Daniela Isac and Charles Reiss have recently published I-Language: An Introduction to Linguistics as Cognitive Science, which apparently has more phonology in it than your typical textbook of this type. (As the book description notes: “Contains phonological parallels to familiar syntactic arguments”.) There’s also a companion website with various resources, including a great page demonstrating Turkish vowels (previously noted by Mr. Verb). The vowels are arranged in a cube-like format that may be familiar to many of us. (This is the way I learned about Turkish vowels from Jorge Hankamer, and it clearly had a lasting effect on me.)

The publisher’s website also includes this sample chapter (Chapter 1, “What is I-language?” — a good place to start), which begins with an autobiographical story about how Charles used his knowledge of Turkish vowels and vowel harmony to save himself and a friend from a near-death experience (hey, read it yourself).

I just so happened to be reading (John Mepham’s translation of) Roman Jakobson’s Six Lectures on Sound and Meaning (that’s the image of a French version off to the right; I couldn’t find an image of my English version). There’s some good discussion of Turkish vowels and phonological features/oppositions there (pp. 79-83 of my English version) which I quote in full here.

Let us try to analyse an example. The vocalic system of the Turkish language is comprised of eight phonemes:

o	a	ö	e
u	y	ü	i

These eight phonemes produce, following the mathematical formula for combinations, twenty-eight distinctions, thus twenty eight binary relations. Ferdinand de Saussure showed us that the phoneme is constituted solely of relations. Now if we follow this Saussurian tradition and take these twenty-eight distinctions as the primary values for Turkish and take the phoneme in itself as secondary and derived, then we are in danger of arriving at a paradoxical conclusion, namely that the number of primary values is much higher than that of the derived values: twenty-eight compared with eight! Thus we are apparently confronted with a second contradiction — the first, remember, being that the ‘opposition’ between phonemes does not conform to the logical rules of opposition.

To remove both contradictions with one stroke it is enough simply to give up one presupposition which it has become traditional to make and which has threatened to lead all phonological research into error. We have been taught that phonemic oppositions, and above all the phoneme as such, are not decomposable. Taking the lead from Baudouin de Courtenay and Saussure, phonological research has accepted the following definition as its starting point: ‘The phoneme is a phonological element which is not susceptible to subdivision into smaller and simpler phonological elements’. Now, this definition (which was submitted twelve years ago to the first Phonological Assembly, in our ‘Project for a standardised phonological terminology’, and which was adopted by that international meeting) has turned out to be incorrect. In the Turkish phonological system the vowels o, a, ö, e are opposed to the vowels u, y, ü, i as open phonemes to closed phonemes; the vowels o, u, a, y are opposed to the vowels ö, ü, e, i as back phonemes to front phonemes, and the vowels o, u, ö, ü are opposed to the vowels a, y, e, i as rounded phonemes to unrounded phonemes. In this way the alleged twenty-eight vocalic oppositions of Turkish can in fact be reduced to three basic oppositions: (1) openness and closure, (2) back and front, (3) roundness and unroundness. It is by means of these three pairs of differential elements, really non-decomposable this time, that the eight vocalic phonemes of Turkish are formed. Thus, for example, the Turkish phoneme i is a complex entity composed of the three following differential elements: closed, front, unrounded.

The reasons why we have just characterised the differential elements in terms which pertain to the act of phonation are firstly because these terms are more familiar, and secondly because the corresponding acourstic definitions, while they would be more appropriate as a way of indicating the salient features of the qualities in question, would stand in need of some explanation, and this would take too much of our time at present. So we will do no more now than emphasise that each differential element exhibits one clear and easily identifiable acoustic feature, and that in analysing phonation precisely in the light of this acoustic effect we are always in a position to separate out from the multitude of phonatory movements a single basic factor which produces the acoustic effect in question.

It is not only the differences between the vocalic phonemes of Turkish which are resolvable into simple and indivisible binary oppositions, but all the differences between all the phonemes of every language. It follows that all the phonemes of each particular langauge, both the vowels and the consonants, can be dissociated into non-decomposable distinctive features. The apparent contradictions are now removed. The oppositions of such differential qualities are real binary oppositions, as defined in logic, i.e., they are such that each of the terms of the opposition necessarily implies its opposite. Thus, the idea of closure is opposed only by the idea of openness; the front and back features mutually imply each other, and so on.

The relation between two phonemes, by contrast, is complex and may be made up of several simple oppositions. Thus in Turkish the distinction between the phonemes u and o is made up of only a single opposition, that between closure and openness; but the distinction between the phonemes u and a is made up in addition of the opposition between the features roundness and unroundness, and the distinction between the phonemes u and e includes, in addition to the oppositions already mentioned, a third one, that between the features back and front. In any given language the number of differences between the phonemes is obviously greater than the number of phonemes, whereas the number of distinctive features is considerably lower. We should recall that the differential elements, while they serve to distinguish between the meanings of the words, do not themselves have meanings, and that it is precisely the fact that these empty entities are limited in number, that there are few of them in each given language, which enables the members of each linguistic community to perceive them, to retain them in memory and to put them to use.

The ‘differential elements’ (or in other terms ‘distinctive qualities or properties’, or finally ‘distinctive features’) appear in language combined in bundles. The phoneme is a bundle of differential elements. But the differential elements have in themselves their own role in the organisation of languages; they operate in language in an autonomous manner. For example, we find in many languages different forms of what is called ‘vowel harmony’. In such languages all the vowels of a word must have a common distinctive quality. For example in the majority of Turkic languages front and back vowels cannot appear together within a word: the vowels are either all front or all back; in Turkish the plural suffix takes the form -ler following a root with a front vowel, and the form -lar if the root has a back vowel: thus, ev-ler ‘houses’ and at-lar ‘horses’. So the front-back opposition operates here in an autonomous manner. In Turkic languages there is moreover a labial vowel harmony: in these languages rounded vowels cannot go together within a word with unrounded vowels. Finally, there are certain languages, for example those of the Manchu group, which do not allow closed and open vowels to both appear within the same word. For example, in Gold, a language of the Amur river region, the closed vowels u-y-i are opposed to the open vowels o-a-e: thus, ga ‘to buy’, bi ‘to exist’, and ga-pogo ‘in order to buy’, bi-pugu ‘in order to exist’. In all cases like this, one of the differential elements takes on an autonomous function, abstracting from the various phonemes of which the element is a part.

Note: the title of this post is for Ed.