Mandarin or Cantonese: which tones are harder?

Cantonese tones are harder to hear and to keep apart. It has six where Mandarin has four, three of the six are level tones that differ only in how high they are, and two of its rising tones are so close that native speakers in Hong Kong are merging them. Mandarin's four tones differ in shape, which is easier to catch, but Mandarin then changes them in running speech in ways Cantonese does not. So the fair answer is that Cantonese has the harder inventory and Mandarin has the harder rules. We found no study that tests the same learners on both, so what follows is put together from research on each.

Updated 14 October 2026 · 10 min read

The two systems side by side

Tones are usually written with numbers from 1 (the bottom of a speaker's range) to 5 (the top). 55 is high and level; 35 starts in the middle and rises.

The Cantonese values are the ones used by Mok, Zuo and Wong in the study described below. Other references write tone 2 as 35 and tone 5 as 13; the tones are the same, and the numbers are different linguists' readings of them. How Cantonese can also be said to have nine tones is in which language has the most tones, and the Mandarin four are set out with a chart in the Mandarin tones explained.

Look at where the tones start. Mandarin's four begin at 5, 3, 2 and 5 and then go in four different directions. Four of Cantonese's six begin at 2. Mok and her colleagues describe the Cantonese tone space as "very crowded in the lower pitch range, with four tones sharing the same low starting point".

Why Cantonese is harder to hear

Three pairs do the damage:

• Tone 2 and tone 5, the two rising tones. Both start low and rise. They differ only in how far.
• Tone 3 and tone 6, mid level and low level. Nothing moves in either; one is a little higher. In the recordings Mok's team measured, the gap was about 30 Hz.
• Tone 4 and tone 6, low falling and low level. They differ by a small fall.

Qin and Mok played pairs of Cantonese tones to people who had never studied the language and asked whether the two were the same or different. These are the error rates on pairs that were in fact different:

The two rising tones defeated everybody: all three groups called them "the same" at least seven times in ten.

Even native speakers are merging them

The strongest evidence that the Cantonese pairs are hard is that people who grew up speaking Cantonese are losing them.

Mok, Zuo and Wong tested 58 Hong Kong university students, 28 of whom had been picked out as merging tones. In a listening test both groups scored over 97%, but the mergers were slower, by about 150 milliseconds on average. In speech the difference was plain. A statistical classifier given the pitch of each recorded syllable sorted two reference speakers' tones correctly 94% of the time, and the mergers' tones 80.5% of the time. One speaker's tone 3 and tone 6 had almost become one: more than half of her tone 3 syllables were classed as tone 6. The team's conclusion was that these speakers still had six tones, squeezed into a smaller space.

A later study of 120 speakers aged 20 to 58, by Fung and Lee, went further. It classed the two rising tones as a full merger, collapsed in both speaking and listening; tone 3 and tone 6 as merged in speaking only; and tone 4 and tone 6 as merged in listening only.

A learner can take some comfort from that. If you cannot hear the difference between tone 2 and tone 5, neither can a growing number of people in Hong Kong.

Where Mandarin is harder

Mandarin's difficulty is not in the inventory. It is in three things that happen when words are put together.

Tone 2 against tone 3. This is Mandarin's confusable pair. Hao tested English-speaking and Cantonese-speaking learners of Mandarin and found that both groups had significant difficulty telling tone 2 from tone 3, and concluded that the pair is hard for learners whatever their first language. Part of the reason is the next point.

The tones change next to each other. When two third tones meet, the first is said as a second tone, and to the ear it is the same as a real one. Bù ("not") becomes second tone before a fourth tone. Yī ("one") has three versions depending on what follows. And the full dipping third tone, the one learners practise, is normally heard only at the end of a sentence or before a pause; everywhere else it is a low tone. So a learner must know the tone a word has in the dictionary and also the tone it has here.

The neutral tone. Many Mandarin syllables are unstressed and have no tone of their own; their pitch depends on the tone before them. One standard description puts 15 to 20% of the syllables in written text in that group.

Cantonese's best-known tone change is of a different kind. In the "changed tone", a word takes a rising tone to make a related word: 糖 is tong4, "sugar", and with the tone changed to tong2 it is "a sweet". It is tied to the meaning of particular words and learned with them; it is not set off automatically by the tones on either side, which is what sandhi is.

If you already speak one of them

Knowing one tone language does not hand you the other.

Mandarin speakers in Qin and Mok's test were much better than English and French speakers at the Cantonese pairs that differ in shape: they confused tone 5 with tone 6 only 10.4% of the time, against 60% and 70%. But on the level pair, tone 3 and tone 6, they were the worst of the three groups. The authors' explanation is that Mandarin tones are told apart by shape, so Mandarin listeners are not practised at listening for height alone.

In the other direction, Hao found that Cantonese-speaking learners of Mandarin did no better than English-speaking learners, and had an extra problem: they confused Mandarin tone 1 with tone 4, which the study traces to the way both fall into overlapping Cantonese tone categories.

How quickly do learners improve?

For Mandarin there is good news on record. Wang and colleagues gave eight American learners eight training sessions over two weeks; their identification of the four tones improved by 21 percentage points and was still up six months later. A separate study by Wang, Potter and Saffran found that listeners improved significantly after one month in a Mandarin class, and that advanced classroom learners heard the tones as well as native speakers.

