1

Introduction

This dissertation deals with one of the central concerns of syntactic theory since Ross 1967: How local is syntax and what are the measures of syntactic locality? These questions lie at the core, for example, of both the theory of movement and the theory of binding. A constant theme has been the issue of whether or not movement and anaphoric relations are governed by a unified concept of locality. I argue here that they are. On an empirical level, I bring together and present the results of three (partly) independent strands of research on German: (i) the study of reduced infinitive constructions, starting with Bechs (1955/57) classic work on COHERENCE versus NON-COHERENCE in German, (ii) the POSSESSOR DATIVE CONSTRUCTION (PDC), with a dative nominal playing the role of both possessor and affectee, and (iii) BINDING, the conditions under which reflexive and non-reflexive pronouns may occur. One of the principal results of the dissertation is that there is a single determinant of locality for these three apparently disparate phenomena. The initial goal is to show how possessor datives and binding bear on the clause structure of various types of infinitival structures, primarily CONTROL and A CCUSATIVUS
CUM

I NFINITIVO constructions. Both the PDC and binding are

phenomena involving relations between two structural positions, and both are

sensitive to the size (in the sense of internal complexity) of intervening material. Exploiting this sensitivity, I propose that possessor datives and binding can serve as probes for infinitival clause size. Building on Wurmbrand 2001, I provide new evidence for a distinction between infinitival complements that is more fine-grained than, and thus goes beyond, Bechs original and traditionally assumed binary distinction between coherence and non-coherence. These results provide the foundation for addressing two larger issues: the typology of clause size and what makes a given phrase-type transparent (constructed coherently) or opaque (constructed non-coherently) to grammatical interactions with elements beyond its boundaries. The category of clauses that are satzwertig (literally of sentence-value) is uniquely important to an understanding of language, since that is the category through which we accomplish core speech acts like the assertion of what is true. It is the category which, since Aristotle, has been taken to express the core semantic relation of predication holding between a subject and a predicate. The two issues above are closely related since one of the properties of sentences is that they are, in a sense, complete in and of themselves. As a consequence, elements trapped within them do not in general enter into grammatical interactions (such as agreement or case-marking) with elements beyond sentence-boundaries. The type of infinitive construction which is my primary focus involves infinitival complements known as Accusativus cum Infinitivo (henceforth AcI), selected by

causative and perception verbs. AcI-complements are a hybrid category in that they have (some of) the properties of sentencehood, but they are also porous, in that they permit (some) grammatical interaction across their borders. By examining this hybrid case, I aim to shed light on the larger question of what it takes for a phrasetype to qualify as satzwertig. Since the opacity property of phrase-types that are satzwertig is the intuition behind Chomskys (2000, 2001) notion of PHASE , and since there are other Minimalist assumptions I adopt, this dissertation provides arguments in favor of the theoretical shift from the traditional Principles & Parameters view, which defined much work in theoretical syntax in the 80s and 90s, to a Minimalist system. I implement my analyses in a framework broadly compatible with Chomskys (1995, 2000, 2001) work, but with at least one crucial difference which is discussed in detail in chapter 3. I ultimately propose the phase as the single determinant of locality which unifies all three of the syntactic phenomena examined here. In a sense, the phase a subpart of the derivation of a sentence, which, once completed, is closed off to grammatical interactions with elements introduced later in the derivation marks a return to the core intuition behind Chomskys (1973) SPECIFIED SUBJECT CONDITION (SSC). Just as the SSC constrained both movement configurations and coreference possibilities, in the system I develop, phases impose the crucial locality restriction on both PDraising and binding. However, while the SSC, i.e. the opacity-inducing presence of

an intervening subject, plays an important role in my accounts of the PDC and binding, I will show that appealing to phasehood covers more ground in that it captures the opacity effects triggered not only by verbal and nominal (vP and DP) domains but also by clausal and prepositional (CP and PP) domains. The dissertation is divided into three major parts, each covering one of the three strands of research I present. Part I (chapter 2) offers an overview of the literature on coherence, paying particular attention to the traditional coherence tests and to Wurmbrands (2001) new take on the binary distinction between coherence and noncoherence. Her restructuring typology provides a more fine-grained set of

distinctions between types of infinitival complementation. Part II is concerned with the possessor dative construction (PDC). Following Landau (1999), I argue for an A-movement (POSSESSOR RAISING) account. Chapter 3 develops the account for German and shows how the German PDC differs from the Hebrew PDC. Chapter 4 presents possessor raising as a new coherence diagnostic. I argue that, in important ways, possessor raising is a more reliable probe for the presence of a phase-defining domain than other coherence tests, in particular SCRAMBLING and LONG PASSIVE. Part III introduces the notion phase. In chapter 5, I present a phase-based analysis of reflexive and pronominal binding, focusing primarily on the context of AcIconstructions. I show that the relevant binding domain for reflexives and
4

pronominals alike is the minimal phase which contains them, and that the apparent long-distance binding ability of the reflexive can be explained by covert reflexive raising to the phase-edge. Finally, chapter 6 extends the new phase-based

understanding of locality to possessor raising and coherence phenomena. I tentatively propose a definition of phase which promises to account for both the well-known opacity effects induced by finite embedded clauses (CPs) and the constraints on movement and binding across sub-clausal categories (vP, DP, and PP) discussed in Parts II and III of the dissertation. The following chapters then take the reader beyond coherence in several respects. First, I provide further evidence for Wurmbrands proposal of a fine-grained distinction between types of complements selected by control verbs the binary coherence/non-coherence distinction is not enough. Second, I show that AcIcomplements, though clearly sub-clausal in a purely structural sense, have properties of Satzwertigkeit that are incompatible with the label coherent construction. Finally, the characterization of phase I arrive at in chapter 6 potentially accounts not only for coherence and non-coherence phenomena in Germanic infinitive constructions but also for transparency and opacity effects cross-linguistically, including contexts other than the verbal domain.

PART I REDUCED INFINITIVE CONSTRUCTIONS

Coherence and Restructuring

One of the three major strands of research this dissertation brings together is the study of reduced infinitive constructions. This chapter reviews the literature on

C OHERENCE and RESTRUCTURING with the goal of conveying what it means for an infinitive construction to be reduced. The traditional picture of coherence versus non-coherence as a binary distinction will be contrasted with Wurmbrands (2001) more fine-grained typology of infinitival clause size.

2.1

(Non)finite complementation and transparency

2.1.1 The phenomenon Consider the English examples in (1). The perception verb saw takes a complement which consists of the pronominal them and the nonfinite predicate leaving town. The cleft-construction in (1b) confirms that the accusative-marked pronominal forms a constituent with the embedded predicate. (1) a. Mary saw them leaving town. b. What Mary saw was [them leaving town]. Given that the embedded nonfinite domain consists of a subject and a predicate them is the subject of leaving town, just as Mary is the subject of saw we expect the nonfinite domain to exhibit clausal behavior (or Satzwertigkeit, as one would say in

German, literally the property of having sentence value). The fact that the embedded subject bears accusative, as opposed to nominative case, however, indicates that it is the matrix verb which case-licenses this nominal, and thus, that case-checking must span the nonfinite clause boundary. As shown in (2), where the embedded subject bears nominative case, case-checking does not span the clauseboundary in the case of a finite complement clause. (2) a. Mary saw that they were leaving town. b. What Mary saw was [that they were leaving town]. The examples in (3) provide even more striking evidence of the contrast in transparency between nonfinite and finite domains. (3) a. They were seen [ __ leaving town]. b. *They were seen [that __ were leaving town]. c. Theyi saw [each otheri negotiating]. d. *Theyi saw [that each otheri were negotiating]. The embedded subject in (3a) can cross the nonfinite domain boundary and become the subject of the matrix clause, bearing nominative case. Movement of the embedded subject into subject position of the (higher) matrix clause is impossible, however, in the case of the finite complement clause in (3b). Similarly, examples (cd) illustrate that the matrix subject can bind the anaphoric element each other across the nonfinite domain boundary in (c), but not across the finite boundary in (d).
8

In fact, this difference in transparency of embedded complement clauses goes beyond the finite-nonfinite distinction. As illustrated by the German data in (4)-(6), there are transparency contrasts even within the class of constructions involving nonfinite complementation. (4) a. Er lie [den Hund laufen].
he let the dog (ACC) run

He let the dog run. b. Der Hund wurde [ __ laufen] gelassen.

the dog (NOM) was (PASS) run let

The dog was allowed to run.

(5)

a. Er hat versucht [den Hund einzufangen].

he has tried the dog (ACC) in-to-catch

He tried to catch the dog. b. Der Hund wurde versucht [ __ einzufangen].

in-to-catch

the dog (NOM) was (PASS) tried

They (impersonal) tried to catch the dog. (6) a. Er hat behauptet [den Hund zu vermissen].
he has claimed the dog (ACC) to miss

He claimed to miss the dog. b. *Der Hund wurde behauptet [ __ zu vermissen].

the dog (NOM) was claimed to miss

They (impersonal) claimed to miss the dog.

All three constructions in (4)-(6) have a matrix verb which takes an infinitival complement. (4) is the German equivalent of the English perception verb
CUM INFINITIVO

construction in (1) and is known as ACCUSATIVUS

(AcI).1 As in

English, the subject of the nonfinite complement is accusative-marked in its base position (a), but can become the nominative-marked subject of the matrix clause (b). Unlike the infinitive in (4), the embedded verbs in (5) and (6), are accompanied by the infinitive marker zu to (in Bechs (1955/57) terms, an infinitive without zu is of the 1st status, and an infinitive with zu is of the 2nd status). Notice also that the accusative-marked nominal in (5) and (6) is the direct object, not the subject of the nonfinite verb. Both versuchen try in (5) and behaupten claim in (6) are traditionally analyzed as SUBJECT CONTROL verbs: the matrix subject controls a null pronominal element (PRO) in the subject position of the infinitival complement. The constructions differ, however, in that the complement of versuchen is transparent, and the complement of behaupten is opaque to long passive movement of the embedded direct object into the subject position of the matrix clause. Since Gunnar Bechs celebrated work on the syntax of German infinitive constructions, the contrast in transparency evident in (5) and (6) has been referred to as KOHRENZ (COHERENCE) versus INKOHRENZ (NON-COHERENCE). Matrix verbs which allow grammatical operations to cross the domain boundary of their infinitival

AcI-constructions will be discussed in detail in later sections of this chapter and in chapter 5. 10

complements are coherence (transparency)-inducing, while matrix verbs which do not are non-coherence (opacity)-inducing.

2.1.2

Accounting for the phenomenon

Since Bech (1955/57) identified the existence of these two classes of infinitive constructions, various attempts have been made to account for the contrast. The phenomenon is a cross-linguistic one, as the coherence/non-coherence distinction holds not only for Germanic languages like German and Dutch but also for many other languages, among them Spanish, Italian, and Japanese (see e.g. Aissen & Perlmutter 1976, 1983, Rizzi 1978, and Miyagawa 1987). While the term

coherence is used primarily for Germanic, other terms for the same phenomenon, in particular C LAUSE UNION and RESTRUCTURING, are used more broadly. The term clause union originated in the Relational Grammar (RG) literature and referred to a process by which all dependents of the embedded verb become dependents of the matrix verb. This process was first recognized in connection with causative

constructions in Romance and was extended to certain infinitival constructions in Spanish by Aissen & Perlmutter (1976, 1983). Their main claim was that the wellknown clitic-climbing phenomenon (where a clitic which is a semantic argument of an embedded verb cliticizes to a higher verb) is a consequence of clause union. Certain trigger verbs optionally cause matrix and embedded domain to merge into a single clause. The universal character of clause union phenomena is confirmed by the

11

fact that, in all languages which have grammatical operations that appear to cross clause-boundaries, there are two classes of verbs, those which can, and those which cannot trigger transparency. While the members of these two classes vary from language to language and even from speaker to speaker, it has been claimed (see e.g. Wurmbrand 2001) that some verbs, like
TRY

(e.g. German versuchen and Spanish

CLAIM/AFFIRM

tratar), are universally transparency-inducing, and others, like

(e.g.

German behaupten and Spanish afirmar), are universally opacity-inducing. Early accounts of the clause union/coherence phenomenon cast in a generative transformational framework were proposed by Aissen (1974) for Turkish and French and by Evers (1975) for German and Dutch. In these works, a clause union/coherent structure is argued to be derived by raising of the embedded verb into the matrix clause, where the two verbs form a unit. This raising process then causes pruning of the leftover embedded structure, so that nominal complements originating in the embedded clause become part of the matrix VP. The term restructuring was originally used by Rizzi (1978), who argued that a restricted class of matrix verbs in Italian govern a rule, the so-called restructuring rule, which optionally transforms an underlying bi-clausal structure into a mono-clausal one and thus creates a unique verbal complex consisting of both matrix and embedded verb.

12

Besides VERB R AISING (Aissen 1974, Evers 1975)2, Rizzis restructuring rule, and other derivational approaches,3 which have in common that a bi-clausal structure is transformed, in one way or another, into a mono-clausal one, there are also nonderivational approaches. Characteristic of the latter type of approaches is that they do not assume that every infinitival complement starts out as a full clause (CP). Haider (1993), for example, argues that infinitival complements in German are transparent to local grammatical operations when matrix and embedded verb are base-generated as a verbal complex with a complex projection base (V Vinf Vmatrix) and blended argument structures. According to Wurmbrand (2001), certain matrix verbs and their complements form a coherent unit because the complements of these matrix verbs (which Wurmbrand calls restructuring predicates) are introduced into the derivation as subclausal, that is they are phrases which lack a tense/agreement and nominative case position and thus must at least be smaller than TP and CP (see also Moore (1991) and Chung (2003) who make similar proposals for Spanish and Chamorro, respectively). Wurmbrands typology of infinitival clause size goes beyond the binary distinction between coherence-inducing and non-coherence-inducing predicates. Since it is a major goal of this dissertation to provide evidence for the existence of (at least part of) Wurmbrands typology, all of section 2.3 is devoted to
2 3

See also Rosengren 1992. Among the derivational approaches proposed for German are processes known as

REANALYSIS (Haegeman & van Riemsdijk 1986, von Stechow & Sternefeld 1988), EVACUATION (Fanselow 1989), and ABSTRACT INCORPORATION (Grewendorf & Sabel 1994, based on Baker 1988). For a cross-linguistic RG account, see Gibson & Raposo 1986. 13

the details of her proposal. Before bringing this proposal into the picture, however, I will present the two major verb classes which have been established for German by the traditional coherence/non-coherence distinction.

2.2

Coherence, non-coherence, and a third construction

2.2.1 Two major verb classes Based on the work of Bech (1955/57), Hhle (1978), den Besten & Edmondson (1983), Haegeman & van Riemsdijk (1986), von Stechow & Sternefeld (1988), and den Besten & Rutten (1989) to name at least some among the many contributors to this research program control predicates which take an infinitival z u (to)complement can roughly be divided into two major classes. One is the class of obligatorily non-coherent control-zu-predicates (Class 14), and the other is the class of optionally coherent control-zu-predicates (Class 25). An infinitival complement which can be extraposed6 (as illustrated in (7a)) is traditionally considered to have CP-status and is thus taken as a sure sign of non-coherence. Since extraposition is an option for all control-zu-predicates, however, it cannot be used to determine whether a given predicate always induces non-coherence or also has the option of entering into a coherent construction with its complement. Rather, the way to identify
4 5 6

Typical examples of Class 1 are bedauern regret and behaupten claim. Typical examples of Class 2 are versprechen promise and versuchen try. Although I use the traditional term extraposition here, I do not necessarily assume that the

position of a postverbal infinitival complement is derived. 14

members of Class 1 versus Class 2 is to test for transparency/opacity effects when the complement is intraposed (as illustrated in (7b)). (7) a. dass wir
that we

bedauert
regretted

haben [INF den Lehrer zu mgen].

have the teacher to like

behauptet
claimed

versprochen
promised

versucht
tried

b. dass wir [INF den Lehrer zu mgen]

that we the teacher to like

bedauert
regretted

haben.
have

behauptet
claimed

versprochen
promised

versucht
tried

2.2.2 Coherence tests It is generally assumed that the absence or presence of an embedded CP is what determines mono- versus bi-clausality and thus whether a construction is coherent or non-coherent. Among the well-established coherence tests, designed to probe for mono- versus bi-clausality, are SCRAMBLING (8), P RONOUN F RONTING (9), the

15

impossibility of CLAUSAL PIED-PIPING7 (see (10)-(11)), COHESION (or the scope of negation) (see (12)-(13)), and VERB-CLUSTER (VC)-FRONTING (see (14)). By section 2.3, where I discuss Wurmbrands (2001) view of coherence, it will have been established that the binary coherence/non-coherence distinction based on monoversus bi-clausality is not sufficiently fine-grained to account for the range of transparency/opacity effects apparent in control verb constructions, but for the purposes of introducing the classic coherence tests listed above, non-coherence can be equated with bi-clausality and thus the presence of an embedded CP. In the remainder of this section, I illustrate how the various tests apply to the control-zuverbs bedauern regret and vergessen forget8 (see also Grewendorf 1987 for an overview of these tests). Scrambling, roughly defined as a local type of argument dislocation, is a common coherence diagnostic because it is generally assumed to be clause-bounded and thus restricted to applying within a CP-domain (see e.g. Grewendorf & Sternefeld 1990). In (8b) and (d), the embedded direct object has undergone scrambling into the matrix clause. (8) a. dass jeder [den Schlssel verloren zu haben] bedauern wrde.
that everyone the key lost to have regret would

that everyone would regret to have lost the key.

