The Phonological Structure of Words: An Introduction (review)

For an introductory text on phonology, this book is noticeably not very long; the body of the book consists of a mere 274 pages (including references and index) and is divided into just four chapters. In comparison, the introductory text by Roca and Johnson (1999) contains 725 pages and 20 chapters. A reason for the difference is provided by the authors in the preface and again in the epilogue: this book is not intended to be a comprehensive introduction to phonology. It is also not intended to be a book that introduces students to the tools of phonological analysis. Instead, it focuses on those aspects of representation that the authors consider to be fundamental to ‘the characterisation of phonological structure’ (xi). Ewen and van der Hulst wish to convey the view that there are basic units of phonology that any theory must recognize. These units at the level of the word include features (Chs. 1, 2), syllables (Ch. 3), and feet (Ch. 4). The focus of the book is therefore on the types of phenomena that justify recognition of such units.

In Ch. 1, E&H are concerned with justifying features as the building blocks of segments. They appeal to the familiar and generally persuasive evidence that segments often display similarities in their patterning (i.e. phonotactics) that can be explained only if certain groups form natural classes; features are crucial to the description of natural classes. This chapter also provides justification for the organization of features into groups; the hierarchical array generally referred to as feature geometry is considered to be just one exemplar of this mode of feature organization. The evidence and arguments used by E&H as support for hierarchical feature grouping are again quite familiar. For example, familiar patterns of place agreement between adjacent segments can be treated as the extension or spreading of the place features as a group. The hierarchical organization of features is exploited in §1.4 to illustrate the behavior of features as autonomous entities. Familiar autosegmental descriptions of Old English umlaut and Turkish vowel harmony are provided. The only new ground covered in this chapter is the indication that similar modes of feature grouping to the familiar feature geometry are found in the phonological literature.

The focus of Ch. 2 is on the nature of features themselves. E&H adopt the standard view that features encode properties as binary oppositions. Hence, this chapter is devoted primarily to consideration of various proposals for representing binary oppositions. In §2.1, E&H establish that for some features (e.g. nasality) the opposition is between presence or absence of the relevant property; such features are single-valued. While E&H appear to be generally committed to single-valued features, they are more concerned with the asymmetry that such a conception of feature values entails. Hence, §2.2 provides a brief overview of how the issue of asymmetric feature values is addressed by the marking conventions of SPE (Chomsky & Halle 1968) and various versions of underspecification theory that emerged in the 1980s. The comparison of contrastive specification theory and radical underspecification theory in this section, although brief, is very informative. Students should find the discussion of the difference between redundancy and markedness considerations helpful. The last part of this chapter, beginning with §2.3, promotes a version of single-valued representation that is not usually encountered in the more popular introductory texts. The single-valued entities (e.g. i, u, a), familiar to government phonology (GP) (Kaye et al. 1985, 1990), are supposed to correspond to a phonetically interpretable elements, but it is not obvious to me that the latter requirement always holds, especially when the representation of consonants (§2.6) is discussed. In relation to the latter, the single-valued features favored by GP as appropriate for the representation of consonant structure seem to yield fewer insights than single-valued vowel features...