ABSTRACT
In the following sections, we shall move on to the position that the CV tier should be seen as an anchoring device (hence the more appropriate name of skeleton) relating the internal content of segments to other types of information. The reader may have noticed, however, that in many respects our approach has remained linear: in particular, the domain of features has been assumed to coincide with that of segment-sized units. But a number of linguistic phenomena suggest that features should not be locked within single segment-sized units but should be allowed to span domains of varying sizes (portions of the syllable, whole syllables, feet, words, etc.). This idea is not new and much work in the Firthian school (or London school: see Palmer 1970; Lass 1984: §10.2) started from this very assumption. In recent years, these insights have been developed and formalized within a
framework known as Autosegmental Phonology. 1 As the original and most intuitively convincing examples of autosegmental analysis have come from tonal phenomena, we shall start from examples in this area and then move on to an integration of these ideas into a more complex geometrical model which we shall call 'multidimensional phonology' - a model far removed from many of the assumptions that we started from in the early chapters of this book.
