Volume 31, Issue 1-2 - Cyclic Linearization of Syntactic Structure / Sound Patterns of Syntax: Object Shift

This paper proposes an architecture for the mapping between syntax and phonology – in particular, that aspect of phonology that determines the linear ordering of words. We propose that linearization is restricted in two key ways. (1) the relative ordering of words is fixed at the end of each phase, or ‘‘Spell-out domain’’; and (2) ordering established in an earlier phase may not be revised or contradicted in a later phase. As a consequence, overt extraction out of a phase P may apply only if the result leaves unchanged the precedence relations established in P. We argue first that this architecture (‘‘cyclic linearization’’) gives us a means of understanding the reasons for successive-cyclic movement. We then turn our attention to more specific predictions of the proposal: in particular, the effects of Holmberg’s Generalization on Scandinavian Object Shift; and also the Inverse Holmberg Effects found in Scandinavian ‘‘Quantifier Movement’’ constructions (Rögnvaldsson (1987); Jónsson (1996); Svenonius (2000)) and in Korean scrambling configurations (Ko (2003, 2004)). The cyclic linearization proposal makes predictions that cross-cut the details of particular syntactic configurations. For example, whether an apparent case of verb fronting results from V-to-C movement or from ‘‘remnant movement’’ of a VP whose complements have been removed by other processes, the verb should still be required to precede its complements after fronting if it preceded them before fronting according to an ordering established at an earlier phase. We argue that ‘‘cross-construction’’ consistency of this sort is in fact found.

In this paper, I explore what a purely phonological account of object shift (OS) involves and what research questions it leads to, in particular what it means for word order to be phonologically motivated and what morpho-phonological primitives are involved. I pursue the possibility that what licenses OS of full DPs in Icelandic is phonological properties, not found in other Scandinavian languages, together with overt case marking. Although the position of the object is determined phonologically, the architecture I propose, in which topic and focus play a central role, allows for an account of the semantic effect associated with OS in Icelandic.

Fox and Pesetsky (henceforth F&P) propose an architecture for the mapping between syntax and phonology which relates a number of different constraints on movement to the way in which phrase structure is linearized. They investigate Object Shift (henceforth OS) and Quantifier Movement (henceforth QM) in Scandinavian and argue that the restrictions on these processes, namely Holmberg’s Generalization (HG) effects on OS and what they call ‘‘the inverse Holmberg effect’’ on QM, reflect a requirement for preservation of the order established in the VP due to the fact that the VP is a Spell-out domain. F&P’s proposal relies on Holmberg’s (1999) formulation of HG which has been challenged by Anagnostopoulou (2002) on the basis of data discussed in Anagnostopoulou (2003) that directly contradict Holmberg (1999). It is my goal here to investigate how these data can be accommodated in F&P’s system. I will argue that even though F&P can, in principle, account for the data in question, the attempt to unify the restrictions on OS, QM with comparable restrictions on passivization under F&P’s architecture fails to express certain crosslinguistic generalizations which are straightforwardly captured in traditional locality accounts.

This volume’s two target articles explore novel approaches to word order alternations, especially Scandinavian Object Shift. They share the common perspective that aspects of linear order long considered the exclusive purview of syntax may be better understood if the burden of explanation is split between phonological and syntactic modules. The two articles differ substantially, however, in how this general hunch plays out, in particular in the amount of the explanation that is attributed to extra-syntactic factors. Fox and Pesetsky’s ‘‘Cyclic Linearization’’ model (hereafter F&P, CycLin ) is compatible with familiar syntactic models, and can be seen as a filter running (cyclically) on the output of syntactic derivations. F&P suggest that their proposal can explain various heretofore stipulated conditions on syntactic operations as consequences of the architecture of their system and a single axiom about linearization. Erteschik-Shir’s proposal in ‘‘Sound Patterns of Syntax’’ (hereafter E-S) is more radical, in the sense that far less of the familiar syntax is retained; where for CycLin movement is still a syntactic process, on E-S’s view a good deal of traditionally syntactic movement must be rethought in linear, rather than hierarchical terms. Both articles are largely exploratory and leave many of the details still to be spelled-out. To engage the ideas on specifics, then, will involve to some degree making some educated guesses about what ancillary assumptions the relevant authors might condone. I will therefore restrict myself to a few comments at a general level, though it will be impossible to do justice to these authors’ ideas in the allotted space.

