So my understanding is that semivowels are phonetically identical (or nearly identical) to their vocalic equivalents, and that the distinction between the two is primarily based on how they behave phonogically. For instance, the only difference between the close front unrounded vowel [i] and the voiced palatal approximant [j] is that [i] functions as the nucleus of a syllable (and is therefore a vowel) whereas [j] does not (and is therefore a consonant). For instance, in the syllable [ja], the [j] would be the onset consonant, with the [a] being the nucleus. But couldn't this syllable also be analyzed as having no onset, with the diphthong [i̯a] as its nucleus? Such an analysis would seem to imply that all sequences of [j + vowel] are really just diphthongs with [i] as the first vowel. Are there any languages that actually distinguish between [ja] and [i̯a]? If so, how would such a distinction be made?

3 Answers 3


In Latin there is qui (nominative singular) and cui (dative singular), presumably something like /kwi/ and /kuj/ respectively.

  • This is not an answer to the question. The OP doesn't deny that the sequences are phonologically different .
    – Colin Fine
    Commented Jan 19, 2022 at 23:25

The conventional phonological difference between the glides [j,w] and [i,u] is that the former is a consonant and the latter is a vowel, in the sense of "not a syllable peak" versus "is a syllable peak". In pre-autosegmental theories of representation, glides are [–syllabic] and vowels are [+syllable], and "syllabic consonants" like [r̩] are also [+syllabic].

In post-SPE theories, vowels are typically distinguished from consonants by choice of some dominating structure, for example "C" versus "V" in Clements & Ford's CV theory, or else as head of N versus non-head in e.g. Levin's X-bar theory. The best that moraic theory can do is posit that if it's dominated by μ it is a vowel and otherwise (if it is directly dominated by σ) it is a consonant = [j,w]. I should also point out that many practicioners of moraic theory (starting with Hyman) redefine "consonantal" to be the equivalent of SPE's feature [syllabic], so that glides are consonantal and vowel of non-consonantal.

Given all of these representational elements, one has a lot of freedom as to how on might represent a given sequence of segments. If you hear something that might reasonably be transcribed as [ja], there are indeed two prima facie plausible analyses: onset [j] plus nucleus [a], or bisegmental monosyllabic nucleus [ia] – a diphthong. If the rules of the phonology include something like "schwa deletes before a vowel of the glide-vowel sequence [ja]", you have a good argument for calling that sequence a "diphthong".

One has to pay close attention to the evidence, if any, that an author gives to justify their transcription. For example, is there a contrast between [ie] and [je], or is their distribution predictable (the former always being preceded by a consonant, the latter by a vowel or Ø)? To check the sufficiency of that evidence, one should also check whether for example there is a contrast between [ie] (one syllable) and a bisyllabic sequence [i.e].

  • Thx for the answer. Just one question: how would [ie] and [je] be contrasted? And are there any languages that are known to have this distinction? Commented Jan 18, 2022 at 1:22
  • 1
    I'd say German has such kind of contrast, albeit marginally: Yacht [yaxt] vs. I-8 [i.axt] where I-8 is kind of a building identifier or proper name, or je [je] vs. I E [i.e] where das I-E refers to a spelling pecularity of the German orthography. Commented Jan 18, 2022 at 13:41
  • 2
    @user6726 Finnish (and I would guess probably Sami as well) does have the distinction word-initially, though. Within a word cases like marja ‘berry’ and Maria are minimal pairs, but can be syllabified differently; initially, the only real minimal pairs I can see would be ies ‘yoke’ and jes ‘yes’, but the latter is a loan word. There are inherited words beginning in both ie- and je-, though. Commented Jan 18, 2022 at 18:35
  • 1
    @jk-ReinstateMonica/ Definitely marginal. But how about "i a" (hee haw) vs."ja" (yes)?
    – fdb
    Commented Jan 20, 2022 at 12:59
  • 1
    @fdb Your example is of the more rustic kind, but definitely marginal, too. Commented Jan 20, 2022 at 13:21

It can be an erroneous, but there are no such counterparts, as [i] and [j], or [u] and [w]. Phonemically (i.e. in //), yes, but phonetically is no. This is the rare thing in languages, the cardinal vowels itself (especially for the /i/ vowel - in most languages it is not so high, and sometimes is retracted - [i̞~i̞˗]. So more correct to say that there are [ɪ] and [ɪ̯]/[j], [ʊ] and [ʊ̯]/[w] counterparts (n.b. phonemically is still /i/-/j/, /u/-/w/).

  • There are exceptions, like Spanish language, where allophonic realisation of /i/ both [i] and [ i̞ ], and the [ai̯] diphthong is possible.
    – T1nts
    Commented Jan 21, 2022 at 12:41

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.