7 8

The traditional term is relative clause pied-piping. I use clausal instead. In (12), where I illustrate the cohesion test, I simplify the example by using versuchen try

instead of vergessen forget. 16

b. *dass den Schlssel jeder [ __ verloren zu haben] bedauern wrde. c. dass niemand [das Zimmer abzuschlieen] vergisst.
that nobody the room off-to-lock forgets

that nobody forgets to lock the room. d. dass das Zimmer niemand [ __ abzuschlieen] vergisst. In the context of bedauern, scrambling of the embedded direct object into the presubject position of the matrix clause is not allowed, whereas it is allowed in the context of vergessen.9 The opacity apparent in (b) is taken as evidence of biclausality and the transparency apparent in (d), as evidence of mono-clausality. This means that bedauern is a member of Class 1, i.e. obligatorily constructs noncoherently, while vergessen is a member of Class 2, i.e. has the option of constructing coherently. Like the scrambling test, pronoun fronting shows that the complement of bedauern is opaque, while that of vergessen is transparent to movement of an embedded argument into the matrix clause. Here, however, it is not a full DP but a pronoun which undergoes the movement.

Note that this pre-subject position is only available with certain types of subjects, more

specifically quantificational pronominals like niemand nobody, keiner not anyone, jemand somebody, jeder each, alle everyone, einer some person, viele many. Other types of subjects (common nouns or proper names), whether definite or indefinite, generally do not allow full DPs to precede them in an embedded clause. 17

(9)

a. dass der Student [ihn

that the student

verloren zu haben] bedauert.

to have regretted

it (MASC) lost

that the student regretted to have lost it. b. * dass ihn der Student [ __ verloren zu haben] bedauert. c. dass der Hausmeister [es
that

abzuschlieen] vergisst.
forgets

the superintendent it (NEUT) off-to-lock

that the superintendent forgets to lock it. d dass es der Hausmeister [ __ abzuschlieen] vergisst. The traditional analysis of data like these is that the pronominal direct object of the infinitive (ihn in (a-b) and es in (c-d)) can only occupy the high position between the complementizer dass and the subject, the so-called WACKERNAGEL position, when the matrix verb is of Class 2 and has entered into a coherent construction with its complement, forming a single clausal domain. If the matrix verb is of Class 1 and the infinitival complement constitutes its own clause, including a CP-projection and thus a closer potential landing site for the pronominal, movement of the pronominal into the matrix clause is blocked. In contrast to scrambling and pronoun fronting, clausal pied-piping, shown in (10), is expected to lead to a grammatical result precisely when the infinitival complement does have CP-status. In this construction, a nominal in the infinitival complement is relativized, and the relative pronoun pied-pipes the whole infinitival complement into
18

Spec CP of the relative clause. As in the case of extraposition, this is only expected to be possible when the infinitival complement is a moveable constituent, namely a CP. Since, on the traditional view, all control-zu-verbs may at least optionally take a CP-complement, this diagnostic cannot be used to distinguish between our test verbs bedauern (a member of Class 1) and vergessen (a member of Class 2) (see (10a-b)). What it can be used for is the distinction between control verbs and RAISING verbs like scheinen seem, which is a member of neither Class 1 nor Class 2. The clausal pied-piping test illustrates that scheinen (see (10c-d)), unlike bedauern/vergessen (see (10a-b)), is incompatible with a CP-complement. A simple extraposition test yields the same results: while the complement of bedauern/vergessen can be extraposed and must therefore have CP-status, the complement of scheinen cannot be extraposed and is therefore taken to be smaller than CP. Examples (a) and (c) show the infinitival complement in its intraposed base position, and (b) and (d) show it in the pied-piped position in Spec CP of the relative clause. (10) a. Der Artikel [REL deni Tim [ ti nicht gelesen zu haben] bedauerte/verga]
the article which Tim not read to have regretted/forgot

handelt von Politik.

deals of politics

The article which Tim regretted/forgot not having read is about politics. b. Der Artikel [REL [den nicht gelesen zu haben]i Tim ti bedauerte/verga] handelt von Politik.

19

c. Der Artikel [REL deni Tim [ti nicht gelesen zu haben] schien]
the article which Tim not read to have seemed

handelt von Politik.

deals of politics

The article which Tim seems to not have read deals with politics. d. *Der Artikel [REL [den nicht gelesen zu haben]i Tim ti schien] handelt von Politik. Unlike control verbs, raising verbs like scheinen then do not have the option of constructing non-coherently (i.e. taking a CP-complement) and thus never allow CP pied-piping. As discussed in later sections, raising verbs are traditionally classified as obligatorily coherent (members of Class 3), and Wurmbrand classifies them as functional restructuring predicates. The following data confirm the validity of clausal pied-piping as coherence test by showing its interaction with pronoun fronting. The prediction is that, once pronoun fronting has occurred, i.e. when a pronominal argument of the infinitive shows up as part of the matrix clause, indicating that the construction is coherent, clausal piedpiping is impossible. The data in (11) illustrate that this prediction is borne out. Examples (a-b) show that the matrix verb vergessen allows clausal pied-piping when pronoun fronting has not happened; (c) shows the result of pronoun fronting; and (d) shows the expected interaction between the two diagnostics: application of one excludes application of the other (see also Grewendorf 1987, 1988).

20

(11)

a. Der Student [REL demi der Professor [es ti zu erklren] vergessen hat]
the student whom the professor it to explain forgotten has

ist verwirrt.
is confused

The student to whom the professor forgot to explain it is confused. b. Der Student [REL [dem es zu erklren]i der Professor ti vergessen hat] ist verwirrt. c. Der Student [REL demi esj der Professor [tj ti zu erklren] vergessen hat] ist verwirrt. d. *Der Student [REL [dem tj zu erklren]i esj der Professor ti vergessen hat] ist verwirrt. In (11d), where it is clear that vergessen induces coherence otherwise the pronoun could not have fronted clausal pied-piping is not an option. The infinitival complement is sub-clausal, i.e. smaller than CP, and therefore cannot be pied-piped into the specifier of the relative clause. The data in (12) and (13) illustrate the cohesion (scope of negation) test. In these examples, the intraposed infinitival complement includes a negative quantificational pronominal, niemanden nobody, which is traditionally analyzed as a cohesive combination of two operators, a negation (corresponding to the lexeme nicht not) and a quantifier (corresponding to the lexeme jemanden somebody). The matrix verb in (12) (here I use versuchen try instead of vergessen to illustrate the test) is optionally coherence-inducing.
21

(12)

dass der Student [niemanden zu mgen] versucht.

that the student nobody to like tries

a. that the student does not try to like anybody. (negation takes wide scope) b. that the student tries not to like anybody. (negation takes narrow scope) Here, the two operators of the cohesive combination can be distributed (in a nontechnical sense) over the matrix and the embedded domain, so that the negation can take (wide) scope over the matrix verb versuchen. This interpretation (see gloss in (12a)) is synonymous with the unambiguous German paraphrase dass der Student nicht versucht, jemanden zu mgen (there is no one x, such that the student tries to like x). The fact that the negation in (12) can take wide scope indicates that the infinitival complement may be smaller than CP, and this in turn is an indication of coherence. As expected with optionally coherence-inducing (Class 2) verbs, an interpretation of the negative quantifier only having scope over the embedded domain is also available. The unambiguous paraphrase of this interpretation (see gloss in (12b)) is dass der Student versucht, niemanden zu mgen (the student tried that there be no one x, such that the student likes x). The narrow scope reading of the negation is evidence of the presence of a CP-boundary and thus non-coherence. When the matrix verb is an obligatorily non-coherence-inducing (Class 1) verb like bedauern in (13), on the other hand, only the non-distributive interpretation of the cohesive combination, and thus only the narrow scope reading of the negation is available.

22

(13)

dass der Student [niemanden zu mgen] bedauert.

that the student nobody to like regrets

that the student regrets not liking anybody. (negation must take narrow scope) Here, the only possible paraphrase is dass der Student bedauert, niemanden zu mgen (the student regrets that there is no one x, such that the student likes x). It is clear, then, that the complement of bedauern obligatorily has CP-status, while the complement of versuchen, like that of vergessen (the Class 2 matrix verb I use to illustrate the other coherence tests) may lack a CP-boundary and can therefore be integrated into the matrix clause. The last of the coherence tests discussed here is verb-cluster (VC)-fronting. As shown in (14), this test centers on the possibility of moving all or at least two of the nonfinite verbs of the verbal sequence at the right edge of an infinitive construction to a left-peripheral position. VC-fronting is another phenomenon that has been argued to be possible only in a coherent construction. The sequence of adjacent verbs is said to form a frontable verbal cluster (traditionally analyzed as a complex head, see e.g. Haegeman & van Riemsdijk 1986) when the matrix verb is a member of Class 2, like vergessen, not when it is a member of Class 1, like bedauern. Examples (14a) and (c) show the verbal elements in their assumed base position before fronting. In (14b) and (d), (part of) the verb sequence has fronted and is followed by the finite auxiliary (here hat has) in C.

23

(14)

a. weil

der Student noch nie [das Zimmer abzuschlieen]

yet never the room off-to-lock

because the student

bedauert/vergessen hat.
regretted/forgotten has

because the student has never regretted/forgotten to lock the room. b. Abzuschlieen *bedauert/3vergessen hat der Student das Zimmer noch nie. c. weil der Hausmeister noch nie [das Zimmer abschlieen zu lassen]
never the room off-lock to have

because the superintendent yet

bedauert/vergessen hat.
regretted/forgotten has

because the superintendent has never regretted/forgotten to have the room locked before. d. Abschlieen zu lassen hat der Hausmeister das Zimmer noch nie *bedauert/3vergessen. The traditional analysis of the VC-fronting phenomenon is that only a mono-clausal structure allows the sequence of nonfinite verbs to form a verb-cluster, (part of) which can be extracted and fronted. The verb sequence (or a subsequence thereof) in (14a) and (c), for example, is argued not to be frontable as a unit when the matrix verb is separated from the infinitive by a CP-boundary (probably because CP forms an impenetrable barrier to head movement). Thus, the ungrammaticality of verb sequence fronting in the context of bedauern is taken to indicate that a CP-boundary is present. Note, however, that a fronted verb sequence in this context is only

24

ungrammatical if the infinitive is transitive and thus leaves behind something when fronted (compare (14c-d) with (15a-b)). (15) a. Zu schlafen bedauert/vergessen hat der Student noch nie.
to sleep regretted/forgotten has the student yet never

Regretting/forgetting to sleep is something the student has never done. b. Rechtzeitig schlafen zu gehen hat der Student noch nie bedauert/vergessen.
timely sleep to go has the student yet never regretted/forgotten

Going to sleep on time is something the student has never regretted/ forgotten.

The verb sequences in (15a-b), where the infinitive is the intransitive verb schlafen, can be fronted even in the context of the non-coherence-inducing matrix verb bedauern. This suggests that, in (14), it is the inability of the direct object to scramble out of the clausal infinitival domain which causes the ungrammaticality. The CP-boundary intervening between the non-coherence-inducing verb bedauern and the infinitive it embeds then only interferes with fronting of the respective nonfinite verb sequence when the infinitive leaves behind an argument which must have scrambled out of the infinitival complement before fronting. This means that the phenomenon here must involve fronting of VP, rather than VC. Analyses which claim that the possibility of verb sequence fronting depends on verb-cluster formation and is only possible when the matrix verb is coherence-inducing have no account of the facts in (15). Here, fronting of the constituent which contains the matrix verb and the infinitival complement is possible regardless of whether or not the complement is
25

a CP. Scrambling of an argument of the infinitive into the matrix clause domain before movement of the matrix VP to Spec CP is a case of remnant fronting (Wurmbrand 2001 uses the term remnant topicalization), and remnant fronting is only possible in a coherent infinitive construction, involving a Class 2 matrix verb. In sum, in order for the fronting facts in (14) to serve as a coherence diagnostic, it is crucial that they be analyzed in connection with scrambling (see also von Stechow & Sternefeld 1988), and what fronts in examples like (14b) and (15a) is not a complex verbal head but rather the matrix VP, which contains both the matrix verb and the infinitive.10 We have seen, then, that the classic coherence tests illustrated here establish a binary distinction between types of control-zu-predicates: bedauern regret is representative of Class 1, matrix verbs which obligatorily induce non-coherence, and vergessen forget (as well as versuchen try, which I used to illustrate the cohesion test) is representative of Class 2, matrix verbs which optionally induce coherence. There are reasons to believe, however, that this binary classification incorporates an oversimplified view of the distinctions that are needed. In particular, the following subsection discusses two constructions which suggest that not all Class 2 verbs are equally coherence-inducing. The possibility that there are more than two classes of control-zu-predicates is in line with current models of phrase-structure which posit

10

See Wurmbrand 2004 and Hankamer & Lee-Schoenfeld forthcoming for recent discussions

of the types of verbal constituents that can front. 26

several clausal categories smaller than CP, namely TP, vP, and VP. In order to probe for transparency/opacity effects, induced by matrix verbs taking one of these reduced categories as their complement, the classic coherence tests, which are sensitive to the presence of CP, are not sufficient. Clearly, other diagnostics need to be found.

2.2.3

Subdivisions in the class of coherent verbs

Reis & Sternefeld (2004) point to two constructions which seem to be compatible with only a subclass of verbs in Class 2. One, known as L ONG PASSIVE (first discussed by Hhle (1978)), is passivization of the matrix verb accompanied by long movement of the embedded direct object into the matrix subject position, and the other is a phenomenon den Besten & Rutten (1989) coined THIRD CONSTRUCTION (recently discussed in depth in Wllstein-Leisten 2001). What distinguishes this latter construction from coherent and non-coherent structures is that it appears to be a combination of both, displaying properties of both mono- and bi-clausality. While at least one argument originating in the infinitival complement is in preverbal position (i.e. to the left of the matrix verb), the rest of the complement, including the infinitive, is in postverbal position (i.e. to the right of the matrix verb). This is illustrated in (16), again in the context of our test verbs bedauern regret and vergessen forget. The examples here are similar to those in (8) and (9), with the crucial difference that the infinitive is preverbal there but postverbal here in (16).

27

(16)

a. dass der Student bedauert [den Schlssel verloren zu haben].

that the student regretted the key lost to have

that the student regretted to have lost the key. b. ?* dass der Student den Schlssel bedauert [ __ verloren zu haben]. c. dass der Hausmeister vergisst [das Zimmer abzuschlieen].
that the superintendent forgets the room off-to-lock

that the superintendent forgets to lock the room. d. dass der Hausmeister das Zimmer vergisst [ __ abzuschlieen]. If the subject in the matrix clause is of the right type (namely a quantificational pronominal like keiner no one), the preverbal element das Zimmer in (16d) can even scramble to the pre-subject position, as shown in (17). This suggests that, despite the postverbal infinitive (in traditional terms, despite the possibility of extraposition), we are dealing with a coherent type of construction. (17) dass das Zimmer keiner vergisst [ __ abzuschlieen].
that the room nobody forgets off-to-lock

that nobody forgets to lock the room. Reis & Sternefeld (2004) confirm that examples like (16) and (17) prima facie present a problem for the standard conception of coherence because, due to the mixture of intra- and extraposition, the infinitive is neither obviously coherent nor obviously non-coherent. These facts in themselves are also a challenge to the

28

standard assumption that extraposition is possible only with CPs (p. 472). Assuming either that the infinitival complement extraposes and then allows one or more of its arguments to move into the matrix clause (den Besten & Rutten 1989) or that the infinitival complement starts out as intraposed and allows movement of an argument into the matrix clause before extraposing (Santorini & Kroch 1990),11 we would have to accept one of two arguably undesirable consequences: either a CP-boundary can sometimes be transparent for scrambling of an argument, or a non-CP constituent can be extraposed. A possible solution to this Third Construction problem is Zwarts (1997) proposal that Dutch (and presumably also German) is actually an SVO language, with complements following verbs, so that seemingly extraposed infinitival complements are in their base-position. SOV order in embedded clauses would then be due to movement of the complement to a position to the left of the verb. If this is true, the possibility of movement out of a post-verbal infinitival complement is not surprising. In traditional terms, as long as the complement is not a CP, movement of an argument of the infinitive across the left edge of the complement boundary into the matrix
11

Other analyses of the Third Construction have been proposed by Bayer & Kornfilt (1989),

Geilfu (1991), Rutten (1991), and Sabel (1994). They all have in common that the infinitival complement is assumed to be extraposed (i.e. has CP-status) and that at least one of its arguments (in one way or another) gets to be in the matrix clause domain (on the other side of the matrix verb). A radically different analysis is Wllstein-Leisten 2001, where it is argued that the Third Construction is underlyingly mono-clausal and the matrix verb can be licensed in front of, behind, or in between the arguments of the infinitive. 29

clause is unproblematic. In Part II of the dissertation, I discuss an instance of argument (A-)movement, namely possessor raising, which can lead to a Third Construction and is dependent on the infinitival complement being no more complex than a VP. Given that VPs are not expected to extrapose, the assumption that infinitival complements originate postverbally becomes inevitable. While instances of the Third Construction are not considered as readily acceptable as clear cases of intra- and extraposition, they are widely attested, and the contrast between examples involving verbs from Class 1 versus verbs from Class 2 (see (16)), is very strong. The Third Construction could then be used as another coherence test: the possibility of one (or more) argument(s) of the infinitive showing up in the matrix clause domain, while the rest of the infinitival complement is separated from this domain by the matrix verb, is an indication of coherence and arguably the absence of an embedded CP-boundary. On Zwarts (1997) view, there certainly could not be an intervening CP-boundary. It is not clear, however, how small a postverbal complement has to be in order to allow an argument to scramble out. Again, new diagnostics are needed to determine the size of complements that are smaller than CP. The question of how small exactly infinitival complements have to be in order to allow scrambling of an argument from a postverbal position within the infinitival domain into the matrix clause becomes particularly urgent in light of data like (18). Different members of Class 2 show different degrees of compatibility with the Third

30

Construction. While the verb beschlieen decide, for example, is diagnosed as a member of Class 2 by the classic scrambling test (18a), it does not allow the Third Construction (18b). (18) a. dass den Schlssel jetzt keiner mehr [ __ zu suchen] beschliet.
that the key now nobody anymore to search decides

that nobody will decide to search for the key anymore now. b. ?* dass den Schlssel jetzt keiner mehr beschliet [ __ zu suchen].12 The data in (17)-(18) confirm, then, that beschlieen and vergessen, while both members of the larger class of coherent verbs (Class 2), induce different degrees of coherence and thus call for a sub-classification within Class 2 based on the exact size of the infinitival complement. The other construction at issue here is long passive, movement of the direct object of the infinitive into the matrix subject (nominative case) position under passivization of the matrix verb. As in the case of the Third Construction, only some members of Class 2 allow long passive. Unlike the Third Construction, however, long passive is considered by many a construction too marginal to be used as a reliable diagnostic (see e.g. Hhle 1978, Kiss 1995, and Reis & Sternefeld 2004). Nonetheless, the
12

