Rules and constraints

Most interactions of phonology with morphology, the part of linguistics which studies how words are made up of meaningful units, like stems and suffixes, although the overlap between the two areas, commonly known as morphophonemics, has been extremely important in the development of phonological theory over the last fifty years. Indeed, the difference between phonetically conditioned allophony and neutralization, which involve only the phonetics and phonology, and cases where we also need to invoke morphological issues, is central to one of the most important current debates in phonology.

Generalizations about the distribution of allophones were stated in terms of rules, the assumption being that children learn these rules as they learn their native language, and start to see that forms fall into principled categories and behave according to regular patterns. Rule-based theories also include constraints – static, universal or language-specific statements of possibility in terms of segment shapes or combinations, and phonotactic constraints. However, since the mid-1990s, an alternative approach has developed, as part of the phonological theory called Optimality Theory. Phonologists working in Optimality Theory do not write rules; they express all phonological generalizations using constraints. Instead of saying that a particular underlying or starting form changes into something else in a particular environment, which is what rules do, constraints set out what must happen, or what cannot happen, as in the examples in (6), which express regularities we have already identified for English.

In most versions of Optimality Theory, all the constraints are assumed to be universal and innate: children are born with the constraints already in place, so all they have to do is work out how important each constraint is in the structure of the language they are learning, and produce a ranking accordingly. For an English-learning child, the two constraints in (6) must be quite important, because it is true that voiceless stops are aspirated at the beginnings of syllables, and there are no sequences of [s] plus a voiced stop; consequently, English speakers will rank these two constraints high. However, for children learning a language without aspiration, or with clusters of [s] plus voiced stop, these constraints will not match the linguistic facts they hear; they will therefore be ranked low down in the list, so they have no obvious effect. On the other hand, a child learning German, say, would have to pay special attention to a constraint banning voiced stops from the ends of words, since this is a position of neutralization in German, permitting only voiceless stops; but a child learning English will rank that constraint very low, as words like hand, lob, fog show that this constraint does not affect the structure of English.

Constraints of this sort seem to work quite well when we are dealing only with phonetic and phonological factors, and may be appropriate alternatives to rules in the clearly conditioned types of allophonic variation we have considered, and for neutralization. However, they are not quite so helpful when it comes to the interaction of morphology and phonology, where alternations are often not clearly universally motivated, but involve facts about the structure and lexical items of that specific language alone. Analyzing such cases using Optimality Theory may require a highly complex system of constraints, as we will have to accept that all the possible constraints for anything that could ever happen in any language are already there in every child’s brain at birth. These issues are likely to lead to further debate in phonology in future years.