Fox and Pesetsky (F&P) make the claim that certain constraints on Object Shift and Quantifier Movement in Scandinavian can be explained within their model of the mapping between syntax and phonology. Among the salient features of this model are: Spell-out occurs phase-by-phase (as in recent work by Chomsky 2000, 2001a,b), and linearization relations established at each Spell-out point must be preserved (F&P’s property of Order Preservation). As F&P demonstrate quite elegantly, this system enforces successive cyclicity in wh -movement, in that non-successive-cyclic derivations will result in an ordering contradiction (violating Order Preservation). Thus, the ‘‘cyclicity’’ of successive cyclicity is shifted to being a by-product of the cyclic spelling-out process. One might still ask why Spell-out itself is cyclic, but this is certainly a welcome simplification.

The article to my mind rightly emphasizes the importance of prosodic and discourse-dynamic factors. Applied to the Norwegian phenomena, these factors may be rendered as: (1) a. A pronoun is incorporated in a prosodic domain P only if its referent is included in the discourse-dynamic domain GIVEN relative to the proposition being expressed. b. Adverbial elements that can occur right-adjacent to a category C, can also occur right-adjacent to the cluster C+pron + when this cluster is a prosodic domain.

What the theories of Fox & Pesetsky (F&P) and Erteschik-Shir (ES) have in common is that they both account for Holmberg’s Generalization (HG) in terms of linearization. That is virtually the only thing they have in common, though, and even that is more a nominal than a substantial resemblance. For one thing, in F&P it is a matter of linearization of VP constituents as part of the narrow-syntactic derivation, in ES a matter of linearization of adverbs in relation to verb and object in a model where phonology / prosody determines word order. They can’t both be right, but they can both be wrong, which, however, we can’t be certain of until the set of hypotheses they propose have been thoroughly investigated and tested. Below I will present some critical remarks on both theories, pointing out some possible false predictions in F&P in section 2, and one or two possible flaws in ES in section 3. In section 4 I will consider how the two theories fare with respect to properties of Object Shift (OS) which they do not themselves take up. In section 5 I will briefly consider the problem of inverted objects.

The goal of this paper is to compare the approach to order preservation in Fox & Pesetsky (2005) with the approach I develop in Müller (2000) on the basis of Williams (1999, 2003) and an earlier study of mine. I will proceed as follows. In section 2, I briefly sketch main features of Fox & Pesetsky’s (2005) analysis. Against this background, I discuss some of the properties of the approach developed in Müller (2000) in section 3. In section 4, the two approaches are compared. Section 5 draws a conclusion.

Fox and Pesetsky (this issue, henceforth F&P) argue that the ‘‘edge effects’’ derived by stipulation within standard phase theory can be explained in their version, where a phase crucially triggers linearization of its constituents. Later phases may add ordering statements to an ordering table in a monotonic fashion, but no information can be erased or altered once it has entered the ordering table. Suppose a phase A contains the constituents x , y , z , and that they are linearized in that order. Then movement within the next phase can’t result in reordering of x , y , z . So x , being the leftmost element, can move leftwards freely within the next phase, while y can only move leftwards provided that x moves even further leftwards. This is, in essence their explanation of Holmberg’s Generalization (Holmberg, 1986, 1999) (HG). To see this, imagine that y in our setup is an object trying to shift, and that x is the verb. They also show that, if y moves to the left edge of A prior to its linearization, this may end up blocking leftwards movement of x in what they term the ‘inverse Holmberg’s Generalization’, and they have empirical support for the existence of that pattern. This gives a version of phase theory where phases are not entirely opaque to outside syntactic probing. Their proposal is highly innovative and elegant, and it succeeds in deriving an impressive range of facts. The following paragraphs present some relatively minor empirical problems with their treatment of HG which unfortunately seem to conspire to uncover a major one.