Note that all the ungrammatical (?*) scrambling examples I present here become at least

marginally acceptable when the moved DP has T-scrambled (see Haider & Rosengren 1998) and is thus part of the so-called Hutkontur (hat contour) intonation. Since T-scrambling can cross CPboundaries, it is not the type of scrambling that is traditionally used as a probe for coherence. Different kinds of scrambling will be discussed more thoroughly in chapter 4. 31

judgments in (19) are relatively clear. Examples (a) and (c) show that both versuchen try and befrchten fear pass the scrambling test, but only versuchen is readily acceptable in a long passive construction. (19) a. dass den Schlssel keiner mehr [ __ zu finden] versucht.
that the key (ACC) nobody anymore to search tries

that nobody tried to find the key anymore. b. dass der Schlssel [ __ zu finden] versucht wurde.
that the key (NOM) to find tried was (PASS)

that they tried to find the key. c. dass den Schlssel keiner [ __ zu verlieren] befrchtet.
that the key (ACC) nobody to lose fears

that nobody fears to lose the key anymore. d. ?* dass der Schlssel [ __ zu verlieren] befrchtet wurde.
that the key (NOM) to lose feared was (PASS)

that they feared to lose the key. To what extent the subdivisions within Class 2 created by the Third Construction and long passive overlap is not quite clear. Since the grammaticality judgments are extremely subtle, different authors have come to different conclusions. This opens the analytical space for one of the contributions this dissertation will make. The classic coherence tests are reliable in probing for an embedded CP but are unable to make more fine-grained distinctions, and those constructions which do seem to subdivide the class of coherence-inducing verbs are either too mysterious structurally
32

or too marginal empirically to be used as reliable diagnostics for the different degrees of coherence which matrix verbs in Class 2 can induce. There is clearly a need for diagnostics based on syntactic phenomena which are both well-understood phrasestructurally and well-documented empirically. Chapters 3 and 5 present the possessor dative construction and binding as constructions which I believe fulfill these criteria. To complete the traditional picture of the coherence/non-coherence distinction, there is a Class 3 of predicates taking verbal complements. This is the class of obligatorily coherent predicates. Its members either take complements with infinitives of the 1st status, i.e. without zu, or, if they do select a zu-infinitive, they do not q-mark their subject, a characteristic typical of raising verbs. The following are some examples of this rather diverse class of verbs:13 modals (e.g. knnen can, mssen must), perception verbs (e.g. sehen see, hren hear), causatives (lassen let, allow, have), aspectuals (e.g. anfangen start, aufhren stop), and raising verbs (e.g. scheinen seem). Since the infinitival complements of these verbs cannot be extraposed,14

13

I am only listing verbs which take infinitival complements here. However, auxiliaries

selecting a participial verb form as their complement are also members of this class.
14

Note that, on Zwarts (1997) view that infinitival complements are underlyingly postverbal,

the impossibility of extraposing a complement of a Class 3 verb has to be accounted for by making movement of the complement (i.e. intraposition) obligatory. This must also be the way to account for object complements, which, in non-root clauses, can never occur postverbally. 33

members of Class 3 are assumed to necessarily induce coherence. There can then not be an embedded CP-boundary.15 As noted by Reis & Sternefeld (2004), the inclusion of perception and causative, i.e. AcI-introducing, verbs in this class is somewhat problematic. Although it is true that AcI-complements look coherent in that they never extrapose (see (20b)), they behave as if they are non-coherent in that they can induce opacity. The embedded direct object in (20c), for example, cannot scramble out of the AcI-complement into the matrix clause.16 (20) a. dass der Professor [ACI die Studenten den Artikel lesen] lsst.
that the professor the students the article read lets

that the professor has the students read the article. b. * dass der Professor lsst [ACI die Studenten den Artikel lesen]. c. * dass der Professor den Artikel [ACI die Studenten __ lesen] lsst. These facts suggest that, in addition to the coherence/non-coherence distinction, we need a distinction based on the presence/absence of an embedded subject. More specifically, the fact that the transitive AcI-subject die Studenten in (20c) blocks movement of the direct object den Artikel into the matrix clause immediately brings
15

See Grewendorf 1987, 1988, however, for an analysis of AcIs as being closed off by an S-

bar boundary and thus as being satzwertig.

16

Interestingly, when the AcI-object is the clitic-like pronominal es, it can precede the AcI-

subject (see Grewendorf 1987, 1988). 34

to mind the SPECIFIED SUBJECT CONDITION (Chomsky 1973). A detailed discussion of transitive/unergative versus passive/unaccusative AcI-constructions, in particular in the context of reflexive and pronominal binding, is provided in chapter 5. Wurmbrands (2001) (non-)restructuring typology is a proposal designed to deal with some of the coherence/non-coherence mismatches at issue here. Analyzing long passive as an instance of A-movement, which is sensitive to intervening A-positions, and using it as a crucial diagnostic for infinitival clause size, Wurmbrand is able to make transparency distinctions that go beyond the traditional mono-versus bi-clausal split. She does not, however, address the problem of the Third Construction, and her typology leaves unresolved the question of how AcI-introducing predicates fit into the picture.

2.3

A typology of (non-)restructuring

Wurmbrand (2001) takes a non-derivational approach to coherence. Within her system, infinitival complements can, in principle, originate as a CP, TP, vP, or bare VP.17 Since this typology postulates the existence of several complement types that are smaller than CP, it looks like a promising approach to those verbs in Class 2 and 3

17

As alluded to earlier, I am making the common assumption that the German infinitival

marker zu does not mark tense, i.e. does not signal the presence of a TP, but is instead interpreted as a marker of verbal case/status in the sense of Bech (1955/57). By bare VP, I mean a zu-infinitive without functional (vP) structure. 35

which remain incompletely analyzed in a theoretical context which assumes only a binary coherence/non-coherence (CP/non-CP) distinction.

2.3.1

Four degrees of (non-)restructuring

Like the traditional coherence/non-coherence classification, Wurmbrand makes a clear-cut distinction between matrix verbs taking a full CP-complement and matrix verbs taking a smaller (in her case, TP, vP, or VP) complement. The former, which Wurmbrand calls (F U L L ) CL A U S A L N ON -R ESTRUCTURING (NR) predicates, correspond to the type of verbs traditionally classified as occurring in non-coherent constructions (Class 1), and the latter can roughly be associated with the class of verbs occurring in coherent constructions (Classes 2 and 3). Wurmbrand subdivides this latter verb class into F U N C T I O N A L R ESTRUCTURING (FR), LEXICAL

R ESTRUCTURING (LR), and R EDUCED N ON -R ESTRUCTURING (RNR) predicates. Members of the FR class are the finite verbs in modal, raising, and other Class 3-type constructions, while LR and RNR predicates are the control-zu-predicates which make up Class 2 in the traditional classification. An overview of Wurmbrands classification with examples and some characteristic properties of each type of (non-)restructuring predicate under her analysis is given in TABLE 1 (see Wurmbrands Table 1, p. 3 and Table 40, p. 327).

36

TABLE 1
Type FUNCTIONAL RESTRUCTURING (FR) Infinitive = main predicate (vP or VP) Properties, Distribution thematic properties are determined by infinitive18 possible with: modal, raising, aspectual, causative, perception, motion verbs LEXICAL RESTRUCTURING (LR) Infinitival complement = VP REDUCED NON-RESTRUCTURING (RNR) Infinitival complement = vP or TP (FULL) CLAUSAL NON-RESTRUCTURING (NR) Infinitival complement: CP no embedded (PRO) subject no embedded structural case possible with: control verbs embedded (PRO) subject embedded structural case possible with: control verbs embedded (PRO) subject embedded structural case possible with: control verbs Examples drfen may, gehen go, hren hear, lassen let scheinen seem vergessen forget versuchen try wagen dare beschlieen decide planen plan
versprechen promise

bedauern regret, behaupten claim vergessen forget

The crucial phrase-structural distinction between LR and RNR predicates is that the LR-type takes just a bare VP-complement, whereas the RNR-type takes a more complex, subject-containing vP or TP-complement. The external argument position (external argument in the sense of subject) is one of the keys to diagnosing Wurmbrands phrase-structural distinctions in complement size. The presence of an embedded subject is argued to entail a vP-projection, and a vP-projection, in turn, is associated with the capability of licensing an object (cf. BURZIO S GENERALIZATION (Burzio 1986)). Thus, in constructions traditionally analyzed as control structures,
18

Notice that this is not true in the case of AcI-introducing predicates like lassen let, allow,

have, which do assign a q-role to the argument in their specifier. As discussed in subsection 2.3.2, AcI-constructions do not quite fit into Wurmbrands typology. 37

where an embedded subject is not overtly expressed, Wurmbrand takes the presence of an embedded object to indicate vP-status of the infinitival complement. The embedded object is then indirect evidence for a PRO-subject in Spec vP. An object that is thematically related to the infinitive but shows up as part of the matrix clause, on the other hand, is argued to indicate that the infinitive is a bare VP, without the functional (vP) layer needed to case-license an object, and thus without a PROsubject position. Wurmbrand motivates this distinction between LR predicates (VP-selecting) and RNR predicates (vP-selecting) by appealing to two diagnostics: scrambling and long passive. Both are movement operations, and, according to Wurmbrand, they are instances of A-movement, possible only when the argument(s) of the infinitive cannot be case-licensed in the embedded domain. Consequently, if a matrix verb allows scrambling from the infinitival domain into the matrix clause, its complement is just a bare VP. If it does not allow scrambling, its complement must be bigger (at least a vP). The examples Wurmbrand uses to illustrate scrambling all have postverbal infinitival complements,19 where movement of the direct object into the matrix clause is never string-vacuous. When the infinitival complement is preverbal, as in my examples (8c-d), repeated here as (21a-b), and the embedded direct object (here das Zimmer the room) precedes the matrix subject (see (21b)), Wurmbrand would

19

Wurmbrand does not use this terminology, but her scrambling examples are in fact Third

Constructions. 38

analyze the complement as a bare VP. When the direct object appears to remain in situ (see (21a)), on the other hand, Wurmbrand would argue either that the embedded domain consists of a CP and the direct object is case-licensed within the infinitival complement, or that the embedded domain consists of only a VP and the direct object checks case by overtly (here string-vacuously) or covertly moving into the matrix clause. Thus, in a restructuring context, direct object movement into the matrix clause is obligatory. If this movement is not overt, then it must happen covertly.20 (21) a. dass niemand
that nobody

[CP das Zimmer abzuschlieen] das Zimmer [VP __ abzuschlieen].

the room off-to-lock

vergisst
forgets

that nobody forgets to lock the room. b. dass das Zimmer niemand [VP __ abzuschlieen] vergisst. A thorough critique of both scrambling and long passive as diagnostics for the presence of a subject-introducing (henceforth agentive or transitive) vP is provided in chapter 4. In the following subsection, I turn to the type of infinitival complement which provides one of the most interesting testing grounds for issues of Satzwertigkeit, but which does not seem to have a clear-cut place in Wurmbrands (non-)restructuring typology.

20

In Wurmbrands (2004) more recent work, the notion of covert movement is replaced by a

static Agree relation between the two case-checking heads. 39

2.3.2 AcI: A hybrid category Verbs in Wurmbrands class of restructuring (LR and FR) predicates have in common that the construction they enter into can have maximally one external argument (or subject) which originates in the functional projection of either the matrix verb (in a control structure) or the infinitive (in a modal or raising-type structure). The former type of structure is characteristic of lexical (LR) and the latter of functional (FR) restructuring. AcI-introducing verbs are analyzed as a type of FR predicate, but unlike other verb-types in the FR-category, they do not form a completely coherent unit with their complement. As is well known from binding facts (see chapter 5), AcIs can induce opacity.21 An AcI-embedded pronominal, for example, can be bound by the matrix subject, as long as the intervening AcI-subject is agentive (i.e. of the PROTO-AGENT type22). In accordance with this distinction between AcI-constructions and purely functional coherent constructions, Wurmbrand argues for another subdivision and analyzes AcI-introducing verbs as SEMI-FUNCTIONAL RESTRUCTURING (SFR) predicates. I interpret semi-functional to mean that AcIs are best thought of as a hybridcategory, sharing properties with both FR and LR predicates. In Wurmbrands
21

See also Grewendorf 1987, 1988 where it is argued convincingly that not all AcIs can be

classified as obligatorily coherent.

22

I am appealing to the notion PROTO-AGENT (versus PROTO-PATIENT) in the sense of Dowty

1991 here. Besides agents, arguments with proto-agent properties are, for example, experiencers and causers (see chapter 5 for a detailed discussion of the syntactic correlate of agentivity). 40

system, AcI-introducing verbs are like FR predicates in that they are introduced as functional/aspectual v-heads and form a syntactically inseparable unit with their complements extraposition is impossible,23 and the complement is not a separate tense or negation domain. Yet they are unlike FR predicates in that they assign an external argument role.24 In this respect, they are more like LR predicates. Unlike LR predicates, though, AcI-introducing verbs can have satzwertige, i.e. semantically complete, complements including a subject (in the sense of protoagent). Thus, AcI-constructions share properties with both FR and LR contexts, but they cannot be classified as either because they potentially contain more than one subject. In both FR (modal and raising-type) and LR (control) constructions, there is only one (pronounced) subject; in FR constructions, it originates in the infinitival complement, and in LR constructions, it originates in the matrix clause. AcIconstructions, on the other hand, can host two referentially distinct subjects one in the matrix clause, and one in the infinitival complement much like nonrestructuring predicates. Although her basic proposal for the organization of German clause structure suggests that SFR v-heads take just a bare VP complement, Wurmbrand deliberately does not take a stand with respect to the complexity and

23

Note, again, that, in a world where complements underlyingly follow the verb (see Zwart

1997), the impossibility of extraposition translates into obligatory movement of the complement into preverbal position.
24

To be exact, in the system I develop in later chapters, it is the agentive light-verb projection

of the AcI-introducing verb which assigns the external argument role (see chapter 5). 41

exact configuration of the complements of SFR predicates. The AcI-complement in (18a), repeated here as (22), clearly has a subject position. (22) dass der Professor [ACI die Studenten den Artikel lesen] lsst.
that the professor the students the article read lets

that the professor has the students read the article. In line with much recent work on verbal phrase structure (see e.g. Hale & Keyser 1993, Chomsky 1995, and Kratzer 1996), I assume that, unlike passive and unaccusative verbs, transitive and unergative verbs project an agentive vP. On this view, v assigns the external argument (proto-agent) role to the argument in its specifier and checks accusative case with the VP-internal argument. Consequently, transitive and unergative AcIs must either be vP-complements, or else contain vP as a proper subpart.

2.3.3 Restructuring, possessor datives, and binding One of the goals of this dissertation is precisely to probe for the presence of AcIembedded vPs and thus diagnose the restructuring status of AcIs in order to correctly classify them with respect to both the traditional picture of coherence/non-coherence and Wurmbrands more fine-grained typology. I appeal primarily to two grammatical phenomena: the possessor dative construction and binding. Both phenomena are sensitive to the transparency/opacity of infinitive constructions and are therefore relevant to diagnosing the phrase-structural distinctions between different types of
42

infinitival complements. Besides serving as much needed additional tools for the classification of control-zu-verbs, in particular for the distinction between Wurmbrands LR and RNR predicates, possessor datives and binding phenomena also shed light on the internal structure of complements to AcI-introducing verbs, which Wurmbrand classifies as SFR predicates but leaves to be further investigated. Part II of the dissertation focuses on the possessor dative construction, and Part III on binding. Both phenomena will be established as coherence diagnostics which are sensitive to the presence of an embedded agentive vP-projection. To conclude this review of the coherence and restructuring literature and to indicate the place Parts II and III of this dissertation have in it, I have shown that the traditional binary distinction between coherence and non-coherence leaves the coherence status of some members in Class 2, the optionally coherent predicates, unaccounted for. While Wurmbrands (2001) more fine-grained (non-)restructuring typology is designed to solve this problem (see TABLE 2 for an overview of how Wurmbrands classification lines up with the traditional coherence/non-coherence verb classes), it does not address the Third Construction and depends crucially on two diagnostics (long passive and scrambling), which will be shown to be problematic.

43

TABLE 2
Wurmbrands Classification Predicate Type FUNCTIONAL RESTRUCTURING (FR) Complement Size VP/vP Class 3: obligatorily coherent (complements diagnosed as smaller than CP by traditional coherence tests) LEXICAL RESTRUCTURING (LR) (verbs allow long passive and scrambling) REDUCED NON-RESTRUCTURING (RNR) (verbs do not allow long passive and scrambling) FULL NON-RESTRUCTURING (NR) CP Class 1: obligatorily non-coherent (complements diagnosed as CPs by traditional coherence tests) vP/TP Subclass of Class 2: optionally coherent (complements diagnosed as smaller than CP by traditional coherence tests) VP Subclass of Class 2: optionally coherent (complements diagnosed as smaller than CP by traditional coherence tests) Traditional Classification of Coherence

I have also drawn attention to the problem AcI-constructions pose to both the traditional system and Wurmbrands. Their hybrid status makes AcIs particularly difficult to classify. In the following chapters, I aim to both supplement and fine-tune the FR, LR, and RNR-parts of Wurmbrands typology, while pursuing the broader goal of investigating what makes different phrase-types transparent or opaque to grammatical interactions across their boundaries and thus what it means for a phrase to be satzwertig.

44

PART II THE POSSESSOR DATIVE CONSTRUCTION

45

German Possessor Datives: Raised And Affected

The goal of this chapter is to show that the German P OSSESSOR DATIVE CONSTRUCTION (PDC) is best analyzed as POSSESSOR RAISING, i.e. an instance of Amovement, and that a dynamic structure-building (Minimalist) system provides a way to account for the fact that possessor datives have to be assigned both a possessor and an affectee role. The proposed analysis will then play an important part in chapter 4, where I argue that an account of POSSESSOR D ATIVES (PDs) as undergoing Amovement makes the PDC a much-needed new coherence diagnostic.

