Wednesday, 21 July 2010

Portuguese

The pronunciation of Portuguese — particularly European Portuguese rather than Brazilian — has a number of unusual and interesting features. The vowel system contains not only the widely found i e ɛ a ɔ o u but also two central vowels, ɨ and ɐ, and furthermore as many as six nasalized vowels, ĩ ẽ ɐ̃ ã õ ũ. (French and Polish, the other two European languages known for nasalized vowels, pale by comparison.) There are also plenty of diphthongs, including nasalized ones such as ɐ̃ĩ̯. The word têm has the superficially improbable pronunciation ˈtɐ̃ĩ̯ɐ̃ĩ̯. Vowels in unstressed syllables are subject to weakening, manifested as raising and (for front vowels) centralization.

The consonant inventory includes the four liquids ɾ ʀ ɫ ʎ.

All of this is admirably set out and illustrated in a new book that has just come into my hands, Fonética do português europeu: descrição e transcrição by António Emiliano (Lisboa: Guimarães).

I am glad to say that the author uses IPA notation throughout. He supplements it by the equivalent SAMPA notation, given alongside. This is perhaps rather wasteful of space, given that a simple table of symbol equivalents would have sufficed. Nevertheless, it is good to see this ASCIIization of the IPA (whose development I oversaw) treated so seriously and in such detail.

As well as discussion and illustration of each sound of Portuguese and its representation in orthography, there is also a seventy-page Vocabulário fonético de geónimos portugueses [Phonetic vocabulary of Portuguese place names], which I shall find very useful. For LPD I shall have to correct the Portuguese phonetics of Coimbra to read ˈkwĩ bɾɐ and will need to adjust the Portuguese name of Lisbon, Lisboa, to read ɫiʒˈboɐ.

Typographically, the book is notable for being set in Gentium, a font devised by Victor Gaultney (blog, 9 May 2006) both for phonetics and for ordinary text. (Distribution and further development of the font has now been taken over by SIL.) As well as being eminently readable, it looks very distinguished. Well done everyone.

44 comments:

  1. Sounds like a treat to read.

    not only the widely found […] a […] but also […] ɐ

    Is that really so, on a phonetic rather than phonemic level?


    (Edit: Word verification for this comment is "comma" - I thought they had software make up probable but non-existing English words.)

    ReplyDelete
  2. Not a very profound contribution, but when I went on holiday to Portugal as a child, I turned on the TV in the hotel and wondered why everyone was speaking Russian.

    But I know Spanish-speakers who say they can understand Portuguese (spoken as well as written) very well after a short acclimatisation, notwithstanding the more numerous vowel distinctions.

    ReplyDelete
  3. Intersting. I associate Portuguese with Bulgarian, quite clearly because of the way the unstressed vowels are reduced. (There are secret pairs between families of languages, for those not in the know. :-) French-Polish, Spanish-Czech, Italian-Russian.)

    ReplyDelete
  4. Russian has reduced vowels as well though doesn't it, which are not always easy to adduce from the English transliteration, e.g. Dosto(y)evsky.

    Speaking of whom, I think he mentioned in House of the Dead how the Polish prisoners spoke French among themselves so the Russians couldn't understand them. I can see how French and Polish might make a pair, both full of nasal vowels.

    ReplyDelete
  5. Ah, but Russian doesn't reduce to the gloomy side.

    ReplyDelete
  6. I have always found the pronunciation of Portuguese to be almost impossibly difficult on paper when I read the Berlitz European Phrasebook over two decades ago. The way it was notated to help English readers nobody could understand it. Both me and my brother laughed whenever I tried to pronounce difficult words like homem.

    ReplyDelete
  7. What about Spanish-Greek? I always thought they sounded similar though I have no knowledge of Greek phonology. Perhaps I should change my name.

    ReplyDelete
  8. Why not? Above, I paired Romance and Slavic languages.

    It's all very impressionist and subjective - for me, Spanish and Greek aren't so parallel. Of course, there's the indifferentiated sibilant and the vowel repertoire is similar, also sharing a lack of length, though in standard European Spanish, stressed vowels in open syllables aren't lengthened.

