Dinka, a major Western Nilotic language spoken in the Sudan, is to a large extent a monosyllabic language. Nevertheless, it has a complex morphology. Thus a significant part of its morphology is non-affixal, being manifested by way of morphophonological alternations in the root. Such alternations involve one or more of the following parameters: vowel quality, vowel length, voice quality, tone and final consonant. While alternations in vowel length, voice quality and tone are treated in Andersen (in press), the present article deals with vowel quality alter-nation. The dialect of Dinka treated here is Agar, more specifically the variety of Agar spoken by people from the area of Pacong, a village about 20 kilometres south-east of Rumbek in Southern Sudan. My principal informants for the present study were Isaac Maker, David Daniel Marial and Peter Gum Panther.

In English, a number of rules affect the realisation of a nasal consonant or a segment adjacent to a nasal consonant. These include rules of Anticipatory Nasalisation, e.g. bean /bin/ [bĩn]; Coronal Stop Deletion, e.g. kindness /kajndnes/ [kãjnnes]; Nasal Deletion and optionally Glottalisation, e.g. sent /sent/ [set] or [set'] (see Malécot 1960; Selkirk 1972; Kahn 1980 [1976]; Zue & Laferriere 1979). These rules, characterised largely on the basis of impressionistic data, are widely assumed to be phonological rules of English. Yet current views of the relationship between phonology and phonetics make the distinction between phono-logical rules and phonetic ones less automatic than once assumed and a reconsideration of the status of these rules is warranted. In the present article, I use phonetic data from English to investigate these rules. Based on these data, I argue that Anticipatory Nasalisation results from phonetic implementation rather than from a phonological rule, as previously assumed. It is shown that the basic patterns of nasalisation in English can be accounted for straightforwardly within a target-interpolation model. I then investigate the phonological status and phonetic realisation of Nasal Deletion, Coronal Stop Deletion and Glottalisation. The interaction of these rules yields some surprising results, in that glottalised /t/ [t'] is amenable to nasalisation.

The German uvular /R/ probably shows more surface variation than any other segment in the language. (1) illustrates that /R/ has a vocalic allophone [A], which can surface either as a glide or a vowel, a sonorant consonant allophone, which is pronounced as a uvular trill or approximant, and two obstruent allophones:

In the present study I focus on the rules producing the consonantal allophones of /R/ in both Standard German and in certain dialects of the Lower Rhineland (henceforth LRG).

Glottal consonants (/h/ and /?/) often have a special status that has led to the hypothesis that they are underspecified for both consonant and vowel place of articulation features. There is an alternative acoustically based explanation for the special status that, while workable, is valued less highly in linguistic theory. This alternative hypothesis can be ruled out only with data where the relevant acoustic phenomena are not contained in the input to be learned. I address four processes of consonant and vowel harmony in the speech of three children learning English, where the alternative hypothesis cannot account for the data. I conclude that glottals are underspecified for consonant and vowel place of articulation features, and that they lack the node in feature geometry that immediately dominates such features (either the Place node or an Oral node).