This article takes as its point of departure an unusual phonological change in a Czech dialect. It then proposes a model of phonological change which would make possible the understanding of structural innovations in the phonology of a homogeneous speech community. The model, which distinguishes two logically different modes of change (abductive and deductive), helps clarify the essential difference between ‘internally motivated’ change and change ‘induced from without’. The model uses our experience of observed phonetic changes, and may consequently have some bearing on our understanding of the structure of phonology.

In Indo-European languages with mobile paradigmatic accent, fixed oxytones and movable athematic stems are accented by the same rule. When applied to syllables as accentual units, this rule generates the Sanskrit accent patterns. When applied to moras, it generates the Greek accent patterns—including, as a special case of mobility, the paradigmatic alternation of acute and circumflex. A modification of the the same rule accounts for the extension of mobility to originally oxytone -o- and -ā-stems in Bal to-Slavic, and for the rise of the Balto-Slavic acute and circumflex accents.

Germanic verbs of the 3rd weak class form presents characterized by an alternation between predesinential ∗ai (e.g. Go. 3sg. habaiþ) and ∗a (e.g. 1pl. habam). These verbs are usually compared with the ‘ē-verbs’ of Italic and Balto-Slavic, but no IE present built on the stative suffix ∗-ē- will account phonologically for the form of the suffix in Germanic. Instead, it can be shown that the characteristic Germanic paradigm results from the ‘activization’ of an older middle paradigm in which a 3sg. in ∗-ai (< IE ∗-oi; cf. Skt. duhé ‘milks‘) was further suffixed by the productive active ending ∗-þi (< IE ∗-ti).

Even with the recent interest in correlating linguistic performance with sociological variables, the tendency in discussions of lexical borrowing is to focus on the words themselves and to ignore the process. This paper argues that the process of borrowing determines what is borrowed. Most borrowings may well be nouns for items new to the borrowing language. However, given sufficiently pervasive cultural contact, speakers may borrow certain types of core vocabulary items in great numbers. Data to support these contentions come from a study of the dialects of Ateso, an East African language.

It is shown that presuppositions and implications associated with verbs distribute themselves relative to the time-axis (action-time) of the verb. Presuppositions pertain to the time preceding this time-axis, while implications pertain to the time following the time-axis. The semantics of two universal verb classes, modality verbs (M-verbs) and cognition verbs (C-verbs), is analysed in light of these findings.

Two widespread diachronic processes of sense extension are studied. One involves the development of sentence modalities from M-verbs. It is shown that a great amount of predictability exists in this process, and that the type of tense aspect marker likely to arise from an M-verb may be inferred, with great accuracy, from the specific presuppositional and implicational structure of the precursor verb. A considerable degree of predictability is also shown in the development of M-verb senses from C-verbs. It is suggested that our temporal and aspectual notions, including our notions of time, negation, certainty, and possibility, are already present in the semantic structure of verbs. Diachronic extensions of verb-senses merely serve to illuminate this.