Vowel patterns in language (review)

This book is an extensive examination of how vowels interact with each other and the role of positional prominence in those interactions. The patterns discussed typically involve a feature or set of features that is restricted to a certain prominent position (e.g. stem or stressed syllable): either the feature(s) cannot appear outside that position at all, or they are permitted in other positions only if they are also realized in the prominent position. Walker develops a formalism within optimality theory based primarily on positional markedness to account for these patterns, and through detailed case studies, she places this framework in a larger context, showing how other phenomena can interact with prominence-based restrictions. Overall, W presents a thorough study of prominence-based vowel patterns, with conceptual soundness and functional explanations as guiding principles. This book represents a significant advance in our understanding of prominence, featural markedness, vowel interactions, and the formal account of these notions.

W begins with the observation that vowel features—especially marked features—are often restricted to prominent positions (stressed syllables, initial syllables, roots/stems, and even final syllables in principle), where prominence is defined in familiar perceptual, articulatory, and cognitive terms. Appearance in these positions enhances the features' perceptual properties. Depending on the language, features subject to such restrictions are either banned from nonprominent positions or permitted in nonprominent positions only when they have simultaneous membership in a prominent position. Strategies for ensuring compliance with positional restrictions vary crosslinguistically; this becomes a central concern in later chapters.

Prominence-based restrictions are enforced by positional licensing constraints, a variety of positional markedness that penalizes features that do not coincide with a specified licensor. This is perhaps the most significant formal contribution of the book: positional licensing is of course not new, but W develops a rigorous GENERALIZED LICENSING schema by which a variety of positional licensing constraints can be projected, and she argues that generalized licensing is superior to alternative approaches to prominence-based phenomena. LICENSE(λ, π) is violated when a feature λ does not coincide with the position π. Only positions independently identified as prominent are eligible for π, and λ may be a single feature or feature value, or a set of features or feature values. (The formalism can also require licensing of λ just when it appears in a marked context.) Generalized licensing achieves two goals: it unifies seemingly disparate prominence-based patterns under a single analytical umbrella, and it imposes restrictions on the content of π and λ to make clear predictions about the kinds of prominence-based phenomena that should be attested. For example, since π must be prominent, generalized licensing cannot produce a system in which a feature is restricted to a nonprominent position. Likewise, if λ includes an unmarked feature value, it must also include its more marked counterpart, so the theory excludes patterns in which only the unmarked value is subject to positional restrictions. These limitations on π and λ, however, are imposed from outside the formalism, and one area for further development might involve building them into the definition of LICENSE(λ, π) directly, perhaps by formalizing prominence and markedness asymmetries more fully. Nonetheless, the predictions are borne out in W's data.

License(λ, π) is satisfied when λ or its correspondent coincides with π. λ may be wholly contained within π, as in 1a (W calls this DIRECT LICENSING; σ_π is a syllable of type π). Alternatively, λ may appear in π and elsewhere as in INDIRECT LICENSING (1b), or λ may not coincide with π but instead stand in correspondence with a feature that does, as in IDENTITY LICENSING (1c). While all of these structures satisfy LICENSE(λ, π), 2 does not.

1. 1.

2. 2.

W reports a typological generalization: a language might permit only direct licensing, both direct and indirect licensing, or all three varieties. This typology stems from other pieces of her formalism. CRISPEDGE constraints produce the first language type: they restrict a feature to having membership in exactly one instance of a particular domain, such as a syllable, and thus ban 1b,c. The second language type comes...