22
1 I am grateful to Elena Anagnostopoulou for valuable comments. This summary profited greatly from a discussion on diagnosing movement with Hamida Dermidache, David Pesetsky and Maria Polinsky at the University of Utrecht in January 2009. In particular section 2.3 incorporates suggestions from David Pesetsky’s handout. All errors, also regarding misrepresentations of the participants’ view points, are my own. GROUP SUMMARY: DIAGNOSING XP-MOVEMENT 1 Winfried Lechner, University of Athens February 2010, v.1.0 The present survey reviews some prominent strategies for detecting phrasal movement by the use of diagnostic tests. Just as in the sciences, applying a diagnostic test requires some preparation. Before conducting the actual experiments on which the diagnostic is based on, it is necessary to answer various questions about a node that is suspected to have moved. What is e.g. the base order of the expressions that this category resides in? Which size is the string that is dislocated (feature, head, phrase, or larger)? Does the node contain descriptive content which can be accessed by tests that evaluate interpretative properties? Which competing analytical options is movement to be pitted against (binding, control, Multiple Dominance, ellipsis among others)? It is these and related questions that will be addressed in the first part of this contribution. The second part catalogues a variety of standard tests that react to properties of dislocation. 1. Movement and diagnosis 1.1. Movement In its most general form, the term syntactic movement denotes a mapping between two distinct linguistic representations r 1 and r 2 , such that (i) r 1 includes an occurrence of a subtree α in position p 1 and (ii) r 2 includes an occurrence of α in position p 2 (where p 1 p 2 ); and (iii) p 1 and p 2 are in a designated graph theoretic relation, usually defined in terms of the first branching node version of c-command. In what follows, ‘position of α’ will be understood as ‘irreflexive mother node of α’. Moreover, ‘landing site of α’ is the ‘landing position of α’, i.e. p 2 in (1)b: (1) a. pre-movement representation r 1 p 1 3 Y M α b. post-movement representation r 2 p 2 3 α ... p 1 3 α In addition, there is a number of criteria that are commonly taken to further restrict the mapping relation M. Even though providing a complete list is difficult, given that theories of displacement are still evolving, a first group of requirements arguably includes the axioms in (2). Thus, each application of movement targets a single node only ((2)a), is subject to bounding and island

Diagnosing XP‐Movement

  • Upload
    uoa

  • View
    7

  • Download
    0

Embed Size (px)

Citation preview

1I am grateful to Elena Anagnostopoulou for valuable comments. This summary profitedgreatly from a discussion on diagnosing movement with Hamida Dermidache, David Pesetsky andMaria Polinsky at the University of Utrecht in January 2009. In particular section 2.3 incorporatessuggestions from David Pesetsky’s handout. All errors, also regarding misrepresentations of theparticipants’ view points, are my own.

GROUP SUMMARY: DIAGNOSING XP-MOVEMENT1

Winfried Lechner, University of AthensFebruary 2010, v.1.0

The present survey reviews some prominent strategies for detecting phrasal movement by theuse of diagnostic tests. Just as in the sciences, applying a diagnostic test requires somepreparation. Before conducting the actual experiments on which the diagnostic is based on, it isnecessary to answer various questions about a node that is suspected to have moved. What is e.g.the base order of the expressions that this category resides in? Which size is the string that isdislocated (feature, head, phrase, or larger)? Does the node contain descriptive content whichcan be accessed by tests that evaluate interpretative properties? Which competing analyticaloptions is movement to be pitted against (binding, control, Multiple Dominance, ellipsis amongothers)? It is these and related questions that will be addressed in the first part of thiscontribution. The second part catalogues a variety of standard tests that react to properties ofdislocation.

1. Movement and diagnosis1.1. MovementIn its most general form, the term syntactic movement denotes a mapping between two distinctlinguistic representations r1 and r2, such that (i) r1 includes an occurrence of a subtree α inposition p1 and (ii) r2 includes an occurrence of α in position p2 (where p1 … p2 ); and (iii) p1 andp2 are in a designated graph theoretic relation, usually defined in terms of the first branchingnode version of c-command. In what follows, ‘position of α’ will be understood as ‘irreflexivemother node of α’. Moreover, ‘landing site of α’ is the ‘landing position of α’, i.e. p2 in (1)b:

(1)

a. pre-movement representation r1

p1 3 YM

α

b. post-movement representation r2 p2 3

α ... p1 3

α

In addition, there is a number of criteria that are commonly taken to further restrict the mappingrelation M. Even though providing a complete list is difficult, given that theories of displacementare still evolving, a first group of requirements arguably includes the axioms in (2). Thus, eachapplication of movement targets a single node only ((2)a), is subject to bounding and island

2For explorations into downward movement see Richards (1999); for sideway movement seeBobaljik and Brown (1997) and Nunes (2001).

3Head raising analysis of relative clauses (Bhatt 2002; Hulsey and Sauerland 2006) andcomparatives (Lechner 2004) represent exceptions to (2)d.

4On covert extraposition see Fox and Nissenbaum (1999). On covert A-movement seediscussion in Polinsky and Potsdam (2007, this volume), and references therein.

5Single output models have been advocated by Bobaljik (1995); Fox and Nissenbaum (1999);Groat and O'Neil (1996) and Pesetsky (2000), among others.

2

conditions ((2)b), proceeds upward2 ((2)c), and is usually taken not to alter the label of thelanding site ((2)d) 3.

(2) Axioms of movementa. Movement is an operation that targets a single node (at a time).b. Movement is bounded by locality conditions.c. Movement proceeds to a position that c-commands the launching site.d. Movement does not change the label of the landing site.

In addition, there are also conditions on the nature of the trees which include α, specifying thatr1 and r2 must not be part of distinct levels of representations (Chomsky 1981). While it is alsopossible to define mappings across different levels (Williams 2003), such operations have beendevised precisely in order to avoid commitment to one or more of the core assumptionsassociated with movement, and are therefore in practice not subsumed under the latter notion.

The extensional manifestations of movement vary widely, depending in particular onspecific choices for four variables: (i) possible dissociations between pronunciation andinterpretation; (ii) model of the grammar; (iii) size of the dislocated category and (iv) ontologicalcommitments. Since the logical space generated by these parameters also defines the alternativeanalytical options that phrasal movement, which forms the central part of the present summary,competes against, each of the four factors introduced above will be briefly addressed in turn.

To begin with, a node that has shifted may be pronounced in its higher, derived or in alower, pre-movement position. The latter results in covert or ‘backward’ movement. Just asregular, audible dislocation, backward movement comes in various different flavors which havebeen argued to include Quantifier Raising (QR), covert extraposition, covert scrambling, andsome less commonly attested manifestations of covert A-movement such as Backward Raisingor backward A-scrambling.4

Second, partially cross-cutting the first determinant, all dislocation may proceed in a singlecycle5, or the operations may be partitioned into an overt and an abstract, covert component. Anode may accordingly undergo silent movement either overtly (overt covert movement) orfollowing pronunciation (LF), resulting in the matrix visualized by the upper four cells of (3)(‘¹’ refers to the content of the left-adjacent cell):

3

(3) Move before Spell-Out Move afterSpell-Out

pronounce higher copy pronounce lower copy

a. Features n/a n/a n/a

b. X° T ? ¹

c. XP wh-movement, topicalization,scrambling, Raising, passive,

HNPS, extraposition,...

QR, covert extraposition, backward scrambling,Backward Raising,,..