3.1

Introduction

3.1.1 External Possession Many languages around the world make use of a construction known as EXTERNAL P OSSESSION (see e.g. Payne & Barshi 1999 and Vergnaud & Zubizarreta 1992). Typical examples from German (a), French (b), and Hebrew (c) are given in (1).
(1)

a. Tim hat der Nachbarin

das Auto gewaschen.

washed

Tim has the neighbor (DAT, FEM) the car

Tim washed the neighbors car. b. Jai coup les cheveux Pierre.
I have cut the hair to Pierre (DAT)

I cut Pierres hair. (Guron 1985: 59)

46

c. ha-yalda kilkela le-Dan

the-girl

et

ha-radio.

spoiled to-Dan (DAT) ACC the-radio

The girl broke Dans radio. (Landau 1999: 3) In such structures, a single dative-marked nominal (for instance der Nachbarin in (1a)) acts simultaneously as a possessor, i.e. a subpart of a larger nominal phrase, and as a complement to a verb (waschen in (1a)). The challenge posed by these structures is to understand the mechanisms that make this dual functioning possible. This is just the kind of situation for which classical generative grammar introduced the device of the movement transformation. One says that the nominal der Nachbarin in (1a) first occupies the usual possessor position within the larger nominal and is thus interpreted as possessors normally are. It subsequently occupies a different position in the structure, a complement position to the verb waschen, and is thus interpreted as such complements normally are. This is the intuition behind accounts that analyze the phenomenon of external possession as P O S S E S S O R ASCENSION/RAISING (see e.g. Allen et al. 1990 and Landau 1999). 3.1.2 The dual role of possessor datives I argue here that the intuition behind possessor raising accounts is fundamentally correct and present some new evidence for its correctness. However, the analysis is difficult to implement within the terms of the standard Principles & Parameters view
47

because that theory is committed to the existence of deep structure a level of representation in which all core semantic role relations are fixed before any movement operations apply. In that conception of how syntax is organized, it is impossible to move a nominal into a derived position in which it will be assigned a new, or additional, semantic role. But in (1a), the possessor is assigned the semantic role of AFFECTEE (more specifically, the person who benefits from the car washing) in its derived position. One of the consequences of making the transition from the traditional Principles & Parameters framework to a Minimalist system (Chomsky 1995, 2000, 2001) is exactly the elimination of the level of deep structure. In this chapter, I explore the ramifications of this theoretical shift in connection with such structures as (1) (the PDC). I argue that the elimination of deep structure opens the way to a much improved understanding of such constructions. This work can be seen, then, as an argument in favor of this crucial theoretical shift. To reiterate, the challenge posed by the PDC is to account for the fact that the dativemarked possessor argument (the PD) also plays the role of an affectee argument. I propose that, in a system where heads with their selectional features are introduced in the course of the derivation, it is in principle possible that an argument which gets merged into the structure to take on one thematic role raises into a newly built sentence domain (or PHASE in the sense of Chomsky 2000, 2001) to fulfill another

48

thematic role (see also Hornstein 1999). This movement and the resulting double qrole assignment are crucially case (i.e. formal feature)-driven; that is they are due to the fact that in its origin site, the raised argument is not case-licensed. Only an additional case-checking head can save the derivation. I argue, in line with much recent work (Anagnostopoulou 2002, Baker to appear, Hole to appear-a, McFadden 2003, McIntyre to appear, Miyagawa & Tsujigka 2004, Pereltsvaig 2003, and Pylkknen 2002), that this head is a malefactive or benefactive (affectee) light verb which assigns inherent dative case to its argument.1 My approach is similar to Landaus (1999) possessor raising analysis but goes beyond the seemingly special case of Hebrew, which, according to Landau, as well as Pereltsvaig 2003, can be analyzed without a syntactic correlate of PD-affectedness, i.e. without assignment of an affectee-role. As confirmed for German by Hole (to appear-a), McIntyre (to appear), and Wegener (1985, 1991) and for Romance by Guron (1985) and Kempchinsky (1992), however, a PD is not only a possessor but must also be an affectee argument of the verb.

McIntyre (to appear), for example, calls this head VDAT and argues that it assigns to its

specifier an interpretation parallel to that found with subjects of English have. For Hole (to appear-a), it is a voice-head (Aff), and for Pylkknen (2002) it is one of the (v-)applicative heads she proposes for the introduction of non-core arguments. 49

3.2

Scope of the chapter

Before I discuss possessor raising and how Landaus (1999) analysis differs from my proposal starting in section 3.3, this section establishes the scope of the chapter. Crucially, PD-movement is not intended to account for the non-PD constructions covered in subsection 3.2.2.

3.2.1 Possessor datives A PD is a dative-marked nominal that is interpreted as the possessor of one of its clausemates. In (2a), for example, Mami is the possessor of das Auto. Unlike genitive-marked possessors (see (2b)), PDs do not show up as part of the same constituent as the possessed DP. While the genitive possessor in (3c), for example, focus-moves as a unit with the possessed DP, the PD in (3a) does not. When the possessed DP gets fronted, the PD must stay in its lower position (see (3b)). (PD and possessee are in bold face; the phrase in focus is given in capital letters.) a. Mein Bruder hat der Mami
my

(2)

das Auto zu Schrott gefahren.

to scrap driven

brother has the mom (DAT) the car

My brother totaled moms car (totaled the car on mom). b. Mein Bruder hat Mamis
my

Auto zu Schrott gefahren.

to scrap driven

brother has moms (GEN) car

My brother totaled moms car.

50

(3)

a. *Der Mami DAS AUTO hat er zu Schrott gefahren.

the mom (DAT) the car has he to scrap driven

Moms CAR, he totaled. b. DAS AUTO hat er der Mami

the car

zu Schrott gefahren.
driven

has he the mom (DAT) to scrap

Moms CAR, he totaled. (The CAR. he totaled on mom.) c. Mamis AUTO hat er zu Schrott gefahren.
moms (GEN) car has he to scrap driven

Moms CAR, he totaled. The fact that the PD der Mami is not licensed without a possessed DP like das Auto (see (4a) and similar examples in Wunderlich 2000), unless the verb takes a dative complement independently of the possessor relation (4b), shows that there is an obligatory thematic connection between the PD and the possessed nominal. a. *Mein Bruder hat/ist2 der Mami
my brother has/is

(4)

gefahren.

the mom (DAT) driven

b. Mein Bruder wollte der Mami

my

helfen.

brother wanted the mom (DAT) help

My brother wanted to help mom. The puzzle is that, despite this obligatory thematic connection between PD and possessed DP, the data in (3) seem to suggest that, syntactically speaking, a PD behaves like an independent argument of the verb. Even in the case of inalienably
2

When used intransitively, fahren drive takes the perfect auxiliary sein be. 51

possessed body parts, so called PERTINENCE

DATIVES

(see (5) and many more

examples in Isac&enko 1965 and Wegener 1985)3, PD and possessed DP do not form a syntactic constituent. Ein guter Ehemann massiert seiner Frau jeden Abend den Rcken.
a good husband massages his wife (DAT) each evening the back

(5)

A good husband massages his wifes back every night. It is clear that the PD (here seiner Frau) and the possessed DP (here den Rcken) can be separated. The definite article of the possessee can be replaced with a possessive pronoun when the possessee is a non-body-part DP. The use of a possessive pronoun is degraded, however, with a body-part DP. This is illustrated in (6a-b). a. Mein Bruder hat der Mami
my

(6)

leider

ihr Auto zu Schrott gefahren.4

to scrap driven

brother has the mom (DAT) unfortunately her car

Unfortunately my brother totaled moms car. b. ?Ein guter Ehemann massiert seiner Frau jeden Abend ihren Rcken.
a good husband massages his wife (DAT) each evening her back

A good husband massages his wifes back every night.

3 4

See Guron 1985 and Vergnaud & Zubizarreta 1992 for similar examples from French. In many dialects of German, particularly in casual speech, the combination of dative

possessor plus possessed DP with a possessive pronoun in Spec DP is not only acceptable but actually replaces the standard genitive construction. Instead of Mamas Auto moms car, speakers of these dialects can say der Mama ihr Auto (to) the mother her car (see e.g. Wegener 1985). Unlike in a PDC, the dative possessor and the possessee in this construction cannot be separated. (Note that (6a) is not an instance of this construction.) 52

As pointed out by Shibatani (1994), this contrast probably stems from the fact that body-part nominals are special in that they are automatically understood to be inalienably possessed by the referent of the dative nominal. Intuitively, the use of a possessive pronoun (which serves the sole purpose of establishing a possessor relation) is then simply redundant. The coocurrence of non-body-part nominals with a possessive pronoun ties in with the discussion provided in the following subsection examples like (6a) are, in fact, non-PD constructions and will be fully explained in section 3.4.

3.2.2 Non-possessor datives There are several types of dative-marked nominals (henceforth non-PDs) which appear to be licensed by neither the presence of a possessed DP nor a verb that selects a dative complement. The dative mir me in (7), for example, corresponds to the so called ETHICAL DATIVE which Borer & Grodzinsky (1986) discuss in their study of dative constructions in Hebrew. Schlaf mir
sleep

(7)

jetzt schn ein, Kleines!

little one

me (DAT) now nicely in

Kindly fall asleep for me now, little one! Here, the person referring to him or herself in the first person expresses an emotional attitude toward the situation of the childs falling asleep. The referent of an ethical dative thus deeply cares about the given situation. As noted by Borer & Grodzinsky,
53

Hebrew ethical datives are obligatorily clitics, i.e. may not be expressed by a nonpronominal dative phrase, and are only compatible with verbs having an external argument. In German, the distribution of ethical datives is not as restricted examples like (7) show that, just like PDs, ethical datives in German are compatible with unaccusative verbs (see subsection 3.3.2) but it is true that the construction is most commonly found in imperatives with the dative first person pronoun mir, which could be argued to exhibit clitic-like behavior. Another type of non-PD, exemplified by (8), is known as the DATIVUS IUCANDIS or ESTIMATIVE DATIVE. Mein Bruder ist der Mami
my brother is

(8)

zu schnell gefahren.
driven

the mom (DAT) too fast

My brother drove too fast for mom. The interpretation here is that my brothers driving was too fast for moms liking. This type of dative typically cooccurs with modifiers like zu too and genug enough. Non-PDs that fall under the rubric of neither ethical nor estimative datives, like (9), are what McIntyre (to appear) calls FICIARY (BENEFICIARY/MALEFICIARY) DATIVES.

54

(9)

Sie hat mir

Bushs Ansprache bersetzt.

translated

she has me (DAT) Bushs speech

She translated Bushs speech for me. (McIntyre to appear: 7) This type of dative is interpreted like a PD but without the possessive aspect. As in the case of PDCs, the well-formedness of all three of the non-PD constructions discussed here does not depend on the presence of the dative DP. The verbs in (7)-(9) do not in any way need a dative argument. The well-formedness of any construction with a non-selected (non-core) dative does, however, depend on the ability of the verb to express some kind of affectedness a negative or positive effect, either physical or emotional on the dative referent (see also Wegener 1985, 1991, McIntyre to appear, and Hole to appear-a). This explains why fahren drive alone, as in (2a), can occur in neither a PD nor a non-PD construction both require a predicate which can assign an affectee role. Getting back to examples like (6a) in subsection 3.2.1, while all the dative constructions discussed above seem to be subject to an affectedness condition, an obligatory possessor relation between the dative DP and another nominal in the sentence only holds for the PDC. The non-PD mir in (9), for example, cannot be the possessor of the speech. It is clear that the genitive nominal Bushs establishes the possessor relation here. Similarly, in (6a), Mein Bruder hat der Mami (DAT) leider ihr (POSS) Auto zu Schrott gefahren, it is not the dative but the possessive pronoun ihr
55

which establishes the possessor relation with the car. The dative in these cases must then be a non-PD, more specifically, a bene/maleficiary. In the case of (6a), this does not make a significant difference in interpretation. Since the possessive pronoun ihr has the same referent as the dative der Mami, the possessor relation established by the possessive pronoun is the same as the possessor relation established by the dative nominal in the corresponding PDC (with a definite article instead of the possessive pronoun). I will occasionally refer back to non-PD constructions for purposes of comparison, but the main focus will be on PDs, which stand both in an affectee and a possessor relation.

3.2.3 Solving the classic puzzle According to Landau (1999), the classic puzzle of the PDC is that an argument in the clause (the PD) can derive its semantic role from another argument (the possessee) but its syntactic behavior from the predicate. I will show that the split between semantic role and syntactic behavior, while compatible with the Hebrew facts, does not accurately describe the more general case of the PDC. In German and the Romance languages at least, PDs derive not only their syntactic behavior but also one of their semantic roles from the verb. The truly puzzling question with respect to the general case of the PDC is thus how to avoid the apparent violation of the qCriterion. Landau (1999) identifies two major paths to take in analyzing PDCs:
56

a. PD is an argument of the verb (male/benefactive). control of PRO in the possessee.

The possessor

interpretation arises through binding of an anaphoric element or through

b. PD is an argument of the possessee. Its misleading syntax is due to syntactic raising to a position typically occupied by verbal arguments (p. 2).

Borer & Grodzinsky (1986) analyze the Hebrew PDC following path (a), while Landau reanalyzes the Hebrew facts according to path (b). Other path (a)-type analyses have been given for German by Hole (to appear-a) and for Romance by Kempchinsky (1992), Vergnaud & Zubizarreta (1992), and Guron (1985).5 As laid out in section 3.3, the argument for syntactic movement (i.e. possessor raising) that Landau proposes for Hebrew generally extends to German. I will therefore follow Landau (as well as Isacenko (1965) and Gallmann (1992) who have proposed & possessor raising analyses for German)6 in taking path (b). Since the class of verbs that allow the PDC in German seems to be more restricted than the corresponding
5 6

See Landau (1999) for a more comprehensive overview of path (a)-type analyses. Isac&enko (1965) proposes a dative transformation rule which turns genitive constructions

like Der Rcken des Mannes schmerzt (The mans back aches) into the corresponding PDC Dem Mann schmerzt der Rcken (literally to the man hurts the back), where the post-nominal genitive des Mannes raises to become the dative dem Mann. Gallmann (1992) argues that, after incorporation of the possessee N into V, the caseless complement of N, namely the possessor-NP, raises and adjoins to V' to get structural dative case from V, which assigns both accusative and dative case in this framework. 57

verb class in Hebrew, however, I will argue that certain aspects of path (a) must be integrated into the analysis. In particular, while I do agree with Landau that a PD originates in the specifier of the corresponding possessed DP and then raises to a specifier position within the verbal domain, I claim, contra Landau, that PDs are also thematically restricted by the verb. Besides acting as the possessor of the DP which is its origin site, PDs must play the role of affectee. Although Landau agrees that [the] PDC is always associated with an affectedness implication for [the] PD (Landau 1999: p. 3, fn. 1), the following examples hint at the generalization that many more verbs imply affectedness, i.e. are compatible with the PDC, in Hebrew than in German. While the German PDC in (10b) is clearly unacceptable, the Hebrew equivalent in (10a) is judged grammatical by Landau. Dispensing with a syntactically encoded affectedness condition then might be an acceptable solution for Hebrew (see also Pereltsvaig 2003, an analysis of the Hebrew PDC as possessor raising to a dative light-verb head which is defective in that it does not assign a semantic role to the argument in its specifier), but not for German. The German PDC, which gets better the more obviously the negative (10c-d) or positive (10e) effect on the dative referent is expressed, clearly involves assignment of an affectee role. a. Gil gar le-Rina ba-xacer.
Gil lives to-Rina in-the-yard

(10)

Gil lives in Rinas yard. (Landau 1999: 4)

58

b. *Tim wohnt Lena

Tim lives

im

Garten.

Lena (DAT) in-the garden

Tim lives in Lenas garden. c. ?Tim steht Lena im Garten herum.

Tim stands Lena (DAT) in-the garden around

Tim stands around in Lenas garden. d. Tim ruiniert Lena

Tim ruins

den schnen Garten.

Lena (DAT) the beautiful garden

Tim ruins Lenas beautiful garden. e. Tim grbt Lena den Garten um.

Tim digs Lena (DAT) the garden around

Tim aerates Lenas garden. After showing in section 3.3 that Landaus arguments for a raising analysis generally hold for German, I will propose in section 3.4 that the affectedness condition can be explained within a dynamic structure-building framework where a second q-role becomes available to an argument after movement.

3.3

Possessor raising

The goal here is to show that several of the basic properties Landau (1999) identifies for the PDC in Hebrew hold for German as well. Like Landaus data, the German facts illustrate that a syntactic movement analysis is superior to a thematic approach that bases the connection between possessor and possessee on binding or control (e.g.
59

Hole to appear-a, Kempchinsky 1992, Vergnaud & Zubizarreta 1992, Borer & Grodzinsky 1986, and Guron 1985).7 After an informal description of three major characteristic properties of the PDC in 3.3.1-3, subsection 3.3.4 introduces Landaus movement analysis and thus provides the basic building blocks for a formal structural account of the PDC and its properties. Finally, subsection 3.3.5 points out the advantages this approach has over non-movement alternatives. Although Landaus proposal is very similar to the dynamic structure-building analysis I present in section 3.4, it will become evident that the latter is superior, at least with respect to the general (non-Hebrew) case of the PDC.

