords vibrate when you articulate the initial sound (you can feel voicing by touching your throat while saying ‘mmm…’).

By the Late Middle Chinese period, however, obstruents, meaning sounds that are made by restricting the airflow either wholly or partially, had lost their voicing distinctions, so that /p/ and /b/ had merged into /p/. But their tones remained the same, and the result was that the number of tones doubled. The high tones are called 陰 ‘Yin’ and the low tones 陽 ‘Yang’. Traditionally, the number of tones is then said to be 8, but if we only count contrasting tones, there were only 6. Regardless of how you prefer to count, all of these categories are preserved in Cantonese, and the high entering tone has split according to vowel length, so that short vowels have a higher tone than long ones, dividing the 入 category into three levels.

Most of the Chinese languages have lost one or more categories: Mandarin has preserved 4 out of 8 categories. Shanghainese has preserved 5. But the interesting thing is how categories were lost: in almost all cases, a category simply merged with another one. In some cases, it split and merged into two different categories, depending on whether or not its initial was an obstruent or a sonorant. In one case, it’s unfortunately random. In addition, there are numerous exceptions to the rules, of course, but the vast majority of syllables/characters follow the rules. Knowing what category a character belonged to in Middle Chinese lets us know what tone a character has today, as long as we know what categories the modern tones correspond to. Because Cantonese has preserved all of the categories, it lets us know what tones a character has in other languages – even in Vietnamese. Let’s look at how the tones of Cantonese lets us predict Mandarin tones:

Tone 1 (陰平): high level tone in both Cantonese and Mandarin. Some Cantonese speakers also distinguish a high falling tone.

Tone 2 (陰上): mid rising tone in Cantonese and low falling/dipping tone in Mandarin. (This is usually called the third tone in Mandarin.)

Tone 3 (陰去): mid level tone in Cantonese and high falling tone in Mandarin. (This is usually called the fourth tone in Mandarin.)

Tone 4 (陽平): low falling/very low level tone in Cantonese and mid rising tone in Mandarin. (This is usually called the second tone in Mandarin.)

Tone 5 (陽上): low rising tone in Cantonese and low falling/dipping tone in Mandarin – just like tone 2 (陰上). Many obstruent-initial tone 5 syllables instead become high falling tones in Mandarin.

Tone 6 (陽去): low level tone in Cantonese and high falling tone in Mandarin – just like tone 3 (陰去).

Obstruent-final Syllables

Sometimes called tones 7 through 9 (because one category split into two in Cantonese), these are actually the same as the other three level tones, 1/3/6, except that they end in unreleased obstruents, -p/t/k. In the most common romanization systems, they are labelled 1/3/6, but it can be useful to think of them as tones 7-9, because it fits the rest of the list:

Tone 7 and 8 (陰入高 & 陰入低): high level tone and mid level tone respectively in Cantonese and random in Mandarin.

Tone 9 (陽入): low level tone in Cantonese and mid rising (if obstruent-initial) or high falling (if sonorant-initial) in Mandarin.

Mandarin doesn’t retain any of the final obstruents, but it’s actually the oddball of the lot.

Final Consonants

Cantonese preserves six final consonants: -m, -n, -ng, -p, -t, -k. Although Vietnamese for example preserves two more, -nh and -ch, the intact tonal categories of Cantonese more than make up for it, because most of the languages have lost those two finals anyway.

Mandarin preserves only two final consonants: -n and -ng. Original -m has merged with -n, so they are predictable going from Cantonese to Mandarin, but not the other way around, just like the tones. The obstruent finals are gone, but as far as I know there is no predictable pattern to go by as far as their traces are concerned.

