My phonetic blog — June 2006

I’ve been doing a consultancy job for a publisher, listening to synthesized speech automatically generated for a handheld computerized dictionary currently under development — not isolated headwords, but phrases and sentences. Speech synthesis is much better these days than it used to be, but it still has some way to go, particularly in the matter of working out a plausible intonation pattern.

I’ve listened to a thousand randomly selected samples. So I can tell you both what the commonest type of error was, and what the worst type was. (How would the software have coped with the sentence I’ve just written? It ought to have been able to recognize the need for contrastive tonicity: both what the commonest type of error was, | and what the worst type was. Would it have succeeded? No.)

OK, then: the worst type of error was the wrong choice between homographs. English has a number of words where the same spelling is used for two words of different sound and different meaning. Thus wind (air movement) is /wɪnd/, but wind (turn, twist) is /waɪnd/; entrance (way in) is /ˈentrəns/ but entrance (bewitch) is /ɪnˈtrɑ:ns/; present (noun or adj.) is /ˈprezənt/, present (verb) is /prɪˈzent/. In the thousand samples I listened to there were five errors of this kind. Obviously, choice of the wrong pronunciation would be seriously misleading for the dictionary user.

However, the most frequent error was failure on the part of the speech synthesis software to recognize compound nouns and therefore to apply compound stress. In the great majority of cases of noun + noun in English it is appropriate to put the primary stress on the first element, as for example adˈventure story, appliˈcation form, ˈmurder charge. (Of course, you need to be aware of the occasional exceptions, such as strawberry ˈjam, gang ˈwarfare.) The software got this wrong in 65 cases, constituting just over a quarter of all errors identified.

You could see this, alternatively, as a demonstration of a fault in English spelling conventions. In the other Germanic languages, compound nouns are written solid, with no space (thus German Abenteuergeschichte, Anmeldeformular, Mordanklage). So you can recognize them as such. English sometimes does this (wintertime, railway) but very often doesn’t. If we wrote adventurestory, applicationform, murdercharge and so on, then the software (and millions of EFL students) might get the stress right.

The bacterium Clostridium difficile has been in the news recently. It is a cause of diarrhoea, usually acquired in hospital.

My interest, as always in this blog, is in how it is pronounced. Like all organisms, it has a Latin binomial name, the first part referring to the genus and the second to the species. There’s no difficulty about how to say the generic name: clearly, /klɒˈstrɪdiəm/. This can also be abbreviated to C., and in this form is often pronounced just as /siː/.

It’s the specific name that raises the interesting question. Apparently this species of Clostridium is called difficile because when it was first discovered it was difficult to grow in the laboratory. In Latin, difficile is the neuter singular nominative of the adjective difficilis ‘difficult’ — neuter because it agrees in gender with the neuter Clostridium. (With a masculine or feminine noun it would be difficilis.) In accordance with the usual rules for pronouncing Latin words in English, you would expect it to be /dɪˈfɪsɪleɪ/ or perhaps /dɪˈfɪkɪleɪ/.

But it so happens that difficile is also the French word for ‘difficult’. And in French it is pronounced /difisil/, anglicized as /ˌdɪfɪˈsiːl/. Because doctors and other health care professionals these days are much more likely to know French than to know Latin, it is perhaps not surprising that the bacterium is generally known in English as /klɒˈstrɪdiəm ˌdɪfɪˈsiːl/.

Or C. diff. /ˌsiː ˈdɪf/ for short.

As a consequence of doing a Google search on svarabhakti (yesterday’s blog entry) I have come across a most informative site dealing with the (Scottish) Gaelic language: www.akerbeltz.org.

It includes an exhaustive account of the remarkable phonetics of this language (go to Fuaimean na Gàidhlig - The Sounds of Gaelic). The site makes extensive use of IPA symbols, and is supported by .mp3 sound files. As far as I can tell, it seems thoroughly accurate and well-informed.