3.3.1 Obligatory possessor interpretation The crucial distinction between a regular dative-marked benefactive or malefactive argument of the verb and a possessor dative is that the latter must cooccur with a clausemate nominal with which it stands in a possessor relation. The possessee is typically an inalienably possessed body part or something that counts as inalienably possessed by extension (house, garden, car, computer, etc.; see Vergnaud & Zubizarreta 1992). In a PDC, the dative-marked nominal is obligatorily interpreted as possessor. If this obligatory possessor relation does not hold, the structure contains a

Note that several authors have specifically argued against a possessor raising analysis for

German (see e.g Hole to appear-a, to appear-b, Pylkknen 2002, and Wunderlich 1996, 2000). As discussed in section 3.4, the account proposed here avoids the main problem these authors have with traditional possessor raising analyses like Landau 1999 and Gallmann 1992. 60

verb which either independently selects a dative argument or is compatible with a non-PD, such as an ethical or estimative dative (see subsection 3.2.2). In a PDC, however, benefactive or malefactive interpretation is superimposed upon but crucially may not replace the possessor relation between the possessed nominal and the dative-marked DP. As Landau puts it, at least transitory possession must hold. Often this possessor relation is not literally one of possessing or owning something but must be interpreted in the broader sense of being responsible for something, even if just temporarily. An important restriction on this aspect of the PDC is that the range of relations that it may subsume is narrower than the range which can be expressed by an overt genitive (pre- or postnominal) possessor. The remainder of this subsection presents evidence for this. Based on the observation that postnominal of-possessors in Hebrew picture noun constructions can be interpreted as either possessor/creator or theme (see (11a)), but PDs can only be the possessor/creator (see (11b)), Landau argues that PDs are obligatorily interpreted as the subject of the corresponding possessed DP, crucially not as an internal (theme) argument.

(11)

a. Gil higdil et ha-tmuna s&el Rina.

Gil enlarged ACC the-picture of Rina

Gil enlarged Rinas picture. [Rina = possessor/creator/theme]

61

b. Gil higdil le-Rina et ha-tmuna.

Gil enlarged to-Rina ACC the-picture

Gil enlarged Rinas picture. [Rina theme] (Landau 1999: 5) This subject-requirement calls for further specification in that it needs to be established whether PDs are the subject (i.e. in the external argument position) of DP or NP. The examples in (12) shed light on this question. If PDs can only originate in Spec DP, not Spec NP, we expect that the PDC should be incompatible with nouns like process nominals which require a non-possessor, namely an agent (in Spec NP), as subject. (12a-b) are the German equivalents of Kempchinskys (1992) Spanish and Landaus Hebrew examples. As these examples are ungrammatical for independent reasons,8 however German sehen see does not express affectedness (see Wegener 1985, 1991), and die Armee the army, a noun referring to an institution, not individual people, does not have a referent that can be affected I appeal to example (12c) to prove the point: process nominals, which require an agent as subject, are indeed incompatible with the PDC. a. *Ich fotografierte der Armee
I photographed

(12)

die Zerstrung der Stadt.

the city (GEN)

the army (DAT) the destruction

I photographed the armys destruction of the city.

8

Note that it is not the genitive DP der Stadt of the city following the head noun Zerstrung

destruction that makes (12a) ungrammatical. As shown by examples like Ottos Beschreibung Marias Ottos description of Maria, German allows for multiple genitive-marked nominals in DP (see Longobardi 1996). 62

b. *Die Journalisten sahen der Armee

the journalists saw

die Hinrichtung von einigen Gefangenen.

of several prisoners

the army (DAT) the execution

The journalists saw the armys execution of several prisoners.

c. *Ich habe Ulli gestern die Wiederaufarbeitung des Kunstwerks fotografiert.

I have Ulli (DAT) yesterday the re-working the artwork (GEN)

photographed

I photographed Ullis remodeling of the artwork.

Given a PDC-interpretation, where the PD Ulli is obligatorily interpreted as possessor of the re-modeling process, (12c) is unacceptable. The ungrammaticality cannot be attributed to the verb fotografieren. As confirmed by examples like Ich musste Ulli (DAT) gestern die abgebrannte Kche fotografieren I had to take pictures of Ullis burned-down kitchen (maybe because he needed the photos for insurance purposes), fotografieren can, in principle, assign an affectee role. Seemingly grammatical examples of process nominals in PDCs, given in Hole to appear-a (see (13)9), do not convincingly disprove the point here. a. Walter hat (?unsx) den Baux
Walter has

(13)

der Mauer befrwortet.

us (DAT) the construction the wall (GEN) supported

Walter supported (for us) the construction of the wall (by us). b. Ed stellte unsx
Ed put

den Bau(x)

der Mauer

in Aussichtx.

us (DAT) the construction the wall (GEN) in prospect

Ed dangled the prospect of constructing the wall before us. (Hole to appear-a: 16-17)
9

Subscript x marks a binding relation between the dative and another nominal in the

sentence. According to Hole (to appear-a), both den Bau and Aussicht are possible bindees in (13b). 63

In (13a), the addition of the dative uns is marked as degraded, and in (b), the dative is clearly part of the expression jemandem (DAT) etwas in Aussicht stellen to dangle the prospect of something before somebody.10 This means that uns in (13b) is not an extra or free dative (as in the other PDC-examples discussed here) and thus cannot originate as possessor of the process nominal Bau. Holes data in (13) then do not undermine Landaus and my claim that the possessor relation in a PDC is not equivalent to the relation established by a genitive nominal functioning as an agent.

More generally, the fact that the PDC imposes restrictions on the type of nominal that a PD can cooccur with not with process nominals corroborates Landaus claim that the possessor relation between the PD and the other involved nominal cannot be replaced by a thematic restriction imposed by the verb. Unlike a beneficiary non-PD, for example, a PD cannot occur as an argument of the verb independently of other DPs in the clause. It is not a grammatical addition to just any verb that has an affectee role to assign. In her thematic binding account, Kempchinsky (1992) is forced to say that examples like (12) are ungrammatical because the dative nominal cannot bear both an agent and

10

The Redewendungen edition of the Duden (Drosdowsky 1992) lists the dative as part of the

expression. To clarify, this expression is a PDC, with the dative being the possessor of Aussicht, but since it is a frozen combination, the dative behaves like it is subcategorized for. It cannot function as possessor of a different nominal, and it cannot be left out like a PD in a regular PDC. 64

a benefactive role. As for non-process nominals, where an agent is not required, however, she argues that their co-occurrence with a PD is acceptable because possessor is not a genuine q-role and can thus be assigned to the PD in addition to the benefactive role coming from the verb.11 As explained in subsection 3.3.4, Landaus movement analysis allows for a more straightforward account of the incompatibility of process nominals with the PDC. The most natural explanation, however, can be given within the dynamic structure-building system I propose in section 3.4. It will become apparent that agents, subjects which originate in Spec NP of the head nominal, have no way of getting to the specifier of the nominals DPprojection. Spec DP, however, is precisely the position from which possessor raising is launched.

3.3.2 C-command restriction In the data considered thus far, the possessee is either a direct object or a PP that is an argument of the verb. As pointed out by both Guron (1985) and Borer &

Grodzinsky (1986), the possessee cannot be the external argument of the verb. This is confirmed by the ungrammaticality of the Hebrew example and its German equivalent in (14a-b). The possessed nominal may only surface in subject position if

11

Kempchinskys non-genuine possessor is similar to what Guron (1985) calls a

secondary q-role, assignment of which is supposed to be exempt from the q-Criterion. 65

it has moved there from inside the VP, as in a passive (15a) or unaccusative (15b) configuration.12
(14)

a. *ha-kelev hitrocec le-Rina

the-dog ran-around to-Rina

Rinas dog ran around. (Landau 1999: 7) b. *Der Hund ist Lena
the dog

herumgelaufen.

is Lena (DAT) around-run

Lenas dog ran around.

(15)

a. Der Hund ist Lena

the dog

berfahren worden

is Lena (DAT) over-driven was (PASS)

Lenas dog was run over (by a car). b. Der Arm ist mir
the

eingeschlafen.

arm is me (DAT) in-slept

My arm fell asleep. In the well-formed examples in (15), the possessee in subject position starts out as the internal argument of the verb. In (14), however, the possessee gets introduced directly as the verbs external argument, i.e. originates above the PD, and the result is ungrammatical. There must thus be a restriction on the PDC that requires the PD to
12

In (15a-b), the clause-initial (Spec CP) position is occupied by the nominative-marked DP.

In the unmarked (neutral) word order, the dative precedes the accusative nominal, as shown in the embedded clauses in (i) and (ii): (i) weil Lena der Hund berfahren worden ist. (ii) weil mir der Arm eingeschlafen ist. 66

c-command (at some point in the derivation) the possessee in its base position. The exact position of PDs will be discussed in subsection 3.3.4.

3.3.3 Locality Guron (1985) observed that the possessor and the possessee must be clausemates. In her non-movement account, the clausemate condition holds at both deep and surface structure. As for the account supported here, the condition is that possessor and possessee must be clausemates, or, more specifically, cannot be separated by a subject-containing category (vP, TP, CP), after PD-raising. In the following French examples, the possessee is an inalienably possessed body part. a. Jean semble [TP lui
Jean seems

(16)

avoir lav les cheveux].

him (DAT) have washed the hair

Jean seems to have washed his hair. b. *Jean lui semble [TP avoir lav les cheveux]. (Guron 1985: p. 48 (18)) As laid out in detail in chapter 4, infinitive clauses with TP or vP-status are opaque for the relation between PD and possessee in German as well. Only a reduced infinitival complement (introduced by a coherent/restructuring predicate; see chapter 2) allows for a PD with a possessee in the infinitival complement to be part of the matrix clause domain.

67

Expanding on Gurons clausemate condition, Landau argues that the locality of the PDC even goes beyond clause-boundedness. In a case where the direct object is a complex DP, as in (17), for example, the PD may only be associated with the larger (containing) DP rather than with the genitive, which is properly contained in (i.e. a subpart of) the larger DP. Put another way, the relation between the PD and the possessor position within the possessee may not extend into the DP (into the domain of its head), but rather may access only its highest (outermost) specifier position.

(17)

a. Gil ripe le-Rina et ha-gur

s&el ha-kalba.

Gil cured to-Rina ACC the-puppy of the-dog (FEM)

Gil cured the dogs puppy which belongs to Rina. b. Gil ripe le-Rina et ha-ima s&el ha-gur.

Gil cured to-Rina ACC the-mother of the-puppy

Gil cured the puppys mother which belongs to Rina. (Landau 1999: 15) In (17a), it is the puppy that must be interpreted as belonging to Lena, whereas in (17b), it is the mother of the puppy. Thus, in both cases, the larger DP corresponds to the possessed DP. It may be pragmatically inferred that the referent of the embedded DP also belongs to Lena, but the syntax does not encode this reading. The German equivalent of Landaus Hebrew examples, given in (18), seems to support Landaus conclusions. The first intuition is that Lena is the caretaker of the foal in (a) and of the mother in (b).

68

(18)

a. Tim pflegte Lena

[das Fohlen [der Stute]] gesund.

foal the mare (GEN) healthy

Tim treated Lena (DAT) the

Tim cured the mares foal which belongs to Lena. b. Tim pflegte Lena [die Mutter [des Fohlens]] gesund.
the foal (GEN) healthy

Tim treated Lena (DAT) the mother

Tim cured the foals mother which belongs to Lena. However, the situation is more complicated than this initial assessment would suggest. First (as pointed out to me by Andrew McIntyre), examples like (19a-b) seem to be exceptions to Landaus generalization, in that they appear to allow for an interpretation of the dative mir as the possessor of the embedded genitive DP. a. Dann stecke ich mir
then stick I

(19)

einen Ring auf [einen Finger [der linken Hand]].

ring on a finger the left hand (GEN)

me (DAT) a

Then I put a ring on a finger of my left hand. b. Mir fiel der Hammer auf [die Spitze [des linken Zeigefingers]].
on the tip the left index-finger (GEN)

me (DAT) fell the hammer

The hammer fell on the tip of my left index finger. Clearly, both the left hand in (19a) and the left index finger in (b) must be interpreted as belonging to the referent of the dative mir. However, since the finger in (a) and the tip in (b) are necessarily possessed by the same person as the left hand and the left index finger, the possessor relation between the PD and the embedded DP need not be syntactically encoded. The data in (20) corroborate this. The key observation

69

regarding examples (b-c) is that the PD cooccurs with a possessive pronoun as part of the embedded DP.
(20)

a. ?Ein guter Ehemann massiert seiner Frau jeden Abend ihren Rcken.
a good husband massages his wife (DAT) each evening her back

A good husband massages his wifes back every night. b. Dann stecke ich mir
then stick I

einen Ring auf [einen Finger [meiner linken Hand]].

ring on a finger my left hand (GEN)

me (DAT) a

Then I put a ring on a finger of my left hand.

c. Mir

fiel der Hammer auf [die Spitze [meines linken Zeigefingers]].

on the tip my left index-finger (GEN)

me (DAT) fell the hammer

The hammer fell on the tip of my left index finger.

All three examples involve a body-part nominal specified by a possessive pronoun, but, while (20a) is clearly degraded (see also section 3.2.1), (b) and (c) are not. Since the possessive pronoun in (b) and (c) can cooccur with the PD mir without degrading the utterances at all, I tentatively conclude that the PD in these examples does not originate in the position occupied by the possessive pronoun, i.e. the specifier of the embedded DP. Rather, it either originates in the possessor position of the larger DP or is really a maleficiary non-PD (not standing in a possessor relation at all). This means that, if there is a syntactic possessor relation at all, mir in (19a) is probably the possessor of the body-part DP einen Finger, not der linken Hand. Just as in expressions like Er hob die Hand He raised his hand and Sie schloss die Augen
70

She closed her eyes, where the logical possessors of die Hand and die Augen are (non-derived) subjects and thus cannot originate in the possessor position of the respective body-part nominals (see section 4), it is plausible that, in (19a), there is a mere pragmatic relation between mir and der linken Hand. As for (19b), I suggest that we may be dealing with a maleficiary non-PD. The reason that the example is degraded when mir is left out is not that the dative is necessary to indicate the possessor of the body part but rather that some human referent is needed in the linguistic environment of a body part. Unlike in (19a), there is no 1st person subject pronoun here to satisfy this need. Other examples which seem to suggest, contra Landau, that PDs can, in fact, be possessors of the embedded nominal of complex DPs are presented in Hole (to appear-a) and shown here in (21a-c).

(21)

a. Man zerriss dem Jungenx die Papiere der Mutterx.

one tore-apart the boy (DAT) the documents the mother (GEN)

They tore apart his mothers documents on the boy. b. Man verweigerte ihmx
one denied

die Auszahlung des Lohnsx.

the wages (GEN)

him (DAT) the payment

They denied him the payment of his wages.

71

c. Man verweigerte ihmx

one denied

die Auszahlung des Lohns

der Mutterx.

him (DAT) the payment

the wages (GEN) the mother (GEN)

They denied him the payment of his mothers wages. (Hole to appear-a: 18) In support of Landaus generalization, it can be argued that (21a) may be explained on a par with (19a-b). The utterance is not degraded when the definite article preceding Mutter is replaced with the possessive pronoun seiner his. I take this to mean that the dative dem Jungen may not be in competition, so to speak, with the possessive pronoun; it could either originate as possessor (here: person who is temporarily in charge) of the larger DP die Papiere or be a maleficiary non-PD. Kinship terms like Mutter are (inherently) relational nouns which, in the absence of a syntactic possessor, are automatically interpreted as related to the speaker or the referent of a nominal in the near linguistic context. As for (21b) and (c), verweigern is a verb that regularly occurs with a dative argument,13 suggesting that, again, the dative in these examples may not be an extra or free argument but rather one that is selected by the verb. No possessor relation between ihm and either of the two DPs then needs to be accounted for. Hole himself states that the definite articles in examples like (13) and (21) (his (34a), (35a), and (37)) can be properly interpreted without the extra dative argument: The possessor variable (which Hole assumes to be
13

Under Phrases and Collocations (3-4 words) for the entry of the verb verweigern, the

dict. cc online dictionary lists jemandem (DAT ) die Erlaubnis verweigern refuse somebody permission and jemandem (DAT) etwas verweigern refuse somebody something, and a Google search confirms the frequent occurrence of datives in the context of this verb. 72

included in one of two lexical entries of every nominal) may be mapped to some arbitrary referent, or be absent altogether, in the absence of the extra dative (Hole to appear-a: 17). All of these matters are subtle and somewhat unclear, and at a later point (in chapter 6), we will encounter examples which seem to pose yet more serious difficulties for the claim that the PD relation cannot reach into the DP domain. At that point, I will be in a position to show that the phenomena are in fact quite compatible with the ultimate definition of locality I arrive at in that chapter. I do not know whether similarly complex effects hold in Hebrew. For present purposes, the principal conclusion to be taken away is that the PD relation in German and the PD relation in Hebrew are subject to apparently similar locality constraints. I will return to a more nuanced discussion in chapter 6, when more of the analytical material has been developed. For now, since my proposal takes Landaus possessor raising analysis as a starting point, I will end this section by giving an overview of Landaus main points, with the goal of highlighting incompatibilities of his approach with the German PDC to be accounted for here.

