The phoneme ŋ

Consider the words pink piŋk and anger aŋgə. Both of them contain a nasal sound, ŋ, which is indicated in spelling by the letter n, which usually represents the sound n. But the transcription of these words shows that the nasal sound is different from n in them, it is not alveolar but velar, ie pronounced at the same place of articulation as the following k or g. It is in fact very common in languages that the nasal sound before k and g is pronounced as a velar nasal, ŋ. Similarly before p and b the nasal is pronounced as a labial nasal, m (eg pamper, timber). However, unlike ŋ, the segment m occurs more freely: before a vowel (summer səmə) or before consonants other than p and b (crimson krimzən, Samson samsən, Camden, chimney, Hamlet, amuse əmjuwz, etc).

Since ŋ occurs before k and g, one might suppose that its distribution is predictable, hence it is an allophonic variant of n. But this is not so, because of an apparently insignificant change that has happened in English, the result of which is present in most accents today. The g in word-final ŋg clusters was lost, so thing, which used to be pronounced θiŋg is now θiŋ or young, which used to be jəŋg is now jəŋ.

In fact, there are parts of England where this word-final postnasal g is retained: around Liverpool, Manchester, Birmingham, also in much of Kent, south-east of London, and in New York City in the US. The map shows these areas of England.

The deletion of the g did not entail the “develarization” of the preceding nasal sound, so it remained velar ŋ, it did not turn into n.^»We do find n in the suffix -ing in many accents: singing siŋin (beside the standard siŋiŋ). This is an unrelated phenomenon, the suffix simply has a different form in these accents. As a result words that previously had a n at the end came to be minimally different from the newly developed words that now had ŋ at the end: thin θin and thing θiŋ, ban ban and bang baŋ, or son sən and sung səŋ became minimal pairs. Accordingly, despite its restricted distribution — it only occurs before g, k, and word finally — ŋ has become a phoneme on its own right, since it may distinguish words when it stands at their end.

Note that in accents of English where the word-final g was not lost, so ŋ only occurs before g and k, but not word finally, the velar nasal, ŋ, is not a phoneme, since thing θiŋg and thin θin are not a minimal pair, just like mill mil and milk milk are not a minimal pair. In such accents, the sound ŋ will be transcribed as n (ie pink pink, anger angə, thing θing, bang bang, sung səng), since in transcriptions we only indicate what is — at least in some cases — unpredictable. The velar place of articulation of tha nasal in such an accent is always predictable, therefore it is not shown in transcriptions. British English and American English are not such accents, so a velar nasal is transcribed as a velar nasal, ŋ: pink piŋk, anger aŋgə, thing θiŋ, bang baŋ, sung səŋ, irrespective of whether it is predictable (as in pink and anger) or not (when it stands at the end of the word). The principle governing this analysis is often cited as “once a phoneme, always a phoneme.”