Opaque Interactions of Merge and Agree: On the Nature and Order of Elementary Operations
Doreen Georgi
June 2014

In this thesis I explore opaque interactions of Merge and Agree in order to gain insights into the nature of elementary operations and their mode of application (simultaneous vs. sequential, extrinsic vs. intrinsic ordering). I investigate a configuration in which the Cycle does not predict an order among operations, i.e. in which a single head triggers both Merge and Agree. I argue (i) for more fine-grained elementary operations, in particular for a distinction between different types of Merge because they can apply at different stages of the derivation relative to Agree, and (ii) for extrinsic ordering of the operation-inducing features on a head. I investigate the configuration in which a single head triggers more than one operation because the Cycle is too weak to make any predictions about the order of operations. The effect in (i) becomes visible when Merge is interleaved with Agree: Empirical evidence suggests that Merge can apply both before and after Agree within a single language because it can feed / bleed Agree and also counter-feed / -bleed Agree: Merge > Agree > Merge. This symmetric order cannot be derived if, as assumed in recent literature in transparent interactions, there is a language-specific parameter that simply orders Merge relative to Agree. Based on the order where Agree “splits up” the Merge operation I show that we need to distinguish between final and intermediate movement steps (two types of internal Merge), external vs. internal Merge, and movement type-specific final and intermediate movement steps. The first distinction is motivated for interactions of Merge with upward and downward Agree. The empirical basis are reflexes of movement, which I take to be the result of a feeding interaction of Agree and internal Merge. I identify four patterns of reflexes and argue that variation is best analysed as a consequence of reordering of operations on a head: Movement steps that apply before Agree can feed or bleed Agree; those that apply after Agree have the opposite effect, i.e. they counter-feed/-bleed Agree. The ordering approach presents the first uniform analysis of these patterns; in addition, it can account for multiple reflexes and optionality of reflexes. I argue against alternative enriched representations analyses of opacity for empirical and conceptual reasons (incompatibility with Minimalist assumptions and modern theories of movement). Given that the approach crucially relies on timing (only a single operation can apply at any given stage of the derivation), it provides strong evidence for a strictly derivational model of grammar. In addition, I argue on the basis of the cross-linguistic variation that (i) final and intermediate movement steps in a long movement dependency are triggered by different features, (ii) that long movement always applies successive-cyclically even if we do not see a reflex of movement in all intermediate landing sites, (iii) that Agree applies in the syntax and is not a post-syntactic operation, and (iv) that the timing of edge feature discharge must be more flexible than proposed in the literature.
Format: [ pdf ]
Reference: lingbuzz/002129
(please use that when you cite this article)
Published in: PhD thesis, Universität Leipzig
keywords: rule interactions, extrinsic vs. intrincis ordering, merge, agree (upward vs. downward), movement, edge features, locality, derivation vs. representation, successive-cyclic movement, patterns of reflexes, wh-agreement, case splits, anti-agreement, morphology, syntax
Downloaded:3041 times


[ edit this article | back to article list ]