The class of phonetic changes in which the loss of a consonant is accompanied by the development of distinctive length in a nearby vowel has traditionally been designated by the term ‘compensatory lengthening’. We first discard cases in which the development of length and the loss of a consonant are in fact independent phenomena, and then analyse remaining examples. It is argued that such changes always involve loss of a consonant immediately adjacent to the vowel in which length subsequently appears, and that the change can be analysed quite generally into the two independently-motivated processes of weakening of a consonant to a glide (either semivocalic or laryngeal) and subsequent monophthongization of a complex syllable nucleus. The category of compensatory lengthening is thus shown not to be an independent mechanism of phonetic change. The fact that synchronic grammars sometimes contain rules of this type is argued to be an instance of the difference, in the general case, between phonetically-motivated processes and phonological rules. Conclusions are drawn for the study of the role of phonetic explanation in phonology.

It is further argued that the monophthongization involved in these examples leads to the development of length if and only if a length contrast is independently motivated in the system of the language concerned. This provides a particularly clear example of the (structuralist) principle that the phonological interpretation of a phonetic change is dependent on the structure of the system in which it takes place.

While tone terracing appears superficially to be a purely phonetic phenomenon, it is frequently subject to syntactic, grammatical, and lexical conditions which are not phonetically definable in surface structure; and this raises the question of how such conditions are to be encoded at the phonetic level to which pitch assignment rules apply. Some studies (e.g. Johnson 1972, Peters 1973, Anderson 1978) have proposed treating downstepped tones as separate ‘tonemes’ (distinguished from other tones in their feature characterization) to which pitch assignment rules apply in a left-to-right iterative mode of application. Such proposals fail to provide an explanation for the phonetic equivalence of basic and derived downstepped tones, and for the equivalence of downstepped and the corresponding non-downstepped tones with respect to following tones. In the alternative treatment proposed here, tone terracing systems are described in terms of a model which provides an independent level of representation to which intonational processes, including register lowering (downstep) apply; such a model is applied to the description of the tone terracing system of Akan. Further evidence for this model is drawn from Kikuyu.

English word stress is analysed in terms of rhythm—of alternating strong (S) and weak (W) syllables. A set of S-placement rules assigns S's to words. A set of rhythmic constraints defines well-formed configurations of S's and W's. The interaction of rules and constraints determines the accentual patterns of most words, whether simple or morphologically complex. The analysis accounts for all ‘stress shifts’ in morphologically related forms (e.g. solid, solidity), as well as certain phrasal stress shifts (e.g. thirteen students). The relation between S/W and degrees of stress is explored. Similarities are noted between the rhythms of discourse and those of verse. The analysis is compared to that of Liberman & Prince 1977, who also make use of alternating strong and weak syllables.

In the Boston dialect, the variable or as in short, forty, has two phonetic variants: dialectal [Də], and more standard [Oə]. Speakers of three ethnic groups—Irish, Italian, and Jewish—showed systematic differences in stylistic and diachronic use of the variants, and in subjective evaluation of them. The observed differences are consistent with a three-stage model of variation and change based on ethnicity and, secondarily, on education. However, an additional factor of ‘cultural force’ is required to explain why the model may operate at different rates. The model has implications for the acquisition of dialectal vs. standard variants, and lends support to the view that the ‘family’ provides a stronger influence than the ‘peer group’ in the development of phonological systems.

Implicational universals of the form ‘If a language has some word order P, then it also has word order(s) Q’ are widely used to explain word order change. A language which develops word order P without having Q is said to be universally inconsistent, and it is predicted that it will re-introduce consistency by subsequently acquiring Q. This paper, however, argues against the validity of such a paradigm. The current assumption that the early western Indo-European dialects are inconsistent is founded on inadequate formulations of the synchronic universals. First, more precise formulations are given for some basic word order universals in a large body of languages ; then their predictions are tested against all the early daughter languages of Indo-European (using data from Friedrich 1975). None of these dialects exhibits inconsistency; hence we have no evidence for the reintroduction of consistency as a valid mechanism of syntactic change. It is shown, using historical data from Germanic, that implicational universals can nonetheless make correct predictions for the relative timing and manner of word order changes, but only on the assumption that languages in evolution obey synchronic universals. Some purely theoretical arguments are provided to support the empirical findings, and an alternative explanatory model of word order change is outlined.

In the light of recent research on Noun Phrase Accessibility (NPA) and relative clause formation (cf. Keenan & Comrie 1977), the main principles of NPA are here clarified and modified. The proposals of Maxwell 1979 for revising the criteria which distinguish relative clause formation strategies, and for stating further generalizations in terms of the strategies thus defined, are considered in detail. It is concluded that Maxwell advances the understanding of NPA by providing further significant generalizations, but that his criteria sometimes lack explicitness, and do not affect the validity of the criteria originally established for distinguishing strategies in the statement of NPA.