Tuesday 26 October 2010

length of Italian vowels

Yes, the transcription I used for Italian yesterday included length marks where appropriate for phonetically long vowels, and yes, vowel length (duration) is predictable in Italian.

The transcription I use for Italian is based on my late colleague Marguerite Chapallaz’s The Pronunciation of Italian (London: Bell & Hyman, 1979) a book long out of print but particularly suited to the practical needs of English-speaking learners.

I think it is helpful to show vowel length explicitly in transcribing Italian. One reason is that segment length is perceptually very salient in this language, making the rhythm of words and sentences sound very different from that of, say, Spanish or Greek. Compare the words for ‘hand’, spelt the same, mano, in Italian and Spanish: ˈmaːno in the first, ˈmano in the second. It does no harm at all to remind foreign learners of this characteristic of Italian at every opportunity.

The other reason is that the rule concerning vowel length is not altogether straightforward. A vowel is long in an open non-final stressed syllable. That means a vowel that is stressed and syllable-final but not word-final, the following consonant (if there is one) being initial in the next syllable. All other vowels are short. (Check out yesterday’s examples.) The matter is complicated by the fact that some consonants are pronounced double even though written single (as a single digraph), which means the preceding vowel is short, e.g. ogni ˈɔɲɲi, paglia ˈpaʎʎa.

We conventionally transcribe long (geminate) consonants by writing them twice rather than by using length marks. This is partly because they typically straddle a syllable boundary. It’s also because, unlike vowel length, consonant gemination in Italian is phonemic. My favourite minimal pair fato vs. fatto is thus transcribed ˈfaːto, ˈfatto, reminding the learner of the difference at two points. It would not be visually distinctive enough if we were to write ˈfaːto, ˈfatːo. (The same arguments apply in transcribing Finnish, where vowel and consonant length vary independently.)

But yes, it would be adequate and unambiguous to omit vowel length marks in Italian, as Italians do themselves. This reminds us that there is more than one way to transcribe any given language phonetically, even remaining strictly within IPA.


  1. It's funny, I was just thinking about the features that made Italian sound the way it does. I concluded that vowel and consonant length had a lot to do with it. Otherwise it would sound more like Spanish, which it sort of does anyway to me at times. I also think there is something about the intonation pattern of Italian that's different from other Romance languages. If you listen to this, I think you'll hear what I mean. I can't really describe it and this is off topic anyway.

  2. Here's the link. I'm sorry. I hate it so much when I screw up on that.

  3. Canepari in his pronunciation dictionary makes the point that vowel-length is stress-dependent, giving the example of the word noi: if he transcribed this as ˈnoːi he would be giving a pronunciation that is only appropriate in a position of emphasis, whereas very often the word carries no sentence stress and is simply ˈnoi.

  4. When I taught myself Esperanto back in the 1980s using only books, I thought the vowels should also be short, just like in Spanish.
    I was therefore surprised when I went to my first Esperanto meeting and realised that most speakers use a vowel length rule similar to the Italian one.

  5. @ Jeff: Argentinian Spanish is usually said to be spoken with Italian intonation. I think that's true and I find it very charming.

  6. >A vowel is long in an open non-final stressed syllable

    According to Mariapaola D'Imperio and Sam Rosenthall (Phonetics and phonology of main stress in Italian. In Phonology, vol. 16(1), 1999, pp. 1-28), a vowel is long in an open penultimate stressed syllable. They found that vowels in stressed antepenultimate syllables were slightly longer than unstressed vowels, but much shorter (as in statistically significant) than those is stressed open penultima. They note that the traditional interpretation is that the vowels in antepenultima are (phonetically) shortened, but argue that in fact the antepenultimate vowels are never lengthened in the first place, i.e. that they are phonologically short.
    On the other hand, the preliminary program for this year's Going Romance conference includes a talk on "what Italian stressed penultima and antepenultima have in common", so I look forward to learning more about that.

  7. Canepari states the rule as "the last phoneme in a stressed syllable is long", which also describes the fact that the /n/ in pinta is longer than the one in pino. OTOH, he needs an exception for diphthongs. (Though I'm not convinced that there are phonemic diphthongs in Italian -- at least not in my accent, and I'd think of noi as underlying /no.i/ with the syllables being merged by some rule quite close to the surface.)

  8. @Pavel: this theory of the vowel in a stressed antepenultimate syllable being shortened immediately put me in mind of the supposed 'rules' of the traditional English Latin pronunciation which was used up to the 19th c. and survives in legal and scientific Latin, proper names, and well-known words and phrases that have entered the English language. An open vowel in a stressed penultimate syllable was always pronounced long regardless of the Latin quantity (e.g. lŏcus was ˈləʊkəs, not ˈlɒkəs) whereas in stressed antepenultimate syllables such vowels were regularly pronounced short (Thūcȳdĭdes = θjuːˈsɪdɪdiːz not θjuːˈsaɪdɪdiːz). There were exceptions for cases involving hiatus (rădius = ˈreɪdiəs, not ˈrædiəs).