    ReplyDelete
  9. @Lipman:
    not only the widely found […] a […] but also […] ɐ
    Is that really so, on a phonetic rather than phonemic level?

    Apparently not. R. C. Willis's 'Essential Course' (a first-class primer) says that [6] is found before nasals and in unstressed syllables, [a] elsewhere. But stress is a tricky differentiator of phonemes, as we saw with schwa and the STRUT vowel on this blog.

    "Handbook of the IPA" lists the two sounds as separate items in the inventory, and the assumption would be that they're therefore phonemic.

    ReplyDelete
  10. Thanks, but I was referring to languages in general, or customary in the trade. With the phonemic thing I referred to the fact that usually a is written for convenience, even if the sound in the language in question is rather ɐ or another a-family vowel, provided there's not another present - that's easier on the eye, and usually corresponds to the language's orthography or common transliteration.

    (Still, the "a" sounds in English are an interesting matter, in my opinion.)

    ReplyDelete
  11. The statement in R. C. Willis's 'Essential Course' is true enough, but not the whole story about a and ɐ.

    You could go along with "Handbook of the IPA" implication that they are phonemic, considering ovelha [oˈvɐʎɐ] and alho [ˈaʎu], though as the spelling suggests it looks as though the ɐ is an allophone of /e/.

    So it's not just an allophone of a. You could call it a homophone here, but there's the complication of restricted distribution and archiphonemes.

    @JW. I think you could probably get away with the β in liʒ ˈβoʌ in LPD, but what was the thinking behind the ʌ? You have ɐ in Coimbra.

    ReplyDelete
  12. And even for Coimbra I don't think you have to abolish the m completely: ˈkwĩ ͫbɾɐ or something would be quite a helpful transcription, wouldn't it?

    ReplyDelete
  13. I meant to say the ɐ is an allophone of /ɛ/. And it's a good thing I noticed that mistake, as I suspect that velho is [ˈvɛʎu], where the ɛ is an allophone of /e/. Another homophone. A sliding scale from one homophonic realization to the next.

    I really must get the book. Not that I ever thought I could fail to do so. It is so obviously something you were right to draw our attention to, John!

    I do not doubt that it has scooped me on this, but I was duty-bound to point out that the business about ɐ being determined by nasals means that although you get it as an allophone of /ɛ/ in functional opposition to a before /ʎ/, as in my above examples ovelha [oˈvɐʎɐ] and alho [ˈaʎu], there is no functional opposition before /ɲ/: tenho is tɐɲu and banho is bɐɲu.

    ReplyDelete
  14. BTW it doesn't refute my counterexamples, and unstressed [o] as distinct form [u] is still I think legit, but another suspicion is that [oˈvɐʎɐ] is in free variance with [uˈvɐʎɐ].

    ReplyDelete
  15. Why [ɫ] in Lisboa? I thought [ɫ] was only used at the end of a syllable, as in su[ɫ] ("south"). They are quite different sounds. (The same with "territorialidade". Territoria[ɫ] ("territorial"), but territoria[l]idade ("territoriality"?), because the "l" joins the "i" in a new syllable.) Also, [ã] would be a regional sound, as in standard Portuguese there are no open nasal vowels. So, in some parts of the country people will say s[ã]to (saint), instead of s[ɐ̃ ]to.

    mallamb: that's not entirely true. In northern accents, people say v[ɛ]lho and t[ɛ]nho. In standard Portuguese, you'll get v[ɐ]lho and t[ɐ]nho. Furthermore, some people, in some parts of the country, still say ov[e]lha (sheep), co[e]lho (rabbit), verm[e]lho (red) (in the North, [ɛ], in standard Portuguese [ɐ]), but I don't think anyone ever says v[e]lho (old), even in remote parts of the country. It's either v[ɐ]lho, v[ɛ]lho or even v[ɛj]lho.

    Lipman: [a] and [ɐ] are separate sounds in Portuguese, eg. pára [paɾɐ] ("stop") / para [pɐɾɐ] ("to"/"for")

    ReplyDelete
  16. mallamb: [ɔ]velha is also quite common.

    ReplyDelete
  17. Pedro,

    that's part of my point - sorry if I hadn't made clear what my question was.

    ReplyDelete
  18. su[ɫ] - wouldn't that be [suw]?

