How Cantonese lyrics work

“Which church do you come from?”

淵維 Vatnið
6 min readAug 16, 2020

The above is a commonly heard phrase in Hong Kong when people are discussing poorly-written lyrics (stick around to find out why!). Specifically, people only say it when the tones of the lyrics don’t match the melody. This could be hard to understand for people who don’t speak tonal languages, and even for some who do speak a tonal language. What does this actually mean?

Crash course on Cantonese tones

(Feel free to skip this section if you’re already familiar)

Cantonese is a tonal language, which means it uses pitch (among other things) to distinguish the meaning of words. Saying the same syllable in a different pitch can yield a totally different meaning, and it’s one of the things learners find hard. (Imagine if see… and see? in English have completely different meanings!)

Cantonese has 6 tones (no, it’s not 9; see this article for more if you can read Cantonese), as shown in the graphic below. The x-axis is time and the y-axis is pitch:

Tones 1 and 3 start from the “high” and “mid” positions respectively and remains flat, whereas tones 2, 4, 5 and 6 all start at the “low-mid” position, and all end up at a different pitches. One way you can indicate the value of a tone is by giving numbers 1–5 to different heights on the chart, with 5 being the highest pitch. For example, Tone 1 starts high and ends high, so its tone value is 55, whereas Tone 2 starts at a low-mid position and ends at a high position, so its tone value is 25. You can find rising tones, flat tones and falling tones, which means one of the most important things to note here is the ending pitch of the tone, at least for the purposes of writing lyrics. Why is this important? You will see in a minute.

What does this mean for Cantonese songs?

Because Cantonese tones are mostly distinguished by pitch, Cantonese lyrics must be carefully written in order to match the pitch of the melody. Failing to do so will result in lyrics that are at best a little funny, and at worst offensive or vulgar (but still funny usually).

So what tones can be put where?

The 6 tones of Cantonese are divided into 4 categories, determined by the ending pitch of the tone. There are various ways of naming the categories, including Dr K.H. Cheung’s terminology of 尖亢下沉 zim1 kong3 haa6 cam4, “sharp; firm; downwards; sinking” named after a descriptive word that happens to represent the sound of that category. For a non-Sinitic-speaking audience, however, I will name them after the ending pitch of each category: high, mid, low-mid, low, and they are as follows (tone values are in brackets):

  • High (5): Tones 1 (55) and 2 (25)
  • Mid (3): Tone 3 (33) and 5 (23)
  • Low-mid (2): Tone 6 (22)
  • Low (1): Tone 4 (21)

Tones 1 and 2 are both end at tone value 5, so they belong to the same category, and by analogy the same is true for Tones 3 and 5. Tone 4 and Tone 6 have their own category. Note that the categories with two tones are both pairs of flat and rising tones. When the actual song is sung, the beginning of the rising note will act as a sort of grace note, creating a special feel for Cantonese songs that not all languages have.

Generally, each note in a melody corresponds to only one category. If you put a syllable in the wrong tone category, it will sound quite strange. However, syllables in a different tone, but in the same category can replace each other. For example, if a Tone 1 syllable is allowed in that note, then so is a Tone 2 syllable. The same applies for Tones 3 and 5. If the melody goes up, then so do the tones of the lyrics. Here’s an example to illustrate:

The first line of the song Happy Birthday goes s s l s d’ t (Hap-py birth-day to you), so the corresponding tone categories are 1 1 2 1 3 2 (remember 1 is the lowest). The lower the pitch, the lower the tone category, and vice versa. Note that both r (birth) and m (you) are different notes, but they both resolve to Category 2 (low-mid), because there are more notes in an octave than there are tone categories, so each category will almost always correspond to more than one note. The below shows what the ending tone value of the song should be (e.g. if the tone value is 25, the ending value is 5):

1 1 2 1 3 2 (Hap-py birth-day to you)
1 1 2 1 5 3 (Hap-py birth-day to you)
1 1 5 5 3 2 1 (Hap-py birth-day dear xxx-xxx)
5 5 5 3 5 3 (Hap-py birth-day to you)

The third line demonstrates the above principle once again, where the birth-day dear xxx-xxx part is downwards melody with five notes, but there aren’t enough tone categories, so the highest category gets repeated in this case (5 5 3 2 1). It may sound strange since it implies the first two notes sound the same, but when sung in tune, with different syllables, it sounds natural to a native speaker.

Once you know what the tone categories are, you can start to fill in actual lyrics. The below shows what tone (Tones 1–6, not the tone values) a syllable has to be:

[4][4][3/5][4][3/5][6] (Hap-py birth-day to you)
[4][4][3/5][4][1/2][3/5] (Hap-py birth-day to you)
[4][4][1/2][1/2][3/5][6][4] (Hap-py birth-day dear xxx-xxx)
[1/2][1/2][1/2][3/5][1/2][3/5] (Hap-py birth-day to you)

Simply (well…) fill in characters that match the melody and you’ll have your song!

Side effects

Which church are you from?

Writing lyrics is already difficult enough, where you have to care about rhyming, rhythm and the actual words. Now I have to care about tones as well? Yeah, it’s hard, and that’s why lyricists are often very well-regarded, much more so than the Western world, at least. This also means that when things go wrong, they’re often made fun of, if the artist’s intention was to write serious lyrics. Some writers chose to use wrong tones on purpose for comedic effects, which brings us to what Hong Kongers call 教會歌 gaau3 wui2 go1 “church songs”. The name comes from the batch of Christian church hymns directly transplanted from Mandarin into Cantonese, where the same lyrics are sung with Cantonese pronunciation, which obviously did not account for the tones. They often sound silly, with 主 zyu2 “Lord” often sounding like 豬 zyu1 “pig”, but this is something that Christians get used to very quickly. This phenomenon, seemingly unique to older Cantonese Christian hymns, has made people associate them with lyrics with tones that mismatch the melody. When someone writes lyrics with tone mismatch (and they’re not doing it on purpose), people will often comment 你邊間教會? nei5 bin1 gaan1 gaau3 wui2? “Which church do you belong to?” as a sarcastic remark.

Song parodies

Since tones are already encoded into the lyrics themselves, it means it also carries the rough melody with it. This property means that it’s very hard to purely read out the lyrics to a song you know and NOT sing it, because the lyrics already sound like music! This, together with the importance lyrics have in Hong Kong music culture, has indirectly contributed to the popularity of song parodies. People can often just read the parodied lyrics to a tune and immediately know what song it is without being told so. This has made it much easier to spread them around and allowed them to quickly gain popularity.

Translated lyrics

Songs often get translated into different languages, and Cantonese is no exception. However, songs often have repeating lyrics with different melodies (like Happy Birthday), so when you translate them into Cantonese, you can’t reuse the same line because the tones would then be all wrong!

Not all tonal languages have this property, notably Mandarin, since their tone system is not complex enough to warrant enough tone categories to make this work. However, many tonal languages still do the same thing with their tones, such as Hokkien. Does your language do the same? Do you think it limits creativity, or is it a constraint that will make people appreciate Cantonese songs more?




淵維 Vatnið

語言學畢業,興趣係成日諗寫咩好但係唔落筆|Linguist whose favourite past-time is to think of what he can write instead of actually writing