May 16, 2004

The International Phonetic Alphabet

Those of us here on Language Log who cite exotic languages or talk about phonetics frequently use the International Phonetic Alphabet. Mark referred to the IPA explicitly a few days ago, but usually we use it without explicitly indicating that that is what we are doing. Since non-linguists may find this confusing, and since the IPA ought to be used more widely than it is, I thought I'd talk about it a little.

The IPA is a system for transcribing speech. It attempts to provide a symbol for every distinct speech sound found in some human language. For instance, the sound at the end of ring is represented by the symbol ŋ, while the sound at the beginnng of this is represented by ð. Since there are several hundred such sounds, some of the symbols are composite. For instance, aspirated consonants are written with a small superscript h after the symbol for the corresponding unaspirated consonant. For example, [p] stands for the unaspirated sound, as in English spot; its aspirated counterpart, as in pot, is [pʰ]. The IPA is intended only to record those differences in sound that are distinctive in some language. However, diacritics are provided that allow finer detail to be recorded. The IPA is defined by the International Phonetic Association, which revises it from time to time as new speech sounds are discovered. The very first version of the IPA dates to 1886; it had attained essentially its current form by 1949. The changes since then consist almost entirely of the addition of symbols for the more exotic speech sounds.

Here is the official chart of the current version. If you're using it on-line, you may find this resizable version more convenient. The IPA proper is intended for describing normal speech, but there are extensions for transcribing disordered speech, described here [PDF]. There is a nice exposition here of a subset of the IPA, with large versions of the symbols, drawings of the articulatory configuration, and audio examples. The fullest exposition of the IPA is to be found in the Handbook of the International Phonetic Association (also in paperback). Audio files containing the illustrations from the Handbook can be obtained here.

Within the field of linguistics, and in related areas such as speech pathology, the IPA is widely used and has been a great success. So long as everyone uses the IPA, if one person writes something down, everyone else has quite a good idea of what it sounds like, whether or not he or she has ever heard the language in question. If someone writes about a certain sound, there is no question what sound is intended. To be fair, the IPA is not so good at representing things like tone and intonation; we still don't understand prosody well enough for there to be a reliable universal system of transcription for it. But so long as we are dealing with the segmental aspect of speech, the IPA does a very good job.

Although the IPA is intended to be an international standard, even people who use it don't always adhere to it strictly. This is partly a matter of idiosyncrasy and of national traditions, but also partly a matter of the use of typewriters. For instance, the IPA symbols for the voiceless and voiced post-alveolar fricatives, ʃ and ʒ, do not appear on normal typewriters, especially English language typewriters, meaning that anyone writing in or about a language with these sounds, which are quite common, had to leave space for them and write them in by hand. Moreover, ʃ is rather difficult to draw. If you aren't careful, it ends up looking like an ordinary <s>, and if you try to distinguish the two, you are likely to make the ʃ too tall, with the result that it sticks up into the line above or down into the line below. As a result, many linguists have used the symbols š and ž instead. These have the advantage that the base is found on the typewriter, so one doesn't have to leave space for them, and the diacritic ̌ is easy to write in later. A list of the typical North American deviations from the IPA can be found here.

Nowadays, with computer word-processing there is no reason not to use standard IPA symbols. The fonts are readily available. If you use a Unicode editor such as yudit or a Unicode-capable word-processor, you might try the Code 2000 font, available here. This font covers pretty much all of the Basic Multilingual Plane, which includes all of the IPA symbols. A smaller font, covering just the IPA, variants of the Roman alphabet, Greek, and a few others, is the Lucida Sans Unicode font. If you use Microsoft Word, you can download one of the free fonts available from the Summer Institute of Linguistics here. Depending on your software and your preferences you will enter the IPA symbols in different ways. It may be necessary to know the numerical codes, either to enter them directly or to set up your keyboard. Lists of numerical codes can be found here, here, and here.