    ReplyDelete
  19. Not canonically, at any rate, Lipman

    Yes, Pedro. I can't believe I didn't mention the standard minimal pair pára [paɾɐ] ("stop") / para [pɐɾɐ] ("to"/"for") in my attempted vindication of the "Handbook of the IPA"! I got carried away with my own examples, I'm afraid.

    But bear in mind what Derek said about stress being a tricky differentiator of phonemes, as with schwa and the STRUT vowel: para [pɐɾɐ] ("to"/"for") is typically unstressed.

    ReplyDelete
  20. I can't believe I didn't see the [ɫ] in Lisboa and "territorialidade", either. What on earth are these ɫs doing here. Perhaps some other symbol is used for the syllable-final l. Suddenly I'm not so sure about this book.

    ReplyDelete
  21. Lipman: [ɫ] becomes [w] in Brazilian Portuguese.

    mallamb: yes, [a] becomes [ɐ] in unstressed syllables: r[a]pido (quick), r[ɐ]pidez (quickness, speed). Other minimal pairs: "dá" [da] ("give!") and "da" [dɐ] ("of the" - feminine), "há"/"à" [a] (there is/at) and "a" [ɐ] (the - feminine).

    ReplyDelete
  22. Lipman: it's interesting you mention Bulgarian, because it seems to me that the foreigners who come pretty close to mastering a European Portuguese accent are the immigrants from Eastern Europe (some of them!). All the other Europeans find it quite hard.

    ReplyDelete
  23. The author calls non-velarized [l] "dialectal". He writes [ɫ] - "lateral laminoalveolar dorsovelar (ou velarizada)" - even in words such as 'liliputiano' [ɫiɫipuˈtjɐnu].

    ReplyDelete
  24. Since John Cowan has not turned up, I will be the one to link to his Language X is essentially language Y under conditions Z page.

    ReplyDelete
  25. @Leo: Every time I hear Greek I think it is Spanish. Except that I can't understand it.

    ReplyDelete
  26. I've been encouraging António to draft an English translation of his exceptional book, which Evertype would be delighted to publish.

    ReplyDelete
  27. This comment has been removed by the author.

    ReplyDelete
  28. Dear Professor Wells, many thanks for your kind reference to this book.
    About velarized /l/: it is standard practice in publications and courses on (European) Portuguese phonetics and phonology to claim the existence of two alveolar laterals, “dark L” and “clear L”. I do not know why. Perhaps it is a matter of ”phonetic tradition legacy”. This distinction does not exist anymore in the standard accent which is Lisbon-based. Just yesterday I received a mail from somenone in Germany who is the coauthor of a Portuguese-German dictionary and they commented that my book was the first to get rid of this nonexisting phonetic distinction, a distinction that their dictionary no longer documents (they just use [ɫ]). The initial [ɫ] in LISBOA is *slightly* more front than in LAGO (lake) or PORTUGAL (as far as the dorsum is concerned). But this is a superficial and predictable instance of coarticulation in phonetic implementation. Universal velarization of /l/ can also be verified in the pronunciation of ES, FR and EN by many Portuguese speakers (it is one of the most noticeable features of EN spoken with a Portuguese accent). “Clear L”, i.e. a “pure” alveolar lateral contoid, is a difficult sound for most Portuguese speakers.
    -----
    Once again, thank you.
    I admire your work very much (and have done so for many years now).
    Best regards from ɫiʒˈboɐ!
    (which is reduced to [ʒboɐ̯] [Zboa_^] in fast-tempo styles; as for PORTUGAL it is pronounced [pɾtʷgaɫ] [p4t_wgaL\] in informal speech, with labialization of the first 3 contoids, or even [ptgaɫ] [ptgaL\]).
    - A.

    PS. I have noticed (for the first time) a nasty typo in the facsimile that you posted above (this book was a nightmare to typeset and to proofread): ‘carrilhão’ (carillon) has unstressed [ɐ] in the first syllable; the IPA and SAMPA forms should read [kɐʀiˈʎɐ̃ũ̯] and [k6R\i"L6~u~_^].