    May just be coincidental, or there may be an underlying phonetic principle common to both languages (English and Italian, I mean)?

  9. John, I now realize why I too thought there was something funny about your transcription [ˈpɔːveɾo]. I saw that you were being more explicit than usual about the realization of /ˈpɔvero/ but I didn't stop to think about it.

    What these two discussions have made me realize is that I must have felt subconsciously that it was less appropriate to write [ˈpɔːveɾo] than to write [ˈpɔːko], for example. The /ɔ/ in /ˈpɔvero/ is only half-long. Thanks to Pavel for giving a reference for that observation.

    (I have not allowed myself to represent the above /e/ and /o/ as archiphonemes in reaction to Mitko's question yesterday of how to treat the unstressed ɛ/e and ɔ/o neutralisations phonemically. To raise it at all he must have already "exorcized the ghost of the archiphoneme" as Martinet put it. But in truth, if we say it is a matter of restricted distribution, and /ɛ/ and /ɔ/ are restricted to the commutation class of the stressed vowels in a phonotagm, we have to say the same of /e/ and /o/, as the closed vowels in unstressed syllables have different phonological identities, not being opposed to their open correlates.)

    And in addition Thomas raised my consciousness of the situation in Esperanto: not only do most speakers use a vowel length rule similar to the Italian one (at that point in the discussion only involving a long-short distinction, all penults being stressed in Esperanto), but I have been doing it myself all my life.

  10. I would just add this, for a good afternoon's reading, and state that if you had been at Monza bridge club in the mid-nineties and heard Sig. P-- scream "Cazzo::! Perché hai giocato il re?", you too would be convinced that /əʊ::/ exists in Italian. E come.

  11. Filippo Marchesani26 October 2010 at 17:47

    I’m Italian and I live near Monza, and yet I’ve never heard “əʊ” by a native speaker, except maybe for the interjection “Oh!” when pronounced by someone particularly angry.

  12. Pavel,
    I hope you have seen from my last post how revelatory your reference to Mariapaola D'Imperio and Sam Rosenthall was for me. However, you say they note that the traditional interpretation is that the vowels in antepenultima are (phonetically) shortened. The only sense I can make of this is that these vowels are shortened with respect to the length they would have when stressed in penultimate position, which analysis betokens a healthy descriptivism in the traditional view. But I cannot make any sense at all of the argument that in fact the antepenultimate vowels are never lengthened in the first place, i.e. that they are phonologically short. No vowel can be phonologically short in Italian, can it? And in what first place are they never lengthened? Is this some sort of "transformational history" they have devised? The first place was Latin, in which they could be phonologically either long or short, and that phonological distinction was lost (except for the differential quality of the vowel in the case of e and o), was it not? And yes, I think Steve's parallel with the traditional English Latin pronunciation which survives in legal and scientific Latin is valid. We know from the metre of mediaeval Latin poetry etc. that the phonological quantity opposition was already lost, and the stress-determined length must have been established before the English Great Vowel Shift.

    Or is the argument so bizarre as to be saying that there is a phonological long-short opposition in the modern language and that that is what determines where the stress falls?

    So army, I think we agree that Canepari has got it the right way round, but how would he even go about convincing you that there are phonemic diphthongs in Italian? And I don't think you need to be so circumspect as to restrict your observations to your own accent: Steve above mentions Canepari's example of the word noi as having stress-dependent vowel length, implying that ˈnoːi does occur in a position of emphasis, and is two syllables, and that when it is simply ˈnoi it is a diphthong. That would indeed exemplify the exception to the rule that "the last phoneme in a stressed syllable is long". Nobody says noiː in any accent, do they? But why should ˈnoi not be simply a purely phonetic diphthong? Or even two syllables with a reduced o in the allegro form? But for a closer correspondence with the /n/ in pinta being longer than the one in pina, nobody says ˈrauːko in any dialect, but why should that make it an exception to the rule? I believe traditionally it is seen as a diphthong, making ˈraˑu.ko an exception, but why should it not be ˈraˑ.u.ko, and as trisyllabic as paˈuː.ra?

  13. The rule about (phonetic) vowel length in Italian is the one that Marguerite Chapallaz explained (the first to discover it was, I think, Amerindo Camilli) and John Wells quotes, namely: a vowel is long if it is stressed and syllable-final, but not word-final; all other vowels are short.

    That is all, and there is no difference between "parole piane" and "parole sdrucciole".