The name of the language, in Gaelic, is Gàidhlig, pronounced [g̊aːlɪg̊ʲ]. The consonant system includes dental and palatal plosives, palatal and velar fricatives, three laterals and three r-sounds. There are no voiced plosives, but fortis consonants are distinguished from lenis by being aspirated or (in non-initial position) preaspirated. Consonants are divided into ‘slender’ (palatalized) and ‘broad’ (non-palatalized), and come in slender-broad pairs.

The vowel system includes back unrounded vowels [ɯː] and [ɤː], as in caol ‘narrow’ and ladhran ‘sandpiper’ respectively.

Like the other Celtic languages, Gaelic has initial mutation (lenition) of consonants in various grammatical or positional environments. Now I understand why the Gaelic for James, written Seumas, has come into English as Hamish. In the vocative case, masculine names undergo both lenition of the initial consonant and ‘slenderization’ of the final consonant. Hence Seumas [ʃeːməs] becomes a Sheumais [ə heːmɪʃ].

One niggle: I wish the author would spell pronunciation correctly. I drew his attention to his “pronounciation” on the front page, which he has corrected; but the error remains elsewhere, e.g. twice on the Vowels page.

And if I were organizing the recordings, I would have insisted that the speaker use the same intonation on all example words. As it is, his tendency towards listing intonation (rise, rise, fall) might make the listener imagine that Gaelic is a tone language.

“Eng-ger-land! Eng-ger-land! Eng-ger-land!” chant the patriotic football crowds, cheering on the English team.

But why do they feel the need to add an extra syllable? Why do they change change /ˈɪŋ.glənd/ into /ˈɪŋ.gə.lənd/?

There’s nothing phonotactically difficult for English speakers in syllable-initial /gl/. Why break up the cluster with an epenthetic schwa? Why the svarabhakti?

(Some speakers don’t have a /g/ in England. They have even less excuse.)

Here is a Scottish gloss on the matter.

There’s only one other instance of the phenomenon in English that comes immediately to mind: the widespread pronunciation of athlete, athletics as /ˈæθəli:t, ˌæθəˈletɪks/ instead of the standard /ˈæθli:t, æθˈletɪks/.

The other day I did hear someone pronounce template as /ˈtempəleɪt/ instead of the usual /ˈtempleɪt/. (A hundred years ago it was a /ˈtemplɪt/. But nowadays the spelling pronunciation prevails.)

The explanation must be that the phonetic C+l sequences could also arise from a process of compression (schwa deletion), as when tingling (from the verb to tingle) is compressed from /ˈtɪŋgəlɪŋ/ to /ˈtɪŋglɪŋ/, /ˈkæθərɪn/ Catherine to /ˈkæθrɪn/ or /ˈdʌbəlɪt/ double it to /ˈdʌblɪt/. (I ignore here the possibility of syllabic [l̩] from /əl/, which is irrelevant to the argument.) Apply this process in reverse, and we get the svarabhakti forms. (But it doesn’t explain why it applies only to a very few lexical items.)

This, together with morphological regularization, presumably explains the occasional /ˈwɪntəri/ for wintry or /rɪˈmembərəns/ for remembrance.

Eng-ger-land fans, don’t let it spread any further. Please! Or as we say nowadays, puhleeze!

“Final alveolar consonants, i.e. /n/, are often subject to assimilation”, writes an undergraduate in an examination answer.

But she doesn’t mean i.e., she means e.g.. The first stands for ‘that is’, the second stands for ‘for example’. The consonant /n/ is not the same as (all) alveolar consonants, it is an instance of an alveolar consonant.

Younger native speakers often confuse i.e. and e.g.. Both these abbreviations stand for Latin phrases, id est and exempli gratia respectively. You could blame their confusion on the decline of classical learning: not many people learn Latin these days. On the other hand there are millions of people who don’t know Latin but who manage to use them correctly.