    ReplyDelete
  29. John, Breton is known for its nasal vowels. It has more than those in French. See the pronunciation section in Breton Grammar, my translation of Roparz Hemon's Grammaire bretonne. (Hemon's transcription isn't as accurate.)

    ReplyDelete
  30. Fernando Lamadrid22 July 2010 19:10

    By the way, why hasn't Gentium been updated? An OpenType version was to be released in 2008; what happened to it?

    ReplyDelete
  31. António Emiliano said...

    it is standard practice in publications and courses on (European) Portuguese phonetics and phonology to claim the existence of two alveolar laterals, “dark L” and “clear L”. I do not know why…

    Neither would I if "existence of two alveolar laterals" meant "existence of two alveolar lateral phonemes", as seems to be implied by the statement "This distinction does not exist anymore in the standard accent which is Lisbon-based." And there is no reason to suppose that such a distinction ever did exist, is there?

    The co-author of the Portuguese-German dictionary who celebrates the abolition of "this nonexisting phonetic distinction" obviously means a purely phonetic distinction, reflected in a transcriptional distinction between allophones, and not a functional phonetic distinction requiring the recognition of two phonemes, and moreover agrees in using the symbol [ɫ] for it. But that does seem overspecified for what I take to be a non-specialist Portuguese-German dictionary. John Wells does not have parallel AmE pronunciations with [ɫ] in his LPD, and the OED doesn't either, although like a lot of dictionaries that give comparative transcriptions for BrE and AmE, the distinctions made often do seem overspecified.

    Your dictionary looks highly specified, but quite appropriately so for a phonetic dictionary, in that it is highly specified for allophones. You don't entirely deny the existence of allophones for /l/ in your examples LISBOA, LAGO and PORTUGAL, for what are allophones but "superficial and predictable instances of coarticulation in phonetic implementation"? And I have certainly always delighted in the very allophony for PORTUGAL which you so graphically transcribe as [pɾtʷgaɫ] and [ptgaɫ].

    When I took up Pedro's point about your use of [ɫ] in syllable-initial position, and before [i] at that, you can see from my comment above that I did allow for the fact that you must be marking (I thought over-enthusiastically) the strikingly "dark" character of the Portuguese /l/ even there, but I supposed that you might nevertheless have some other, perhaps modified, symbol in such a precise transcription as we see in the facsimile for what both Pedro and I thought was a characteristically darker syllable-final /l/.

    So I have indeed always been struck by the "universal velarization of /l/… by many Portuguese speakers", and observed that '“Clear L”, i.e. a “pure” alveolar lateral contoid, is a difficult sound for most Portuguese speakers', but so it is for most AmE and even BrE speakers, and I am surprised that you confirm that you do not recognize any gradation in this Portuguese velarization worth noting even in such a precise phonetic transcription as yours. Unfortunately I have not been able to contact a NS informant to seek confirmation that this purely phonetic difference "does not exist any more in the standard accent which is Lisbon-based."

    A pity about the typo, but although we were talking about [a] only occurring in stressed syllables, none of us seems to have noticed it, and I certainly still intend to get your book. I can see that my life depends on it, even though you have confirmed my worst suspicions that the ɫ was not a typo.

    ReplyDelete
  32. This comment has been removed by the author.

    ReplyDelete
  33. I can confirm that all of the speakers of European Portuguese whom I have heard speaking English fairly uniformly substitute [ɫ] for English [l] and [ɫ]. So what António is saying is that the allophone in LISBOA, while a bit [l]-ish is still heavily velarized; it doesn't become [l]. I should think one would put a diacritic on the [ɫ] in a close transcription, rather than to use [l].

    ReplyDelete
  34. Michael, I think you mean he could have put a diacritic on the [ɫ] to show the relatively lighter velarization we all seem to agree on.

    Well we can see he didn't, but we couldn't see on this page what he did for the relatively more heavily velarized allophones, so I surmised in my earlier posts that he might put some diacritic (which is what I meant by "modified") on the [ɫ] for those. Now we know he doesn't.

