What characteristics of a language tend to maximize the average number of allophones per phoneme?

Would a small phonemic inventory tend to increase the number of allophones, e.g. in a language with a phonemic inventory comprising /a/, /i/, and /u/, would /a/ more allophones than it would in a language with a 10 vowel system?

Would a language with few or no free morphemes have more allophones on account of the greater number of phonetic contexts in which a give realization of a phoneme might occur?

What other factors might maximize the average number of allophones per phoneme?

  • Why do you assume that there is a set number of allophones to a phoneme, and that this is the same for all speakers? And what good is an average over all the different phonemes? English /h/ has as many allophones as there are vowels that can occur after it; as far as I can see, that says nothing at all about the structure of the language, only about English /h/.
    – jlawler
    Mar 23, 2013 at 22:10
  • 1
    Well, I don't assume that there is a set number of allophones per phonemes; nor do I assume that the number of allophones per phonemes doesn't vary across individual speakers. I do assume that some allphones of a given phoneme are the same for most speakers of a given variety of a language (e.g. /n/ is realized as [ŋ] before a velar stop and is realized as [n] in most other contexts). Also, I'm not necessarily interested in the structure of a particular language--I just want to know what factors influence the number of allophones per phoneme in general. Mar 24, 2013 at 6:03
  • Well, averaging isn't gonna tell you much. Manjusri has a lot of good points; and you might consider that in computerized speech production, the key to success is finding more and more phonemes, each specialized to a particular niche, because the phones are much better attuned to their context and sound more natural. The more phonemes, the better the system, just like megabytes of RAM. Of course that sense of "phoneme" is the diametric opposite of the linguistic sense. But the linguistic sense is suspect, because it's so personal and unverifiable.
    – jlawler
    Mar 24, 2013 at 14:36
  • Oh, btw, the case of /h/ is instructive. It used to be that /h/ was the only velar fricative in English, but it had a lot of allophones because it was sometimes voiced and sometimes heavily fricated, in different environments. Then /h/ went silent in every environment except prevocalic (we still spell the silent /h/'s as GH's, though), leaving only one allophone. What happened then? It simply adapted to its new environment and developed into a simple voiceless vocal onset, which means that the allophone of /h/ that precedes /i/ is voiceless [i̥], the one that precedes /u/ is vl [u̥], etc.
    – jlawler
    Mar 24, 2013 at 14:59

1 Answer 1


I think it is a matter of presence/absence of a phonetic system within a given language as a milestone for comparison-contrast procedure, either historically or diachronically. Or maybe there is no relation between a phonetical inventory and a number of allophones at all.

The point is, if number of allophones might be regarded within the scope of same inventory (consonants/vowels only), or does it cover the whole phonetical system of a language.

E.g. Sanskrit has three primal vowels (a/i/u), which produce long vowels and diphtongs, yet its consonants ṭ,ṭh, t and th (= aspirated t) are regarded as separate consonants, not as allophones. In fact, its vowel phonetic inventory might not be taken as really 'minimal' together with (semi)vowels ṛ,ṝ,ḷ or ḹ.

The languages like English or Chinese has no inner comparison/contrast systems, but the number of consonant and vowel allophones in English is bigger as compared to Chinese, perhaps due to the polycentrism of English. Chinese almost has no vowel allophones, and position of phonemes zh, j, x and q is dubious. Well, to a Westerner's ear, at least.

One of the languages with a relly minimal vowel inventory is Abkhaz, which has only two distinctive vowels, but a number of Abkhazian consonants equals 58. They are, however, regarded not as allophones of, say, t/tʼ/tʷ, d/dʷ, or h/χʲ/χ/χʷ/χˤ/χˤʷ/ħ/ħʷ sounds, but as separate consonants.

In Arabic, (which has not just a/æ/e and i/y allophones, but also a special [o:] phoneme for its Egyptian variety), there is a big set of consonants. There are tˤ/t and x~χ/ħ~ʜ/h phonemes, which are also recognised as different consonants.

Let us take as an example the Estonian language which has all the possible dyphtongs and three-grade vowel length (short, long and overlong). All this abundence of phonetics, and overlong vowels as well, happens in initial syllables only, while the rest of Estonian word has either short or long back vowels and long or short consonants. Not too much of vowel allophones, but a system of front/back vocals and long/short vowels and consonants instead.

Finnish has the same system (long/short vowels and consonants and front/back vowels together with vowel harmony). Not too much of allophones, either (with the exception of nasalised vowels and n-allophones).

Modern Irish has broad/slender comparison paradigms for consonants together with a number of consonant allophones. A number of vowel allophones has survived in Ulster dialect up to day. Additionally, there has been lenis/fortis comparison-contrast paradigm for sonorants.

Finally, Russian has palatalised/unpalatalised dychotomy for its comparison-contrast paradigm and a large set of vowel allophones.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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