Word Order Variation: A Constraint-Based Approach (review)

Most contemporary generative theories of syntactic structure predict a single configuration for each syntactic combination, the result of the interaction of necessary principles and parametric settings. Optimality theory (OT), however, is predicated on the notion that all constraints can be violated on a language-specific basis—although OT has not always been successful in analyzing apparently free variation in language. The book under review combines both theoretical paradigms in an attempt to derive multiple word orders, largely involving adverbs, focused elements, and other adjuncts. Most of the examples come from European Portuguese and English, with occasional specimens from other languages. The book contains brief introductory and concluding chapters and six substantive chapters of which the first three deal with structural representations and the latter three outline a model of constraint interaction.

Following an introduction, the book begins with a chapter on the distribution of adverbs. Using comparative data from several languages, Costa concludes that adverbs may be adjoined to different projections but may not be right-adjoined; apparent right-adjunction results from the movement of elements into [Spec, AdvP]. Prosodic factors that affect adverb placement are also introduced, allowing for apparent deviations from canonical adverb placement.

Ch. 3, ‘The distribution of arguments’, concentrates on Portuguese, beginning with the position of subjects. Preverbal subjects are claimed to occupy [Spec, IP] as topics while postverbal subjects are in [Spec, VP] (the base position) and are focus elements. Prosodic factors including the sentential stress assignment algorithm require that focus (main-stressed) elements occupy sentence-final position. Similarly, focused objects remain in their base-generated position while defocused objects undergo scrambling and left-adjoin to VP.

Ch. 4, ‘Focus in situ’, elaborates the notion that focus targets rightmost rather than leftmost sentence constituents. Rejecting the notion that focus involves movement to Spec FP or the specifier of some other functional projection, even at LF (due to violation of c-command requirements), C builds an in-situ focus model in which prosody plays a key role.

Part 2, ‘Constraint interaction’, contains three chapters plus a concluding section. Ch. 5, ‘Word order and constraint interaction’, introduces the main OT constraints which will account for cross-linguistic variation in word order. According to C’s analyses, arguments (subjects and objects) may not occur superficially in the positions in which their case is licensed. The potentially violable constraint case (differentiated for subjects and objects) requires that NPs be licensed in the Spec position of an AgrP. Alignfocus requires that the unmarked focus of the sentence be the rightmost position (in accordance with the branching structure of the language). Stay disallows movement operations. The first two constraints account for most instances of subject position, while stay is involved with scrambling and object movement. Ch. 6, ‘Word order typology in OT’, presents a unified account of differing word orders based on permutations of the above-mentioned constraints. For C, unmarked word orders do not arise from lexical features (as suggested, e.g. within the minimalist paradigm) but rather through the ‘emergence of the unmarked’ when normally dominated constraints are allowed to emerge and shape word order patterns. In the case of word order, this occurs when discourse constraints are lifted, allowing purely syntactic considerations to dictate word order. C assumes no universal basic word order but presents all occurring patterns as the result of constraint interaction. This requires some problematic assumptions along the way, e.g. that there are two fundamental varieties of Spanish, one in which SVO word order is basic and one with basic VSO order.

In Ch. 7, ‘Comparing frameworks: Parameters vs. soft constraints’, the competing models are juxtaposed. C presents OT as a means of accounting for apparent exceptions in the application of syntactic principles such as movement and case-assignment, distinguishing between a high-ranking constraint and a positively set parameter as well as low-ranked constraints and negatively set parameters. The observed violability of even high-ranked constraints in OT (for example...