Affricates

The affricates ʧ, ʤ are, phonetically, composed of a plosive followed by a fricative. It is possible to treat each of the pair ʧ, ʤ as a single consonant phoneme; we will call this the one-phoneme analysis of ʧ, ʤ. It is also possible to say that they are composed of two phonemes each - t plus ʃ, and d plus Ʒ respectively - all of which are already established as independent phonemes of English; this will be called the two phoneme analysis of ʧ, ʤ. If we adopted the two-phoneme analysis, the words 'church' and 'judge' would be composed of five phonemes each, like this:

t -  ʃ  -  з: - t -  ʃ           d -  Ʒ  -  Λ  - d -  Ʒ

instead of the three phonemes that result from the one-phoneme analysis:

ʧ  -  з: -  ʧ         ʤ  -  Λ  -  ʤ

and there would be no separate ʧ, ʤ phonemes. But how can we decide which analysis is preferable? The two phoneme analysis has one main advantage: if there are no separate ʧ, ʤ phonemes, then our total set of English consonants is smaller. Many phonologists have claimed that one should prefer the analysis which is the most "economical" in the number of phonemes it results in. The argument for this might be based on the claim that when we speak to someone we are using a code, and the most efficient codes do not employ unnecessary symbols. Further, it can be claimed that a phonological analysis is a type of scientific theory, and a scientific theory should be stated as economically as possible. However, it is the one-phoneme analysis that is generally chosen by phonologists. Why is this? There are several arguments: no single one of them is conclusive, but added together they are felt to make the one-phoneme analysis seem preferable. We will look briefly at some of these arguments.

i) One argument could be called "phonetic" or "allophonic": if it could be shown that the phonetic quality of the t and ʃ (or d and Ʒ) in ʧ, ʤ is clearly different from realizations of t, ʃ, d, Ʒ found elsewhere in similar contexts, this would support the analysis of ʧ, ʤ as separate phonemes. As an example, it might be claimed that ʃ in 'hutch' hΛtʃ was different (perhaps in having shorter duration) from S in 'hush' hΛʃ or 'Welsh' welʃ; or it might be claimed that the place of articulation of t in 'watch apes' wɒt ʃeɪps was different from that of t in 'what shapes' wɒt  ʃeɪps. This argument is a weak one: there is no clear evidence that such phonetic differences exist, and even if there were such evidence, it would be easy to produce explanations for the differences that did not depend on phonemic analyses (e.g. the position of the word boundary in 'watch apes', 'what shapes').

ii) It could be argued that the proposed phonemes tS, dZ have distributions similar to other consonants, while other combinations of plosive plus fricative do not. It can easily be shown that tS, dZ are found initially, medially and finally, and that no other combination (e.g. pf, dz, tθ) has such a wide distribution. However, several consonants are generally accepted as phonemes of the BBC accent despite not being free to occur in all positions (e.g. r, w, j, h, ŋ, Ʒ), so this argument, although supporting the one-phoneme analysis, does not actually prove that ʧ, ʤ must be classed with other single-consonant phonemes.

iii) If ʧ, ʤ were able to combine quite freely with other consonants to form consonant clusters, this would support the one-phoneme analysis. In initial position, however, ʧ, ʤ never occur in clusters with other consonants. In final position in the syllable, we find that tʃ can be followed by t (e.g. 'watched' wɒtʃt) and ʤ by d (e.g. 'wedged' wedƷd). Final ʧ, ʤ can be preceded by l (e.g. 'squelch' skweltS, 'bulge' bΛldƷ); Ʒ is never preceded by l, and ʃ is preceded by l only in a few words and names (e.g. 'Welsh' welʃ, 'Walsh' wɒlʃ). A fairly similar situation is found if we ask if n can precede ʧ, ʤ; some BBC speakers have ntʃ in 'lunch', 'French', etc., and never pronounce the sequence nʃ within a syllable, while other speakers (like me) always have nʃ in these contexts and never ntʃ. In words like 'lunge', 'flange' there seems to be no possible phonological distinction between lΛndƷ, flændƷ and lΛnƷ, flænƷ. It seems, then, that no contrast between syllable-final lʃ and ltʃ exists in the BBC accent, and the same appears to be true in relation to nʃ and ntʃ and to nƷ and nʤ. There are no other possibilities for final-consonant clusters containing ʧ, ʤ, except that the pre-final l or n may occur in combination with post-final t, d as in 'squelched' skwelʧt, 'hinged' hɪnʤd. It could not, then, be said that ʧ, ʤ combine freely with other consonants in forming consonant clusters; this is particularly noticeable in initial position.

How would the two-phoneme analysis affect the syllable-structure framework? Initial ʧ, ʤ would have to be interpreted as initial t, d plus post-initial ʃ, Ʒ, with the result that the post-initial set of consonants would have to contain l, r, w, j and also ʃ, Z - consonants which are rather different from the other four and which could only combine with t, d. (The only alternative would be to put t, d with s in the pre-initial category, again with very limited possibilities of combining with another consonant.)

iv) Finally, it has been suggested that if native speakers of English who have not been taught phonetics feel that ʧ, ʤ are each "one sound", we should be guided by their intuitions and prefer the one-phoneme analysis. The problem with this is that discovering what untrained (or "naive") speakers feel about their own language is not as easy as it might sound. It would be necessary to ask questions like this: "Would you say that the word 'chip' begins with one sound - like 'tip' and 'sip' - or with two sounds - like 'trip' and 'skip'?" But the results would be distorted by the fact that two consonant letters are used in the spelling; to do the test properly one should use illiterate subjects, which raises many further problems.

This rather long discussion of the phonemic status of ʧ, ʤ shows how difficult it can be to reach a conclusion in phonemic analysis.

A number of other phonological problems will be discussed comparatively briefly. We have already seen  problems of analysis in connection with the sounds usually transcribed hw, hj. The velar nasal ŋ, also raises a lot of analysis problems: many writers have suggested that the correct analysis is one in which there is no r) phoneme, and this sound is treated as an allophone of the phoneme n that occurs when it precedes the phoneme g. It was explained that in certain contexts no g is pronounced, but it can be claimed that at an abstract level there is a g phoneme, although in certain contexts the g is not actually pronounced. The sound ŋ is therefore, according to this theory, an allophone of n.