¹

d. λ-binders n/a n/a n/a

The missing lower portion of the table is provided by the third factor, which regulates the sizeof the entity that moves. Apart from features, heads and phrases movement has also beenhypothesized to affect categories that lack descriptive content all together ((3)d). In particularthree operations, all of which are semantically motivated, are relevant in this domain: derivedpredicate formation by silent raising of pied-piped relative pronouns (see (4)), movement of PROin order to provide control predicates with suitable property-type arguments ((5); Heim andKratzer 1998), and the creation of intensional expressions by raising of λ-binders ranging oversituations ((6); Percus 2000).

(4) a. a person [about which you might have heard]b. a person [λ1 [about t1 you might have heard]]<e,t>

(5) a. Sam wanted PRO to callb. Sam wanted [λ1, PRO t1 to call]<e,t>

(6) Sam wanted [λ2 [λ1, PRO t1 to call(s2)]]<s, <e,t>>

In all three cases, the analysis starts from the premise that movement is interpreted as λ-abstraction, and is motivated by the idea to strengthen this relation also in the other direction,rendering contexts which are interpreted by λ-abstraction amenable to a movement account.

The size of a moved node α is of particular significance for present purposes, as it influencesthe choice of test. This is so because the size of α covaries with the descriptive content of α, andmost diagnostics for movement react to the position of the descriptive content relative to othernodes in the tree. As a result, while head and phrasal movement can in many cases be detectedby inspecting word order or interpretive properties of the lexical content of α and its parts (fordetails see section 2), movement of features and λ-binders is uninformative in this respect.Furthermore, due to their limited capacity to embed lexical material, heads that move remain inall but very few circumstances invisible to the semantic component. Thus, movement of nodeswithout descriptive content (features and λ-binders) must be diagnosed by different, more finegrained instruments than head movement, which is in turn less easily detectable than phrasalmovement. Consequently, diagnosing movement of this type is contingent on detecting tracesof the movement procedure itself such as island conditions or ordering restrictions on movement(see section 2.3 for detailed discussion).

6The distinction between the literal vs. extensional interpretation of the grammar is borrowedfrom Phillips and Lewis (2009), who provide extensive discussion of that issue.

7Movement competes with ellipsis e.g. in the analysis of Right Node Raising (Sabbagh 2007).

4

Finally, theories of movement may vary across a fourth dimension which is defined byconsiderations concerning the ontological status of the operation. Notably, movement denotesa method for modeling a relation between two nodes, as in (1). Whether this model is taken tobe ontologically grounded or remains a meta language description of the mapping operation in(1), much like colored beans are used to model DNA, depends on one’s theoretical inclinations.On a strict literalist interpretation6, movement is an integral part of the language system,understood as a cognitive module with concrete biological manifestations. For adherents of thisposition, the mapping operation in (1) is actually part of the object language, and can thereforealso be diagnosed on the basis of its functional physiological correlates which can bequantitatively analyzed with the help of brain imaging techniques, Event Related Potentials andothers methods. By contrast, it is also possible to conceive of movement as a mere tactic toprovide an adequate description of a system of data, with no commitment as to how well thelinguistic model reflects underlying biological properties of the language module. Under thisextensionalist view, incorporating extralinguistic, experimental evidence into the theory requiresan explicit hypothesis as to how operations of the linguistic model translate into mental statesand their measurable physiological correlates. These and related considerations need to be takeninto account when evaluating psycholinguistic, patholinguistic and other external evidence formovement (see Dermidache, this volume.)

Ontological commitment (of a different type) also inform the long standing debate betweenderivational vs. representational interpretation of movement. A strictly derivational theory ofmovement incorporates features that prevent a representational re-interpretation (i.e. such thatr1 = r2 in (1)).To defend this position, it accordingly needs to be shown that the grammargenerates opaque representations which are subject to conditions that cannot be replicated at laterstages of the derivation. Since such evidence has so far by and large proved evasive (but seeLechner, this volume), ‘movement’ is in practice often broadly construed as an operation thatprovides a characterization of certain dependancies in derivational as well as representationaltheories. In line with this tradition, such an inclusive, liberal definition of the term will beadopted here, which contrasts movement with binding, ellipsis7, control, or treatments of thedislocation property in monostratal categorial models of the grammar (‘wrapping’).

1.2. Diagnosing movementIn the prototypical cases, output representations in which a category has been dislocated (r1

above) do not match their input (r1). Movement can therefore be interpreted as a distortion, anaberration which adds new properties to - or removes them from - linguistic representations.Using medical terminology, one might also conceive of movement as a condition affecting treeswhich reveals itself on the basis of various associated symptoms. These symptoms are causally

8For details see (11) in section 2; for non-linguistic markers see Dermidache (this volume).

9Alternations that have resisted a derivational analysis include the double object construction(give John a book /a book to John) and attributive modifiers (visible stars vs. stars visible).

5

related to movement and can therefore be used in the procedure of diagnosing dislocation.

In order to expose more transparently the evolution and internal composition of diagnostics

it is instructive to consider more closely the nature of symptoms first. The changes introducedin the transition from r1 to r2 systematically correlate with the emergence of specific propertiesin designated location of the raw data which, adopting the nomenclature of bioscience, will bereferred to as markers. It is by virtue of such markers that symptoms express themselves. For theconcrete study of syntactic movement, three different types of linguistic markers are of particularrelevance, each associated with a distinct effect of movement.8 The first two classes emerge asthe result of changes in form or interpretation and are (modulo reconstruction) visible in theoutput representation r2. The third group of markers tracks conditions which regulate themapping from r1 to r2, recording the sensitivity of movement to islands, Subjacency, thecycle/extension condition, the Phase Impenetrability Condition (Chomsky 2000), orderpreservation constraints (Superiority), the prohibition on improper movement and others (seesection 2.3.5 for details).

Yet, not all markers are symptoms because the correlation between a marker and anassociated condition might either be accidental or caused by other, unrelated conditions. Forinstance, while word order alternation is often the product of movement, there are notableexceptions to this generalization that do not lend themselves to a derivational analysis.9 Thus,changes in linearization do in isolation not provide a reliable symptom of movement. Evidently,demonstrating that a marker actually expresses a symptom requires to establish a causal linkbetween the condition (movement) and the marker. This is attained by the application of(diagnostic) tests under experimental conditions. Concretely, a test involves two components:(i) systematic modification of the raw data which controls for spurious correlations betweenmarker and condition (the experiment), and (ii) the formulation of predictions as to whichchanges in the data this modification will trigger. If the changes reflected in the marker matchthese predictions, the marker can be employed as a symptom in a diagnostic test.

Conditions such as movement generally express themselves in more than just a singlesymptom. Furthermore, many symptoms can be due to a variety of causes. In practice, obtaininga successful diagnosis therefore necessitates the deployment of more than one test in order to beable to eliminate all but one cause for the convergence of symptoms under consideration. Sucha conspiracy of multiple symptoms is for instance responsible for the classic paradigm in (7),

which documents the interaction between object fronting and overt participle morphology in

French (Kayne 1989).

10For arguments that all agreement takes place in specifier-head configurations see Koopman(2006), among others.

11The information that features are eliminated on the higher copy must be transferred to thelower copy. An elegant solution to this long standing conundrum consists in adopting a MultipleDominance theory of movement (Bachrach and Kazir 2007; Gračanin-Yuksek 2008).

