Making generalizations

Several examples of allophonic variation were considered. In one case, we found that /k/ has two variant pronunciations, namely velar [k] in cupboard and palatal [c] in kitchen. Another involved /p/, /t/ and /k/, which have aspirated allophones, with a perceptible release of air, in pill, till and kill, but unaspirated allophones in spill, still and skill, or sip, sit and sick.

However, providing a list of words where the relevant allophone appears is only our starting point. Phonologists are interested in generalizations about the language they are working on, and indeed in generalizations about language in general; and generalizations are not best expressed simply as lists, as these do not reveal the factors which the forms in the lists have in common. Identifying these factors will help us to understand why the particular allophone appears in that context and not elsewhere, and to predict what will happen in other words with a similar context.

As an example, recall the [c] and [k] allophones of /k/. English speakers (with the exception of New Zealanders and Australians) will have palatal and velar pronunciations distributed as in (1).

If you were asked to predict the pronunciation of the initial sounds of keep, cool, ceilidh (for non-Scots, pronounced exactly as Kayleigh) and koala, you would not get very far by considering (1) as just two lists of words: how could you tell whether each of these examples fitted into the [c] list or the [k] list? The key is to consider what connects the words where each allophone appears: and the answer is that [c] appears before a front vowel, while [k] precedes a back vowel. It follows that keep and ceilidh will also have [c], since the boldfaced vowels are front, while cool and koala will have [k], as the boldfaced vowels are back. Since front vowels are made roughly at the hard palate, and so is palatal [c], while back vowels are produced at the velum, as is velar [k], the pairs of vowels and consonants ‘match’. It is extremely common for sounds to become more similar, or to assimilate to one another, in this sort of way. The vocal organs undergo very complex, coordinated movements during speech, and anything that simplifies the gymnastics involved while not jeopardizing comprehension is understandably very welcome to speakers. Specifying what the different examples have in common therefore allows us to understand the results we find, and make predictions about the behavior of other forms with the same environment. And as we might expect, / /, which matches /k/ in every respect except voicing, behaves in exactly the same way, being palatalized before the same set of vowels as /k/ in the same varieties.

In the case of /p/, /t/ and /k/ aspiration, the relevant conditioning factor is not the shape of an adjacent segment, but rather position in the word.

 What pill, till and kill have in common (along with peel, pass, play, pretty and many others) is that the /p/, /t/ or /k/ is right at the beginning of the word. In spill, still, skill, sip, sit and sick, and many others, it is not right at the beginning of the word; either another consonant precedes it, or it is word-final. We can test this hypothesis by finding lots of other examples where /p/, /t/ and /k/ appear word-initially, and checking whether there is aspiration. So long as we keep finding aspirated allophones there, and nowhere else, our generalization holds. If we find counterexamples, where either aspirated forms appear in other contexts, or word-initial allophones of /p/, /t/ or /k/ are not aspirated, we have to modify our generalization to include them. After a while, when we keep finding data that agree with our observation and not finding data that disagree, we can feel more confident that our generalization is the right one, and regard our hypothesis as confirmed.