    There's a clear difference, for instance, between "abile" ("able, capable") /'abile/ ['a:.bi.le] and "abbile" ("have [imperative] them [fem. pl.]") /'abbile/ ['ab.bi.le]. (Well, in some regional pronunciations -- in Rome, e. g., or in the South -- every intervocalic /b/ tends to be realized as [bb], thus ['ab.bi.le] for both words; but that's another matter.)

    Also, "sdrucciolo" is ['zdrutʃ.tʃo.lo], "truciolo" ("wood shaving") is ['tru:.tʃo.lo]. Etcetera.

    "Povero" is ['pɔ:.ve.ro], with a long stressed vowel, no less clearly than "poco" is ['pɔ:.ko]. The "a" in "papero" ("duck") ['pa:.pe.ro] is as clearly long as the first "a" of "papa" ("pope") ['pa:.pa] (but all the vowels of "pappa" ["pap, baby food"] and of "papà" ["daddy"] are short).

    Italians aren't usually aware of this difference, obviously because it isn't a distinctive (meaningful, phonematical) one. But, as John Wells says, it is "perceptually very salient" for a non-Italian (with phonematical vowel-length differences in his language), and also for a trained Italian ear.

  14. In the examples of ogni and paglia where we have "silent g", as I like to call it, I suppose I subconsciously think of this "g" as a placeholder for the first part of the underlying double consonant. It also seems to me that the only reason why "silent g" is used here is due to a lack of special characters for these palatal phonemes, /ɲ/ and /ʎ/. These are just impressions of mine and I'm only moderately familiar with Italian.

  15. @mallamb:
    Yes indeed, they are working within what you might call a transformational theory, or, in their case, Optimality Theory. You will note that the paper is titled "Phonetics and *phonology* of main stress in Italian". From the traditional structuralist/distribution-based perspective it does not indeed make much sense to talk of distinctive vowel length in Italian phonology. However, some notion of length (or rather moraicity) can be leveraged to explain a number of facts in Italian; specifically:

    - Long vowels never precede geminates (so [fa:to] and [fat:o] are OK, but [fa:t:o] is not) and divers other segments such as the palatal sonorants (note that palatal sonorants are also like geminates in that they are said "not to undergo" raddoppiamento fonosintattico when initial, which makes sense if we assume that they are in some sense already long)
    - Conversely, when stress falls on one of the two final syllables, the stressed syllable always has a long vowel or is closed, often by a geminate if it can help it (i.e. with the proviso that final long vowels are excluded). Thus, we have the following subcases:
    -- Penultimate stress:
    --- Long vowel in an open syllable
    --- Short vowel in a closed syllable
    -- Final stress
    --- Short vowel + geminate due to raddoppiamento sintattico, if it is available
    ---- Short vowel if RS is unavailable
    Now these restrictions do not contribute much to distinctiveness, but they are quite robust, and, perhaps more importantly, seem to operate with the two distinct categories of length rather than a continuum which we would expect if lengthening were purely phonetic. So it makes sense to try to explain them in terms of discrete categories, i.e. phonology. In this case, an elegant solution is available using the device of the bimoraic trochaic foot. The bimoraic foot avoids parsing the final syllable unless forced to by the location of the stress, which is lexical. If stress falls on the penultimate syllable, the bimoraicity will have to come from some source within that syllable, so something has to lengthen. If, on the other hand, stress falls on the antepenultimate syllable, the foot can get the second mora from the nucleus of the penultimate without further ado, and there is no incentive to lengthen.
    This was the bare bones of their analysis. A similar analysis can be extended to many Northern Italian and Friulian varieties (where vowel length is undoubtedly phonological because minimal pairs exist), where vowel length in final stressed syllables has non-trivial connections to the nature of the following consonant, whereas vowels in penultimate stressed syllables are almost always simply short. (For an oveview, see Lori Repetti (1992) Vowel length in northern Italian dialects. In Probus 4(2), pp. 155-182

  16. Tommaso said...

    >a vowel is long if it is stressed and syllable-final, but not word-final; all other vowels are short.

    >That is all, and there is no difference between "parole piane" and "parole sdrucciole".

    So what of the finding of Mariapaola D'Imperio and Sam Rosenthall that stressed vowels are much shorter (as in statistically significant) in open antepenults than in penults? And what of the "traditional interpretation" that they are (phonetically) shortened (scil with respect to the length they would have when stressed in penultimate position, as I explained my understanding of it above)?

    Of course there's a clear difference between "abile" and "abbile", "sdrucciolo" and "truciolo". Obviously this whole question has been about open antepenults, not closed ones. But the latter difference is not between ['zdrutʃ.tʃo.lo] and ['tru:.tʃo.lo] in any known dialect of Italian. Would it convince you of this if I pointed out that that would be uncomfortably like the difference in English between the "putsch choice" […ʊtʃ.tʃɔ…] and the "coup choice" [..u:.tʃɔ…] for a would-be dictator. I suggest ['zdrut.tʃo.lo].

