John Wells’s phonetic blog — archive 15-31 March 2007

email:

To see the phonetic symbols in the text, please ensure that you have installed a Unicode font that includes all the IPA symbols, for example Charis SIL (free download).

Browsers: Some versions of Internet Explorer and Safari have bugs that prevent the proper display of certain phonetic symbols. I recommend Firefox (free) or, if you prefer, Opera (also free).

RSS feed

The late great

Not everyone knows about Olle Engstrand’s Phonetic Portrait Gallery, Pioneers in Phonetics and Speech Research.

This page offers pictures of nearly 150 phoneticians, the pioneers of our subject. Here you can find photos of my late UCL colleagues Daniel Jones, Hélène Coustenoble, Dennis Fry, A.C. Gimson, and J.D.O’Connor, — people I am proud to have studied under — as well as of such luminaries as Paul Passy, Ferdinand de Saussure, Henry Sweet, Kenneth Pike, David Abercrombie, Shiro Hattori, and Edward Sapir. The only restriction seems to be that no living person is shown.

The gallery has been “under construction” for years and years. It may be some time since it was updated, because there is no picture of, for example, Peter Ladefoged, despite his having died over a year ago.

The page is very slow to load, since it has all the photos on a single enormous page. (To prevent the same thing happening with this blog, I show only ten or so days at a time, relegating older entries to the archive pages.)

All the pictures I display here appear to be in the public domain. Some of the others are acknowledged to a particular photographer.

    
A.C. Gimson, 1917-1985; Dwight Bolinger, 1907-1992; André Martinet, 1908-1999; Paul Passy, 1859-1940

Now you can see why I like, wherever possible, to show pictures of the (still living) colleagues who have written to me for this blog or whom I write about in it. I hope that when we are all dead and gone the compiler of this gallery or its successor will make use of some of the pictures.

To see more, go to www.ling.su.se/fon/phoneticians/Gubbar.html

Prince Nikolai Trubetzkoy, 1890-1938

Otto Jespersen, 1860-1943

An American tragedy

I found this clipping in Private Eye.

It just shows what t-voicing leads to.

(Do you know the story of Atom and Eve?)

Sounds familiar?

The British Library has made available a new collection of sound recordings under the above title. According to their website,

You can listen to 71 sound recordings and over 600 short audio clips chosen from two collections of the British Library Sound Archive: the Survey of English Dialects and the Millennium Memory Bank. You’ll hear Londoners discussing marriage and working life, Welsh teenagers talking with pride about being bilingual and the Aristocracy chatting about country houses. You can explore the links between present-day Geordie and our Anglo-Saxon and Viking past or discover why Northern Irish accents are a rich blend of seventeenth century English and Scots. You can study changes in pronunciation among the middle classes or find out how British Asians express their linguistic identity.

It sounds very useful. Obviously, this is where I must send those would-be writers of essays and dissertations on English accents who ask me for help because unfortunately, being overseas, they don’t have access to actual speakers.

The leader writer in yesterday’s Guardian newspaper got quite carried away.

In praise of... regional accents

The world is shrinking in lots of ways but happily not in all of them. The past is a foreign country with a long "a" in Dorking but it retains a short vowel in Huddersfield. In Stoke-on-Trent people still take only short baths, while in the Medway towns, he who laughs last laughs longest. If we are to believe the British Library, England, like Gaul, is divided into three parts: a south-east dominated by long vowel pronunciation of words like France and class stretching up to about Northampton, a broad west-to-east strip of flatter vowels stretching from Cornwall across to Norfolk, and an impregnable redoubt of short vowels in the north. There is something splendidly obdurate about the fact that, according to the library's new Sounds Familiar? website, there is a sort of linguistic Hadrian's Wall somewhere to the south of Birmingham. North of this invisible line of defiance against estuary English and received pronunciation, millions keep the long grass short, while to the south even the shortest grass is always long. There are enclaves and minorities on each side of the boundary, the library says, with long vowel headmasters reported from Keswick and short vowel dancers overheard in Peterborough. The persistence of all these differences is a tribute to an indomitably unbiddable spirit and may even be a useful metaphor against conformism in other fields too. The south-east controls the nation in many ways but, depending on where they are from, an English person's accent is still their castle - or indeed their castle.