73

3.3.4 Landaus possessor raising account Based on the characteristic properties of the PDC discussed above and drawing on possessor raising analyses that have been proposed for a variety of other constructions (Szabolcsi 1983, Kubo 1990, Ura 1996, Keach & Rochemont 1992), Landau (1999) proposes the case-driven movement analysis summarized in (22) and illustrated in (23). a. The possessor is generated in a caseless Spec position14 within the possessee. b. It is generated with dative case features. c. It then raises to check its case features with V. (Landau 1999: 9)

(22)

14

Landau (1999) does not discuss why this position should be caseless. As I explain in

section 3.4, a Minimalist framework allows for caseless Spec DP positions because the operation which creates the lexicon for a given language is not subject to well-formedness conditions. In other words, a lexicon containing Ds that lack case-licensing ability is not in any way ruled out. In fact, this is precisely where the distinction lies between languages that have the PDC and languages that do not. While the lexicon for German and Hebrew, for example, includes non-case-licensing Ds, the lexicon for English does not. 74

(23)

Possessor raising:
vP 3 DP v' Subject 3 V+v VP 3 DP V' Possessor 3 tv DP 3 tPD D' 3 D NP Possessee

This version of possessor raising works for both Hebrew and German in as far as the basic PDC properties of the two languages overlap. Since German and Hebrew seem to differ, however, as far as affectedness is concerned, Landaus analysis cannot be directly extended to German. The goal of this subsection is to show how the basic properties of the PDC discussed in 3.3.1-3 fall out from Landaus analysis but also how this analysis is at odds with the German affectedness condition. I give a preview here of how the dynamic structure-building system I appeal to in section 3.4 resolves these difficulties. In Landaus framework, the first basic property discussed above, namely the obligatory possessor interpretation of PDs, falls out from the definition of chains. Since PDs form a chain with their possessee-internal trace, and since a chain may only bear a single semantic role, PDs must bear the q-role they receive in their base
75

position. As the base position is Spec DP, i.e. the topmost subject position of the possessed DP, PDs must be assigned the role of possessor (or creator, in the case of Hebrew picture nouns), not the internal theme role, and not an agent role, which is assigned in Spec NP. Furthermore, the possessor role may not be overridden by an affectee role from the verb. Although the obligatory possessor interpretation also holds for the German PDC, Landaus chain-based explanation precludes an account of the fact that the German PDC must allow for the assignment of an affectee role in addition to the possessor role. The analysis I propose in section 3.4 allows for double q-role assignment while still ruling out PD-raising from a DP-internal (agent or theme) position. In a dynamic structure-building system, where movement proceeds via phase edges (in this case, Spec DP),15 and is driven by formal (including case, but not q-) features, the origin site of a PD is necessarily Spec DP. Any lower origin site would violate restrictions on movement operations. As for the second and third basic PDC properties, the c-command and the locality condition, the former straightforwardly falls out from Landaus version of possessor raising because PDs originate in the specifier of, that is higher than, the lexical shell of the possessee. Consistent with the facts discussed in subsection 3.3.2, the structure

15

See McCloskey 2000 and Svenonius 2004 for analyses which suggest that DPs are phases

in the sense of Chomsky 2000, 2001. A detailed discussion of the notion phase with respect to locality restriction on syntactic operations is provided in Part III of the dissertation (see in particular subsection 5.4.1 and section 6.2). 76

in (23) therefore forces the PD to c-command the possessee in its base position.16 The locality condition results from the fact that possessor raising is an instance of Amovement. Subject-containing categories like vP and DP prevent an argument from moving across their left edge to another A-position. In the case of vP, the PD argument is prevented from moving out of the clause, and in the case of complex DPs, it is unable to raise out of the possessee if it starts any lower than the specifier of the larger (containing) DP. The locality restriction correctly predicts the incompatibility of the PDC with process nominals, as described in subsection 3.3.1. If PDs must start out in Spec DP, and process nominals need an agent-subject which is thematically related to N and therefore occupies Spec NP, then a PD cannot be an appropriate subject for a process nominal. Note, however, that this is where Landaus determination of Spec DP as the base position of PDs involves some degree of stipulation. Although it is clear that PDs cannot move out of DP across an occupied subject position, nothing in

16

In a passive or unaccusative constellation (see (15)), the entire possessor-possessee

complex raises to Spec TP, the subject position of the sentence. As noted by Landau (1999), this leads to an unbound trace, a case of remnant movement (Mller 1996) where a constituent containing a trace moves outside the c-command domain of the antecedent of that trace (p. 12). A Hebrew example is shown in (i). (i) [IP [DP t1 ha-kovec ha-xadas]2 nimxak [VP le-Gil1 [v' tv t2]]]
the file the new was-erased to Gil

Gils new file was erased. (Landau 1999: 12) This is not a problem for the c-command condition since the now standard assumption is that the ccommand condition on movement is satisfied derivationally, not on post-movement structures. 77

Landaus version of possessor raising predicts that PDs could not start in Spec NP or as the complement of N and then move out via Spec DP if this specifier is unoccupied. Landau argues that the complement of N is the domain of dative case checking in Hebrew, and that a PD generated in Spec NP would have no need to move. However, if certain Ds can be caseless and thus force the argument in their specifier to check case elsewhere, why, in Landaus system, should there not be instances of caseless N, forcing agent arguments to raise for case-checking purposes? Again, a dynamic structure-building approach, which incorporates the requirement that movement be strictly formal-feature-driven and proceed via the closest phase edge, has the benefit of naturally ruling out PD-raising from Spec NP or the complement of N to the caseless Spec DP position. While N certainly is caseless in this framework, there is a non-stipulated reason for why a DP in Spec NP is unable to move out of DP, and that is the idea of DERIVATION BY PHASE (Chomsky 2001). In order to move or be accessible to positions higher in the syntactic object, an element needs to be at the edge of the minimal phase containing it. Even if this edge, here Spec DP, is unoccupied, the complement of N or an agent nominal in Spec NP cannot reach this position because in a PDC-context, D is defective and thus cannot attract a case-seeking nominal. Only a nominal that originates in Spec DP (i.e. the topmost specifier of the possessee) then has the chance of moving out of the DP-phase. As compared to non-movement alternatives, the main theoretical advantage of a possessor raising approach is that it only makes one PDC-particular claim: PDs are
78

generated in a caseless position. Everything else, the nature of PD-raising and the interpretive consequences, follows from general tenets of the respective theoretical framework (Landaus Principles & Parameters framework and the Minimalist system proposed here). The next subsection presents some specific evidence against control and binding accounts.

3.3.5 Arguments against control and binding Gurons (1985) control/PRO-analysis is inherently incompatible with (and thus relies on a PDC-particular stipulation in order to account for) the local nature of the PDC because the relation between PRO and its controller is not, in general, subject to the strict locality requirements that characterize A-movement. As shown by the French data in (24) and similar examples from German in (25), it is perfectly grammatical for a vP/TP boundary and more than one DP boundary to intervene between PRO and its controller. a. Jean1 a promis au directeur de [TP PRO1 venir la soire]. John promised the director to come to the party. b. Jean1 a [DP une liste de [DP livres PRO1 lire]]. John has a list of books to read. (Landau 1999: 12)
(25)

(24)

a. Jan1 hat dem Direktor versprochen [vP/TP PRO1 zur Party zu kommen].
Jan has the director promised to-the party to come

Jan promised the director to come to the party.

79

b. Jan1 machte [DP eine ganze Reihe von [DP Versuchen PRO1 den Artikel zu lesen]].
Jan made a whole sequence of tries the article to read

Jan made a whole bunch of attempts to read the article.

Both the possessor raising analysis and the standard non-movement alternatives have in common that there is an empty category inside the possessed DP. As for the type of empty category, however, Landau presents convincing evidence that it is a trace, rather than PRO or a null anaphor. This evidence stems from argument-adjunctsensitivity. While it is possible in Hebrew to extract a PD from locative and source PPs, typical adjunct PPs expressing, for example, cause and opposition block possessor raising out of the prepositional object. As shown by the examples in (26)(29), German is similar to Hebrew in this respect: the PDC works as expected when the possessed DP is associated with direction, locative, or source PPs that are selected (at least as optional arguments) by the verb, but a possessor relation is impossible to establish in the case of non-selected PPs headed by prepositions like wegen because of, trotz despite, and ohne without. In all of the following examples, the verb does not license a dative-marked argument independently of the possessor relation. a. Eine Katze kommt meinen Eltern nicht [ins Haus].
a cat comes my parents (DAT) not in-the house

(26)

A cat is not allowed in my parents house. (idiomatic) b. Das Kind legte sich dem Papa [auf den Bauch].
the child lay self the dad (DAT) on the belly

The child lay down on the dads belly.

80

(27)

a. Er stand der Braut [auf der Schleppe].

he stood the bride (DAT) on the train

He stood on the brides train. b. Das Buch lag Tim

the

[direkt vor

der Nase].

book lay Tim (DAT) directly in-front-of the nose

The book was lying directly in front of Tim (literally: in front of Tims nose). Er hat der Frau etwas [aus der Handtasche] genommen.
purse taken

(28)

he has the woman (DAT) something out the

He took something from the womans purse. a. *Tim musste seiner Schwester [wegen der Katze] aufrumen.
Tim had-to his sister (DAT) because-of the cat up-tidy

(29)

Tim had to clean up because of his sisters cat. b. *Tim a der Mama [trotz der Bitte] nicht auf.
Tim ate the mom (DAT) despite the plea not up

Tim didnt eat up despite moms plea. c. *Tim hat der Mama [ohne Geschirrsplmittel] abgewaschen.17
Tim has the mom (DAT) without dish-soap off-washed

Tim did the dishes without moms dish soap. In the unacceptable examples (29a-c), the ungrammaticality is two-fold. First, it is impossible to interpret the respective dative-marked nominal as possessor (of the cat,

17

Unlike in Hebrew, where, according to Landau (1999), INSTRUMENTAL PPs are compatible

with the PDC, example (29c) would also be bad if the preposition were mit with (cf. Landaus (35a)). 81

the plea, and the dish soap), and second, the verbs aufrumen clean up, aufessen eat up, and abwaschen do the dishes (in their unergative use, which lacks an overt direct object but implies it) cannot license a dative-marked nominal, at least not in the given context.18 Crucially, when used transitively, i.e. with a direct object instead of an adjunct PP, these same verbs do license the dative case on the extra nominal, allowing it to be interpreted as affectee. This is shown by the grammatical examples in (30). a. Tim musste Lena
Tim had-to

(30)

die Kche aufrumen.

Lena (DAT) the kitchen up-tidy

Tim had to clean up Lenas kitchen. b. Tim a Lena netterweise den Spinat auf.
the spinach up

Tim ate Lena (DAT) nicely

Tim was nice enough to eat Lenas spinach. c. Tim hat Lena den Teller abgewaschen.
off-washed

Tim has Lena (DAT) the plate

Tim rinsed off Lenas plate. The generalization is then that only argument PPs are compatible with PDCs. Assuming that arguments are transparent and adjuncts are opaque to extraction (Chomsky 1986, Huang 1982, Longobardi 1985), Landaus movement analysis
18

The examples in (29) could pass as ethical dative constructions but for that interpretation,

the preferred context is an imperative with a first-person pronoun as the PD, as in Iss mir jetzt erstmal brav auf! Be good and eat up for me! In this context, the verbs in (29a-c) then express affectedness, even when used unergatively. 82

makes perfect sense of this generalization. The empty category inside the possessee must be a trace, not PRO. It cannot be a null anaphor either because, while a binding analysis can account for the locality of the PDC, it has no explanation for the fact that the construction is not licensed across a PP-adjunct boundary. As shown in (31)-(32), neither control nor anaphoric binding is sensitive to the argument-adjunct distinction. Brittai hat sich [beim PROi Warmlaufen] vor dem Spiel verletzt.
Britta has self at-the warm-running before the game injured

(31)

Britta injured herself while warming up before the game.

(32)

a. Timi lsst Lena nicht [neben sichi] essen.

Tim lets Lena not next-to self eat

Tim doesnt let Lena eat next to him. b. Der Direktori lsst die Versammlung [ohne sichi] anfangen.
the director lets the gathering without self start

The director lets the gathering start without him. Both PRO and the anaphor sich can be coindexed with the respective matrix subject despite the intervening PP-adjunct boundary.19 In contrast, the (b)-examples of the minimal pairs in (33)-(34) confirm that the PD may not be separated from its associated possessee by boundaries of PP-adjuncts like the neben or ohne-PPs here.

19

In (32), the anaphor is even bound across a vP-boundary (see chapter 5 for an analysis of

binding in AcI-constructions). 83

(33)

a. Timi lsst Lena nicht [neben sichi] essen.

Tim lets Lena not next-to self eat

Tim doesnt let Lena eat next to him. b. *Tim musste Lena [neben dem Sessel] auframen.
Tim had-to Lena next-to the armchair up-tidy

Tim had to clean up next to Lenas armchair. a. Der Direktori lsst die Versammlung [ohne sichi] anfangen.
the director lets the gathering without self start

(34)

The director lets the gathering start without him. b. *Tim hat der Mama [ohne Geschirrsplmittel] abgewaschen.
Tim has the mom (DAT) without dish-soap off-washed

Tim did the dishes without moms dish soap. In short, non-movement accounts of the PDC simply do not have a natural explanation for restrictions on the construction. Not assuming movement of the PD from inside the possessee means that the PD should be licensed independently of the presence of a possessed DP. Restrictions on the cooccurrence of PDs with other nominals in the clause would thus be unexpected.20 Whether the possessee is embedded in an argument or adjunct, and, in fact, whether there is a possessee at all should not interfere with the PDC. It is clear, however, that the PDC is sensitive to
20

Hole (to appear-a), for example, who proposes a Kratzer-style voice account of dative

binding in German, where PDs are introduced by an Aff voice-head and bind a variable in the possessee, can only speculate on a generalization that explains why DPs should block possessor-related binding. (On Holes view not all DPs are opaque to the possessor relation between the non-core dative and the possessee.) 84

these factors. As discussed in subsection 3.2.2, dative-marked DPs that are not possessors are only allowed to occur if the verb selects a dative-marked complement or if the DP is what I call a non-possessor dative. Disregarding non-PDs for now (section 3.4.2 offers a more thorough discussion), example (35) is ungrammatical if the dative-marked DP does not cooccur with a possessee.21 Tim a seiner Schwester *(den Spinat).
Tim ate his sister (DAT) (the spinach)

(35)

Tim ate his sisters spinach. Summing up but also looking ahead to the following section, there are two basic choices for the analysis of PDCs: the binding/control approach and the movement approach. I follow Landau in arguing for the latter. The PD raises from Spec DP into a verbal argument slot. What Landau does not discuss at all is the connection between POSSESSOR and DATIVE embodied by the PD. Why should the possessor come with a dative case feature? This is where my proposal crucially differs from Landaus. The true puzzle of the German PDC is that the possessor receives a new qrole after raising. The PD is not only an argument of the possessee but also a malefactive or benefactive argument of the verb (i.e. is subject to the affectedness condition discussed in section 3.2). I will propose that the dative form of the PD and its interpretation as an affectee participant are linked in the way that such facts
21

The star in (35) indicates that the example is bad without the content of the parentheses.

Since it is impossible to translate the possessee-less version of this sentence into English, the given translation includes the parentheses. 85

usually are. Dative (in this use at least) is an inherent (or thematically-linked) case which is associated lexically with the assignment of a particular semantic role (affectee in the present case). Landau avoids the problem of double q-role

assignment, an apparent violation of the q-Criterion, by focusing on the Hebrew PDC, which is tolerated by a wider range of verbs than its German counterpart. In Landaus Hebrew examples, affectedness seems to be an implication of the given verbs in a particular context, which does not need to be encoded syntactically. Evidence from Spanish and French (see e.g. Kempchinsky 1992 and Guron 1985), where the PDC is as restricted as in German, suggests that Hebrew is untypical in this respect. In order to explain the more general case of the PDC, the task is then to find a theoretical framework that, under certain circumstances such as inherent casechecking, allows double q-role assignment. The goal of the following section is to integrate the basic ideas of Landaus possessor raising analysis into such a framework.

3.4

Possessor datives in a framework of dynamic structure-building

In this section, I continue to motivate (3.4.1) and then present in detail (3.4.2) a dynamic structure-building approach to account for the German PDC. I claim that this framework allows for double q-role assignment (3.4.4) and thus enables PDs to play the role of both possessor and affectee. This makes the current approach more attractive than traditional possessor raising analyses which have been criticized for
86

not taking the obligatory affectedness of the dative referent into account (see especially the criticism in Hole to appear-b and Pylkknen 2002). I will also make reference to non-PDC examples of double q-role assignment provided by Lpez (2001) and show that the system I propose has a better solution to the problem of overgeneration than Lpez account does.

3.4.1 The facts: Hebrew versus German If Landaus (1999) possessor raising account were applied to the German PDC without modification, the contrast between Hebrew and German apparent in (36)-(39) would be unexpected. Like examples (10a-e) in subsection 3.2.3, these example-pairs suggest that, unlike in German, the PDC in Hebrew requires no appeal to a syntactically encoded affectedness condition. A more concise (but still informal and descriptive) statement of what I mean by affectedness condition is given in (36).22 The proposal to be developed in this section aims to account for this condition on the German PDC.

22

Interestingly, the native speaker of Hebrew I asked to confirm Landaus judgments rejected

all the examples that do not satisfy the affectedness condition. This discrepancy may be due to speaker variation or contextualization effects. Landaus judgments could be based on possible but

pragmatically unusual situations. More native speakers need to be consulted to clarify this. If it turns out that the Hebrew PDC is subject to the same syntactically encoded affectedness condition as the German PDC, Landaus data truly represent a special case of the PDC, and his classic puzzle (see (41)) solves itself with respect to the more general case. The remaining question would be why the Hebrew dialect described by Landau (and also Pereltsvaig (2003)) should be so different. 87

(36)

PDC affectedness condition: In German, a PD is licensed only if the verb which takes the possessorpossessee complex as its complement can accommodate two internal arguments. One of these arguments must be assigned an affectee role. a. Gil lakax le-Rina et ha-tik.
Gil took to-Rina ACC the-bag

(37)

Gil took Rinas bag. (Landau 1999: 8) b. *Tim nahm Lena

Tim took

die Tasche.

Lena (DAT) the bag

Tim took Lenas bag. a. Gil histakel le-Rina al ha-bayit.

Gil looked-at to-Rina on the-house

(38)

Gil looked at Rinas house. (Landau 1999: 26) b. *Tim schaute Lena
Tim looked

das Haus an.

Lena (DAT) the house at

Tim looked at Lenas house. a. Gil maca le-Rina et ha-taba at.

Gil found to-Rina ACC the-ring

(39)

Gil found Rinas ring. (Landau 1999: 27) b. ?Tim fand Lena den Ring.

Tim found Lena (DAT) the ring

Tim found Lenas ring.

88

(40)

a. Gil caxak le-Rina ba-mitbax.

Gil laughed to-Rina in-the-kitchen

Gil laughed in Rinas kitchen. (Landau 1999: 28) b. *Tim lachte Lena in der Kche.

