Phonetic similarity

In the vast majority of cases, applying our phoneme tests will provide results in keeping with native speakers’ intuitions about which sounds belong together; very often, as we have seen, allophones of a single phoneme will not in fact be distinguishable for a native speaker at all, without a certain amount of phonetic training. However, there are some cases where sticking to those tests too rigidly can have quite the opposite consequence.

One of the best-known and most obvious examples of this kind in English involves [h] and [ŋ]. The minimal pairs in (5.1) show that [h] contrasts with a number of English consonant phonemes word-initially; but there is no minimal pair for [ŋ]. Conversely, in word-final position, it is straightforward to find contrasts for [ŋ], as in rang, ran, ram, rat, rack, rag, rap, rash; but there is no equivalent minimal pair for [h]. The generalization extractable from this is that [h] appears only before a stressed vowel (or at the beginning of a syllable), as in hat, ahead, apprehensive, vehicular (but not vehicle, where appears in the spelling, but there is no [h], as the stress here falls on the first vowel). On the other hand, [ŋ] is not permissible syllable-initially: it can appear only at the end of a syllable, either alone, as in rang, hanger, or before a velar plosive, either [k] or [ ], as in rink, stinker, finger, stronger.

What this means, in purely technical terms, is that [h] and [ŋ] are in complementary distribution. One appears only syllable-initially, where the other never does; and in consequence, there is no possible minimal pair which will distinguish the two. If we take only predictability of occurrence and invariance of meaning into account, we will be forced into setting up a phoneme which we might symbolize as, which is realized as [h] in one set of environments, and [ŋ] in another.

It is not going to be easy to convince native speakers of English that this is the right solution – not because we have to work on bringing previously subconscious intuitions to the surface, but because those intuitions suggest strongly that [h] and [ŋ] are entirely separate and unrelated. There is some evidence in favour of that view, too. First, although we have seen that the English spelling system is not absolutely and reliably phonemic, different spellings are never consistently used for different allophones of a single phoneme, as would be the case for [h],and [ŋ]//. Second, native speakers can easily tell the two sounds apart, which would not be true, for instance, of clear and dark variants of /l/, or aspirated and unaspirated allophones of /p/. Since our core criteria for allophony very generally give the right results, it is probably unwise to mess about with them much; but we can add a further condition on determining allophony, which applies both to the ‘normal’ cases and to the situation of [h] and [ŋ].

In brief, this additional criterion for allophony states that all the allophones of a phoneme must be phonetically similar. Using distinctive features allows this rather vague notion to be quantified: but there is still no straightforward equation for determining what counts as phonetically similar and what does not. However, although we cannot draw a dividing line which will be universally applicable, for instance requiring that the allophones of a single phoneme must be different by no more than three features, we might at least hypothesize that two sounds are highly unlikely to be allophones of the same phoneme if the number of contrasting feature values is higher than the number of shared ones. For [h] and [ŋ], this produces an unambiguous result: both are consonants, but there the similarity ends. [h] is a voiceless fricative, while [ŋ] is a voiced stop; [h] is oral, while [ŋ] is nasal; [h] is glottal, while [ŋ] is velar; [h] is an obstruent, while [ŋ] is a sonorant. On almost every parameter which could distinguish the two, they are in fact distinct. Rather than setting up a single phoneme with two such bizarrely different realizations, invoking phonetic similarity allows us to justify regarding /h/ and /ŋ/ as distinct phonemes, despite the lack of minimal pairs.

Phonetic similarity also helps in cases where a single allophone could theoretically be assigned to more than one possible phoneme, a situation commonly encountered when members of a natural class of phonemes undergo the same rule. For instance, we have seen that in Old English, the voiceless fricatives /f θ s/ were voiced between voiced sounds. It follows that all the voiceless fricative allophones were in complementary distribution with all the voiced ones, since [v ð z] could appear only between voiced sounds, and [f θ s] could appear only elsewhere. Purely on the grounds of predictability of occurrence and invariance of meaning, there is no guidance on which of these we should assign to which phoneme: in theory we could set up one phoneme with allophones [f ] and [z], a second with [θ] and [v], and a third with [s] and [ð], if all that matters is for one allophone to be voiceless and the other to be voiced.

We might also feel that this solution would make Old English speakers turn in their graves: their intuitions are highly likely to have favoured grouping the two labial sounds together, the two dentals, and the two alveolars. Again, this intuitive solution is supported by a requirement of phonetic similarity, this time involving the assignment of the two most similar allophones, those sharing a place of articulation, to a single phoneme in each case. In Modern English, a precisely similar problem and solution arise with the voiceless stop phonemes and their aspirated and unaspirated allophones.