On long-distance theta-role assignment

Highlights

• Locality and activity in thematic theory.

• Derivation of some complex paradigms involving interactions between SE and causative constructions in Spanish.

• Arguments against radical locality in thematic theory.

Abstract

In this paper, the long-standing assumption that (External or Internal) Merge is a necessary condition for theta-role assignment is challenged. It is argued, then, that a theory of thematic assignment which allows for long-distance theta-role assignment under conditions of locality and activity is empirically preferable when it comes to some complex interactions between se and causative constructions in Buenos Aires Spanish. It is also shown that recent arguments against long-distance theta-role assignment do not hold in a theory as the one defended here, which conceives of theta-roles as regulated by conditions generally available in the realm of A-dependencies.

Introduction

A pervasive assumption about thematic theory is that Merge is a necessary condition for thematic assignment. The strongest version of such an approach is that only External Merge can satisfy thematic requirements (Chomsky, 1995, Chomsky, 2000, Chomsky, 2001, Miyagawa, 2010, among others). A weaker version – mostly represented by the proponents of the movement theory of control (MTC, Hornstein, 1999 and much subsequent works) – claims that Internal Merge also targets thematic positions (i.e., movement into theta positions is allowed). The common assumption is, again, that Merge is a necessary condition to establish thematic relations. Indeed, Sheehan (2012) has formulated this as a principle of UG:

In this paper, I challenge this long-standing assumption in current minimalist theory by showing that theta-role assignment can proceed in a long-distance fashion provided that conditions on activity and locality are met (Chomsky, 2000, Chomsky, 2001 and subsequent work).

The notions of locality and activity that are at play for thematic assignment will be defined and illustrated in detail below. As we will see, my claim that activity and locality are the relevant conditions for thematic assignment does not force us to accept that theta-roles are features checked under some version of the so-called Agree operation; indeed, through this paper I will remain neutral about this possibility (see section 3).

It is also important to emphasize that the crucial difference between (1) and (2) is locality. As I will define this notion, a given argument will be local with respect to a given thematic head if it is contained within the projection of such a head. Evidently, such a definition is more liberal than its competitor in (1), which requires a stricter notion of locality, i.e., Merge in any of its variants. The consequence of this difference is that, according to the theory informally presented in (2), Merge is not a necessary condition for theta-role assignment.

As argued in section 3, both theories seem to be extensionally equivalent in several empirical domains (i.e., well-known contrasts between reflexive and impersonal se constructions, intervention effects in Spanish double object constructions, and basic control structures). However, they differ with respect to a basic fact, namely, the ban of reflexivizing the causee subject in analytical causatives in Spanish and other Romance languages (Baauw and Delfitto, 2005, Saab, 2014). See (3b):

In section 4, I argue that the ungrammaticalities in (3b) are totally unexpected under some version of (1), in particular, under the attract-based theory mainly defended by the proponents of the MTC. On the contrary, as shown in Saab (2014), the same basic fact is directly derived in a theory that allows for theta-roles to be assigned in a long-distance fashion under activity and locality conditions (as defined below). If this is correct, we will have a new argument for dispensing with the Merge condition on theta-role assignment. Section 5 will reinforce the core argument by showing that analytical causatives of the hacer ‘to make’ type cannot be seen as ECM-constructions (pace López, 2001, López, 2012, among others), i.e., constructions in which the infinitival subject has to vacate the main clause for Case reasons. Indeed, the behavior of causatives and ECM constructions in Spanish drastically differ with respect to the main facts to be discussed here.

In section 6, I will address a putative argument presented by Sheehan (2012) against long-distance analyses of theta-role assignment and show that such an argument does not hold in the theory I will defend. Finally, I will also demonstrate that Reinhart and Siloni's (2005) long-distance theta-role assignment approach is incapable of accounting for the basic facts discussed in section 4.2 (see Appendix).¹

Before entering into the details of the arguments to be developed in this paper, I will first introduce some basic assumptions about argument structure based on previous works by Pujalte and Saab, 2012, Pujalte and Saab, 2014, Pujalte (2013) and Saab (2014). The main idea in those works is that argument structure effects have to be seen as epiphenomena resulting from the interactions between the operations Merge and Agree and the formal make up of functional heads. The theory will be illustrated with reference to se constructions in Spanish (mainly, reflexives and impersonals), an empirical domain that will be crucial for the main arguments developed in detail in the rest of the paper.