In this short review I address certain issues and predictions that arise from the position papers by Erteschik-Shir, and by Fox and Pesetsky. The two proposed analyses are radically different, so I have made no attempt to relate them in terms of their properties. In section 1 I discuss some underlying issues in the architecture of grammar assumed by Erteschik-Shir, and in section 2 I evaluate some of the predictions generated by Fox and Pesetsky’s approach. In section 3 I draw some more general conclusions about what consequences Object Shift shows us concerning the organization of grammar, and the nature of grammatical principles. While my remarks mainly address problems or drawbacks that I see with the proposals in the two position papers, they both offer a stimulating account of OS and explore the wider implications of the natures of their enterprises, and deserve fuller consideration beyond the present venue, consideration that they will surely receive.

Both target articles, in conformity with minimalist ideals, aim to discover to what extent syntax itself can be relieved of descriptive burden, and just how much of word order can be attributed to mechanisms of linearization at the PF interface. I point out two paradoxes the treatment of V2 gives rise to in Erteshik-Shir’s paper, and dispute the status and effectiveness of processing constraints in accounting for (un)grammaticalities in various word order patterns. I discuss the relation of Fox and Pesetsky’s cyclic linearization algorithm to ‘tucking in’ as well as to the ‘Attract Closest’ property of movement. Qualms are raised regarding several apparently stipulative and / or conceptually unattractive properties of the linearization mechanism, and ostensible violations of order preservation are examined.

The papers here represent very different analyses of Object Shift (OS), and take very different positions on the interaction of phonology and syntax, an area in which there are many unsettled questions. Mainstream work tends to hold that syntax is blind to phonological content, with certain exceptions, for example sometimes phonetically null elements require special syntactic licensing (Chomsky 1981), or certain syntactic rules only apply to nodes with phonetically visible features (Holmberg 2001). Basically falling within the mainstream are proposals that syntactic movement can be blocked by or driven by requirements that have phonological effect at the output, such as adjacency (Bobaljik 1995, Kidwai 1999) or rules matching prosodic structure with focus structure (Zubizarreta 1998). Such accounts generally describe movements in the syntactic terms of specifiers and feature checking and so on, and do not rely on the visibility to syntax of strictly phonological features. We can call these all Syntactic accounts. Most accounts of OS are Syntactic in these terms, for example those of Holmberg (1986), Holmberg and Platzack (1995), and Bobaljik (1995, 2002).

‘‘Cyclic Linearization’’ (CL) is similar in its intent to my own ‘‘Representation Theory’’ (Williams 1998, 2003) (RT) in that it seeks to explain some parallelism constraints in terms of general architecture, and in fact both were inspired by Holmberg’s generalization and the general properties of Object Shift in the formulation of central tenets. But here, I will explore a difference in the models that has more to do with WH movement.

Our proposal is concerned with the relation between an aspect of phonology (linearization) and syntax. In the picture that we had in mind, the syntax is autonomous – ‘‘it does what it does’’ – but sometimes the result maps to an unusable phonological representation. In this sense, linearization acts logically as a filter on derivations. We know of no evidence that the syntax can predict which syntactic objects will be usable by the phonology, and we know of no clear evidence that the phonology communicates this information to the syntax. In this sense, our proposal fits squarely into the tradition that Svenonius characterizes as the ‘‘mainstream’’. We thus attempted to identify certain deviant configurations that are not plausibly excluded for syntax-internal reasons, but are filtered out in the linearization process.

In this response, I address the main points of my article raised by the reviews in this volume. I argue that the properties responsible for word order are in fact phonological. I clarify the properties of the identification constraint, revise my account of V-2, and elaborate on some of the phonological properties involved in incorporation.