You might also think that we ought to have English abbreviations for English phrases. Foreign learners, in fact, sometimes suppose that to be the case. I have several times come across “f.e.” or “f.ex.” in English essays written by Germans. After all, the German equivalent of for example is zum Beispiel, abbreviated as z.B.. I have to explain that “e.g.” is the only way of abbreviating this expression in English.

Another English abbreviation I have seen invented by Germans is “a.s.o.”. What does it stand for? Obviously, “and so on” (German und so weiter, u.s.w.). But the English for that is etc.. It’s Latin again: et cetera.

A popular (mis)pronunciation of this latter phrase is /ɪkˈsetrə/, as if it began ex-.

How do you pronounce artisanal?

That was a question a correspondent asked me two or three years ago. At the time I had to reply that I did not know the word: it wasn’t part of my vocabulary and I’d never heard anyone say it. However, despite the fact that artisan is /ˌɑ:tɪˈzæn/, I thought the regular stress effect of the suffix -al ought to yield /ɑ:ˈtɪzənəl/. Adjectives in -al, of three or more syllables, have penultimate stress if the penultimate vowel is long (archetypal, primeval, universal) or followed by a consonant cluster (dialectal, incidental), but otherwise have antepenultimate stress (personal, industrial, medicinal). (There are one or two exceptions in which the penultimate vowel lengthens on adding -al, as adjectival /-ˈtaɪvəl/.)

The word artisanal is not in the OED. It seems to have come into use quite recently as a kind of antonym of industrial, referring to small-scale production methods, agricultural or other. On Google it has four million hits.

Answers.com would have us believe that it is /ˈɑ:tɪzənəl, ˌɑ:tɪˈzænəl/. I don’t think so. That looks like someone’s guess &mdash someone who hasn’t fully absorbed the English stress rules (blog, 21 June).

Early this morning I was jolted out of my half-sleep by hearing someone on my bedside radio, on the farming programme, use the word not once but twice. I can’t remember what they were talking about, but my semi-conscious mind did note that the pronunciation used was indeed /ɑ:ˈtɪzənəl/. Result!

I hope you like the minor redesign of this page. The narrower columns should make the text more readable.

I hope too that you can read all the various characters correctly: not only the IPA symbols but also the odd bits of Greek or other writing systems. There are two requirements here: you must have a Unicode font with the relevant symbols installed on your system, and your browser must be set up in such a way as to make use of them.

My own preferred browser is Firefox, running under Windows 2000/XP. It has the distinction of being much more tolerant of faulty HTML code than Internet Explorer or Netscape. This is generally an advantage to the user, who sees the intended symbols even if they haven’t been coded exactly correctly. But it is a disadvantage to the web page author, who seeing that everything is correctly displayed on his own browser may suppose that other browsers too are displaying it correctly even when this may not be the case.

In the style sheet (.css) that accompanies this page I specify the font Trebuchet MS for the text. I like this font. I find it clear and legible and aesthetically more pleasing than the ubiquitous Arial. In particular, I like the fact that a lower-case Ll looks different from an upper-case Ii. However, if you haven’t got Trebuchet MS on your system then you’ll see the text in whatever is the default font for you.

Trebuchet MS does not contain IPA symbols. So in the style sheet there is a line of code as shown alongside. This should cause the browser to select one of the named fonts whenever I mark a bit of text as phonetic by tagging it with the markup code <span class="phon">. These fonts do contain the IPA symbols, and your browser should go through them in turn until it finds one that is installed, and then use that one.

span.phon {font-family:"Charis SIL", "Doulos SIL", "Gentium", "Lucida Sans Unicode", "Arial Unicode MS", "MS Mincho"; color:darkgreen;}

ˈhɪər ə səm fəˈnetɪk ˈsɪmbəlz

Similarly, I mark up bits of Greek with the code <span class="gk">. This actually doesn’t matter for Modern Greek, the symbols for which are included in Trebuchet MS. But it is vital for classical Greek, which requires ‘polytonic’ symbols involving quite complex combinations of diacritics (rough and smooth breathings; acute, grave and circumflex accents; iota subscript). Not many fonts are equipped for polytonic Greek, and if you haven’t got one of the fonts named alongside your system probably cannot display it.

