Locality in vowel harmony (review)

Andrew Nevins’s ‘Linguistic inquiry monograph’ is misnamed in a way that understates the scope and importance of his work. Although the bulk of the data that N considers in this monograph is from the phenomenon of vowel harmony, the fundamental proposals about locality that N makes have implications for all of phonology in general. An additional welcoming aspect of this work is the demonstration of how to pursue formal phonology in the twenty-first century in a computationally grounded way. There are a total of six chapters with notes, references, an index of terms, and an index of languages (over sixty languages are discussed as examples). An additional aspect of this book is an online tool (http://mitpress.mit.edu/vowel_harmony), which demonstrates all of the examples in the book and allows for the reader to investigate and test other data as well.

Ch. 1, ‘What is vowel harmony, how does it vary, and why study it?’, lays the foundation of the monograph. N carefully circumscribes vowel harmony to only range over phonological instances of harmony. Phonological vowel harmony is distinct both from morphological vowel harmony, where distinct allomorphs are chosen based the phonological content of the root/base/stem, and from coarticulation, which is the phonetic phenomenon where both local and long-distance phonetic targets affect the articulation of a particular phone. N does not deny the existence and importance of these other phenomena but instead carefully narrows his proposals to exclude these other cases.

At the core, N proposes that vowel harmony is the result of a principles-and-parameters (PP) type operation that copies phonological features from a source to a target that can apply in a nonlocal fashion. A PP approach is well suited to this task since the parameters that N posits directly define the symbolic nature of phonological harmony, because it is phonological features that are being manipulated. The crosslinguistic variation is directly limited to what can be achieved through the manipulation of these features. The final component in this chapter is the justification for this new approach to vowel harmony, which is the question of whether locality in phonology is computed in an optimal manner. N presents a prelude to the answer to this question by invoking parallels in syntax where it is clear that ‘distance’ is computed in a relativized manner (Rizzi 1990), so it should not be surprising that phonology also has this characteristic.

Ch. 2, ‘The search principle’, presents the core mechanisms that account for harmony. A search-and-copy process occurs if a segment is idiosyncratically marked through the omission of a distinctive feature (see Lightner 1972 for discussion of localizing diacritics on segments, not morphemes). The search process will mechanically move through each segment (direction is set via a parameter, R, L, or both), checking whether the search conditions have been met. If the search conditions are not met, thus triggering copying of the feature, then search moves onto the next segment. This continues until either a feature is copied or the end of the word is met. The final novel aspect is that the phonological operation of copying is ‘from’ rather than ‘to’, meaning that the search originates from the segment that will receive the copied feature value, as opposed to previous approaches where a feature is copied or spread from the source to the target.

To make N’s basic proposals explicit, consider his analysis of the classic Turkish genitive (gen) and plural (pl) suffixes. The gen suffix in Turkish alternates between four different surface forms (i.e. -in, - i n, -ün, -un) based on the values of the phonological features [+/–back] and [+/–round] of the preceding vowel. The pl suffix alternates between two different surface (i.e. -lar, -ler) forms based on the value of the phonological feature [+/–back] of the preceding vowel. The representation in 1 shows the phonological representation before any search-and-copy procedure occurs. The pl and gen suffixes differ in what phonological features are sought by...