Movement Operations after Syntax DavidEmbick Rolf Noyer Wedevelop a theoryof movement operations that occur after the syntacticderivation, in the PF component,within the framework of DistributedMorphology. The theory is anextension of whatwas called MorphologicalMerger inMarantz1984 and subsequent work. A pri- maryresult is that the locality properties of a Mergeroperation are determinedby the stage in the derivation at whichthe operation takes place:specifically, Merger that takes place before Vocabulary Inser- tion,on hierarchical structures, differs from Merger that takes place post–Vocabulary Insertion/ linearization.Specific predictions of the modelare tested in numerous case studies. Analyses showing the inter- actionof syntactic movement, PF movement,and rescue operations areprovided as well,including a treatmentof English do-support. Keywords: morphology,syntax, Morphological Merger, adjacency, PF movement,Distributed Morphology Theproperties of syntacticmovement have been studied extensively in linguistic theory, both in termsof localityconditions and in termsof thetypes of constituentsaffected (phrases, subparts ofphrases,heads). Despite differences in particular analyses or frameworks,the locality conditions onmovement operations are a centralconcern of current research. Here we addressmovement operationsas well, but operations of a differenttype. In particular, we examineand analyze movementoperations that occur after the syntactic derivation, in the PF component,and provide atheorythat makes proposals concerning (a) thelocality conditions on such movements, (b) the typesof constituentsthey affect, and (c) theposition of suchoperations in the sequential derivation from theoutput of syntaxto phonologically instantiated expressions. From asomewhatabstract perspective, the fact on which we baseour study is that not all structuresand strings are the result of operations that occur exclusively in thesyntacticcomponent ofthegrammar; this observation stems from abodyof prior research investigating the relationship betweensyntactic structure and phonological form. Theobservation covers two domains: one dealingwith linear sequences that are syntactically opaque, the other with movement operations. Inthe first domainit has been demonstrated that the internal ordering of clitic clusters cannot Wewouldlike to thank Artemis Alexiadou,Karlos Arregi,Rajesh Bhatt,Jonathan Bobaljik, Morris Halle, Heidi Harley,Tony Kroch, Alec Marantz,David Pesetsky, Beatrice Santorini,and audiences at NYU, MIT,and ZAS/ Berlin, where partsof this article were presented.We wouldalso like to acknowledge the Penn Research Foundationfor financial supportfor our research. Linguistic Inquiry, Volume 32, Number 4, Fall 2001 555–595 q 2001 by the Massachusetts Institute of Technology 555 556 DAVIDEMBICKANDROLFNOYER followfrom syntacticprinciples (see Perlmutter1971; also see Bonet 1991 and Noyer, to appear, for explicitdiscussion in the framework assumedhere). In the second domain, which we explore here,it has also been recognized that syntactic movement cannot be responsiblefor certainmove- mentoperations. One of the most familiar comes from theinflectional system of English. While ithas long been recognized that English main verbs do not move to T(ense) in the syntactic derivation(Emonds 1978, Pollock 1989), it isneverthelessthe case that tense morphology appears ontheverb in the surface string: Johnkick-ed the ball to Mary. Thepositioning of tensemorphol- ogyon the verb has been attributed to Lowering, a movementoperation that ultimately derives from theAffix Hoppingtransformation of early generative grammar. Subsequent attempts to identifyand analyze ‘ ‘mismatches’’ betweenmorphology and syntax can be found in the work ofMarantz(1984, 1988) and Sadock (1991). Our goalin thisarticle is todevelop a morecompre- hensivetheory of the types of morphological movement operations in the grammar andto situate theseoperations within an explicit architecture for thePF derivation.The focus is on cases in whichsome sort of displacementmust apply to generate the proper phonological form from an identifiablesyntactic structure. We thusexclude from considerationcases in which, for example, thesyntactic distribution of clitic clusters is exactly what one would expect given the proper syntacticanalysis. 