span.gk {font-family:"Palatino Linotype", "Gentium", "Arial Unicode MS", "DejaVu Sans";}

ἐν ἀρχῇ ἦν ὁ λόγος

For things to be displayed correctly in Internet Explorer, all this extra coding is essential. It doesn’t matter for Firefox, because clever Firefox, if it comes across a symbol not found in the current font, searches for a font that does contain it. IExplorer merely displays a blank square, and Netscape a boxed-in question mark. If you see a strange string of symbols like &#x1f10, it means that I have forgotten the final semicolon in the code for a Unicode symbol I can’t keyboard directly (but Firefox would nevertheless display the correct character).

Kensuke Nanjo reminds me that the question of s → ʃ / _ r was addressed in three of the items in Bert Vaux’s North American dialect survey, available on-line. Of 10,981 respondents who voted there on the s in anniversary, 94% identified it as [s], 6% as [ʃ].

As you can see from the map (reproduced, reduced in size, from the University of Wisconsin Milwaukee website), there was no particular geographical basis to the distribution of preferences: the blue dots, representing [ʃ], are scattered thinly across the east, midwest, and west. The red dots, [s], are everywhere. (Go to detailed maps.)

For the s in nursery, the figures were similar, but with a few more votes for [ʃ]. 88% thought the fricative was [s], 11% thought it was [ʃ], and 1% thought it was something else.

For the c in grocery, however, things were much more evenly divided: 52% thought it was [s], 45% [ʃ], and 2% ‘other’. It is not at all clear why the figures for grocery are so different from those for anniversary and nursery. I wonder how the respondents would have voted on miserable (not in the survey).

If you go along with Chomsky and Halle in The Sound Pattern of English (1968), native speakers of English have as part of their grammar (= their implicit knowledge of the language) a complex set of rules allowing them to produce the appropriate word stress and vowel qualities for English words, while the words themselves are specified in our mental lexicon without indication of stress placement and possible vowel reduction.

Chomsky and Halle were wrong, of course. The fact that anyone was uncertain about how to pronounce hypernymy (19 June) bears this out. The relatively familiar words anonymous, eponymous, synonymous, synonymy show us clearly that the -nym- element has a short vowel. This plus the single consonant /m/ gives us a so-called weak cluster, so that when -y or -ous is added the stress is thrown back onto the antepenultimate. So the same thing must happen in hypernymy (despite its being ill-formed).

I have recently noticed another counter-example to the SPE principle. Every adult native speaker knows the words paralysis and analysis. Some also know one or more of catalysis, dialysis, electrolysis or some fifty other technical words with the same ending. All have antepenultimate stress: /pəˈræləsɪs, əˈnæləsɪs, kəˈtæləsɪs, daɪˈæləsɪs, ˌelekˈtrɒləsɪs/ etc.

Two years ago I was diagnosed with a heart condition and underwent an angioplasty procedure. Since then I have attended the Cardiac Support Group run by my local hospital. One of the techniques for treating a heart attack is known as thrombolysis. But I have noticed that the cardiologists, cardiac surgeons, paramedics and nurses don’t call this /θrɒmˈbɒləsɪs/, as you would predict from all the other -lysis words. They call it /ˌθrɒmbəʊˈlaɪsɪs/. Ah well, language doesn’t obey rules as much as some linguists would like.

In my lectures I sometimes talk about allophony. Guess how I pronounce it. Remember that everyone agrees that telephony is /təˈlefəni/ and cacophony is /kəˈkɒfəni/, and that I am someone who plays by the rules.

Kensuke Nanjo writes from Japan, “I have a question about the pronunciation of the word classroom. If I remember correctly, Jennifer Jenkins pronounced the word classroom like claashroom ([ʃ] instead of [s] before [r]) when she gave a lecture a few years ago in Osaka. Although this pronunciation isn’t recorded in any of the dictionaries, including pronunciation dictionaries, is it a common pronunciation in RP in particular and British English in general?”

