Volume 44, Issue 5

1. Introduction to OT syntax General tendencies that hold in or across languages are easy to find. For example, sentences in English usually have a subject. The subject often is the first element in the sentence. In many cases, the subject is the agent of the action expressed by the verb. Crucially, these statements are mere tendencies, not absolute laws. It is very difficult to find observable properties that hold without exception across languages. If we were to formulate these general statements as rules, these rules would often have to be broken because of a number of exceptions. In standard generative syntax, constraints are assumed to be inviolable, i.e., they must be satisfied in a grammatical sentence. In order to “save” generalizations (rules) from apparent violations (counterexamples) in the linguistic data, generally three strategies are applied: 1. assume empty structure to satisfy the constraint (invisibly); 2. assume an abstract level at which the constraint is satisfied (invisibly); 3. modify the constraint, making it less general, so that it is satisfied by the data.

The hierarchy of grammatical functions currently employed by harmonic alignment (Aissen 1999a) makes use of the distinction between the core relations of subject and object. Evidence from languages with unregistered objects indicates that this hierarchy must be expanded to include a restricted set of objects as well. For example, a certain type of derived intransitive clause in Halkomelem Salish results in a ranking paradox under the normal conception of harmonic alignment due to the unexpected presence of an object in an intransitive construction. To relieve the pressures that these objects place on accounting for valency distinctions, it is proposed that constraints refer to both objects and to a class of restricted objects in order to make the distinction between transitive constructions and intransitives with objects.

The central question I address in this article is a theoretical one: is it possible to integrate binding into a local derivational syntactic approach, and what would the theoretical consequences of such an enterprise be? The answer to the first question is positive: I thus develop an approach to binding that takes into account both the crosslinguistic variation we encounter in the field of reflexivity and universal generalizations that can be observed with respect to binding. The theory I propose basically works as follows: (i) Binding corresponds to feature checking between binder and bindee (=:x); (ii) The concrete realization of x is determined during the course of the derivation in an optimality-theoretic competition. In the beginning, x is equipped with a realization matrix that contains its possible realizations (anaphoric or pronominal specifications); (iii) After the completion of each phrase, optimization takes place and might restrict x's realization matrix depending on the respective language and the domain that has been reached. As a result, anaphoric specifications might be deleted; (iv) When checking takes place, the optimal realization matrix of x is mapped to PF, where the concrete realization of x is finally determined in a postsyntactic process.

In this article we show that cumulativity is necessary to account for probabilistic variation found in actual language use, and we compare the accuracy of the predictions that different versions of stochastic OT make. We distinguish two versions of cumulativity, namely ganging-up cumulativity and counting cumulativity . We will compare how Paul Boersma's (1998) version of stochastic OT on the one hand and Maximum entropy models on the other hand deal with cumulativity. The second part of the article reports empirical data on English genitive variation. It turns out that both versions of cumulativity obtain in the empirical data. In the last part of the article, we compare the predictions of the two theories with respect to this empirical domain. The maximum entropy model proves to be clearly superior, both with respect to the accuracy of its predictions and to its learnability properties.

The present article deals with constraints on transcategorial processes such as nominalization. In particular, it addresses the issue whether one can predict the order in which verbal categories are lost and nominal categories are acquired in nominalization. It is argued that the disruption/acquisition of categories in transcategorial processes is determined by functionally based hierarchies of nominal and verbal categories, as suggested in the functional-typological literature. The hierarchy constraints in turn are shown to arise from the interaction of F unc F aith constraints forcing decategorization/recategorization and L ex F aith hedging these processes. Structural factors such as morpheme order and category cumulation can also interfere with the hierarchy constraints. These structural factors can be derived from conditions on output-output correspondences (OOCs) between the morphological structure of nominalizations with that of finite verbs, on the one hand, and nonderived nouns, on the other. Thus the outcome of nominalization processes is determined by an interaction of the function-based hierarchy constraints and OOC-related structural constraints.

Aissen (1999, 2003) argues that prominence hierarchy effects in morphosyntax are governed by fixed rankings of markedness constraints related to the hierarchies themselves by harmonic alignment (Prince and Smolensky 1993). In this article, I analyze the effects of prominence hierarchies on agreement in Dumi, an endangered Kiranti language spoken in Eastern Nepal (van Driem 1993), and argue that the empirical facts can be captured best if hierarchy effects follow from freely rankable binary preference and markedness constraints, not from the fixed ranking of markedness constraints. The analysis further reveals the relevance of number hierarchies for agreement, which are largely neglected in the literature. Thus Dumi resorts to the hierarchy plural > dual > singular that competes with the parallel effects of the more standard person hierarchy (1st > 2nd > 3rd person) on agreement control and also triggers an as yet undocumented number inverse marking, which indicates if one argument of a verb is more prominent for number than the other, but less prominent for person.

The fact that object shift only affects weak pronouns in mainland Scandinavian is seen as an instance of a more general observation that can be made in all Germanic languages: weak function words tend to avoid the edges of larger prosodic domains. This generalization has been formulated within optimality theory in terms of alignment constraints on prosodic structure by Selkirk (1996) in explaining the distribution of prosodically strong and weak forms of English function words, especially modal verbs, prepositions, and pronouns. But a purely phonological account fails to integrate the syntactic licensing conditions for object shift in an appropriate way. The standard semantico-syntactic accounts of object shift, however, fail to explain why it is only weak pronouns that undergo object shift. This paper uses an optimality theoretic model of the syntax-phonology interface which determines linear order by the interaction of syntactic and prosodic factors. The account can successfully be applied to further related phenomena in English and German.

This article shows that word-order freezing is not an adequate application of bidirectional optimality theory and that what it attempts to account for presents a problem for bidirectional optimality theory. Max(X) constraints allow moderate bidirectionality within a monodirectional production-oriented optimality theory and give better results on freezing. Since freezing is the only argument for bidirectional syntax that is syntactic in nature, the article also reviews the other arguments for full bidirectionality in syntax and concludes that they too are not compelling.