    I don't think we can tell from Portuguese pronunciations of English whether his otherwise extremely precise and informative phonetic transcription benefits from the total abolition of the established practice of acknowledging by means of the most readily available allophonic symbols what are admittedly ranges at the two extremes of the continuum I referred to above as a gradation of velarization, even if the whole of that continuum is shifted towards the velarized end of the spectrum. These are IPA symbols, not Canepari's, who was discussed on a recent thread, and the IPA allows for phonetic as well as phonemic symbols to permit of language-specific definition.

    ReplyDelete
  35. I'm intrigued by the use of SAMPA: surely I can see its merit when IPA is not possible to use, but I would think that anyone who is familiar with it would at the very least be equally familiar with IPA, so what is the point in providing both?

    ReplyDelete
  36. I don't know anyone who reduces Lisboa to [ʒboɐ̯]. Surely Prof. Emiliano meant [lʒboɐ̯]? Also, dropping the "r" in "Portugal" would sound very strange. As for the [l]-[ɫ] distinction, it seems to me that sometimes (if not always) initial "l" is pronounced quite differently from [ɫ]. I'll have to buy the book and analyse all the arguments, I suppose.

    ReplyDelete
  37. @mallamb: it seems we may have a conceptual or terminological misunderstanding here.
    «superficial and predictable instances of coarticulation in phonetic implementation" are NOT allophones. I don’t mean to start a discussion here about the nature of allophones. There was such a discussion back in 1999 in the LINGUIST list (“What Exactly Are Allophones?”) and it raged on for a full month before a moderator finally put an end to it. I printed the whole thing in Hoefler 10pt (1.5 space) and it turned out to be 60 pages long! No consensus was ever reached, which was, I suppose, to be expected.
    I very clearly state in my book that systematic phonetics and phonetic implementation are different levels of analysis. In the words QUILO [ˈkiɫu], AQUELE [ɐˈkeɫ(ɨ)], AQUELA [ɐˈkɛɫɐ], CURA [ˈkuɾɐ], COR [ˈkoɾ], CORA [ˈkɔɾɐ], CARA [ˈkaɾɐ] the initial contoid is *phonetically* a [k]. However there are “considerable” implementational differences. In some words the [k] is more front/palatal, in some it is labialized, according to the features of the vocoid that follows.
    Allophones are discreet segments, i.e. phonetic categories. That’s what phonetic transcription is supposed to capture IMHO, no matter the degree of comparative-ness one chooses to adopt. Cooarticulation *may*, as you know, result in allophony and even in some cases phonemic split. What was once a palatalized [k] in Late Latin is today [s] in PT.
    I see no need to represent types/degrees of velarized/velar [ɫ]. The minimal difference in frontness or backness of the lateral is *always* a function of the following vocoid. In a broad procotol of transcription there is no problem in representing (as for instance the pronouncing dictionaries of EN do) the [ɫ] phone with the simple symbol. I adopted a narrow comparative solution here because an alveolar velar(ized) lateral and an alveolar lateral are different phonetic categories. Different segments. Different entities, concepts, “things”.

    (note: a considerable number of Portuguese children fail to master the articulatory configuration for [ɫ] (I can’t provide any quantitative data re this phenomenon); most will eventually figure out how to produce a lateral with two C-places; some do not; what most do is substitute a labiodental approximant [ʋ] — SAMPA [v\] — (with differing degrees of lip rounding) for [ɫ]; in syllabic coda a velar vocoid [u̯] may be substituted for [ɫ] (which then becomes part of a diphthong) by children who use [ʋ] in syll. onsets. When you try to “force” such a child to pronounce a lateral (a dark L) in both syllabic contexts what happens is that they tend produce a clear L. They clearly have a motor coordination problem with the two places of articulation and the open approximation of [ɫ]. Some parents take these children to speech therapists. No comments. End of note.)

    The Portuguese phoneticist Gonçalves Vianna, who was a penpal of Henry Sweet, with whom he maintained lengthy discussions, admirably recognised and transcribed more than 30 oral vocoids in the Lisboa-dialect. Today we make do in our descriptions with 9 or 10 oral vowels and 6 nasal vowels (yes, [ã] does exist in Standard pt-PT in sandhi contexts; cf. “a antologia” [ãtuɫuˈʒiɐ], “da antiga” [dãˈtiɡɐ]). Are our descriptions of PT poorer or less accurate because of this? I think not. We simply acknowledge that some distinctions are not categorial phonetic distinctions and we choose to work with a reduced set of vocoids. We filter/interpret our data in a different manner than Vianna.