6

(7) Participle agreement in French as a symptom of movementa. J’ai construit/*construites les maisons

I have constructed the housesfem

b. les maisons1 que nouns avons construites t1

the housesfem which we have constructedc. Ils les1 ont construites t1

they themfem have constructed‘They constructed them’

d. Elles1 ont été construites t1

theyfem have been constructede. Quelles maisons1 as tu construites t1

whichfem houses have you constructed?

(7) involves two symptoms which in collaboration result in the diagnosis of successive cyclicmovement. Concretely, inversion in (7)b-e offers a first symptom of displacement. Overt

participle morphology represents another marker, which signals abstract agreement.

Furthermore, on a prominent definition, agreement is a reflex of (that is, caused by) feature

valuation in a local specifier head relation. For the relevant examples of (7), such a constellation

can only be obtained by movement. Since agreement is induced by feature valuation, which is

in turn contingent upon movement, it follows by transitivity that movement causes agreement.10

Agreement therefore constitutes a second symptom of movement.11

That agreement and dislocation have a common source is also reflected in the standard

theoretical account for participle agreement in (7), sketched in (9), on which fronted objects

temporarily enter a local specifier-head relationship with the participle in an intermediate

projection (Kayne 1989):

(8) [object [Φ-features] ... [XP tobject, [Φ-features] [X’ participle[Φ-features] [... tobject ...

The finding that agreement indicates movement finally also provides a diagnostic that can be

used in the analysis of new data. For instance, the test confirms the hypothesis that the subject

in (9) originates in a VP-internal position and passes through XP, furnishing independent support

for the unaccusativity hypothesis.

(9) Elles sont venues (with audible reflexes in some varieties; Belletti 2005 syncom)theyfem are come‘They have arrived’

As was seen above, diagnosis is a complex procedure that initially starts from specific areas

in the raw data (markers), proceeds to a theoretical interpretation of these markers, yielding

12Elena Anagnostopoulou (p.c.) points out that the absence of agreement with the higherparticiple is probably due to the fact that the relative pronoun reaches the relevant specifier by Ā-, andnot by A-movement.

7

symptoms, and collects a sufficient number of symptoms in order to reach at a diagnosis.

Moreover, tracing the actual cause of a symptom proceeds by process of elimination, which

exposes alternative explanations to falsification with the help of suitable tests. Tests thereby act

as filters for probable causes, linking causes to their symptoms. In reality, it is hard to devise a

simple deterministic diagnostic algorithm for this procedure, though. Even the brief discussion

above revealed that each filter is itself an artifact that needs to be designed, and design is subject

to choices. Among others, this renders tests vulnerable to what has been called observationselection effects. The size of the smallest fish that can be sampled from a pond e.g. depends on

the choice of net, to use an example from Bostrom (2002). Tests, qua being filters, therefore

reflect the nature of ancillary assumptions which enter into the analysis and determine, among

others, which symptoms or properties are actually held to constitute a valid test. To illustrate,

the agreement symptom introduced above is not fine-grained enough to detect all instances of

movement through XP. This can be inferred from the observation that gender agreement is only

triggered on the lowest participle (Chomsky 1995: 325):12

(10) la jupe1 qu'il a [XP t1 dit/*dite que Pierre a [XP t1 faite/*fait t1]]the skirtfem which he has said that Peter has made

Depending on the particular choice of assumptions, one and the same set of data can therefore

be assigned radically different epistemological values. As such interdependencies between tests

and theoretical assumptions are by no means isolated cases, many components of diagnosis are

open to interfering factors. It follows that conditions very rarely reveal themselves when

approached by mechanical, step-by-step diagnostic procedures.

2. Tests for syntactic movementIn the previous section, symptoms were grouped into three classes, depending on the type ofmarker they are associated with. This classification finds a corresponding tripartite taxonomyfor the standard syntactic tests which react to movement, as shown in (11)a-c. (11)d and (11)eadd two new groups of quantitative psycholinguistic diagnostics that look for symptoms ofmovement in child language, aphasia and related areas, as well as in physiological correlates:

(11) a. Tests which correlate with changes in form (phonology or morphology)b. Tests which correlate with changes in interpretationc. Tests which reflect properties of the movement procedured. Tests from ontogenetic language development e. Tests which react to mental and/or physiological correlates of movement

Below, some of the main exponents of the first three classes will be briefly introduced (on (11)dand (11)e, see Dermidache, this volume). As the internal mechanics of the tests is moreimportant than their extensional manifestations, I will not attempt to provide an exhaustive

8

catalogue of diagnostics, though, which can be found in introductory texts and handbooks.

2.1. Movement and formFirst, displacement is known to induce changes in form which can either affect (i) the displaceditem α itself or (ii) nodes that α has entered a local relationship with. The former type ofmodification is most immediately observable in changes in precedence relations, while classicmanifestations of the latter property are found with movement into intermediate positions andinclude French participle morphology (see (7)), quantifier float in West Ulster English(McCloskey 2000), shown in (12), wanna contraction, Irish complementizer agreement(McCloskey 2002) and many others. The view that movement proceeds successive cyclically isfurther confirmed by doubling phenomena in child language ((13); Crain and Diane Lillo-Martin1999):

(12) What1 did he say [CP all t1 (that) he wanted t1](13) What1 do you think [CP what1 Cookie Monster eats t1] (age 5;0)

Finding any of the markers above signals the presence of a symptom of movement.

2.2. Movement and interpretationSecond, as movement effects changes in c-command, it implies consequences for c-commandsensitive principles of interpretation such as quantifier scope, variable binding, coreference,polarity licensing, idiom interpretation and ellipsis resolution. Given this rationale, local scopeambiguities that do not correlate with word order changes as in (14) can be modeled by optionalsilent movement (QR) of the object across the subject:

(14) A critic attacked every movie. › ™ œ/ œ ™ ›

More specifically, once a node α has crossed to the left of node β, there are three attestedinteractions between α and β, schematically depicted in (15) (the symbol ‘™’ subsumes allsemantic relations sensitive to c-command). Detecting any of these three combinations in thedata indicates that α has reached its overt position by movement. (The target output is well-formed, except for (15)c, where unacceptable strings also provide evidence for movement).

(15) a. /α ™ β and /β ™ αb. /α ™ β and *β ™ αc. *α ™ β and /β ™ α

2.2.1. Optional reconstructionTo begin with, configuration (15)a obtains if movement generates new meanings but at the sametime preserves previously established interpretations, resulting in ambiguity. This option isexemplified by inversion in scope rigid scrambling languages such as Hungarian, Korean orGerman, in which quantifier scope ambiguity is contingent upon overt reversal of the scopebearing items (see (16)).

13As with movement, lowering can be achieved by different strategies, i.e. in semantics, insyntax or by mapping representations to more than one interpretations (e.g. by the Scope Principle;Aoun and Li 1993; Frey 1989).

9

(16) a. weil irgendeiner jedes Buch zu Hause hat › ™ œ / * œ ™ ›“since somebody read every book at home”

b. weil [irgendein Buch]1 jeder t1 zu Hause hat › ™ œ / œ ™ ›“since everybody read some book at home”

It follows that the emergence of ambiguity in non-canonical word orders furnishes a test thatsignals movement (or, to be precise, the presence of one of its symptoms in scope rigidlanguages).

