Let's say a language uses two vowels /A/ and /B/ which differ only by one relevant phonological feature [+/- X] such that /A/ is [- X] and /B/ is [+ X]. Now let's say there's a consonant phoneme /C/ which is also [+ X]. Finally, let's say there's a phonological rule such that all vowels in the environment of /C/ may be realized phonetically with [+ X]. In this language, how do we phonemically analyze a sequence that includes a vowel phone sharing the features of /B/ preceding /C/? Are /BC/ and /AC/ equally valid analyses?

To make this a like less abstract, the confusion that prompted this question is nasal vowels in languages like French, but it's probably applicable to many other languages and possibly other phonological features. I'm really thinking of /A/ as /ɛ/ and /ɔ/, /B/ as /ɛ̃/ and /ɔ̃/, and /C/ as /n/ and /m/. I know that it's super common in world languages for vowels to be realized [+ nasal] before consonants that are [+ nasal], and I believe this may be true for French as well. Thus, /ɛn/ could be realized [ɛ̃n], which would be same realization as hypothetical /ɛ̃n/. If this is true, we have to either say: 1) that the distinction [+/- nasal] is neutralized for vowels before /n/, or 2) that /ɛ̃/ is not allowed to occur before /n/ (making /ɛn/ the right analysis of [ɛ̃n]). What justification is there (French-specific or otherwise) for choosing one claim over the other? Additionally, if claim 1 is chosen, what justification is there for choosing one phoneme symbol over another in transcription?

For example, there is a word pair "chien" and "chienne." I would have assumed that these words only differ phonologically by the presence of /n/ in the latter (and if the next word begins with a vowel, there would be no difference at all due to liaison.) However, I noticed that online resources actually transcribe "chienne" with ⟨ɛ⟩ rather than the ⟨ɛ̃⟩ used for "chien," which strikes me as very bizarre. Either the transcribers are choosing claim 2 or are choosing claim 1 and have some justification to prefer the symbol without the diacritic. As I'm not an expert in French phonology, I'm not sure what this justification might be - it could either be for simplicity's sake (people don't like to use diacritics), for diachronic reasons, or for psychological reasons (which would be by far the most interesting).

  • I think some of my writing may have been unclear. If you agree, I would appreciate any edits to the question's wording.
    – Graham H.
    Oct 7, 2023 at 21:18

2 Answers 2



In French, there is not actually neutralization between vowel pairs like /ɛ̃/ and /ɛ/ before a nasal consonant: the pronunciation of tînmes /tɛ̃m/ is clearly distinct from that of thèmes /tɛm/ "themes".

Phonetically, French /ɛ̃/ is not [ɛ̃]. See the discussion here: Why is the nasal vowel /ɛ̃/ transcribed as it is if it is closer to [ã]? Aside from the difference in nasality, there is a difference in quality: /ɛ̃/ in a Parisian accent is a lower and backer vowel than [ɛ] is. The transcription of French nasal vowels is conservative (or could be interpreted as a "deep" phonemic transcription, along the lines of 'Sound Pattern of English' analyses where the vowel in English time is considered to be a 'tense' counterpart of the vowel in 'tim', rather than being an underlying diphthong).

Phonetically, chien = [ʃjæ̃] or even approaching [ʃjã] whereas chienne = [ʃjɛn]. Some phonetic nasalization might be present in the second case, but that is not sufficient for the vowel to be perceived as the same as the nasal vowel /ɛ̃/.

In general

In general, it isn't uncommon to treat nasality in nasalized vowels as a separable feature, rather than an inherent feature of the phoneme. This analysis is fairly questionable for languages like French where nasal vowels have different qualities from oral vowels and there are fewer distinctions in quality among nasal vowels compared to oral vowels (although some analysts nevertheless attempt to "explain away" French nasal vowels as being underlying vowel + nasal consonant sequences).

But in a language where there is no possible contrast between a nasal vowel and an oral vowel before a nasal consonant, an analyst might prefer to simply treat apparent cases of "nasal vowels" not followed by a surface nasal consonant as underlying /VN/ sequences where "N" is a "placeless nasal" phoneme.

Here's an invented example:

  • [bãn] = /ban/
  • [bã] = /baN/
  • [bãt] = /baNt/

This achieves a certain consistency, although there may be additional arguments for or against this type of "bisegmental" analysis of surface nasalized vowels.

  • Regarding your last statement - if chienne = [ʃjɛn], does that mean oral vowels are not actually nasalized before nasal consonants in French like they are in English? If so, that would kind of render the whole question faulty, at least for French (though the situation I described could exist in other languages).
    – Graham H.
    Oct 7, 2023 at 21:37
  • 1
    @GrahamH.:It's definitely the case that oral vowel phonemes are not pronounced the same way as nasal vowel phonemes before a nasal consonant in French. I'm not sure about the phonetics. For comparison, British English has a phonemic vowel length distinction, but its vowels also separately have different phonetic length depending on their context: something similar could be true of nasality in French. Oct 7, 2023 at 21:39
  • I know /ɛ̃/ is typically realized lower than /ɛ/ (and I think /ɔ̃/ may be realized higher than /ɔ/), but clearly the creators of standard descriptions don't find those height distinctions to be phonologically relevant enough for phonemic transcription (which may or may not be the best choice).
    – Graham H.
    Oct 7, 2023 at 21:40
  • @GrahamH.: Like any IPA transcription, French IPA transcriptions are influenced by phonemic considerations. As with English, an additional factor reducing the phonetic character of IPA transcriptions of French is that the conventional transcriptions were established something like a century ago. Oct 7, 2023 at 21:52
  • Sorry, one last question for your great answer. If the standard transcriptions are based on historical pronunciations, does that mean there might have been a time when some speakers did phonetically merge the oral and nasal vowels before /n/ and /m/? Might there be some minority of speakers today for whom that’s still the case?
    – Graham H.
    Oct 8, 2023 at 1:16

If two vowels differ only in X where B is X and A isn't, consonant C has X, and all vowel become X next to C, then there is contextual neutralization between A and B next to C – you only get B, and A never exists next to C. For any surface [BC], you can posit /AC/ or /BC/. Then the question is, in each case of surface [BC], which analysis does one pick? There is no general answer, but there is a general approach.

Since the directly observable fact is always [BC], you can either posit that the underlying form is exactly what you directly hear, the phonetic form, or is is something different. Some evidence is required to justify claiming that the two forms are different. Therefore, [mũ] should be analyzed as /mũ/ unless you have a good reason to say that it is /mu/. You have posited that A and B both exist and they differ in this feature – i.e. /ũ/ and /u/ both exist – so the "avoid an additional phoneme" argument fails.

There can be direct evidence for /mu/, for example (a) /m/ deletes after /s/ and when prefixed with [s], you find an alternation [mũ] ~ [su]. This would be a good argument that the underlying form is /mu/, not /mũ/. Frequently, we find such contextual neutralizations and we can easily tell whether the surface segment has been changed because of some contextual property. But what if there are no phonological rules or morphological processes that tell you what the underlying form is?

The approach I outlined above simply says "what you see is what you started from", nore exactly, "assume that the surface and underlying forms are the same unless there is good reason to say otherwise". There are approaches which claim that you should make the underlying form be as "unmarked" as possible, in which case you would always avoid nasal vowels in underlying forms when possible, always avoid voiced consonants if possible, etc. I don't follow that approach, but it's out there.

A third answer (which I also disagree with) is "it doesn't matter". A less reprehensible way of looking at this type of non-decision is to say "you can't reach a principled decision, so the child picks a solution at random" (still, I think phonetic transparency is a perfectly valid reason to marking a decision).

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.