I’m afraid that this still needs a phonetician to explain that ‘flatter’ means a front as opposed to a back vowel. The southeast and RP have [ɑː], the Cornwall-to-Norfolk strip (which is discontinuous) has [aː], and the north has [a].

Neutralization again

Thomas Widmann, who works for Collins dictionaries and is the author of a multilingual blog, writes:

“ I read your blog posting about Neutralization (19/3) with great interest. It made me wonder whether schwa should be used at all in a phonemic description of English, given that it only occurs in the weak-vowel system and all other vowels in that system are already members of the strong-vowel system. In other words, isn't schwa equivalent to your /T/ in /sTæmp/, and wouldn't it make sense to write 'panda' as either /'pændæ/, /'pændɒ/ or /'pændʌ/ rather than /'pændə/? ”

This is essentially what the first edition of the OED did, using notations such as ă. It was abandoned in later editions.

In some words (not panda) this kind of analysis is supported by the use of strong vowels in singing. Transcribing angel as /ˈeɪndʒĕl/ (where ĕ = weakenable e) could be justified on these grounds, even though in all kinds of speech other than singing we say /ˈeɪndʒəl/.

But it is precisely cases such as the second syllable of panda that present an interesting case. There is no style of speech in which we use anything other than /ə/, and no grounds other than orthographic for choosing one of Thomas’s alternatives rather than another. As long as it remains a weak vowel, the final vowel of panda contrasts only with what I write /i/ and /u/, i.e. the vowels of happy and thank you. Compare panda and handy. Hence, as long as we know we are in the weak vowel system, schwa does indeed represent a neutralization of /e, æ, ɒ, eə, ɑː, ɔː, ɜː/, as we see in the strong/weak alternation of them, at, of, there, are, for, sir.

Actually, there’s a problem with /e/, which in the endings -ed, -es, -est weakens not to schwa but to /ɪ/ for most speakers of RP etc. (In closed syllables there is a slightly larger weak-vowel system, which includes /ɪ/.) Such speakers formerly weakened it to /ɪ/ in -less, -ness, but now mostly weaken it to /ə/. So we have the awkward anomaly that /e/ can have two different neutralization forms, even for the same speaker.

ə

However, the main argument against this line of thought is rather different. It is that we never know from the general structure of a word whether a given unstressed syllable will select its vowel from the strong system or from the weak system. Alongside words like gymnæst (strong system) we have words like modɪst (weak system). Think about the unstressed final syllables of phoneme, Kellogg, cuckoo, syntax, torment (n.) — all strong. No neutralization there! Chomsky and Halle have some pointers to what’s going on, but in many cases it remains arbitrary whether or not vowel weakening occurs.

In this respect English differs strikingly from, say, Russian, where weakening is highly predictable once you know where the stress is located.

PS: Thomas also tells me that he read my book Lingvistikaj Aspektoj de Esperanto in a Danish translation, and that it was his first introduction to IPA and also to Greenberg’s language typology, which he then took as the topic for his MA dissertation. He says ‘Ĝi vere ŝanĝis mian vivon’ (it really changed my life). That’s the kind of comment that makes any author purr.