While the surface interpretation of (16)b (left-hand side of (15)a) can be directly read offovert syntax, the inverse reading (right-hand side of (15)a) is produced by reconstructing thesubject into a position below the object.13 More generally, reconstruction refers to a set ofoperations that repair level ordering conflicts which arise as the result of movement maskinginformation of previous stages of the derivation (see discussion of opacity and (27) in section2.2.3). Hence, finding traces of reconstruction in the data represents an important criterion forconfirming the diagnosis of movement. Conversely, pervasive absence of reconstruction effectsis commonly held to constitute an argument for base generation. A textbook illustration of howthis divide discriminates between two related constructions comes from control and raising. OnlyDPs that have undergone A-movement behave as if they are part of the lower clause w.r.t. thecomputation of scope, as exemplified by the well-known contrast in (17):

(17) a. Some unicorn is likely to kill every knight. › ™ œ / œ ™ › b. *Some unicorn is willing to kill every knight. › ™ œ / *œ ™ ›

Other tests for movement have been based on markers which react to anaphor licensing (see(18)), selectional properties ((19)), idiom interpretation ((20)) and referential opacity ((21)). Thelatter pair is informative inasmuch as (21)a can be used as a notional de dicto report conveyingthe information that John misidentified his new neighbor, whereas (21)b exclusively expressesa false de re proposition. Since de dicto readings are contingent upon reconstruction, thisindicates that control subjects can not be generated in the complement clause, and that, as aconsequence, control must not implicate movement:

(18) a. [Friends of each other1]2 seemed t2 to amuse the men1 (exx. from Bailyn 2009)b. *[Friends of each other1]2 promised/wanted PRO2 to amuse the men1

(19) a. There2 seemed t2 to be a commotion b. *There promised PRO to be a commotion

(20) a. All hell2 is likely t2 to break looseb. *All hell is eager PRO to break loose

(21) a. A Martian seems to John to have moved in next door. /de dicto//de reb. A Martian threaten John to move in next door. *de dicto//de re

10

Polinsky and Potsdam (2007, this volume) furthermore demonstrate that the interpretivecontrasts between raising and control also obtain if it is the lower, and not the higher, positionof the dependency which is spelled out. In the North West Caucasian language Adyghe, for one,inverted scope readings are attested with Backward Raising, i.e. contexts structurally identicalto (17)a, except that the notional subject is construed in-situ. Similarly, Backward Controldisplays the properties familiar from ‘forward’ control. This corroborates the view that theinterpretive characteristics of the construction (raising vs. control) covaries with the semanticclass of the superordinate predicate, and is not contingent upon superficial properties ofpronunciation. In addition, Polinsky and Potsdam’s findings strengthen the case againsteliminating the distinction between raising and control, as suggested by Hornstein (1999).

Another important set of diagnostics for derivational syntactic relations that has not beenaddressed so far is supplied by ellipsis. Concretely, there are two ways in which ellipsis can beused as a sign of movement. On the one hand, movement provides a strategy for escaping thephonological consequences of phonological deletion. For instance, in the structures generatedby pseudogapping in (22), at least one category has been evacuated from VP prior to theapplication of ellipsis:

(22) a. John sent flowers to Lucy before Max did chocolates1 to Sue2

- = [VP sent t1 t2]b. John sent flowers to Lucy before Max did chocolates1

- = [VP sent t1 to Lucy]c. John sent flowers to Lucy before Max did to Sue2

- = [VP sent flowers t2]

(23) illustrates that this type of movement process reconstructs, thus falls into group (15)a.

(23) *While some believed him1 everything, others did only the story that John1 had met aliens.(Sauerland 1998: 114; (78))

On the other hand, movement also feeds ellipsis resolution, as most prominently seen in theinteraction between (covert) QR and VP-ellipsis in antecedent contained deletion ((24)a). Thecontrast between (24)a and (24)b also documents that whatever process is responsible fortransporting the object into its scope position obviates disjoint reference effects. This propertyis well-known from overtly fronted relative clauses such as (24)c:

(24) a. You showed him3 every book [CP that Sam3 wanted you to -]- = [VP show him3]

b. *You showed him3 every book [CP that Sam3 wanted you to show him3]c. Which book [CP that Sam3 wanted you to show him3] did you show him3

Thus, not all types of movement reconstruct - and those that don’t instantiate manifestations ofanti-reconstruction effects, schematized by the combinatorial option (15)b.

14To be precise, the analysis is not compatible with a strictly surface oriented, monostratal,model of the grammar, but can be recast in representational terms.

11

2.2.2. Anti-reconstruction While anti-reconstruction obfuscates the relation between the overt position and its trace, it alsosupplies a test for unmasking derivational dependencies which is qualitatively different from thediagnostics introduced so far. More precisely, on an influential strand of analyses, both (24)a and(24)c lend themselves to a uniform treatment in terms of an sequential ordering of operationswhich is only compatible with a movement analysis.14 On this account, documented by (25) for(24)c, the object is not completely assembled in-situ, but moves first ((25)a) and is counter-cyclically combined with the CP embedding the name only afterwards by an operation knownas Late Merge ((25)b). This has the consequence that at the point where the name is added((25)b), it no longer resides within the c-command domain of the pronoun.

(25) a. Step I: Move head of relative clauseWhich book1 did you show him3 t1

b. Step II: Late Merge relative clauseWhich [book1 [CP that Sam3 wanted me to -]] did you show him3 t1

Thus, a further symptom of movement can be isolated from its ability to feed the application ofLate Merge.

Note in passing that there are also exponents of class (15)b which are, at least at first sight,not particularly suitable for the purpose of detecting movement, as their analysis does not dependon Late Merge or related operations. Raising with negative quantifiers, shown in (26)a, typifiessuch context in which reconstruction fails for different reasons. (26)b paraphrases the readingwhich is unavailable for (26)a (Partee 1971; Lasnik 1972):

(26) a. Nobody is certain to pass the test. ¬› ™ certain / *(¬)certain ™(¬)›b. It is certain that nobody will pass the test

The contrast in (26) will be taken up again later (see (49)).

2.2.3. Obligatory reconstruction The critical property of the third and last possible combination (15)c is that in selectedenvironments, a moved node is necessarily interpreted in a position below the one this nodesurfaces in. Empirically, such constellations are exemplified by obligatory reconstruction and,to some extent, crossover effects. The former type comes in two flavors, both of which attest tothe fact that there are operations which render accessible aspects of pre-movementrepresentations to the output, resulting in syntactic opacity.

Opacity itself subdivides into two different classes. In some contexts, a positive principlesuch as anaphor licensing in (27) imposes necessary conditions on an expression α which cannotbe satisfied in the surface position of α (counter-bleeding opacity). On the other hand, there are‘negative’ statements like Principle C, which specify the sufficient conditions that an acceptable

15Obligatory reconstruction can also be enforced by certain types of NPI licensing (see (i) and(ii)) and selectional properties such as idiom interpretation; on criticism of arguments based on thelatter see Runner (1997). (i) [A doctor who knew anything about acupuncture]1 was *(not) t1 available.

Uribe-Etxebarria (1994)(ii) a. ?Pictures of anyone aren't likely to be at the exhibit Lee (1993)

b. *Anyone isn't likely to be at the exhibit

16For discussion see Ruys (2000). Movement reconstructs only for the purposes of variablebinding, but not for question semantics and other interpretive aspects. Note also in this context thatcrossover can also be seen as an instance of the cases that fall into group 2.3 (‘properties of themovement procedure’).