Japanese has only one final nasal, /ɴ/ (ん). Original -n and -m correspond to this (loaned as /nu/ and /mu/), but -ng doesn’t. Instead, -ng was loaned as final /u/, and is (usually) written う even today, though usually in the combination おう, signifying long /oː/. Of the obstruent finals, -t and -k are preserved usually as /t(s)u/ (つ) and /ku/ (く), and sometimes as /t(ɕ)i/ (ち) and /ki/ (き). The last one, -p, was borrowed into Japanese as /pu/, which has since undergone a change to /hu/ (ふ) and then deletion of the /h/, meaning -p is also /u/ う in most cases. Examples: 十 ‘ten’ (-p) /ʑipu/ => /ʑihu/ => /ʑiu/ => /ʑuː/; 白 ‘white’ (-k) /paku/ => /haku/; 工 ‘work’ (-ng) /kou/ => /koː/.

Korean also preserves six final consonants, but -t was borrowed as -l (possibly due to weakening in Old Mandarin). Otherwise, the endings are the same, except for a few labial-initial syllables, because Cantonese has undergone a process of dissimilation where a labial final consonant turns into a dental one if the initial consonant is also labial. But these syllables are few and far between, so don’t worry about it. Examples: 法 ‘law’ (-p, but -t in Canto) /pʌp/; 發 ‘prosper’ (-t) /pal/; 南 (-m) /nam/.


Although Cantonese (and Mandarin) lost their voicing distinctions in initials, the low tones of Cantonese still lets us predict their presence in e.g. Shanghainese and Japanese. However, it should be noted that the two main categories of Sino-Japanese readings, 吳音 (Go’on) and 漢音 (Kan’on), differ in this respect, with 吳音 retaining voiced initials and the later 漢音 having only voiceless obstruent initials, so it’s not always predictable for each word. Example: 上 ‘above/rise’ (-ng) has the Go’on /ʑoː/ (じょう), but the Kan’on /ɕoː/ (しょう).

Problems with Cantonese

Recently (that is, in the last couple of centuries), Cantonese has undergone, and is still undergoing, certain sound changes that sadly make it less suited to be used as a key. However, all but one of these are preserved in every major romanization scheme:

  • Merger of (null)- and ng-. Originally, the three high tones had null-initials and their corresponding low tones had ng- (with a few exceptions; 啱 had both ng- and a high tone). Nowadays, they are used interchangeably by most people, with ng- being preferred in formal contexts and null otherwise. Not a big problem comparatively, since the tones reveal where ng- was originally.
  • Merger of n- and l-. This is common in large parts of China, especially the south. Unlike the ng-/null merger, though, this one is not otherwise predictable, so you should take care to learn which syllables start with n-, even if you decide to pronounce them l-.
  • Merger of ng and m. With the exception of 唔 ‘not’, every single instance of “m” as its own syllable in Cantonese was originally ng. Easy to predict.
  • Merger of gwo-/kwo- and go-/ko-. The same goes with this one: learn which syllables have a -w-, even if you decide not to pronounce it.
  • Merger of -n/-ng and -t/-k. These also have to be learned even if you decide not to distinguish them.
  • Merger of sh/ch/zh- and s/c/z-. This is an old merger, and is the only one not usually distinguished in romanization. However, given the choice between tone categories and consonants, I choose tone categories any day of the week.
  • Merger of -om/-op and -am/-ap. This is another old merger, but one that affects a relatively small number of words, and isn’t that useful comparatively anyway. Cantonese syllables ending in -om are -an in Peking and those ending in -am are -in in Peking.

Cantonese has also lost a fair number of medials. Vowels may not be too helpful (though you’ll notice some patterns even when they are different, such as Cantonese short /a/ corresponding to Mandarin /i/, and /oː/ corresponding to Mandarin /ɑ/. Mandarin vowels tend to be slightly more similar to Japanese and Korean ones, but vowels seem to be the least predictable feature.

So now you know why you should learn Cantonese to make everything easier for yourself. Remember that you don’t need to understand all the technical stuff in this article; your brain is great at recognizing patterns, so learning Cantonese will give you most of the benefits automatically. Have fun!