There are 214 (by one count) Mandarin phonemes. How can I learn

  1. Which of those are not shared by English?
  2. Which of the remaining number are in the most common use (assuming they follow the Pareto Principle — 80/20 or whatever — like everything else)?

I'm doing research and need to be pointed in the right direction.

  • If you're looking for some similar research, look at aclweb.org/anthology-new/C/C82/C82-1060.pdf which seems to be spot on. Jun 14 '12 at 17:37
  • 8
    Who has proposed 214 phonemes in Mandarin? I think it's a slight exaggeration! Jun 14 '12 at 21:56
  • 6
    Are you sure you're not confusing phonemes with the 214 traditional radicals in Hanzi? The number seems a huge coincidence.
    – Colin Fine
    Jun 14 '12 at 23:18
  • 6
    Formally, no phoneme is shared between two language, because by definition a phoneme is a set of phones which are indistinguishable (or freely substitutable) within a language (or variety). You can obviously informally identify phonemes in different languages, but if you try to count the correspondences you might get quite arbitrary answers. If one language systematically ignores voicing (like Chinese) and the other systematically ignores aspiration (like English), the number of correspondences will depend how you choose to group them.
    – Colin Fine
    Jun 14 '12 at 23:26

This isn't really an answer, but I'd like to poke at a few different aspects of the question and this is too long to be a comment.

First, where are you getting 214 phonemes? Are you sure you aren't thinking of the 214 Kangxi radicals, as Colin Fine suggested? Most analyses I've seen have between 20 and 30 phonemes, depending how you interpret the vowels and the palatal consonants.

Second, phonemes don't necessarily follow the Pareto principle. Phoneme distributions are still an active area of research, but the number of phonemes in a language is generally small enough that even the least common are still quite important. In other words, even if they follow a power-law distribution, there aren't generally enough phonemes to get very far down the tail.

Third, it's not clear what it means for languages to "share" phonemes, because phonemes are sort of like equivalence classes. Punjabi, for example, has three types of stops, /tʰ t d/, while English has only two, usually transcribed /t d/. Which of those do the languages share, if any? The English "voiceless" stop is usually aspirated and the English "voiced" stop isn't always voiced. And besides that, the Punjabi stops are generally dental, while the English ones are generally alveolar—does that mean they're fundamentally different, even though neither language makes a distinction between dental and alveolar stops?

Mandarin specifically is usually analyzed with a two-way aspiration distinction /tʰ t/, while English is usually analyzed with a two-way voicing distinction /t d/. But the English voiceless stops are often aspirated, and the Mandarin unaspirated stops are sometimes voiced. How do you decide if they're shared? In the end, this will have to be to some extent an arbitrary decision—and that means your final result will be somewhat arbitrary too.


The question makes the standard assumption that language sounds have objectively comparable phonetic properties, which are described by some set of features, and that the "phonemic value" is whatever is underlyingly present, for example in the case of English surface [k kʲ k̚ kʰ] the presumed underlying phoneme is /k/, or the features describing a "voiceless unaspirated velar stop".

The phonemes of Mandarin are /m, n, ŋ, pʰ, tʰ, kʰ, p, t, k, tsʰ, ʈʂʰ, ts, ʈʂ, f, s, ʂ, ʐ , x, l, i, u, y, ə, a/. There are more surface sounds, but the question is about phonemes which is a reduction of surface sounds to an abstraction, eliminating contextually predictable variants. The "common currency" in comparing between between languages is (expert) IPA standards, in case you are not sure what [ʂ] sounds like and how it differs from [ʃ]. [m, n, ŋ] sound the same in Mandarin and English, and there is nothing to the pronunciation of these sounds in either language that suggests that these phonemes are different: the nasals are "shared" (the same between the languages). In the case of Mandarin /ʂ/, that is phonetically different enough from English [ʃ] that they should be counted as different sounds. The stops also match pretty well, though there is debate over English <b, d, g>, namely whether they are "basically" voiced or "basically" unaspirated. The "b is unaspirated" theory has the problem that it then claims that English has two degrees of aspiration, on that is allophonic and one that is phonemic, so I reject the aspiration analysis. However, there isn't a compelling phonological argument that Mandarin b,d,g are voiceless unaspirated, that is based on (a) historical linguistics and (b) phonetic closeness. I accept that Mandarin b,d,g are in fact voiceless unaspirated stops: so, the phonemes are different (they have different feature analyses, under the theory of universal features). The retroflex sibilants and affricates do not match English, and English also doesn't have /ts, tsʰ, x/. English has no /y/, otherwise the vowels of Mandarin are also found in English (but again, they are pronounced slightly differently, though we're not completely invested in a comparision of pronunciation, we're looking at the feature analysis of the two languages.


I'd say do it by hand. First write down all 214 Putonghua phonemes and convert them to symbol tokens in your universal phonetic alphabet. Now write down all N English phonemes and convert them to symbol tokens in the universal phonetic alphabet. Write it so it's two columns, and if the same symbol is used in the universal phonetic alphabet, then they are shared. That's how you do the first part.

Now for the second part, set up a correspondence between your 214 phonemes and their equivalents in pinyin, then take a largish sized representative Putonghua corpus (a million characters or so should work), convert it to pinyin, then convert it to your 214-phoneme system, then tokenize it. Now cross out all of the ones which were counted as "shared" in part I. Now rank them by decreasing token frequency.

That's it! The only tricky part is in deciding on your universal phonetic alphabet, or put another way, what counts as "the same thing" between English and Putonghua.

Not the answer you're looking for? Browse other questions tagged or ask your own question.