12

string needs to satisfy (counter-feeding opacity). (28) fails to do so:

(27) [Which pictures of each other1]2 do you think that they1 like best t2

(28) *[Which pictures of John1]2 do you think that he1 likes best t2

Both flavors of opacity signal that interpretive properties of the source position of α matter.15

Moreover, since obligatory reconstruction reveals itself in solid intuitions of (un)acceptabilityinstead of notoriously difficult ambiguity judgements characteristic of optional reconstruction,the strategy represents a particularly useful tool for tracking dislocation.

Turning to crossover, the weak and strong contrasts in (29) and (30), respectively,demonstrate that pronominal variable binding is subject to the syntactic condition dictating thatA-traces of variable binders must c-command the pronouns they bind (Postal 1971; Wasow1972; see Ruys 2000 for recent discussion). Thus, it looks as if options for pronominal variablebinding are calculated on the basis of pre-wh-movement configurations. In addition, (30) inducesa disjoint reference effect subsequent to reconstruction.16

(29) a. Who1 t1 likes his1 motherb. ?*Who1 does his1 mother like t1

(30) a. Who1 t1 thinks she likes him1b. *Who1 does he1 think she likes t1

The crossover condition has also been recruited for the purpose of diagnosing movement. Forinstance, Chierchia (1992) argued that crossover - and therefore movement - underlies thedistribution of functional and pair list readings, as shown by (31) and (32).

(31) What1 did everyone1 buy t2(1) (single answer/pair-list or functional answer)i. Possible answer (single answer): Everyone bought applesii. Possible answer (pair-list/functional): Ann bought apples, Ben bought beer, ...

(32) a. Who1 t2(1) bought everything1? (single answer/*pair-list or functional answer)i. Possible answer (single answer): Ann bought everythingii. Impossible answer (pair-list/functional): Ann bought apples, Ben bought beer,...

b. Who everything1 t1(2) bought t1?

13

In (31), the lowest occurrence of the universal subject c-commands the object trace and maytherefore bind an implicit variable inside that position (represented as a bracketed suffix to t2),legitimizing the pair-list answer (31)ii. By contrast, object QR in (32)a yields the crossoverconfiguration (32)b, in which the trace of everything fails to c-command the internally complexwh-trace. This explains the absence of a pair-list reading for (32)a.

In sum, both obligatory and optional reconstruction effects are strong indicators ofmovement. This holds at least if movement is understood as an operation relating twooccurrences of a category that is under experimental conditions (i.e. in the right contexts)interpreted by other means than plain variable binding. These observations do not decide,though, between derivational and representational accounts of movement, as the analysis ofopacity employs devices such as copies that can also be given a representational reinterpretation.For factors that might aid in deciding between these options see Lechner (this volume).

2.3. Properties of the movement procedureThe third large group of linguistic symptoms for XP-movement does not reflect changes in formor meaning, but is caused by properties of the movement operation itself. Tests for this type ofsymptoms exploit the existence of constellations which have been concluded to involvemovement on independent grounds, and which have then been shown to be subject to restrictionsthat can neither be attributed to aspects of form or interpretation alone. For instance, the basiccontrast in (33) indicates that the object of has moved.

(33) a. You asked me, [which pictures of each other1]2 they1 like best t2b. *You asked me, [which pictures of each other1]2 I1 like best t2

Once it is known that the wh-phrase has reached its overt position by movement, it is natural toattribute the ill-formedness of (34) to features of the movement operation itself:

(34) *[Which pictures of each other]2 did [the claim that they like best t2] surprise you.

In this way, sensitivity to syntactic islands, a property which itself cannot be exclusively tracedback to changes in form or meaning, becomes a symptom of (overt) displacement.

More specifically, it is possible to distinguish among at least three larger groups ofdiagnostics for the movement operation itself: dominating islands, intervention effects, andeconomy. Fine-tuning the parameters of (a fragment of) locality theory in turn generatesadditional criteria for movement that partially cross-cut among these three groups. Superioritycan e.g. be related both to intervention effects or economy. Still, starting with a tripartitetaxonomy turns out to be convenient for expository reasons.

2.3.1. Dominating islandsThe first group comprises of tests which respond to purely configurational properties of trees andcharacteristically involve a particular combination of nodes dominating the launching site of acategory to be dislocated. Syntactic objects that fail to abide by these strong islands, exemplified

17Two notes. First, this presupposes that Θ-properties - i.e. the A vs. Ā- distinction - can bemapped to featural properties. Second, it is not known at the moment whether all syntacticintervention effects can be reduced to economy. Moreover, at least some intervention effects such asInner Islands should be given a semantic analysis.

14

by the Subject Condition ((34)), the Adjunct Condition, the Complex NP Constraint and theCoordinate Structure Constraint ((35)), display symptoms of sharp unacceptability:

(35) a. *I wonder which picture1 Sally [was abroad [while you bought t1]]b. *I wonder which picture1 you met [the painter [who sold t1]]c. I wonder which picture1 [[you bought t1] and [I will have to pay the bill *(for t1)]]

Closely related to the analysis of islands is the question of how legitimate instances ofapparently unbounded movement as in (36) are made compatible with the dictate of localitytheory in general, and the Phase Impenetrability Condition (PIC) in particular (Chomsky 2000):

(36) Which picture did you [vP t1 say [CP t1 that Mary [vP t1 bought t1]]]

Under the prevalent view, the PIC requires that movement proceeds successive cyclically insmall incremental steps that leads through intermediate landing sites currently thought to includeCP and vP. Since the PIC implies that all but a few movements create intermediate landing sites,finding traces of these positions consequently adduces strong evidence for the claim that a treewas assembled with the help of movement. For some tests which have been designed to recordsuch intermediate steps see discussion in section 1.2 ((7)) and section 2.1 (quantifier float in

Ulster English (12)); for evidence for successive cyclicity from interpretation see Fox (1999) andLebeaux (1990), among many others.

2.3.2. Intervention effectsExpressed in terms of an attract model, the relevant substructure of strong islands in (35)includes a single attracting feature (e.g. a wh-feature on C°) and a single node bearing amatching feature (e.g. a wh-phrase), such that movement proceeds to a fixed position. Addinga third feature into the tree generates a second, more complex class of phenomena whichfundamentally differ from strong islands in that they reflect properties of admissible featurecombinations in addition to purely structural determinants. In their most general form, theseintervention effects arise as the result of movement across intervening nodes with similar featurespecification, possibly in the presence of other attractors (i.e. α — β — tα , where α and β bear‘similar’ features). Since it is possible to assign different values to the notions ‘intervening node’(c-command vs. domination), ‘similar feature’ (identity vs. subset relation) and to the numberof attractors, the matrix projected by this class is internally complex. The discussion to followwill briefly address some aspects of c-commanding configurations and proceed from there toeconomy, finally turning to dominating interveners.

Assume initially that intervention is restricted to c-commanding nodes and ‘similarity’ isunderstood as identity in X’-theoretical status (X° vs. XP) and Θ-properties (A- vs. Ā-position).17

18The discussion follows Pesetsky (this volume). For defective intervention see also Hiraiwa(2000). For a feature geometric version of RM see Starke (2001).

15

Suppose moreover that only a single attracting head is present. Such combinationsyield configurations that fall under classic Relativized Minimality (RM; Rizzi 1990) illustratedby (37).

(37) a. *How1, [wh] did he wonder [why[wh] she behaved t1] wh-island)b. *Someone1, [A-property] seems that [ it[A-property] [is likely t1 to win]] (Superraising)c. Who meet1 will t1 you (Head Movement Constraint)

Sensitivity to RM effects has been widely used as a gauge for the analysis of filler - gaprelations.