    I don't think anyone here disagrees that "povero" has a long stressed vowel, no less clearly than "poco", or that the "a" in "papero" is "as clearly long" as the first "a" of "papa". All of them typically have long stressed vowels as compared to the short stressed vowels in the corresponding closed syllables. The question is whether the "a" of "papero" etc is really phonetically clearly AS long as the first "a" of "papa" etc, of phonetic necessity, or whether the phonologically determined ranges of phonetic length are different, as Mariapaola D'Imperio and Sam Rosenthall found.

  17. mallamb

    "I suggest ['zdrut.tʃo.lo]."

    Yes, you're right, sorry. "['zdrut.tʃo.lo]" is what I meant to write; just a moment of inattention.
    I usually write [ttʃ] and [dʤ] for the "double" palato-alveolar affricates; but there are some – particularly Canepari – who prefer to write [tʃtʃ] and [ʤʤ] (or something similar).
    We all know, of course, that "putsch choice" and "sdrucciolo", "rich choice" and "ricciolo" ['rit.tʃo.lo] sound differently.

    "The question is whether the 'a' of 'papero' etc. is really phonetically clearly AS long as the first 'a' of 'papa' etc., of phonetic necessity, or whether the phonologically determined ranges of phonetic length are different."

    Yes, I'm sure the "a" of "papero" is clearly as long, phonetically, as the first "a" of "papa".
    Let me put it like that, empirically: if I start saying "pa(pero)", but change my mind after the first syllable and decide to say "(pa)pa", and pronounce the second syllable of this word, there is nothing strange in the resulting ['pa:.pa]; and vice versa.
    In other words, the first syllables of these two words (pronounced in isolation) are phonetically identical.

    I don't think that phonological considerations have anything to do with vowel length in Italian.

    So, remaining on the safer ground of mere phonetics, are there in Italian any appreciable differences in length among long vowels, so that a vowel, that is long, may nevertheless be called shorter (or even much shorter) than another long (and longer) vowel?

    Let's consider for example the phoneme /a/.
    It has two allophones, or (as I like to call them, using a word that was contrived by Arrigo Castellani) "phonotypes" ("fonòtipi" in Italian), the short [a] and the long [a:]; and their distribution in the different words is always predictable, as we know.
    Can we go further, and introduce, at an even lower level of abstraction, other categories, such as a "shorter [a:]" and a "longer [a:]"? Or are we violating Occam's razor?

    Well, of course, there may be – there are, obviously – differences among individual phonetic realizations of [a:], by different people or by the same person in different moments: they can be measured accurately in centiseconds. But they very often depend on extralinguistic (psychological, etc.) reasons, and are therefore substantially meaningless even from a strictly phonetic point of view. That's why it would be unwise to build theories based on such data. (I'm not suggesting that this is what has been done in some particular case; I'm speaking in general terms.)

    To my ears, the phonotype [a:] (for example) sounds substantially very much the same in every word (pronounced in isolation) and phonetic context and position where it occurs.

    I can't entirely rule out the possibility of there being some (slight) differences that are not completely meaningless from a phonetic point of view: somehow regular differences (as a working hypothesis, one might perhaps consider, for instance, the possibility of long vowels being a little less long before certain consonant clusters). But, even if it were so, these differences would be (very) slight, in the sense at least that the two phonotypes ([a] and [a:], etc.) would still be clearly recognizable and distinguishable from each other; and at any rate – to go back to "papero", "papa" and the like –, these differences, even if they exist (which I doubt), can't be related to the position of word stress.

    Compare "seme" ("seed") ['se:.me], "semina" ("he sows; sowing") ['se:.mi.na] and also "seminano" ("they sow") ['se:.mi.na.no] (which is an example of a "parola bisdrucciola"): the first syllable, ['se:-], is the same in the three words.

  18. I am not a linguist, so I don't understand your jargon. 'Phonemes', 'allophones' etc, I don't understand. However, I understand what it is for a language to have vowel quantity. German, Czech, Finnish have it, for instance. English, maybe; I mean 'sheet' and 'sh*t', 'beach' and 'bitch' are different not just by dint of length, there is something diphthong-like about the first vowel in both pairs.

    Italian, however, does not seem to have vowel length at all. You read three degrees of vowel-length into it, OK... . To my ear, stressed vowels, if the word is under logical or emotional emphasis, or if the word is pronounced separately, say as a demonstration for a foreigner, are usually slightly lengthened, especially if in an open syllable (but also sometimes in a close one). 'È mio paadre', say, while pointing to someone on a photograph. But if not under emphasis, stressed vowels are not lengthened, but 'just stressed'. Someone who drawled them in every position, in the middle of a sentence, say, would sound ridiculous in Italian.


Note: only a member of this blog may post a comment.