Structural Markedness and Syntactic Structure

A Study of Word Order and the Left Periphery in Mexican Spanish

Rodrigo Gutiérrez-Bravo

• Éditeur: Routledge
• Année: 2013
• ISBN:
  9781135500276
• Édition: 1
• Sujet: Languages & Linguistics
• Sous-sujet: Spanish Language

This book investigates a number of word order phenomena in Spanish, concentrating on this language's unmarked word order and the perturbations of this order that result from topicalization and wh-movement.

Chapter One
Introduction

The purpose of this work is to investigate some of the grammatical properties that lead to word order variation in Spanish, and to argue that these properties are inherently related to markedness restrictions on syntactic structure. In this first chapter I introduce the problems that will be addressed in the chapters that follow and the theoretical assumptions that I will adopt for their analysis. I first present an overview of recent analyses of word order and the data from Mexican Spanish that renders these analyses problematic. Then I discuss the OT framework that I adopt throughout this work. I conclude by laying out some of my general assumptions about the syntax of Spanish and by characterizing the variety of Mexican Spanish that provides the data for this work.

1.1 The Study of Word Order

1.1.1 General overview

Costa (1998), one of the most influential works on word order variation developed in recent years, summarizes the fundamental issue that needs to be addressed when studying word order phenomena. As noted by Costa, the study of word order variation provides two different, but presumably closely related areas of investigation. The first one is cross-linguistic word order variation, the study of why constructions with the same interpretive properties and discourse status can differ with respect to word order across different languages. The second one is language-internal word order variation, that is, the study of why, in any particular language, constructions with different interpretive and discourse properties can show differences with respect to word order. Although most of the research that follows will concentrate on the latter issue, it is worth noting that the proposal that I will develop here stems from what I believe are two serious limitations of most cross-linguistic analyses of word order.

In order to characterize what these two problems are, consider the analysis in Alexiadou & Anagnostopoulou (1998), another very influential work on word order variation. Simplifying somewhat, this analysis takes as its starting point the observation that there are strict SVO languages, strict VSO languages, and languages that show an SVO/VSO alternation with respect to unmarked word order. English would correspond to the first group, Welsh to the second, and Greek and some varieties of Spanish to the third group.

On a first approximation it looks like these word order differences can be accounted for by appealing to the interaction of a number of different parameters, set to different values in the languages under consideration. This is precisely the proposal developed in Alexiadou & Anagnostopoulou (1998). The first parameter they propose has to do with the satisfaction of the Extended Projection Principle (EPP) of Chomsky (1981, 1982). Alexiadou & Anagnostopoulou argue that VSO and SVO/VSO languages without expletives satisfy the EPP via verb raising because they have verbal agreement morphology with the categorial status of a pronominal element.² From this it follows that (i) pre-verbal subjects in SVO/VSO languages are not in an A-position, and (ii) VSO orders never involve a covert expletive. Concretely, Alexiadou & Anagnostopolou propose, following work by Rizzi (1982) and Contreras (1991), that in these languages verbal morphology agreement includes a nominal element ([+D, +interpretable phi-features, potentially +Case], so V-raising in these languages checks the EPP feature in the same way that XP-movement does in non-pro-drop languages. Consequently, it is not necessary for the subject DP to move to satisfy the EPP in null subject languages and the VSO order that these languages can display follows automatically. On the other hand, this parameter is set to the opposite value in languages like English. Accordingly, in the absence of expletive insertion, movement of the subject XP to [Spec, AgrS] is the only option available to satisfy the EPP. This explains the strict SVO order attested in these languages.

The second parameter has to do with whether or not [Spec, T], TP being the phrase immediately subjacent to AgrSP, is projected for Case-theoretic reasons. Alexiadou & Anagnostopolou propose that [Spec, T] is projected in Welsh and other Celtic languages. Following the analysis of Irish in McCloskey (1996), the VSO order in these languages is then derived by movement of V through T, ultimately to land in AgrS, and by movement of the subject DP to [Spec, T] to satisfy Case requirements. Consequently, the subject has a fixed position in these languages (the strict VSO order). In comparison, this parameter is set to a negative value in Greek and Spanish: [Spec, T] is not projected in this case, and so the subject does not need to move from its VP-internal position to satisfy Case requirements. Given verb raising, it does not need to move to satisfy the EPP either, and so the SVO/VSO alternation can now be understood as an instance of optionality, presumably related to discourse considerations. The essentials of Alexiadou & Anagnostopoulou’s proposal can be summarized as follows:

Impressive as this proposal is, it faces two major challenges that result from broader empirical considerations. On a first approximation it is not evident why a parametrical account of word order variation should be inherently problematic. But this is the result of the fact that too narrow a set of phenomena is being considered in the first place, namely, active clauses with transitive predicates. When we expand this narrowly defined set, a completely different picture emerges. This is because it is not unusual that in any particular language, the relative order of the arguments of a predicate may be different depending on the specific class of verbs to which the predicate belongs.³ The relevant facts are well-known from the descriptive literature on word order (see for example Arnaiz (1998) for Romance languages and Holmberg & Rijkhoff (1998) for Germanic languages), and Spanish is precisely a case in point. Following the standard assumption that unmarked word order is displayed by sentences that can be felicitous answers to questions like what’s happening?, in (5) a clear asymmetry can be observed when transitive, psych, and unaccusative predicates are all taken into account.⁴

Clearly, any analysis of word order that takes into account only transitive clauses and derives the movement of the subject to a preverbal position as a consequence of a property that the subject must satisfy there, will not derive the right result, since the subject DP does not occupy the same position in different classes of predicates. Whatever condition is met by fronting the subject to a preverbal position in transitive constructions is clearly not being met in unaccusative constructions like (5c), where the subject remains in a post-verbal position and the preverbal position remains empty. Similar complications arise with Psych predicates. Even if we assume that the preverbal oblique experiencer⁵ of a construction like (5b) satisfies the same condition that the transitive subject does in (5a), it is not clear why it is the oblique experiencer, and not the grammatical subject, that is fronted to this position.

This observation is particularly troublesome for analyses of word order in transitive constructions where the subject undergoes movement to the preverbal position in order to satisfy a condition specifically defined with respect to the argument of the verb that bears the subject grammatical relation, such as transformational analyses of Spanish where Case is assigned in the preverbal position. It also represents a serious problem for Optimality-theoretic analyses where the subject moves to this position to satisfy a structural subject condition (typically the SUBJECT constraint of Grimshaw 1997) as in Samek-Lodovici (1996), Grimshaw & Samek-Lodovici (1998), and/or a structural Case requirement (Costa 1998).⁶

It is not immediately evident how this criticism applies to the proposal in Alexiadou & Anagnostopolou (1998), though, where fronting of the subject in transitive constructions is an optional operation. It could be argued that this option is taken in (5a), but not in (5b) or (5c). In either case, no syntactic condition is satisfied by fronting of the subject, so it comes as no surprise that the absence of this fronting operation has no effect on the grammaticality of the relevant examples.

Although Alexiadou & Anagnostopolou’s analysis fares better than Subject-condition analyses (either transformational or Optimality-theoretic), it still raises two important issues. First, if fronting is entirely optional in Spanish, it is not obvious why speakers have such clear intuitions about unmarked word order (i.e. IOVS for Psych predicates instead of SVIO, for example). And second, if fronting is entirely optional, this means that for every sentence in Spanish where a constituent is fronted there should be an alternative option, where the preverbal position is empty.

This second issue leads us to the second problem faced by parametrical accounts. Such accounts are based on the assumption that once some parameter is set for a given language, the language will show a fairly uniform set of grammatical properties that are related to the parameter in question. For example, a language is either pro-drop or it is not, and a number of (sometimes mutually exclusive) properties follow from this setting of the parameter. In other words, parameters represent an all-or-nothing approach to the characterization of grammar and grammatical constraints. However, as research in Optimality-theoretic syntax has pointed out since its earliest days (see for example Costa 1998; Sells et.al. 1996; Grimshaw 1997; Samek-Lodovici 1996; Grimshaw & Samek-Lodovici 1998) there is considerable empirical evidence that the all-or-nothing approach to word order cannot be entirely correct.

In relation to the parametrical analysis of the satisfaction of the EPP in Alexiadou & Anagnostopolou (1998), Mexican Spanish is particularly illuminating. Mexican Spanish, like any other variety of Spanish, displays all characteristics of languages where the EPP is satisfied through verb raising; null subjects, lack of expletives, lack of definiteness restrictions on post-verbal subjects. Initially, the data in (5) would appear to further provide evidence in favor ol this analysis: since the EPP has been satisfied through verb raising, it is not surprising that the presence or absence of a preverbal XP in the examples in (5) has no effect on grammaticality. But, once again, an entirely different picture emerges as soon as we expand the range of constructions in which we expect to find these same effects. In particular, all the examples from Mexican Spanish below, where the preverbal position is empty, are extremely marked, if not downright ungrammatical:⁷

Example (6a) is particularly important, since it shows that the SVO/VSO alternation is absent in Mexican Spanish,⁸ but all the examples pose the same problem for Alexiadou & Anagnostopolou’s proposal. In their analysis, the unacceptability of the data in (6) goes unaccounted for. The verb has raised to I⁰ in all cases, carrying the [+D] feature that characterizes verbal agreement morphology in null subject languages, so EPP checking should have taken place successfully. By contrast, all the examples in (...