In the contexts regulated by RM, the intervener (why in (37)a and it in (37)b) is ‘defective’in that it is not eligible for further movement (Chomsky 2000). More generally, defectiveinterveners block other dependencies, but do not themselves partake in chain formation. Theclassic illustration of this poorly understood phenomenon comes from the interaction betweendative experiencers and raising in Italian in (38):

(38) Maria sembra (*a Gianni) essere felice Mary seemed to John to be happy

As a group of markers that has only recently been explored, incomplete intervention provide animportant new class of tests for movement which have for instance been used to gain a newperspective on the perennial puzzle of tough constructions and their status w.r.t. movement.18

Hartmann (2009) noticed that the well-formedness of (39) inversely correlates with the presenceof a matrix experiencer. This strongly suggests that the surface subject has indeed reached itsovert position by movement.

(39) Cholesterol1 is important (*to Mary) [PRO to avoid t1]

2.3.3. EconomyApart from dominating islands and intervention effects, movement has also been hypothesizedto be controlled by a variety of economy constraints. In fact, already phenomena captured byclassic RM lend themselves to a re-interpretation in terms of an economy metric which favorsshorter over longer movement paths (henceforth ‘Shortest’; Chomsky 2000; Richards 1997). For(37), this constraint blocks e.g. movement across the closest possible target (embedded SpecCPand SpecTP in (37)a and (37)b, respectively), or, alternatively, attraction across a closer goal(why and it in (37)).

Recall at this point that the interveners in classic RM contexts are defective, which entailsthat such objects are frozen in place. As a result, there are no alternative orders of movement forexamples such as (37)a. Thus, none of the derivations results in a well-formed output, leadingto what has been called ineffability. More generally, RM constellations do not produce a

16

‘winning strategy’ and therefore mask the inherently comparative nature of Shortest. This aspectof economy more transparently manifests itself in Superiority phenomena, documented by (40).

(40) Superioritya. Who1 did John persuade t1 [PRO to visit whom2]b. *Who2 did John persuade whom1 [PRO to visit t2]

Superiority by definition involves more than one potentially mobile (‘active’) wh-phrase, theclosest of which is selected by Shortest, and therefore also yields a grammatical output. If asecond node is attracted by the same feature (or another feature in the same position), Shortestcreates the expectation that multiple movement leads to the formation of crossing, orderpreserving paths (α — β — tα — tα; see Richards 2001). This prognosis is confirmed by theobservation that order preservation effects are cross-linguistically attested in a wide range ofconstructions, among them multiple overt wh-movement in Bulgarian ((41)), Scandinavianobject shift, multiple scrambling, clitic placement, and others (see Abels 2007 for recentdiscussion).

(41) Multiple movement: crossing patha. Koj1 kogo2 vižda t1 t2 (Rudin 1988:449)

who whom sees‘Who sees whom’

b. *Kogo2 koj1 vižda t1 t2 whom who sees

I will return to the issue of order preservation shortly.It has also been suggested that in some contexts, economy is calculated relative to a fixed

interpretation. This concept is commonly referred to as interface economy.

(42) Interface economya. Who remembers [what2 who1 t1 bought t2] (Baker 1970)b. Who remembers [who1 t1 bought what2]

To illustrate, the surface Superiority violation (42)a escapes the verdict of Shortest on thereading in which the embedded subject (who1) is construed with matrix scope. This is so becausethe target interpretation cannot be generated by alternative means. Concretely, the requirementthat interrogative CPs contain an overt wh-expression (in English) renders the matrixinterpretation of who1 unavailable for the alternative serialization (42)b. Thus, interface economysuspends the Superiority effect for (42)a. Similar effects of Scope Economy with covertmovement are discussed in Fox (2000).

In general, finding reflexes of (interface) economy in the data is indicative of movement,given that comparable binding relations are generally not controlled by economy principles suchas Shortest (for exceptions see e.g. (44) below).

19See Pesetsky (1982), Williams (1998), among many others.

17

2.3.4. C-command and containment, nesting and crossingSo far, the discussion of intervention effects was restricted to contexts in which the intervenerc-commands the launching site of movement. If the definition of ‘intervener’ is loosened so asto include dominating as well as c-commanding positions, though, the economy based analysisof Superiority can also be extended to nesting configurations as in (43), which are on a par withthe standard Superiority violations in (40) (Takano 1994; Kitahara 1995; Müller 1998):

(43) Dominating intervenera. [Which book of who2]1 did you buy t1?b. *Who2 did you buy [which book of t2]1?

(43)b and (40)b are both excluded by a general economy principle which bans movement of anwh-expression across another (c-commanding or dominating) node bearing the same featurespecification. Furthermore, (43) parallels (40) in that multiple movement leads in both cases tothe creation of dependencies which preserve the pre-movement order of the expressions.19 Thus,attraction by a single head correlates with order preservation effects and the formation ofcrossing paths.

It is interesting to note at this point that movement appears to differ in this respect frombinding, which generally results in nesting, instead of crossing dependencies. This restrictionexpresses itself e.g. in the distribution of bound variables inside elliptical VPs (Dahl’s Puzzle;Dahl 1973; Fiengo and May 1994; Fox 2000):

(44) Binding: ‘nested’ dependenciesMax1 said he saw his mother and Oscar2 did, too.a. Max1 said he saw his mother and Oscar2 said that he2 saw his1 mother b. *Max1 said he saw his mother and Oscar2 said that he1 saw his2 mother

(44) can be assigned the ‘nested’ reading (44)a, in which the higher pronoun is locally identified(2 — 2 — 1), but lacks the crossing dependency interpretation (44)b (2 — 1 — 2). It therefore istempting to employ this distinction between crossing vs. nested path formation as a criterial testfor movement vs. binding. On this view, movement elicits symptoms of crossing dependencieswhich can in turn be identified by order preservation effects.

Although this diagnostic method can be applied to a limited set of contexts (see Lechner,this volume), it cannot be generally used to expose the full range of properties associated withmovement. This is so as it is well-known that the characterization above, on which multiplemovements lead to the creation of crossing dependencies is not entirely pervasive (Pesetsky1982). In all cases of order preservation discussed so far, movement was induced by attractingfeatures in a single position. However, if a second attractor is introduced into a different part ofthe derivation, as shown by the classic examples (45) - (47), the results attest to the emergenceof nested, order reversing dependencies.

18

(45) Moving intervener (subject): nested patha. ?What1 did you wonder [who2 t2 saw t1] (only weak island violation)b. *Who2 did you wonder [what1 t2 saw t1]

(46) Moving intervener (indirect object): nested patha. ?What subject1 do you know [who2 PRO to talk to t2 about t1]b. *Who2 do you know [what subject1 PRO to talk to t2 about t1]

(47) tough movement: nested patha. ?Which violin1 are these sonatas2 hard [PRO to play t2 on t1]b. *Which sonatas2 is this violin1 hard [PRO to play t2 on t1]

As noted by Richards (2001), this is expected on an account that combines economy with thecycle, because Shortest ensures that the order in which (active) nodes are attracted mirrors theorder in which they are first merged.

Finally, it becomes now also possible to use the systematicity noted above in designing adiagnostic tool which exposes defining characteristics of movement. Unlike before, the test doesnot simply react to order preservation, but also takes into account the number of attractors, suchthat each relative output order is now symptomatic of movement, but with a differentderivational history. Crucially, for this method to succeed, one must be able to find independentevidence in the data for the number and position of attracting heads involved.