Tim laughed Lena (DAT) in the kitchen

Tim laughed in Lenas kitchen. The only difference between well-formed German PDCs and the (b)-examples in (37)-(40) appears to be that the verbs in the latter do not imply an obvious effect on the referent of the PD (here Lena). I thus conclude that the ungrammaticality of the German sentences is due to precisely this lack of PD affectedness, a violation of the affectedness condition. In order to improve the German PDCs, nehmen take could be changed to tragen carry; anschauen look at to anstreichen paint; finden find to verbaseln lose (colloq.) and lachen laugh to herumstehen stand around. All these changes implicate that the action the verb expresses must have immediately obvious physical or emotional consequences for the referent of the possessor which can be viewed either negatively or positively. This confirms that, besides being a possessor, PDs must play a malefactive or benefactive (affectee) role, at least in German. (Again, see also Wegener 1985, 1991, McIntyre to appear, and Hole to appear-a.)

89

At this point, two related challenges arise. The first is to understand the PDC affectedness condition (36) in a deeper way to derive it from independent principles rather than stipulate it. The second is to understand double q-role assignment, that is how to allow for a doubly q-marked DP without relying on dubious distinctions like primary vs. secondary (Guron 1985) or genuine vs. non-genuine (Kempchinsky 1992) q-roles. In a dynamic structure-building framework like Chomskys (1995, 2000, 2001) Minimalist Program, the Projection Principle and to an extent also the q-Criterion lose their relevance because the levels of deep and surface structure are eliminated. The crucial property of this framework is that heads with their selectional requirements are introduced in the course of the derivation. This opens the possibility that selectional features can in principle be satisfied either by drawing material from the lexicon (or numeration (see next subsection))23 or else by way of movement. I will show that a slight revision of Chomskys system makes it possible for an argument to first receive a semantic role in situ and then, under certain conditions, raise to a higher head, check its formal features and take on another semantic role. As will be explained in subsection 3.4.4, the two q-roles of the chain linking the initial position of the raised constituent to its landing site do not violate any principles that
23

Note that complex phrases do not come directly from the lexicon or numeration. The

terminology is misleading here. Drawing material from the numeration should be interpreted as parallel construction of a phrase in a so-called work-space followed by its merger with the topmost head in the larger syntactic object (sentence) that is being derived. 90

hold for the revised system. I therefore propose that a dynamic structure-building framework provides the answer to what Landau calls the classic puzzle of the PDC (mentioned in subsection 3.2.3 and restated here in (41)). An argument in the clause (the possessor) derives its semantic role from another argument (the possessee) but its syntactic behavior from the predicate. What is the possessor dative an argument of? (Landau 1999: 2) My claim is that, at least in German, a PD is both an argument of the possessee and of a verbal head that gets merged as the sister of the possessor-possessee complex. Since this means that PDs derive not only their syntactic behavior but also one of their semantic roles from the predicate, the German PDC (which is probably the general case) ceases to be anomalous. The actual task here is to find a theoretical framework that allows for a consistent account of the German facts without allowing for unwanted derivations. The following subsection explains how case-checking and crucially the double q-role assignment work in the framework I propose.

(41)

3.4.2 Possessor datives as both possessed and affected Assume that in the numeration24 that is the source of a particular instance of a PDC there is a nominal that gets introduced as the possessee. Assume further that the Dhead of the possessee comes without genitive or dative case licensing ability. This is

24

A numeration is a random array of feature-bundles selected from the lexicon, which is in

turn made up of all legal lexical and functional feature combinations existing in the language. 91

plausible because the content of a numeration is not controlled by well-formedness conditions. In fact, it is no surprise that the lexicon of languages like German and Hebrew, which have possessor datives (unlike the lexicon of English which does not), contains non-case-licensing (or defective) Ds. If the elements of a numeration are not combined correctly, i.e. if (some of) their features cannot be matched to form A GREE relations, the derivation crashes. If, on the other hand, an element like the defective D-head at issue here is matched with another element that makes up for the defect, the derivation can still converge and yield a grammatical sentence. Continuing with the derivation at hand, if a DP with a dative case feature is introduced as the specifier of the defective D, it can receive a q-role from the possessee,25 but its dative case feature must get checked by an element introduced later in the derivation. I propose that this element is a functional (or light) v-head (henceforth affectee v) which gets merged with the VP containing the possessor25

By saying that the possessor role is assigned by the possessee, I do not mean that it is the

lexical head of the possessed DP, namely N, that assigns the role. I follow Landau (1999) in assuming that, unlike process nominals (see subsections 3.3.1 and 3.3.4) and fear-type nouns whose subjects must be based in Spec NP because there is a clearly identifiable thematic relation between agent and process or experiencer and sensation, possessed nouns have their subjects in Spec DP. The possessorrole (which Kempchinsky (1992) appropriately calls a non-genuine q-role) must then be assigned by D. The nature of the actual thematic relation involved is context dependent. It seems plausible to assume that there is a semantic rule applying in the syntactic context in (i) which has the effect that the referent of DP1 stands in some relation R (specified by context) to the referent of DP0.
(i) DP0 / \ DP1 D' / \ D NP [poss]

92

possessee complex. While the lexical V has a theme role to assign, the affectee v comes with the need for an argument to which it can assign an affectee (male/benefactive) role. The DP sister of V, i.e. the possessor-possessee complex, fulfills the theme role. Then, if the numeration does not provide for another nominal suitable for assignment of the male/benefactive role, a previously introduced phrase can be raised from a position within the same syntactic object that contains the verbal head which assigns the male/benefactive role. In particular, this movement (or INTERNAL MERGE) operation targets the possessor DP (the PD) and raises it from the specifier position of the possessor-possessee complex to the specifier of the affectee vP. In its post-raising position, the PD both checks its dative case feature and receives the male/benefactive role. In other words, the affectee v assigns inherent dative case to the argument it attracts into its specifier. In contrast, the DP out of which the PD has moved (the possessee) enters into a non-movement Agree relation and checks structural accusative case with the topmost functional (light verb) head. As explained in chapter 2 in connection with Wurmbrands (2001) restructuring typology, this topmost verbal projection, headed by an agentive v, assigns a protoagent role to the (external) argument in its specifier.26 It is considered the

26

Following Hale & Keyser (1993), Chomsky (1995), and Kratzer (1996), I assume that

transitive/unergative verbs project a double-layer verbal structure with the outer shell being an agentive vP, while unaccusative verbs just consist of a bare VP (see chapter 5, section 5.3.1 for a more thorough discussion). Other v-projections may be added to these basic structures because, following Pylkknen (2002), I assume that a number of other thematic role-types, especially those involved in 93

instantiation of Burzios Generalization (Burzio 1986) in that it links the presence of an external argument to the assignment of accusative case. To illustrate all of this, the approximate configuration of a sample PDC, including PD-movement and q-role assignment, is diagrammed in (42). After completion of subject and verb

movement,27 this will yield Er ruinierte mir die Wohnung. He ruined my place.
(42) vP 3 DP v' Subject 3AGENT Er vP v 3 [ACC] DP v' ruinierte [DAT] 3MALE/BENEFACTIVE mir VP v 3 <arg> V' [DAT] 3 tv DP V THEME 3 <arg> tPD D' tv POSSESSOR 3 D NP [] | [ACC] N die Wohnung

applicative alternations, are also introduced by light verbs; hence my proposal of PDs being licensed by affectee vPs (see subsection 3.4.3 for a critique of Pylkknens analysis of the German PDC).
27

V ERB-SECOND word order will be spelled out correctly after the verb (which has already

raised via head movement from V to the highest v) moves from v to T to C, and the subject moves via Spec TP to Spec CP (see Vikner 1995 but Haider 1993 for a different view). 94

The affectedness condition and thus the ungrammaticality of the German examples in (37)-(40) fall out from this analysis in the following way. If the lexical verb does not project an affectee v, which needs an argument in its specifier to which it can assign its q-role, the PD will not find a head to check dative case with. This follows from the standard assumption that checking of inherent case goes hand in hand with the assignment of a designated q-role. Without an affectee v, the PD then ends up in situ with an unchecked [DAT]-feature (or, taking a slightly different view, an unvalued case feature), and the derivation crashes. In (37b)-(40b), then, the source of ungrammaticality is the absence of that crucial light verb projection. Without assignment of an affectee role, i.e. satisfaction of the affectedness condition, dative case cannot be licensed. If the verbal shell does include an affectee v and thus provides a dative case licensor, and if, in addition, the numeration happens to provide the elements needed for another DP, the raising operation is blocked because introducing an element from outside the syntactic object that is being built (EXTERNAL MERGE) is preferred over the more complex operation involving movement of an element from inside this object (INTERNAL MERGE). The additional DP then gets introduced directly into the specifier of the affectee vP, and no possessor raising occurs. In this case, the derivation is successful if the possessor DP comes with a genitive case feature that can be licensed by D. The result is a non-PD (ethical, estimative, or ficiary dative), as shown, for example, in (43a). Based on the data presented in subsection 3.2.2, it has
95

already been established that a non-PD construction does not depend on the presence of a possessee (another example confirming this is given in (43b)). Now we see why this is so. In a case like (43a), the specifier of the possessor-possessee complex is filled with a genitive possessor. The dative-marked nominal never started out in this position and was thus never assigned the possessor role. Instead, it got externally merged into the affectee argument position.28 There is no raising involved here. a. Du hast der Mama gemacht.
made
29

(43)

doch hoffentlich nicht Omas

hopefully not

Geschirr kaputt
broken

you have the mom (DAT) but

grandmas (GEN) dishes

I sure hope you did not ruin Grandmas dishes on mom.

28

Note that case-checking with an externally merged argument goes against Chomskys

(1995) early assumptions concerning the non-interaction of Checking and q-Theory. It is clear that inherent case checking represents a necessary area of checking and q-overlap. This will be discussed in subsection 3.4.3.
29

In contrast to (43a), where the possessor is not forced to and therefore does not move,

possessor raising is compatible with a genitive construction if the genitive is expressed as a PPcomplement to the possessee. A Hebrew example is given in (i). (i) Gil s&avar le-Rina et ha-mis&kafayim s&el Sigal.
Gil broke to-Rina ACC the-glasses of Sigal

Gil broke Sigals glasses on Rina. (Landau 1999: p. 7 (9)) Here Spec DP is available for the PD, and the possessee must be interpreted as possessed by two different individuals. As Landau puts it, Rina has transitory possession of the glasses. The German equivalent of Landaus Hebrew example is given in (ii). The Hebrew genitive construction is translated as a von (of)-PP. (ii) Tim hat Lena die Brille von Silke zerbrochen.
Silke broken Tim has Lena (DAT) the glasses of

Tim broke Silkes glasses on Lena. 96

b. Die Kinder schlafen mir

the children sleep

nicht schnell genug ein.

fast enough in

me (DAT) not

The children dont fall asleep fast enough for me. The mechanisms of dative case checking, inherently connected with the assignment of an affectee role, in the context of both PDCs and non-possessor-related dative constructions30 are thus accounted for by the analysis I sketched in (42). As for the nature of the affectee light verb projection, my proposal is compatible with the assignment of a male/benefactive role in applicative contexts (see e.g. Baker to

30

As for datives that are neither PDs nor non-PDs, i.e. core dative arguments that are selected

by the verb (e.g. by verbs like helfen help, gratulieren congratulate, gefallen please, fehlen lack), Maldonado (2002) argues that the dative-marked participant is always affected, either positively or negatively. If all instances of dative case-checking are indeed tied to a particular type of q-role and thus inherent, there is a natural explanation for the incompatibility of the PDC with dative verbs (see examples above) shown in (i) and (ii). (i) *Ich habe unseren Nachbarn der Tocher
I have our neighbors (DAT)

geholfen.

the daughter (DAT) helped

I helped our neighbors daughter. (ii) *Er hat seiner Freundin dem Vater gratuliert.
he has his girlfriend (DAT) the father (DAT) congratulated

He congratulated his girlfriends father. Assuming there can be maximally one affectee v-head per clause, the affectee role can only be assigned to either the PD or the core dative argument, and consequently, only one of the two can check its dative case feature. In German, two datives can marginally cooccur only if one is the first person pronoun mir in an ethical dative construction. This makes sense with respect to the Hebrew facts. As noted by an anonymous JCGL reviewer, PD-raising out of a possessee that is itself a dative is possible if the PD is a clitic that does not absorb the case of the predicate. It could be that German mir in examples like Mach mir der Lena bitte nicht die Brille kaputt Please do me a favor and dont break Lenas glasses behaves like a clitic. A more thorough investigation of the connection between PDs and core datives is certainly worthwhile but is not my focus here. (See Wegener 1985, 1991 for a discussion of the cooccurrence of different datives.) 97

appear and Pylkknen 2002). In many languages, the male/benefactive aspect comes in the form of a functional marker, a morpheme. The lexical verb is argued to incorporate the applicative morpheme by raising to the light verb head. To summarize, the general case of a PDC arises, on this view, from the presence in a given lexicon of two elements: (i) an instance of D which defines the semantics of possession but happens to lack the ability to case-license the possessor, and (ii) a light verb head which assigns an affectee role and has the means to case-license the argument that receives this role.31 The problem posed by (i) is how to case-license the possessor. In effect, there is one more DP to be licensed than there are heads to license DPs. The solution to this problem is partially provided by (ii). The affectee v has an extra case feature to check. In order to take advantage of this, however, the other part of the solution is the relatively costly one of using a movement operation. The ultimate outcome is a doubly q-marked DP. Subsection 3.4.4 deals with the legitimacy of double q-role assignment and addresses the question of how to prevent overgeneration of Internal Merge structures. First, however, in 3.4.3, we will take a

31

Since the cooccurrence of these two elements in a numeration is random, we expect

affectee light verbs to show up independently of defective Ds. This expectation is borne out in the case of non-PD constructions which can but do not need to coincide with a possessed nominal. One of the examples I give in subsection 3.3.5, however, seems to contradict the independence of affectee light verbs and the PDC. In (35), Tim a seiner Schwester den Spinat Tim ate his sisters spinach, the possessee den Spinat cannot be left out, i.e. Tim cannot simply eat for his sisters benefit. This goes back to the aspectual restriction that the predicate in non-PD constructions must express a result or imply a consequence which obviously affects the person referred to by the PD. 98

closer look at Pylkknens (2002) analysis of the PDC as a low applicative construction which avoids the problem of double q-assignment by maintaining that PDs are like the dative in English DOUBLE OBJECT CONSTRUCTIONS and that there is no direct relation between the verb and the dative nominal at all.

3.4.3 Arguments against a low applicative account In Pylkknens (2002) typology of non-core argument introducers, there are three types of applicative heads. Pylkknen labels them HIGH, LOW RECIPIENT, and LOW S OURCE applicatives. A high applicative construction, as illustrated by the Chaga benefactive example in (44), expresses a thematic relation between (the referent of) the applied argument and the event described by the verb. The applicative head is merged above the verb and thus part of the verbal shell.
(44) VoiceP 3 he 3 Voice 3 wife 3 APPLBEN 3 eat food

He is eating food for his wife. [Chaga] (Pylkknen 2002: 19) VoiceP hosts the external (proto-agent) argument, the subject he, and the applicative (APPLBEN)-head introduces the dative benefactive argument, the wife. As the English

99

gloss indicates, the eating event happens for the benefit of the wife. In contrast to high applicative constructions like (44), low applicatives express a transfer of possession relation between two individuals, namely the direct and the indirect (applicative) object. Crucially, since the applicative head is merged below the verb, there is no direct relation between the applicative argument and the verb. In low recipient applicative constructions, like the English Double Object Construction (DOC) in (44), the assertion is that the direct object is to the possession of the indirect object.
(45) VoiceP 3 I 3 Voice 3 bake 3 John 3 APPL cake

I baked a cake and the cake was to the possession of John. [English] (Pylkknen 2002: 19) Here, the verb expresses a cake-baking event, and the applicative object is the recipient of the cake. The only difference between low recipient and low source applicative constructions is the direction of transfer of possession. In the case of a source applicative, the direct object is from the possession of the indirect object. This from-transfer of possession is precisely the relation Pylkknen claims holds between

100

the direct object and the dative non-core argument in PDCs. Her structural analysis of the Korean PDC is illustrated in (46).
(46) VoiceP 3 thief 3 (NOM) Voice 3 3 steal Mary 3 (DAT) APPL ring
(ACC)

The thief stole a ring and it was from Marys possession. [Korean] (structure corresponds to Pylkknen 2002: 21) While a PDC account based on transfer of possession works for this Korean example and can probably be extended to a number of others, it makes incorrect predictions about the German PDC. The following are three arguments against Pylkknens account of the German PDC as a low applicative construction; the first two take issue with the transfer of possession relation, and the third is based on the incompatibility of low applicatives with PP-internal DPs. One of Pylkknens diagnostics for the high-low distinction between types of applicatives appeals to verb semantics. Since the event in a low (as opposed to a high) applicative construction must result in transfer of possession, the verb cannot be stative. In the case of the English DOC, the diagnostic leads to the desired result. The stative verb hold in (47), for example, is incompatible with a DOC and thus correctly classifies the English DOC as low.
101

(47)

*I held him the bag. (Pylkknen 2002: 24)

When applied to German, however, the diagnostic does not support Pylkknens claim. The fact that (48) with the stative verb halten (the German equivalent of hold) is grammatical indicates that the transfer of possession relation characteristic of low applicatives mischaracterizes the German PDC. The diagnostic classifies the German PDC as high. 3Ich habe ihm
I

(48)

die Tasche gehalten

held

have him (DAT) the bag

I held his bag for him. Pylkknen also appeals to the transfer of possession relation between the applicative argument and the direct object to account for the obligatory possessor relation which holds between the PD and the possessed nominal (see subsection 3.3.1). In

Pylkknens system, this possessor relation must always coincide with a possession transfer. The notion of possession transfer is compatible with the obvious loss of possession expressed by the Korean example in (46) the ring was taken away from Mary. It can also be extended to account for the Finnish examples in (49). In (49a), it is possible to interpret the event of Riikkas seeing the undershirt as a loss of privacy for Sanna, the possessor of the shirt. Pylkknen argues that the transfer of possession relation is reflected in the privacy of the undershirt being taken from Sanna. In contrast, in (49b), where the direct object is an overcoat and thus publicly
102

possessed, no loss of privacy is involved, and hence, no transfer of possession has taken place. As desired, Riikkas seeing Sannas overcoat cannot be expressed as a PDC; (49b) is judged to be awkward. a. Riikka nki Sanna-lta aluspaida-n.