Nothing comparable has been published for English-speaking learners of Cantonese over the long term, as far as we could find. Short training studies show improvement, and a 2021 study of Mandarin speakers learning the Cantonese level tones found they got better at naming a tone without getting better at telling two apart.

Children take years over both

Older studies said children had their tones by the age of two. Newer ones, which filter the recordings so that judges cannot guess the tone from the word, find otherwise, in both languages.

• Mandarin: Wong and Strange recorded 44 children aged two to six. Their mothers' tones were identified correctly 97 to 100% of the time; the children were significantly below that on every tone at every age, with one exception, and the authors concluded that learning Mandarin tones "takes more than six years".
• Cantonese: Mok, Fung and Li tested 111 children aged two to six and found that neither speaking nor hearing any of the six tones was adult-like by six. In a study of three-year-olds by Wong, Fu and Cheung, judges identified the children's tone 6 correctly only 38% of the time, and their mothers' tone 6 only 67% of the time.

So neither system comes free even to native speakers, and the tones that are late for children are the ones that are hard for adults: Mandarin's third tone, and Cantonese's rising and level pairs.

Does tone matter equally in both?

Yes, as far as it has been measured. Surendran and Levow calculated how much of the work of telling Mandarin words apart is done by tones, and found it as high as the share done by vowels. Oh and colleagues compared the two languages directly and reported the same balance in Cantonese, with tone carrying, if anything, slightly more. In neither language can a learner treat tones as decoration.

So which is harder?

• To hear: Cantonese. More tones, and three pairs separated by small differences in height or rise.
• To say clearly: Cantonese, for the same reason. Even native speakers' level tones are sometimes mistaken for each other.
• To use in a sentence: Mandarin. Its tones change with their neighbours and many syllables lose their tone; Cantonese's main tone change belongs to particular words.
• Overall, for an English speaker: the United States government's language school puts both in its hardest category, at 88 weeks of full-time study each, and does not rank one above the other. On tones alone, Cantonese is the bigger job.

If you are choosing between the two, tones should not decide it; Mandarin vs Cantonese compares the things that should. If you are starting Mandarin, the Mandarin course, Verbavia (our own course), teaches the tones from the first lesson with recordings, free on the web, and the most common Mandarin words are listed with pinyin. Whether Mandarin as a whole is hard is in is Mandarin hard to learn. For a third Chinese language whose tones work differently again, with whole words sharing one melody, see Shanghainese tones and tone sandhi.

To hear the two side by side in phrases you can use, see Chinese New Year greetings in Mandarin, Cantonese and Shanghainese.

Sources

• Mok, P. P. K., Zuo, D. & Wong, P. W. Y. (2013). Production and perception of a sound change in progress: Tone merging in Hong Kong Cantonese. Language Variation and Change, 25(3), 341–370. cuhk.edu.hk
• Fung, R. S.-Y. & Lee, C. K. C. (2019). Tone mergers in Hong Kong Cantonese: An asymmetry of production and perception. Journal of the Acoustical Society of America, 146(5), EL424–EL430. doi.org (abstract)
• Qin, Z. & Mok, P. P. K. Discrimination of Cantonese Tones by Speakers of Tone and Non-tone Languages. Kansas Working Papers in Linguistics. journals.ku.edu
• Hao, Y.-C. (2012). Second language acquisition of Mandarin Chinese tones by tonal and non-tonal language speakers. Journal of Phonetics, 40(2), 269–279. doi.org (abstract)
• Wang, Y., Spence, M. M., Jongman, A. & Sereno, J. A. (1999). Training American listeners to perceive Mandarin tones. Journal of the Acoustical Society of America, 106(6), 3649–3658. doi.org (abstract)
• Wang, T., Potter, C. E. & Saffran, J. R. (2020). Plasticity in second language learning: The case of Mandarin tones. Language Learning and Development, 16(3), 231–243. doi.org (abstract)
• Qin, Z., Zhang, C. & Wang, W. S.-Y. (2021). The effect of Mandarin listeners' musical and pitch aptitude on perceptual learning of Cantonese level-tones. Journal of the Acoustical Society of America, 149(1), 435–446. doi.org (abstract)
• Wong, P. & Strange, W. (2017). Phonetic complexity affects children's Mandarin tone production accuracy in disyllabic words. PLoS ONE, 12(8), e0182337. journals.plos.org
• Mok, P. P. K., Fung, H. S. H. & Li, V. G. (2019). Assessing the link between perception and production in Cantonese tone acquisition. Journal of Speech, Language, and Hearing Research, 62(5), 1243–1257. doi.org (abstract)
• Wong, P., Fu, W. M. & Cheung, E. Y. L. (2017). Cantonese-speaking children do not acquire tone perception before tone production. Frontiers in Psychology, 8, 1450. frontiersin.org
• Surendran, D. & Levow, G.-A. (2004). The functional load of tone in Mandarin is as high as that of vowels. Proceedings of Speech Prosody 2004, 99–102. isca-archive.org
• Oh, Y. M., Pellegrino, F., Coupé, C. & Marsico, E. (2013). Cross-language comparison of functional load for vowels, consonants, and tones. Proceedings of Interspeech 2013, 3032–3036. isca-archive.org
• Wikipedia: Standard Chinese phonology (tone sandhi, the half third tone, and the share of unstressed syllables), Cantonese phonology and Cantonese changed tones.
• US Department of State, National Foreign Affairs Training Center: foreign language training, which lists "Chinese – Cantonese" and "Chinese – Mandarin" separately in Category IV.