2.3.5. Improper Movement (C-command and containment II)The previous discussion suggested that movement is not only susceptible to feature interventionby c-command, but is also potentially trackable in contexts where one goal contains the other.Generalizing the notion of intervention from c-command to containment thereby considerablyexpands the range of possible contexts which can be screened for movement. Similarlypromising results of extrapolating from (principles defined in terms of) c-command to (principlesdefined in terms of) containment have also been reported in the study of improper movement(IM). Following some general comments on orthodox manifestations of IM, these findings willform the closing part of this survey.

It was seen in section 2.3.1 that movement can also be inferred from the presence ofintermediate landing sites and properties of these positions. Such intermediate landing sites alsoreveal themselves indirectly, by way of interaction with other movement procedures. Arguablythe most prominent interaction of this type falls under the reign of the ban on IM (Chomsky1973), which prohibits Ā-movement from feeding A-movement ((48)).

(48) a. *Who1 is believed [CP t1 that it was told t1 [that he will win the race]]z----------mz-------m

A-movement Ā-movement

b. *Who1 seems that [CP t1 it is likely [t1 to win the race]] (Saito 2002)

Crucially, the IM condition relies on the assumption that apparently non-local relations are theresult of multiple applications of local movement steps. Constellations that display sensitivityto IM therefore alert to the possibility that they have been generated by movement. Deploying

20Successive cyclic movement can be captured by the assumption that each movement stepgenerates a secernate chain to be evaluated by the generalized IM condition. See Abels (2007) fordiscussion.

19

this diagnostic, it turns out that one plausible candidate for IM comes from the prohibition onscope reconstruction with negative quantifiers (26)a, repeated below.

(49) a. Noone is certain to pass the test. ¬ ™ certain/ *certain ™ ¬b. noone is certain [NegP t1 Neg° [noone[Neg] to pass the test]]

z----------mz-----m A-movement Ā-movement

If negative QPs bear features that - in analogy to wh-features - need to be licensed in a localrelation with a negative head, as detailed by (49)b, the unattested narrow scope reading of (26)acan be blocked by the requirement that raising must not follow Ā-movement. In this way, theapplication of a diagnostic gained from inspection of the IM phenomenon suggests a plausibleanalysis for an unrelated, previously unaccounted for conditions on interpretation.

The IM restriction applies to contexts in which a single node has been subjected to morethan a single movement operation. There is interesting evidence suggesting that this conditioncan be further generalized and extended into two directions (Abels 2007). First, prohibitionssimilar to IM have been identified for constellations where the two movement processes targettwo distinct categories in a containment relation. Müller (1998) and Takano (1994) observe, forone, that subextraction of α out of a larger node β which itself moves yields well-formed resultsonly if α and β are not targeted by the same movement type. It is for this reason that scramblingout of a node that is later topicalized, as in (50)a is allowed, while scrambling out of a scramblednode, as in (50)b, is not:

(50) a. [tα zu lesen]β hat keiner das Buchα versuchtto read has nobody the book tried

b. *...daß [tα zu lesen]β keiner das Buchα versucht hat that to read nobody the book tried has“Nobody has tried to read the book”

A generalization in a second direction expands the typology of movement operations whichfall under IM to include scrambling, topicalization and other processes not covered by theoriginal prohibition (Takano 2000: 144; fn 5). Abels (2007) suggests that a more completeevaluation metric has to distinguish among at least four different types of movement, given in(51):

(51) A-movement ™ Scrambling ™ wh-movement ™ topicalization

The ordering of these values in (51) moreover expresses the hypothesis that movements whichare lower on the scale precede movements that are located higher (see also van Riemsdijk andWilliams 1981; Williams 2003).20 Thus, A-movement may feed scrambling or wh-movement,but not vice versa. For empirical justification of (51) and further discussion see Abels (2007).

20

The picture emerging from the discussion above looks as follows: First, movementproperties that are sensitive to intervention effects reveal themselves in contexts of c-commandas well as containment. The core cases of both are captured by economy (Reinhart 2006).Second, intervention effects themselves react to the composition of the features involved (Starke2001). Third, it is possible to envision a theory that expresses restrictions on possiblecombinations of movement operations such as IM (and generalizations thereof) in terms offeatures and properties of attractors. On this view, locality falls into two groups: configurationalconditions (strong islands) and conditions on intervention such as (52), which subsume classicRM, intervention by containment, conditions on subextraction and standard instances of IM.

(52) A node can be attracted by F1 across F2 as long as F2 is lower than F1 on the hierarchy ofmovement operations in (51).

3. ConclusionThere is a clear correlation between the theoretical advances in a field on the one side, whichlead to a better understanding of causal relations between phenomena and the primitives of thetheory, and the precision and ease with which it is possible to arrive at a diagnosis for certainunderlying causes of these phenomena. At earlier stages of the development in a field, it is oftenpossible to encounter generalizations that organize data by mere surface appearance, groupingtogether what at a later stage turn out to be only epiphenomenally related properties. Medicalsciences usually reserves the term syndrome for these lesser understood combinations ofsymptoms, which cannot be traced to a single (set of) cause(s) yet, to distinguish them - withvariable degrees of terminological consistency - from diseases, illnesses and conditions.Generalizing these notions, progress in a field can then be said to be marked by advancing fromsyndromes to conditions. It seems as if the concept of syntactic XP-movement is proceeding inthis direction. The set of properties associated with phrasal displacement are arguably betterunderstood at present than ever before. At the same time, it was seen above that linguists haveby now at their disposition a comprehensive catalogue of diagnostic tools which can be deployedas tests for particular symptoms of phrasal movement.

Yet, many interesting and important issues still remain unresolved. Empirically, the debatewhether tough movement, the various guises of resumptive strategies and Backward Raising and,for some, control involve movement or not is e.g. far from settled. And theoretical questionswhich have so far evaded a satisfactory answer abound. What are for instance the principlesdetermining bounding nodes, phases and illicit combinations thereof (strong islands)? To whichextent is movement motivated by semantic considerations? Can all (weak) islands be given asemantic reanalysis? Which conditions regulate the distribution of phonologically overt copies,and ensure feature transfer after checking or valuation (see fn. 11)? While recent incarnationsof Multiple Dominance theories offer new perspectives on at least the last two puzzles fororthodox conceptions of movement (Bachrach and Kazir 2007), solid evidence that supports onealternative over the other has in this - as in so many other domains - not materialized so far.

21

ReferencesAbels, Klaus. 2007. Towards a Restrictive Theory of (Remnant) Movement. Ms., Universitetet i Tromso,

and University College LondonAoun, Joseph, and Li, Yen-hui Audrey. 1993. Syntax of Scope: Linguistic Inquiry Monograph 21.

Cambridge, Massachusetts: MIT Press.Bachrach, Asaf and Roni Katzir. 2007. Right-Node Raising and Delayed Spellout. Ms., MIT. [Available

at LingBuzz: http://ling.auf.net/lingBuzz/000314]Bhatt, Rajesh. 2002. The raising analysis of relative clauses: evidence from adjectival modification.

