I have often marvelled at the ingenuity displayed in some of the transcription systems devised by early phoneticians, such as those proposed by Bell and by Sweet, using completely original symbols to represent sounds, or Kenneth Pike's articulatory coding system that comes at the end of his book 'Phonetics', requiring a string of 80 symbols to represent the vowel [o]. Having long been out of touch with what is going on at the cutting edge of phonetics, I had rather assumed that people had decided to settle for what the IPA offers for transcription purposes, and be glad of the standardization of usage that has been achieved. I suppose I have also come to assume that the IPA framework more or less covers the range of discriminations that the average phonetician is capable of making. After all, there would be no point in having diacritics to indicate four different degrees of aspiration if nobody could use them reliably, or such a complex set of fine-grained vowel differences that nobody would be able to remember which sound corresponded to which symbol.

So it came as rather a surprise to me, browsing through the undergrowth of Wikipedia's many phonetics articles, to discover that people are still developing new ways of transcribing sounds. The first one I came across was mentioned in a recent edit to the article Phoneme, which noted that "Navlipi, a new universal script published in 2013, claims to deal with phonemes and phonemic idiosyncrasies in a practical, novel way." I had never heard of Navlipi before, so I followed it up. It turns out that there is now a Wikipedia article on Navlipi, a system which has been devised, published and patented by an Indian chemist called Prasanna Chandrasekhar. Reading the WP text, I was not impressed by this observation: "As more peculiar examples, we can cite [x] (uvular/velar fricative, the famous “throaty r“ of Parisian French and also much modern German, a sound coming from deep within the throat) ...". It seems the main selling point of Navlipi is that it is based on Roman letters and doesn’t use diacritics; apparently it makes it possible to show the Iinks between phonetic and phonemic representations in all the world's languages. There is a lot of information available on the Navlipi website here, including four (mixed) reviews. The claims made for this system are rather bold: “In addition to being a phonemic script, NAVLIPI is also a precise phonetic (phonic) script that very accurately transcribes the sounds and features found in all the world’s languages ... It is far more thorough, complete, distinct and practical than the alphabet of the International Phonetic Association (IPA)”. So far, though, my dwindling brain cells have not been able to figure out how it scores over IPA notation.

 As the WP article points out, there is also a two-volume book setting out the full Navlipi system with a vast number of examples. I was rather intrigued to see that the book contains a Foreword written by Dr. Nicholas Ostler, a well-known specialist in endangered languages (whom I met many years ago when he was evaluating DTI-funded speech technology projects). Out of curiosity, I looked at the past contributions of Zinko101, who contributed the recent note drawing attention to Navlipi. The history shows that before he or she did this edit, their only other contribution to Wikipedia has been the full article on Navlipi. That, plus a sentence just added to the Wikipedia article on Dr. Nicholas Ostler noting that he wrote the Foreword to the Navlipi book. Sometimes I think WP gets a bit incestuous.


Another recent WP discovery for me was the vast IPA-with-everything-added system devised by Italian phonetician Luciano Canepari.. This system, modestly named CanIPA, can be read about here. It is claimed to contain 500 basic, 300 complementary and 200 supplementary symbols, though I haven't counted them. Inevitably, I wonder if anyone, apart possibly from the author himself, could ever learn to make use of such a complicated system. There is a huge amount of material, published and unpublished, on the impressive-looking website, but I have found it very tough to follow the rather idiosyncratic English and grasp what the point of CanIPA is. As Alex Rotatori has pointed out very cogently in his blog, there is something strange in Canepari’s work on the pronunciation of English, in which he constructs an accent which is a fusion of “neutral American” (GA) and  “neutral British” (RP) (i.e. an accent which doesn’t exist), and then presents an account of the phonetics of this accent. 

 

I think I’ll stick with IPA.

 

 

Quite often, topics are covered several times in different Wikipedia articles, and this is certainly true of English phonology. You get a general overview in English Phonology, but you get a similar (though briefer) coverage in the Phonology section of English Language. I’ve been studying the latter and finding a lot of things that need fixing (by the time you read this, I may have fixed some of them). The coverage can be split into two areas: Vowels and consonants, and Stress, rhythm and intonation

 In addition, I'm not happy with the section much lower down called Dialects, accents and varieties, but I’ll get round to that some other time.

Consonants

We are told that “Most English dialects share the same 24 consonant phonemes”, and then (without any supporting evidence) that “Consonant pronunciation varies less between dialects than that of vowels”. I can’t think of any scientific way of measuring whether there is more inter-dialectal variation in one class of sounds than another, so the statement is meaningless. This paragraph says that some accents have an extra phoneme in “voiceless w” (as in ‘whine’), but it makes no mention of cases where accents have fewer phonemes in their inventory, e.g. where /h/, /ŋ/, /θ/ or /ð/ are not in the inventory.

The statement that “Within the same syllable, a vowel before a lenis stop is longer than a vowel before a fortis stop” brings up an old issue: the statement as it stands is correct, but the implication is that lenis consonants lengthen a preceding vowel within the syllable. This is a very widespread misunderstanding. As far as I know, the correct version is that fortis consonants shorten the preceding vowel, a process nowadays graced by the term pre-fortis clipping. Vowels which are not followed by any consonant tend to have roughly the same length as vowels followed by a lenis consonant. WP actually has a tiny article on clipping here. If anyone out there cares about pre-fortis clipping (come on, you know you do!) it would be a good idea to expand this article with some facts and references – there has been discussion about deleting it. In spite of that, the WP article on vowel length sets out a very clear statement that lenis consonants make vowels longer (using what seem to me to be completely unnecessary generative rules), so I think some harmonization is needed.