    I say 9 or 10 oral vowels because of the problem of [ɐ]. Stressed [ɐ] and unstressed [ɐ] are in fact different vowels. The latter should more accurately be transcribed as [ə]. I do discuss this in my book and do favour the transcriptional distinction (although I did not adopt it). Perhaps I will in my “Introdução à Fonologia do Português Europeu” (in progress). I think that one should be tactful regarding legacy matters even in phonetic transcription.
    Thanks for the comments.
    Cumprimentos cordiais. - A.

    ReplyDelete
  38. @António: I don't really think we do have a conceptual or terminological misunderstanding here. I think it's just a matter of what degree of allophony we can agree to mark for different purposes.

    I did ask what "superficial and predictable instances of coarticulation in phonetic implementation" were if not allophones, intending it of course to be a rhetorical question, but you have answered it to my satisfaction. I realize "systematic phonetics" has historically been a term for phonemics or phonology, variously defined or not properly defined in different schools and that "phonetic implementation" is an uncontentious term for an allophonic level of analysis. It was just that not having yet read your book, I did not find that this characterization of what I would call allophones in the examples you gave as such an "instance of coarticulation in phonetic implementation" enlightened me very much. And I am agog to discover what you do mean by phonology in “Introdução à Fonologia do Português Europeu”. In a recent blog entry John Wells said of French nasalization "Back in the days of generative phonology, people analysed bon bɔ̃ as underlyingly #bɔn# and bonne bɔn as underlyingly #bɔn+ə#. The masculine form then underwent a rule changing Vn into a nasalized vowel: Vn → Ṽ / _{#,C}. How this is handled in these days of Optimality Theory someone else will have to tell us." And someone did. I thought I was reading aright between the lines of all that, but I am keeping an open mind about your work in progress.

    There is of course no tradition or legacy, however otiose, of marking the k allophones in your examples QUILO [ˈkiɫu], AQUELE [ɐˈkeɫ(ɨ)], AQUELA [ɐˈkɛɫɐ], CURA [ˈkuɾɐ], COR [ˈkoɾ], CORA [ˈkɔɾɐ], CARA [ˈkaɾɐ], and it would make no more sense than marking that degree of allophony in English, for example. For as you say, allophones are discrete segments, i.e. phonetic categories, which phonetic transcription is supposed to capture, but it can only do this by bringing this discreteness into being within the framework of the relationship between phonology and phonetics, which are themselves systems of relationships between relationships. But it can go on to make ever more precise refinements of the observations of the purely realizational phenomena, as appropriate to the purposes of particular analyses. And I wouldn't include diachronic ones any more than you apparently would.

    And again like you, I see no need to represent types/degrees of velarized/velar [ɫ], and I never have seen any, because by "need" you appear to mean "absolute need", or what I would call functional necessity. Of course the minimal differences in frontness or backness of the lateral are a function of the following vocoid. I would add "and of the preceding one", but not to the same extent, I suspect.

    And by saying you adopted a narrow comparative solution in this context, I can only think that you mean cross-linguistically, to reflect the fact that a significant range of the allophones of this phoneme (though not I think all of them) are significantly darker than in comparable languages, rather than adopting a language-specific solution. I still haven't heard back from any NS informant other than you about all this. I think your solution may have caused some consternation further afield than on this blog!

    Thanks for the entertaining note about Portuguese children who use [ʋ] as a convenient get-out! I wonder if any parents whose children use it for English r take them to speech therapists. I can hardly believe any ever did. Though if they took them for anything else, I'm sure there would have been speech therapists who would have had a go at the [ʋ] as well!

    ReplyDelete
  39. @António ctd.

    I think we can all celebrate the fact that for all practical purposes we filter/interpret our data in a different manner than the admirable Vianna.

    I look forward to the discussion in your book of whether the unstressed [ɐ] should more accurately be transcribed as [ə], though [ə] is not usually thought of as a model of accuracy!

    ReplyDelete
  40. I think "d" between vowels is pronounced more like a "th", like the "th" in "this".

    ReplyDelete