Natural Language Semantics 10: 43-90.Bobaljik, Jonathan. 1995. Morphosyntax: the syntax of verbal inflection. Doctoral dissertation, MIT.Bobaljik, Jonathan David, and Samuel Brown. 1997. Interarboreal Operations: Head Movement and the

Extension Requirement. Linguistic Inquiry 28.2: 345-356.Bostrom, Nick. 2002. Anthropic Bias: observation selection effects in science and philosophy: Studies

in Philosophy - Outstanding Dissertations. New York and London: Routledge.Chierchia, Gennaro. 1992. Questions with Quantifiers. Natural Language Semantics 1.2: 181-234.Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht, The Netherlands: Foris

Publications.Chomsky, Noam. 1995. The Minimalist Program: Current Studies in Linguistics. Cambridge,

Massachusetts: MIT Press.Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by step: Essay on minimalist

syntax in honor of Howard Lasnik, ed. by Roger Martin, David Michaels, and Juan Uriagereka,89–155. Cambridge, Mass.: MIT Press

Crain, Stephen and Diana Lillo-Martin. 1999. An Introduction to Linguistic Theory and LanguageAcquisition. Blackwell Publishers: Oxford

Dahl, Osten. 1973. On so-called Sloppy Identity. Synthese 26: 81-112.Dermidache, Hamida. this volume. Arguments for LD movement in LD questions in child language.Fiengo, Robert. 1977. On Trace Theory. Linguistic Inquiry 8: 35-61.Fiengo, Robert, and May, Robert. 1994. Indices and Identity. Cambridge, Massachusetts: MIT Press.Fox, Danny. 1999. Reconstruction, Variable Binding and the Interpretation of Chains. Linguistic Inquiry

30.1: 157-196.Fox, Danny. 2000. Economy and Semantic Interpretation. Cambridge: M.I.T. Press.Fox, Danny, and Nissenbaum, John. 1999. Extraposition and Scope: A Case for overt QR. Proceedings

of WCCFL 18.Frey, Werner. 1989. Syntaktische Bedingungen für die Interpretation. Doctoral Dissertation, University

of Stuttgart. Groat, Erich, and John O'Neil. 1996. Spell-Out at the LF Interface. In: Werner Abraham, Samuel David

Epstein, Höskuldur Thráinsson and Jan-Wouter Zwart (eds.), Minimal Ideas, 113-139. Amsterdam:John Benjamins.

Heim, Irene, and Kratzer, Angelika. 1998. Semantics in Generative Grammar. Oxford: Blackwell.Hiraiwa, Ken. 2000. Multiple Agree and the Defective Intervention Constraint in Japanese. MIT Working

Papers in Linguistics 40, 67-80.Hulsey, Sarah, and Sauerland, Uli. 2006. Sorting out Relative Clauses: A Reply to Bhatt. Natural

Langauge Semantics 14.1: 111-137.Kayne, Richard. 1989. Facets of Romance Past Participle Agreement. In Beninca, Paola (ed.) Dialect

Variation and the Theory of Grammar, pp. 85-104. Dordrecht, Holland: Foris Publications.Kitahara, Hisatsugu. 1995. Target alpha: Deducing Strict Cyclicity from Derivational Economy.

Linguistic Inquiry 26.1: 47-78.Koopman, Hilda. 2006. Agreement: in defense of the “Spec head configuration”. in Agreement Systems,

Cedric Boeckx (ed.), Amsterdam: John Benjamins. Lasnik, Howard. 1972. Analyses of Negation in English. Ph.D. dissertation, Department of Linguistics

and Philosophy, MIT. Lasnik, Howard, and Mamoru Saito. 1992. Move α. Cambridge, Massachusetts: MIT Press.Lebeaux, David. 1990. Relative clauses, licensing and the nature of the derivation. In Proceedings of

NELS 20, 318–332. GLSA, University of Massachusetts, Amherst.Lechner, Winfried. 2004. Ellipsis in Comparatives. Berlin, New York: Mouton de Gruyter.Lechner, Winfried. this volume. Diagnosing covert movement with the Duke of York and reconstruction.

22

McCloskey, Jim. 2000. Quantifier Float and Wh-Movement in an Irish English. Linguistic Inquiry 31:57–84.

McCloskey, Jim. 2001. The Morphosyntax of WH-extraction in Irish. Journal of Linguistics 37: 67–100.Nunes, Jairo. 2001. Sideward Movement. Linguistic Inquiry 32.2: 303-344.Partee, Barbara. 1971. On the requirement that transformations preserve meaning. In: Studies in linguistic

semantics, ed. Charles J. Fillmore and Terence Langendoen, 1-21. New York: Holt, Rinehart andWinston.

Percus, Orin. 2000. Constraints on Some Other Variables in Syntax. Natural Language Semantics 8.3:173-231.

Percus, Orin, and Sauerland, Uli. 2003. Pronoun Movement in Dream Reports. Ms., Univerity ofTübingen.

Pesetsky, David. 1982. Paths and Categories. Doctoral Dissertation.Pesetsky, David. 2000. Phrasal Movement and its Kin. Cambridge: MIT Press.Polinsky, Maria and Eric Potsdam. 2007. Expanding the scope of control and raising. In W. Davies & S.

Dubinsky (eds.) New horizons in the analysis of control and raising. Springer.Polinsky, Maria and Eric Potsdam. this volume. Diagnosing covert A-movement.Postal, Paul. 1971. Cross-Over Phenomena. New York, New York: Holt, Rinehart and Winston.Phillips Colin and Shevaun Lewis. 2009. Derivational Order in Syntax: Evidence and Architectural

Consequences. Ms., University of Maryland. Richards, Norvin W. III. 1997. What Moves Where When in Which Language? Doctoral Dissertation,

Cambridge, Massachusetts.Richards, Norvin. 1999. Dependency Formation and Directionality of Tree Construction. In MIT

Working Papers in Linguistics #34: Papers on Morphology and Syntax, Cycle Two. Vivian Lin,Cornelia Krause, Benjamin Bruening, and Karlos Arregi (eds.)

Richards, Norvin. 2001. Movement in Language. Oxford: Oxford University Press.Riemsdijk, Henk van, and Williams, Edwin. 1981. NP-structure. The Linguistic Review 1: 171-217.Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Massachusetts: MIT Press.Runner, Jeff. 1997. BT Exempt Anaphors: an Argument from Idiom Interpretation. Proceedings from

WCCFL 16.Ruys, Eddie. 2000. Weak Crossover as a Scope Phenomenon. Linguistic Inquiry, 31.3: 513–539.Sauerland, Uli. 1998. The Meaning of Chains. Doctoral Dissertation, MIT.Starke, Michal. 2001. Move dissolves into Merge: a Theory of Locality. Doctoral Dissertation, University

of Geneva.Takano, Yuji. 1994. Unbound traces and indeterminacy of derivation. In Current topics in English and

Japanese, ed. Masaru Nakamura, 229–253. Tokyo: Hituzi Syobo.Takano, Yuji. 2000. Illicit Remnant Movement: An Argument for Feature-Driven Movement. Linguistic

Inquiry 31.1: 141–156Uribe-Echevarria, Myriam. 1994. Interface Licensing Conditions on Negative Polarity Items: A Theory

of Polarity and Tense Interactions. Doctoral Dissertation, University of Connecticut-Storrs.Wasow, Thomas. 1972. Anaphoric Relations in English. Doctoral Dissertation, Cambridge,

Massachusetts.Williams, Edwin. 1998. Economy as Shape Conservation. In Keyser, Jay and Ken Hale (eds.), Chomsky’s

70th Birthday Celebration Web Page.Williams, Edwin. 2003. Representation Theory. Cambride, Mass: MIT Press.Gračanin-Yuksek, Martina. 2007. About Sharing. Doctoral dissertation, Massachusetts Institute of

Technology.