Txtin tym agen

In connection with my discussion of text messaging and its possible reflection of the writer’s accent (blog, 13 March), my former student Georgina Foss writes:

“I'm not sure you can assume anything about the sender's accent / dialect. Having supervised dozens of student investigations of texting, I've noted that the prime constraint is speed and space, although in a message like the one you quoted there may well be the desire to project a certain image. In my own text messages (usually to my children, from whom I've undoubtedly picked up a few texting tricks) I use an eclectic mixture: I usually g-drop as in GOIN, MAKIN; possibly also h-drop as in AVIN A GD TIME (neither of which represent my pronunciation!) The reduction of th to d as in DA or DE (the) and DIS, DAT is rather a matter of economy than th-stopping for street cred or chatting black, at least in my case; similarly WIV, NEVA and BRUVA are not my pronunciations, but are useful in text messaging. I use the standard letter homophones and number homophones as in C U L8R 2DAY, and have seen 3 used for "free" as in R U 3 2NT? The word "text" itself is interesting, as it appears to have acquired the past tense "texed" implying a simplification of the consonant cluster.”

This all strikes me as pretty sensible. Probably I built an unwarranted mass of inference on the basis of insufficient evidence.

However, Michael Ashby mentioned to me today that there are forensic linguists who use analysis of txting style as evidence of speaker identity.

And in India and Pakistan, where /v/ and /w/ are not generally distinguished, they use the letter v in text messaging to represent ‘we’. R V goin 2 do dat, 2?

txtin

The rule of three

There are plenty of words in English that seem to change their stress depending on the phonetic context. Typical examples are afternoon, unknown, sixteen. We say the 'late after'noon but an 'afternoon 'nap, 'quite un'known but an 'unknown as'sailant, 'just six'teen but 'sixteen 'people.

The usual explanation of this is that the words in question are lexically double-stressed. Dictionaries show them with a secondary stress on the early syllable, a primary stress on the later one, thus for example /ˌɑːftəˈnuːn/ or àfternóon.

I think that really the two stresses are of equal lexical status. The supposed difference between secondary and primary merely reflects the fact that when we say one of these words aloud, in isolation, the intonation nucleus necessarily goes on the last lexical stress, making it more prominent than the first.

The alternation goes by various names, including ‘stress shift’ and ‘iambic reversal’, but I call the general principle involved the rule of three. This means that when there are three successive potential accents (= syllables that could be realized with pitch prominence plus a rhythmic beat), the middle one can be, and often is, downgraded, losing its pitch prominence and possibly its rhythmic beat too.

Thus a 'nice 'old 'dog becomes a 'nice old 'dog, and 'very 'well de'signed becomes 'very well de'signed. The 'B'B'C becomes the 'BB'C, and our 'after'noon 'nap becomes an 'afternoon 'nap. Likewise an 'un(')known as'sailant, 'six(')teen 'people.

Anyhow, the point of all this is that when I was in Italy last week I got caught out through applying the same principle, wrongly, to Italian. My room number in the hotel was 202, which in English is 'two 'hundred and 'two, which by the rule of three becomes 'two hundred and 'two. In Italian it’s duecento due, which I discovered does not become *'duecento 'due. It has to be due'cento 'due. That’s because (most) Italian words can have only a single lexical stress. So ‘two hundred’ is due'cento, not 'due'cento.

The consequence is that English people sometimes put accents in the wrong place in Italian, as I did; and conversely Italians find it difficult to apply the rule of three to English double-stressed words.

202

Neutralization

In his recent email, David Deterding continues “[...] I wonder if it isn’t true that all vowels have a indeterminate, semi-reduced version in pre-vocalic environments. Now, LPD uses /i/ for the first syllable of create, which has always troubled me because it is not strictly a phoneme of English and I always believe that dictionaries should show phonemes (though of course I acknowledge that it is phonetically accurate). I further note that you are proposing to adopt this intermediate /i/ vowel for the first syllable of predict and becalm (blog, 29 January). But if other vowels [...] also have weak forms, can this non-phonemic treatment really be justified?

“To be truly consistent, if /i/ is adopted as a neutralisation of /iː/ and /ɪ/ in some environments, we should really find an intermediate symbol between /t/ and /d/ for the plosive in words such as stay and storm.”