I can’t find any discussion of this possibility in the literature, although I think it is by no means unknown. I do mention the possibility in one of my lecture handouts, where I list various environments in which the assimilation of /s/ to [ʃ] may occur.

You can get this wherever /s/ and /r/ abut: across word boundaries as in this reason, between the parts of a compound as in the classroom that Prof Nanjo heard, or indeed within a word as in the compressed version of nursery [ˈnɜːs(ə)ri, ˈnɜːʃri] or grocery.

There can be assimilation of /z/ to [ʒ] in the same environments, e.g. miserable [ˈmɪʒrəbɫ̩].

I have no statistics about how widespread assimilation before /r/ is. I don’t do it myself, and it’s certainly not a mainstream pronunciation.

I suspect that the assimilation product may in some cases not be identical with an ordinary [ʃ, ʒ], but rather a kind of retroflex [ʂ, ʐ], anticipating the place of articulation of the /r/.

Rather more widespread, and on the increase, is /s/ → [ʃ] before /tr/, as in strong. Many of my native-English students have this at least as an optional variant.

The str → ʃtr assimilation joins the stj → stʃ → ʃtʃ assimilation as phonostylistic phenomena we shall soon have to teach our EFL students. I call them s-affricate assimilation.

Does anyone know of any studies of these phenomena?

One of our Modern English Language MA students has chosen hypernymy as a dissertation topic. There is some hesitation about how to pronounce this word: where does the stress go? I would say that following the pattern of synonymy /sɪˈnɒnəmi/ it must clearly be /haɪˈpɜːnəmi/.

This got me thinking about the etymology of this group of words. The classical Greek for ‘name’ is ὄνομα ónoma, stem ὀνοματ- onomat-, which we see in onomatopoeia and onomastics. So why isn’t it *synonom, *synonomy etc? Where does the y in -nym come from?

Dictionaries say that synonym is actually from an adjective συνώνυμος synó:nymos ‘having the same name’, which is composed of syn- ‘with’ and the ‘name’ element. I haven’t found a clear answer to the question why -onom- changes to -onym- in the course of adjective formation: but it appears that ὄνυμα ónyma was the Aeolian or Doric dialect version of ὄνομα ónoma. Using this dialectal variant obviates a possible confusion that might otherwise have arisen with -nom- ‘law’, as in metronome, astronomy, economy.

According to the OED, the earliest meaning of hyponym is ‘a name made invalid by the lack of adequate contemporary description of the taxon it was intended to designate’ (1904), i.e. a biological name that is unusable because its meaning is unclear. As a linguistic term — with a quite different meaning — it is rather recent, the earliest citation being from John Lyons in 1963. Hyponymy is slightly earlier (1955), and in linguistics refers to the relationship between such pairs as scarlet and red, tulip and flower. Scarlet is a hyponym of red because anything that is scarlet is also necessarily red, although the converse does not hold.

Hypernym and hypernymy are so new that they are not included in the second edition of the OED. They describe the converse relationship to hyponym and hyponymy. Thus red is a hypernym of scarlet, flower is a hypernym of tulip. A synonym of ‘hypernym’ is ‘superordinate term’.

Hypernym and hypernymy are obviously modelled on hyponym(y). However, from the point of view of etymological morphology, they are actually not well formed. The elements of hyponym are not hypo- and -nym but hyp- and -onym, as indeed we see when we replace hyp(o)- by syn- to give syn-onym. (So also an-onym-ous, ‘nameless’.) Like other Greek prefixes ending in o, hypo- loses the o when followed by a stem beginning with a vowel, thus hyp-onym. So when we change hyp(o)- to hyper- we ought to get not hypernym but hyperonym, not hypernymy but hyperonymy. And I think we would all agree on how to pronounce the latter. It would have to be /ˌhaɪpəˈrɒnəmi/.