You'd think that with a successful standard system for representing speech sounds in place for over half a century it would be widely used in contexts in which people want to indicate pronounciation, but outside of specialist publications, that isn't so. Take dictionaries. Both monolingual dictionaries and bilingual dictionaries intended for speakers of languages other than English usually indicate the pronounciation of words. Do they use the IPA? Almost never. It seems as if every dictionary has its own system for indicating pronounciation. That makes it painful to move from one dictionary to another, but what is worse, it means that if you don't happen to be familiar with the variety of English, or language other than English, that is used to explain a symbol, you have no way of knowing what it means. If dictionaries used the IPA, it would only be necessary to learn and keep track of one system, and when you used a dictionary, you would either know what a symbol meant or have a straightforward way of finding out.

The situation is similar with language textbooks. I recently had occasion to use John Mason's Tigrinya Grammar (Lawrenceville, NJ: The Red Sea Press, 1996). Tigrinya is a language whose sound system is quite exotic from an English speaker's point of view. It has a series of ejectives, voiced stops that are truly voiced even in word-initial position, phonemic glottal stop, and several fricatives absent from English, including a pair of pharyngeals. The English speaker desirous of learning Tigrinya is going to need some assistance with the sound system. This is not provided by the textbook. It uses the Ethiopic script, in which Tigrinya looks like this: ሰዋስው, throughout, without transliteration. That's very hard on the learner, who needs some practice with an unfamilar writing system, and some confirmation that he or she is reading it and writing it correctly. What is worse, the explanation of the Tigrinya writing system uses an ad hoc transcription system. Some of the ejectives are simply not indicated as being ejective; others are incorrectly labelled "plosive". Several Ethiopic letters are transcribed by the letter q, alone or in combination. If this means anything to an English speaker, it will probably be taken to indicate [kʷ]. In fact, it represents an ejective velar stop; the combinations represent a glottalized velar fricative and labialized versions of the first two. There is no way that anyone could learn this from the textbook.

As it happens, Mason didn't just invent this transcription with q. This is a traditional usage of Semitic specialists, based on the fact that in some Semitic languages the cognate sound is a voiceless uvular stop, whose IPA symbol is [q]. But this is a purely historical notation, irrelevant to the pronounciation of Tigrinya, and in any case is known only to Semitic specialists. To be fair, the book contains the disclaimer that to learn the pronounciation properly one needs to listen to a speaker. But many people aren't going to have access to a speaker, at least not when they start learning and not on a regular basis, and even if they do, without at least a foundation in phonetics, they will be hard put to know what to make of what they hear.

Now, I don't mean to give Mr. Mason a hard time. I imagine that he did his best. He's probably not a linguist. I suspect that he's a missionary. My point is that this is actually quite typical of language textbooks. Few authors of textbooks, and few publishers, seem to understand that it would be a good idea to provide the reader with an unambigous description of the sounds of the language, and that there is a simple way to do this: use standard linguistic descriptive terminology and use the IPA.

An example of a textbook that does use the IPA is R. E. Asher and E. Annamalai's Colloquial Tamil (New York: Routledge, 2002). Where it deviates from the IPA it is arguably for a legitimate reason, namely to reflect the native Tamil writing system. Furthermore, even where it deviates from the IPA, it explains its transcription in terms of the IPA. This book, as it happens, comes with audio recordings, from which the actual sounds can be learned, but even when such recordings are provided, a comprehensible text helps to prepare the student to make use of the recordings. (And of course, if the book is bought used or borrowed from a library, there is a good chance that it will come without the recordings.) It seems that Routledge has perhaps taken the idea of using the IPA to heart. Another recent entry in their series of language textbooks is Daisy L. Neijmann's Colloquial Icelandic (2001). This also uses the IPA to explain the sound system of Icelandic.

Although the use of IPA is still distressingly uncommon, these are by no means the first. Just rummaging through books I have on hand, I see that 小沢重男's モンゴル語四週間 (Mongolian in Four Weeks), published in 1986, uses the IPA to explain the Cyrillic writing system for Mongolian. So it isn't as if the IPA is a recent creation or that publishers and textbook authors have been unaware of it.

A topic that has come up repeatedly here is that most people, even highly educated people, know very little about language, and in particular, that they have no descriptive vocabulary for speech sounds. Mark has described several examples of this recently. What is more surprising is that people who do have a particular interest in language, such as the authors and publishers of dictionaries and language textbooks, fail so often to make use of the appropriate tools.

Posted by Bill Poser at May 16, 2004 07:54 PM