David has a point here. Behind the use of /i/ in words such as glorious, and at the end of words like happy, lie two main considerations. One is saving space. If I pronounce /ˈhæpɪ/ but the great majority of my students say /ˈhæpiː/ — i.e. like Jones and Gimson I identify the final vowel with that of kit, they identify it with that of fleece — then we save space by using a special symbol, /i/, distinct from both, rather than by transcribing each such word twice (and there are a very large number of them).

But the other reason, the one to which David alludes, is a more sophisticated one. English arguably distinguishes two vowel systems, strong and weak. Weakening means switching from the strong system to the weak, making a strong vowel weak. Thus [ə] is the weak counterpart of strong [æ, ɒ, ʌ] (and various other vowels), as we see in the strong and weak forms of at, of, us. In exactly the same way, [i] and [u] may be seen as the weak counterpart of strong [iː, uː], as seen in me and prevocalic you, to. Because the weak-vowel system is much smaller than the strong-vowel system, in these positions we have a neutralization (in Trubetzkoy’s terminology, an Aufhebung, ‘annulment’) of some of the phonemic oppositions present in the strong-vowel system.

However, as David points out, these are not the only neutralizations we find in English. The opposition between /t/ and /d/, exemplified in tamp vs. damp, is neutralized after tautosyllabic /s/, as in stamp. So logically we might introduce a new symbol, say /T/, to show it, and instead of /stæmp/ write /sTæmp/. What applies to the alveolar plosives also applies to the labials, as in pin - bin - spin, and to the velars, as in core - gore - score.

As a further example, within the strong-vowel system, I could add the neutralization of /iː/ (bee) vs. /ɪə/ (beer) in the environment _rV, as in the first syllable of serious.

But the practical needs of EFL learners inhibit us from going any further down this path. Even the happY vowel is puzzling enough for many of them.

Torino Photo: Papariku

Opening and closing glides

Following yesterday’s discussion, here’s Masaki Taniguchi:

“The glides in j and w are opening (gliding from closer to opener position), whereas the post-central elements of such English diphthongs as aɪ and aʊ are closing (gliding from opener to closer).

“Can we use the same symbol for both an opening glide and a closing glide?

“I think the essence of j and w is their short duration (quickness) and opening glide, which distinguishes them from vowels. The post-central elements of the above diphthongs are not quick and not opening.”

He wrote a short paper describing an experiment on this: Masaki Taniguchi (1986), ‘On the use of j or (y) and w for the post-central elements of English long vowels and diphthongs’, Bulletin of the Phonetic Society of Japan.

However, what we find in experimental phonetics doesn’t necessarily determine what phonological analysis we adopt. If the opening and closing glides are in complementary distribution, we are free (if we see fit) to regard them as allophones of the same phoneme. Are they in fact in complementary distribution? It depends on whether we recognize syllable boundaries as a conditioning environment. We must be able to distinguish the beginning of RP oasis from away. If both are analysed as having /əwej/ we might be in trouble with David Deterding’s analysis. And that’s one of the reasons I abandoned the bipartite analysis myself.

Masaki Taniguchi

Bipartite long vowels and diphthongs

David Deterding writes from Singapore “I would like to offer a slightly different perspective on the choice of /oʊ/ or /ow/ for GOAT (blog, 9 March). It seems to me that /ow/ makes better sense in American English than RP British English, and the reason for this is rhoticity.

“Let me explain. RP has /ɪə/, which simply has to be represented as a diphthong. In contrast, AmE does not have centering diphthongs because NEAR is /ɪr/ (or maybe /ir/), so all the vowels of American English can easily be represented as a pure vowel followed by a glide. In many ways, won is just now backwards, so there is an argument that both should be shown as a nasal + vowel + glide (or the other way round). And this is why so many American dictionaries use the glide notation while British dictionaries do not.

“I have written it up in a bit more detail here.”

Something that David perhaps doesn’t know is that my first published article, in the Maître Phonétique back in 1962, describing my own idiolect, analysed all long vowels and diphthongs as vowel plus glide. (But since then I have changed my mind, or at least changed my practice.) Here’s a fragment.

David Deterding