Syntax guru, Gereon Müller will present “Buffers in Syntactic Derivations,” at the department colloquium this Friday in Machmer E-37. Because of Open House events, the time of Gereon’s talk is later than the usual colloquium time. His talk starts at 4:30.
An abstract for Gereon’s talk:
Phase-based minimalist syntax can be characterized as a localderivational approach to grammar: Syntactic structures are generatedbottom-up, by alternating operations like Merge, Move, and Agree, andthe accessible window of a derivation is quite small throughout -- itis standardly assumed to be confined to the minimal phase (PIC,Chomsky (2001)). In such an approach, all long-distance dependenciesmust be modelled locally. Thus, unbounded wh-movement is assumed to becomposed of a series of smaller movement steps to intermediate phaseedges, and similar local analyses postulating a decomposition ofseemingly non-local syntactic operations into sequences of smallersteps have been given for other non-local phenomena, likelong-distance reflexivization, non-local case assignment, andlong-distance agreement.
Against this background, I will discuss three movement-relatedphenomena that have so far proven recalcitrant from alocal-derivational, phase-based perspective: improper movement,remnant movement, and resumptive movement. The empirical evidence willmostly come from German. In all three cases, the dilemma is that itseems that information from a syntactic domain A must be used in asyntactic domain B even though it should not be accessible in B --either because A is not present yet (the look-ahead problem), orbecause A is too deeply embedded (a kind of backtracking problem).
First, what is arguably the simplest and most elegant approach toimproper movement (viz., the Williams Cycle; Williams (1974; 2003),Sternefeld (1992), Grewendorf (2003; 2004), Abels (2008), Neeleman &van de Koot (2010), Bader (2011), Keine (2014)) presupposes that inorder to determine whether a given long-distance movement operation islegitimate or not, information about landing sites in the embeddedclause is available (backtracking). Second, it seems that in order todistinguish legitimate and illegitimate cases of remnant movement(involving Freezing, Anti-Freezing, and what has been called theMueller-Takano Generalization; see Mueller (1993; 1998), Takano(1994), Kitahara (1994), Koizumi (1995), Sauerland (1999), Pesetsky(2012)), decisions must be made early in the derivation that depend oninformation which is not available yet (look-ahead). Third, withresumption, the problem is again that at crucial stages of thederivation (viz., when an island can -- or, indeed, must -- becrossed; Sells(1984), Pesetsky (1998), Boeckx (2003), McCloskey(2005)), the relevant information (about whether or not there is aresumptive pronoun in the base position) does not seem to be availableanymore (backtracking).
My goal in this talk is to present a unified analysis of these threephenomena (this implies treating the remnant movement problem as abacktracking rather than a look-ahead problem) that makes crucial useof buffers which temporarily store minimal aspects of the derivationalhistory on a moved item. The locus of this storage is themovement-related feature of the moved item (e.g., [wh]), moreprecisely, its value, which is viewed as a first-in-first-out listthat constantly changes throughout the derivation but must qualify aslegitimate (essentially: respect the functional sequence of heads,f-seq) in criterial positions.
