Introduction

In this chapter, I propose both ‘incomplete’ and ‘overcomplete’ structures for specific segment types. Incomplete structures are structures that miss one of the element classes. We have already seen examples of this, in the sense that a non-tonal language does not use the laryngeal node for vowels. We have also seen that the place class can be missing, as in central vowels and pharyngeal and laryngeal consonants. What cannot be missing is the manner class, because this class, being the head class, is obligatory. This, of course, is a theoretical decision that follows the idea that headedness plays a role and the claim that the manner class is the head class. One might ask whether RCVP allows a formal difference between the non-presence of a class and the presence of a class without any element. This issue has already been discussed in § 6.4.1. The answer is no for dependent classes, that is, the laryngeal and place class, and also no for secondary classes: a segment cannot have a secondary element in the absence of a specified primary element in any class. The simplest and most general answer would of course be that empty class nodes are not allowed at all, but I will consider below whether the manner class, being obligatory, can be ‘empty’. I will, however, provide an approach that does not require this kind of option, which means that, indeed, RCVP bars empty class nodes entirely. I then turn to overcomplete structures, which are necessary for various classes of so-called complex segments such as clicks, consonants with multiple articulations, short diphthongs and some other types of segments.

Incomplete structures

No content at all

Let us first consider the possibility of allowing a so-called ‘empty nucleus’, that is, a nucleus position that is not associated with any segmental structure. In GP it is assumed that every ‘syllable’2 must have a nucleus position, this being the obligatory unit in the syllable. In this theory, this allows for the option to have no content in this nucleus position, which is then a ‘skeletal point’ that does not dominate segmental content. Such an empty nucleus will remain silent (if governed) and must otherwise be filled and thus realised, or ‘just realised’ (in the phonetic implementation). This proposal underlies the treatment of vowel/zero alternations in this theory (KLV90; Scheer (2004); van der Hulst (2006b)), which will be discussed below.