Vowels

There is a table of vowel symbols for RP and GA, and for once we have a WP article that doesn’t bother with the oddball symbols used by Oxford University Press. We are told that “Vowel length varies between dialects and between words. RP has long vowels, … but in GA they are typically shortened”, which seems to me to imply that GA has underlying long vowels like RP but something happens to them to make them shorter. When we get on to regional variation, we find a confusing couple of paragraphs that talk about how we pronounce a combination of a vowel and a letter <r>, an example of the muddle to be found elsewhere when writers mix up spelling and phonemic representation.

 

Stress, rhythm and intonation

 This section, which duplicates a lot of material in other articles, contains some badly written stuff. We are told that “English is a strongly stressed language”, a statement that needs some explanation. Later we read that “Stress in English is phonemic” – I’m not sure if anyone in phonetics nowadays would agree, but it’s certainly not how I use the word phonemic. Then we get “As concerns intonation, the pitch of the voice is used syntactically in English; for example, to convey whether the speaker is certain or uncertain about the polarity” – is this an appropriate use of the word ‘syntactically’? I doubt it. Then we get the uncritical generalization that “Rhythmically, English is stress-timed”, despite a more critical treatment of the stress-timed/syllable-timed distinction to be found in other articles.

 Plenty more to work on in this very large article.

 

 


 

 

Not really a Wikipedia matter, but I can't resist recommending the Uncyclopedia's ridiculing of WP's use of IPA phonetics here. Or if you prefer, you can read it in transcription (sort of) here.

The subject of coarticulation is a really big one in phonetics, and it has dominated speech production research and theory-making for decades. Wikipedia’s treatment of coarticulation is, therefore, really disappointing, and needs a thorough overhaul. For a start, there is only a single reference at the end of the article, and that is to a book on historical linguistics, so the few generalizations made about the topic are almost completely unsupported. The text implies that coarticulation is only concerned with the influence of a single segment upon another, adjacent, single segment, whereas one of the most interesting things that coarticulation research has shown is how its effects can spread over many segments. We get a list of names of different coarticulation theories with no explanation of what they are, and the first bullet point in the discussion of the topic only mentions assimilation of place. The reader is given no guidance on the relationship between assimilation and coarticulation. So it’s a very unsatisfactory little article. Writing a proper article on the subject will be no easy task.

At the end we get a short explanation of the difference between coarticulation and co-articulated consonants (as in the voiceless labial-velar plosive /k͡p/ found in many West African languages). This confusing terminology is not Wikipedia’s fault.

One of the Wikipedia phonetics pages that I find troublesome is the one intended to give guidance on how English is/should be transcribed phonemically in WP articles. It is called Help:IPA for English.  The problem with it lies, I believe, in the fact that it tries to be all things to all people. In order to be able to represent all accents of English using a single set of quasi-phonemic symbols, it has opted for a set of diaphonemic symbols that allow for different interpretations according to which accent is being transcribed. I won’t go into all the difficulties that arise with this here, but at the bottom of the tables provided there is a section called “Reduced Vowels” which contains a box for two symbols for “vowels that are frequently dropped”. One is superscript schwa, which is used in several pronunciation dictionaries in cases like ‘bottle’: the transcription /bɒtəl/ can be pronounced with a schwa between /t/ and /l/, or with a syllabic /l/ immediately following the /t/. That seems OK. However, I have not until now been able to see the point of the superscript /i/, for which the WP article gives the example ‘nasturtium’. I have always been sure that this word would never have an /i/ in its pronunciation – to me that would sound like a caricature of a Victorian pedant. However, having made a note to that effect on the article’s Talk page, I have been put in my place by an editor who looked it up in OED and found that the pronunciation given there is /næˈstɜːʃ(ɪ)əm/. Oh dear. 
This raises the question of whether we find optional /i/ or /ɪ/ in other English words. In the discussion on WP, I suggested that 'sentient' might be an example, as both CEPD and LPD list two-syllable and three-syllable pronunciations, but if the /i/ is elided, the preceding consonant becomes /ʃ/ or /tʃ/, so the two pronunciations don’t differ solely in the presence or absence of the vowel. The only other possibility I could come up with was the increasingly common (it seems to me) pronunciation of 'create', 'creating' as /kreɪt/, /kreɪtɪŋ/: this word could be transcribed with a superscript /i/ after /kr/. However, as far as I know, this pronunciation has not been noted in published work, so to use it here would presumably count as Original Research, which is not allowed in Wikipedia material. 

As I said I would, I wrote a short article for Wikipedia explaining what practical phonetic training is. I thought I had set this out in a factual and neutral way, with references from Jones, Abercrombie, Catford, Ladefoged etc., to back up what I wrote. However, the WP reviewer has rejected it for being too subjective and for pushing my own point of view. When I queried the decision I was told that WP doesn’t want experts writing on their own fields, but pieces written by non-specialists who just report on what other people have said. I am invited to revise it and resubmit it, but I really can’t be bothered. So I have put the article on my website here in case anyone wants to read it.

From now on, I will confine my activity on Wikipedia to making changes to existing articles that contain errors or omissions.