In lieu of an abstract, here is a brief excerpt of the content:

Reviewed by:
  • A theory of phonological features by San Duanmu
  • Jeff Mielke
A theory of phonological features. By San Duanmu. Oxford: Oxford University Press, 2016. Pp. xiv, 178. ISBN 9780199664962. $99 (Hb).

In this book, San Duanmu proposes a minimal set of distinctive features to account for the segmental contrasts observed in phoneme inventories of about 1,000 languages. This is the first large-scale investigation of this type, and I think phonologists are likely to be surprised by how efficiently these contrasts are accounted for and by which phonological distinctions are not necessary. The book provides a wealth of setting-off points for further work investigating the sound systems of the world’s languages. D has written extensively on prosodic and segmental phonology. Pursuing a minimal set of phonological distinctive features has been a thread through much of his work, culminating in this book.

The project is reminiscent of Jakobson, Fant, and Halle’s (1952) Preliminaries to speech analysis: The distinctive features and their correlates, because it addresses segment inventories [End Page 477] but not classes of sounds involved in phonological rules. Jakobson and his colleagues had the spectrograph and information theory as new sources of insight, and D has about 1,000 phoneme inventories reported in UPSID (Maddieson 1984, Maddieson & Precoda 1990) and P-base (Mielke 2008), enabling him to examine segmental contrasts on a large scale.

The heart of the book is the analysis of inventories, documented in Chs. 2–5. Ch. 6 proposes the feature system motivated by this investigation. Chs. 1 and 6–8 address several major phono-logical topics, such as the granularity of speech segmentation, tone features, underspecification, phonetic realization, and the representations of allophones.

It is helpful to consider the potential implications both internal and external to the study of segmental contrasts. The intrinsic value of positing a universal feature set to account for observed segmental contrasts is greatest when the features have clear phonetic definitions. Strictly enforced phonetic definitions are what separate a restrictive proposal of nineteen binary features from one that is only falsified by an inventory with more than 524,288 (219) segments. Further, if a feature system that is motivated by inventories can account for other types of phonological observations, then it involves a fundamental claim about phonology rather than a claim about the description of inventories. D hypothesizes that the proposed feature system could account for classes of sounds involved in phonological patterns and provide a model of possible categorical allophones.

D’s starting point is the principle of contrast, which states that every pair of contrastive sounds in every language must be distinguished by at least one feature. His method for identifying necessary distinctive features is to search the inventories to find out how many degrees of contrast are required in each phonetic dimension (the maxima first principle). For example, searching the inventories for vowels that appear to differ primarily in backness turns up pairs of sounds in most languages (which is solid evidence that the feature system minimally needs a binary backness contrast), but it also yields several apparent backness triplets such as [i ɨ ɯ] and [e ə ɤ], which suggest the necessity of three degrees of backness. If each of these can be reanalyzed in a way that requires only two degrees of backness, then just a single binary feature is posited. Some of the major results of applying this method are the conclusions that all features are binary, that vowel quality can be described with four binary features for height, backness, rounding, and tongue root advancement, and that phonation differences can be handled exclusively by [stiff] and [spread]. D points out that binarity is not necessarily predicted by innatist or functionalist approaches to phonology; it is simply a result of the method he has applied.

Ladefoged (2007) also proposed a feature set meant to account for all segmental contrasts, based on his own phonetic data and other phonetic descriptions of languages exhibiting rare contrasts. These two feature proposals demonstrate opposite approaches to a splitter-lumper problem: Ladefoged sought to represent phonological contrasts and phonetic differences, and D seeks only to represent phonological contrasts. Ladefoged examines phonetic data and reports five degrees of contrastive...


Additional Information

Print ISSN
pp. 477-481
Launched on MUSE
Open Access
Back To Top

This website uses cookies to ensure you get the best experience on our website. Without cookies your experience may not be seamless.