1 1Syntacticversus Postsyntactic Our initialdiscussion of movement in PFhastwo components. The first concernswhy postsyntac- ticdisplacements are needed in addition to syntactic movement. The second addresses the idea thatsyntactic movements of the more familiar sort are part of thePF derivationas well. 1.1Clitic Placement Theneed for nonsyntacticmovements has been clearest in the domain of clitic placement —in particular,in the analysis of second-position clitics. In some cases the question that was posed was whetheror not syntactic movement could possibly account for patternsof cliticplacement, orinparticular second-position phenomena. We takethis question to be ill formed. Syntax is a generativesystem, and assuming that one is willing to loosen many of the constraints on syntactic movement,the ability of such a systemto capture certain linear orders should never have been inquestion. Rather, the question that we taketo be central is whether or not it is desirable to havesyntax perform such operations. The two positions that we wishto contrast,along with the meansby which each approach captures apparent postsyntactic movements, are as follows: (1) a. Syntaxonly: Syntaxperforms operationsthat are explicitly executed so astoresolve amorphophonologicalproblem. Patterns of apparent postsyntactic movement are reducibleto the effects of these‘ ‘special’’ syntacticprocesses. b. PFmovement: Syntaxgenerates and moves terminals according to itsown principles 1 Athirdtype of difference instructure between PFandsyntax proper involves the addition of morphemes at PF tomeet language-particularrequirements; this will be outlined below. MOVEMENTOPERATIONSAFTERSYNTAX 557 andis oblivious to morphophonological concerns. PF takesthe output of syntax and resolvesmorphophonological dependencies according to its own principles. Theposition we adoptendorses the latter possibility. 2 Inessence, the idea in thedomain of cliticplacement is that unless syntax incidentally provides a hostfor aclitic,PF canperform movementoperations to satisfy a cliticdependency (see, e.g., Marantz 1984, 1988, Halpern 1992b, Schu¨tze1994, Embick and Izvorski 1995). Although the discussion of this section might seem somewhatperemptory, the literature on cliticizationis vastand cannot be addressedhere. Argu- mentsagainst syntactic treatments of various clitic placement phenomena are abundant in the literature.In the body of thearticle we willnot make a ‘‘notsyntactic’ ’ argumentfor eachcase studywe present,although such arguments could certainly be derived from theexamples we analyze.Rather than focusing on this aspect of the phenomena, we willprovide a theoryof the localityconditions on postsyntactic readjustments of the type outlined above. 1.2Syntax in PF? Recentdevelopments in syntactic theory, particularly those associated with the Minimalist Pro- gram,attempt to circumscribe the operations that syntax proper is supposed to perform.Although we donot adopt some of the more extreme versions of this view, such as Chomsky’ s (2001) positionthat head movement does not takeplace in syntax,we findthe generalpicture provided by suchtheoretical contexts compelling, particularly to theextent that they acknowledge movement operationson thePF branch.This having been said, the nature of themovements that we intend todiscuss must be clarified somewhat. Chomsky (2001) takes the more extreme position that a greatdeal of apparently phrasal movement also takes place in the phonological component (in somecases this is whathas been referred toas ‘ ‘stylisticmovement,’ ’ inother cases not). Unlike themovement operations to bestudied here, which have an immediately local character and which aremotivated by the satisfactionof primarilymorphological or morphophonologicalrequirements, themovements Chomsky relegates to PFseemto have many of the properties of straightforward syntacticmovement. We takethe position that the grammar includesonly one syntactic component: that is, every- thingthat looks like syntactic composition or movementtakes place in syntax;movement of this typeis not distributed across PF aswell. In part this position is motivated by parsimony: two modularlydistinct syntactic systems should not be posited unless absolutely necessary. Until it hasbeen conclusively demonstrated that a syntaxlikemovement system is requiredat PF, we will assumethe more restrictive option. 2BackgroundAssumptions Inthis section we sketchour background assumptions about the structure of the grammar. We assumea theoryin whichmorphology interprets the output of the syntactic derivation (Distributed 2 Muchof thediscussion of these issues centers aroundsecond-position cliticization effects, especially intheSlavic languages.Franks and Holloway King 2000 contains many relevant references totheliterature. 558 DAVIDEMBICKANDROLFNOYER Morphology;Halle 1990, Halle and Marantz 1993, 1994, Noyer 1997, and related work). The basicelements of the syntactic computation are abstract features, which appear in bundles as terminalnodes. Universal as well as language-specific well-formedness conditions determine in whatmanner these features may combine to form syntacticcategories, the atoms of syntactic representation,which we callhere morphemes. We reservethe