Introduction

In this chapter, I argue for a model of phonology which does not canonize the distinction between predictable and contrastive features in phonology with phonemic underlying representations. It is proposed instead that individual tokens of experience are stored and organized into categories without redundancies removed, as suggested in Chapter 2. In the following, I present evidence for this position by examining the following types of cases: (i) cases of phonetic variation that are specific to particular words (Section 3.4.1), (ii) cases of distributionally predictable features that are lexically contrastive (Section 3.4.2), and (iii) cases showing the involvement of variable sound change with morphology (Section 3.6).

In this chapter I also demonstrate the use of schemas for phonological generalizations and present a model for the effect of sound change on the lexicon that allows for both phonetic and lexical gradualness as well as for the tendency for high-frequency words to change faster than low-frequency words.

The Phonemic Principle

A principle basic to our understanding of phonology and, in fact, of all aspects of linguistic structure, is the phonemic principle – that objectively different sounds form relations with one another and that some of them can be considered at some cognitive level to be very similar, or even the same. This basic insight can be expressed in different ways by different models.

The structuralist models we are all familiar with, probably based on the success of alphabetic orthographies, postulate that mental representations of sound structure are written out in phonemes; only phonemes exist in mental representations and variants of phonemes emerge only in surface forms.