(49)

Riikka.NOM saw Sanna I-ABL undershirt-ACC

Riikka saw Sannas undershirt. (Lit: Riikka saw an undershirt from Sanna.) b. #Riikka nki Sanna-lta pllystaki-n.

Riikka.NOM saw Sanna I-ABL overcoat-ACC

Riikka saw Sannas overcoat. (Lit: Riikka saw an overcoat from Sanna.) (Pylkknen 2002: 47) The abstract notion of privacy loss as an instance of possession transfer, however, does not work for German examples like (50a-b). The car in (a) is publicly possessed, yet compatible with a PDC, and the husbands massaging his wifes back in (b) is surely not an event that should be associated with a loss of privacy. a. Tim hat der Nachbarin gestern das Auto gewaschen.
washed

(50)

Tim has the neighbor (DAT, FEM) yesterday the car

Tim washed the neighbors car for her yesterday. b. Ein guter Ehemann massiert seiner Frau jeden Abend den Rcken.
a good husband massages his wife (DAT) each evening the back

A good husband massages his wifes back every night.

103

Pylkknens notion of transfer of possession relation then does not fully capture the German PDC. It seems that transfer of possession is a possibility, not a requirement, and transfer of possession is not at all representative of PD-affectedness. As thoroughly discussed in section 3.2, a PDC (in any language) implies that the PD is somehow affected by the event the verb expresses, and I have shown for German that this PD-affectedness condition must be syntactically encoded. There must then be a direct (q-)relation between the PD and the verb. This is the case in high, not low applicative constructions. Another argument against the PDC as a low applicative construction is that, in Pylkknens system, a low applicative cannot stand in a relation to a DP that is embedded in a PP. This is illustrated by the ungrammatical English DOC in (51). *John sat Mary in the car. (Intended meaning: John sat in a car which was to Marys possession.) (Pylkknen 2002: 56) Based on the ungrammaticality of (51), Pylkknen claims that the Hebrew example in (52), which looks like a PDC with a PP-embedded possessee, is really a different type of construction. Gil gar le-Rina ba-xacer.
Gil lives to-Rina in-the-yard

(51)

(52)

He lives in Rinas yard. (Landau 1999: 4)

104

The dative, le-Rina, here is argued to be introduced by the verb be, as in simple possessor constructions like Jon has a son, literally (there) is to Jon a son. This line of argumentation leads to the prediction that only languages, like Hebrew, which have be-possessor constructions, i.e. use be, not have, in simple possessor constructions, have PDC-look-alikes like (52), where it is possible for the possessee to be embedded in a PP. Standard German does not have be-possessor constructions32 and is thus predicted to be incompatible with a PP-embedded possessee. This is counter to fact, however. The examples in (53) are perfectly grammatical.

(53)

a. Er stand ihr

[auf dem Fu].

the foot

he stood her (DAT) on

He stood on her foot. b. Das Kind legte sich dem Papa [auf den Bauch].
the child lay self the dad (DAT) on the stomach

The child lay down on dads stomach. c. Eine Katze kommt meinen Eltern nicht [ins
a cat comes my parents (DAT) not

Haus].

into-the house

A cat is not allowed in my parents house. The one German example provided by Pylkknen, given here in (54), is ungrammatical for independent reasons.

32

Utterances like Das Buch ist mir (DAT) The book is mine, which are common in some

dialects of German, are completely unacceptable in Standard German. 105

(54)

*Hans hat ihm

in seinem Hof gewohnt.

yard lived

Hans has him (DAT) in his

Hans had him living in his yard. (Pylkknen 2002: 57) The cooccurrence of the dative ihm with the possessive pronoun in the specifier of the possessed nominal indicates that ihm must be a non-PD (possibly a maleficiary), but regardless of whether the dative is a PD or non-PD, the verb wohnen live, unlike its Hebrew equivalent (see (52)), does not imply affectedness and is thus incompatible with the dative-case-licensing affectee vP projection. All the German facts discussed here are correctly accounted for if the PDC is analyzed as a high applicative construction, the possessive aspect of its meaning derived by possessor raising rather than a transfer of possession relation. A possessor raising, i.e. A-movement, account of the PDC makes the right prediction with respect to examples like (53). PD-movement out of PP is expected to be grammatical, as long as the PP is an argument. There must then be a distinction between genuine low applicatives, like the English DOC, which really do involve transfer of possession, and high applicatives in which the applicative argument happens to originate as a possessor. This distinction eliminates the dubious notion of privacy loss and avoids incorrect predictions about the German PDC, while maintaining the core of Pylkknens otherwise attractive typology.

106

Having established now that an accurate analysis of the German PDC must provide an account for both the possessor relation between the PD and the other nominal involved and the direct thematic relation between the PD and the verb, double qassignment is inevitable, and the following subsection will show that this is not necessarily a problem.

3.4.4 Double q-assignment According to Landaus (1999) possessor raising analysis, PDs get their q-role from the possessee and check dative case with V after raising to Spec VP. q-assignment and feature-checking thus happen in different positions. This is in fact consistent with Chomskys (1995) claim that the modules of feature-checking (Checking Theory) and q-role assignment (q-Theory) are complementary. There should be no interaction between q-theory and the theory of movement. q-roles are not formal features in the relevant sense; typically they are assigned in the internal domain, not the checking domain, and they differ from the features that enter into the theory of movement in numerous other respects. q-relatedness is a base property, complementary to featurechecking, which is a property of movement (Chomsky 1995: 312-313). While I agree that movement should be driven exclusively by formal features (case and agreement), not q-roles, the analysis I am developing is incompatible with the assumption that the domains of feature checking and q-assignment cannot overlap.

107

One obvious counterexample to this assumption is inherent case-checking. When case-marking is inherent, i.e. q-related, there must be overlap of the two domains.33 In addition to inherent case-checking, the counterexample I am concerned with here involves the assignment of a q-role to an argument due to movement, and this is an even more serious offence (see also Rizzi to appear). Although the movement is still case-driven and therefore in accordance with Chomskys assumptions, my proposal is clearly incompatible with the claim that a raised element cannot receive a q-role (Chomsky 1995:113). The motivation for ruling out q-assignment after movement is to prevent illicit double q-marking as in (55) (Chomskys (113)), where the possibility of receiving a q-role after movement seems to allow for these derivations to converge, although the verbs HIT and BELIEVE, which share the q-structure of the actual hit and believe but lack case features, should not exist. a. John [VP t' [HIT t]] b. John [VP t' [BELIEVE [ t to be intelligent]]] (Chomsky 1995: 313) The actual verbs hit and believe would cause these derivations to crash because they (or, to be more exact, their agentive v-projections) have accusative case-licensing ability and must therefore eliminate their case-features by checking them with John, causing John to be inactive and thus frozen in place. The derivations with the impossible verbs HIT and BELIEVE seem to converge, however, because John is
33

(55)

I assume here that inherent case is like structural case in requiring checking in the syntax. 108

able to pick up both the theme and the agent role and then move on to Spec TP to check case and agreement features. Immediately after bringing up the configurations in (55), Chomsky actually solves the problem of unwanted convergence without having to appeal to checking and q-complementarity. Surely no strong feature of the target is checked by raising to the [Spec, HIT] position, so overt raising is barred; in fact, no checking relation is established. The only possibility is direct raising to [Spec, I]. The resulting sentences John HIT and John BELIEVES to be intelligent are therefore deviant, lacking the external argument required by the verb (Chomsky 1995: 313). The fact that the first move of John to Spec HIT or, to make this compatible with (little) v-projections, to the specifier of the agentive v, is illicit because it is not driven by formal features this specifier is not a case-position is enough to rule out these unwanted configurations. In (55), double q-assignment is thus illicit, but if the only well-formedness requirement is that movement be driven strictly by formal features, assignment of a second q-role is not banned when it coincides with feature-checking. This latter scenario is precisely of the type I am dealing with here. In PDCs, the PD raises to the specifier of the affectee v, not for q-purposes but because the v-head is an active probe (attractor) and the head of the PD is an active goal, both in need of checking their uninterpretable case-feature. The PD receives the affectee role as a by-

109

product of the feature-driven movement.34 The restriction given in (56) should therefore be sufficient to prevent overgeneration of double q-assignment constellations. Restriction on Internal Merge: Movement is driven by formal features, i.e. a legal operation only if, as a direct result, formal features of the goal and probe are satisfied.35 But this, in turn, is just the requirement of LAST RESORT, which has been a feature of the Minimalist Program, in one form or another, from the very start. It means that a DP can be an active goal for movement only if it has an uninterpretable formal
34

(56)

Note that, since it is driven by formal features (case), PD-raising is an instance of A-

movement. In order for this move to be legal, the origin site of the PD, the Spec of the possessed nominal, must itself be an A, not an A-bar position. As pointed out to me by Marcel den Dikken (p.c.), the analysis is then incompatible with Kaynes (1993) account, which, following Szabolcsi (1981, 1983), draws a parallel between DP and CP and therefore takes Spec DP to be an A-bar position. To make DP-movement out of a possessed nominal via the Spec of this nominal to an A-position legal, Kayne proposes that Spec DP of a possessed nominal becomes an A-position via incorporation of D into the verbal head be. The derived Spec of D+ BE counts as an A-position. Since nothing in my account hinges on have being derived from D+BE (see Kaynes analysis), I will not follow Kayne and Szabolcsi in assuming that Spec DP is an A-bar position. No incorporation is needed then.
35

Satisfaction or checking of formal features does not necessarily mean feature deletion.

Interpretable features, like [f] (agreement) or [WH ] on a DP, are involved in the matching relation (Agree), get checked, but do not get deleted as a result of the movement. Thus, in cases of unbounded or successive-cyclic movement like Wh-movement or Subj-to-Subj raising, movement to intermediate specifier positions results in checking of interpretable features on the goal but does not inactivate it because its uninterpretable (case or [Q]) feature has yet to be deleted. The goal establishes as many [f] or [WH] Agree relations with intermediate probes as it needs to reach a position where its uninterpretable feature can be deleted. 110

feature, e.g. case, and if there is no other way to satisfy the matching formal features of the probe. Besides preventing configurations like (55) from converging, this movement restriction also rules out illicit possessor raising constructions like the German example in (57a) with the meaning given in (b).
(57) a. *Chrisi bewunderte [DP ti den Bruder]
Chris admired the brother

b. Chris bewunderte seinen Bruder

Chris admired his brother

Chris admired his brother. Here the possessor has been moved into external argument (subject) position. Without the restriction in (56), we have the following scenario: If the DP Chris happens to come with a nominative case feature which it cannot check in Spec DP and there is no other nominal to fill the subject position of the sentence, Chris will receive the possessor role in situ, raise to the specifier of the agentive vP, take on the proto-agent role, and end up in Spec TP to check its nominative case and agreement features. Convergence of this derivation is an undesired result because (57a) does not force the possessor reading given in (b). It is not clear whose brother is being admired. With the movement restriction in place, however, this type of nominativemarked possessor constellation has no chance to converge. As already discussed in connection with Chomskys examples in (55), in order to get the proto-agent q-role, the PD Chris has to move to the specifier of the agentive vP, a non-case position. It is

111

commonly assumed that the first specifier of an agentive v is never a position in which formal features are checked.36 Thus, with the case feature of the PD not being checked as a result of this move, the operation is illegal. Direct movement of the PD to Spec TP would allow for nominative case checking but would prevent the agentive v from assigning its q-role. In fact, before T can even be introduced, the selectional requirements of v must be met. Besides entering into an accusative case-checking Agree relation with a DP in the structure that has already been built, the agentive v must be supplied with an argument in its specifier to which it can assign its q-role. In a system where Internal Merge is strictly formal-feature-driven, only a DP newly (i.e. externally) assembled from elements in the numeration can be merged into this specifier. Once this happens, direct movement of the PD to Spec TP is impossible because T will find the newly introduced DP in Spec vP to check nominative case with. T cannot look further down to find the PD. The derivation fails as desired.37

36

Note that it is a slight simplification to say that Spec vP is not a case position. There are

subjects that are generated in Spec vP and stay there because they are not forced to move by an EPP feature but enter instead into a static Agree relation with T (see Wurmbrand 2004). Spec vP then is a case position, but only when higher structure, including the nominative case-checking head, has been built. Since no higher structure has been built at the point when the agentive v looks to assign its qrole, v cannot attract an argument to move into its specifier. It should also be noted that Ura (1996) has argued for v being able to assign ergative case.
37

It is worth pointing out that the restriction in (56) rules out movement of the possessor not

only into subject position but also into the position of the direct object. Since the agentive, accusativecase-checking v has not yet been merged in at the point when the verb looks to assign its internal argument role, the direct object position is not a case position. Despite the occurrence of examples like 112

To reiterate, the restriction in (56) does not rule out PD-raising. Since PDs move for inherent dative case-checking with the affectee v-head, the operation is legal. Having given up on strict checking and q-complementarity, nothing prevents the affectee v from assigning its male/benefactive role to the argument that has been merged into its specifier. Furthermore, as hinted at in section 3.3, allowing movement only for formal-feature-checking provides a natural explanation for why the PD must originate in Spec DP, as opposed to Spec NP or the complement of N. If DP is a phase (and movement must proceed via phase-edges), movement to Spec DP from either Spec NP or the complement of N is illegal because, being headed by a defective D, Spec DP is not a case position. It is thus possible to account for the cooccurrence restrictions on the PDC with process nominals (see subsection 3.3.1) without having to stipulate that the PD is based in Spec DP (though see chapter 6 for a characterization of phase which may complicate this matter).
(i) as apparent PDC-alternatives (see also Hole 2005), I maintain that (56) makes the right prediction here. (i) Er trat (ii) Er trat mich mir in den Magen. in den Magen.
he kicked me (ACC) in the gut he kicked me (DAT) in the gut

In (i), the accusative mich can be argued to not have originated in the specifier of the Magen-DP. The accusative is then an ordinary direct object, and the PP in den Magen is an adjunct. This is in line with Hole 2005. Use of the accusative, i.e. the construction in (i), expresses a different kind of affectedness. Unlike in (ii), the body part in (i) represents the whole person. It is the referent of the accusative pronoun which plays the internal argument (patient) role, not the particular body part added by the PP. Hence, constructions like (i) do not involve possessor raising. (Note, however, that, on this view, the verb schieen shoot, as in Er hat ihn in den Rcken geschossen He shot him in the back has to be treated as an exception because it is not grammatical with just a direct object. It also selects a PP.) 113

Other counterexamples to Chomskys strict separation of Checking (i.e. Movement) and q-Theory are provided by Lpez (2001). He reports on cases of double qassignment in Spanish and English. In the Spanish causative construction in (58) (Lpez (21)), for instance, the agent argument of the verb in the embedded clause is argued to additionally receive a q-role from the matrix verb. Yo le hice reparar mi coche a mi mecnico favorito
my car to my mechanic favorite (DAT)

(58)

I cl. (DAT) made repair

I made my favorite mechanic repair my car. (Lpez 2001: 705) According to Lpez, the causee mi mecnico favorito is both the agent of reparar and the affectee of hice make. The causee thus receives a second q-role after being copied and raised into the position where it surfaces as the dative clitic le. Lpez English examples are ECM constructions like (59a). Since, in Lpez system, accusative case is checked by the lexical V-head, the matrix Spec VP in an ECM construction is a position that allows both checking of accusative case and assignment of a q-role. a. Sue estimated Bills weight to be 150 lbs. (cf. Sue estimated Bills weight.) b. *Sue estimated Bill to weigh 150 lbs. (cf. *Sue estimated Bill.) (Lpez 2001: 703)

(59)

114

The ungrammaticality of (59b) is supposed to show that the verb estimate imposes thematic restrictions on the raised ECM-subject. Bills weight then not only gets a qrole from the head of the embedded VP be 150 lbs but also from the matrix verb estimate. Since the judgments given in (59) are extremely subtle, it is not clear that the ECM-subject really does receive a second q-role. According to Chomskys (1995) account of ECM-constructions, the ECM-subject does not raise overtly. Still, Lpez has a point in noting that at least for the Spanish causative example in (58), there must be a way to allow for case-checking and q-assignment overlap. In order to prevent instances of illicit double q-role assignment, as in *Chris saw me hit, meaning Chris saw me hit myself (see (60)), where the embedded direct object me receives both the internal and the external q-role of the embedded verb hit, Lpez has to appeal to his somewhat controversial move of taking accusative case-checking ability away from the external-argument-introducing v. He claims the structure in (60a) cannot result because the object is never in the checking domain of the embedded external q-role assigner v. a. *Chris saw [vP me [v v + hit] [VP tv tobj ]] b. Chris saw me hit myself. (Lpez 2001: 698) With V being both q-assigner and accusative case checker, the ECM-subject me does not enter into a potentially pied-piping (or movement-triggering) Agree relation with
115

(60)

v, i.e. it is not in a position to raise to Spec vP. The ECM-subject stays within VP and is thus not in danger of receiving the external q-role from v. The controversial side effect of this solution to the problem of illicit double q-role assignment is that the function of the agentive v is reduced to the introduction of the external argument. This means that agentive v s are no longer the formal element capturing Burzios Generalization: if v is absent, the possibility of having both structural accusative case and an external argument will be absent too (see Lpez section 4 for a discussion of this consequence). Within the system I develop here, Lpez problem case in (58) is taken care of without having to give up Burzios Generalization. The movement restriction in (56) makes raising of the ECM-subject me to Spec vP an illegal operation because, again, the first specifier of this vP is not a case position. No formal features are checked as a result of this move. Note also that Lpez does not address how his revision of Chomskys system would handle the type of overgeneration scenario I show in (57). Any dynamic structure-building framework that does not stipulate strict complementarity of Checking and q-Theory, i.e. a framework that allows q-assignment after movement, must find a different way of preventing the operation Internal Merge from raising an argument via the specifier of the agentive vP (picking up the external q-role) to Spec TP (checking nominative case). In sum, I have shown that there is reason to believe that the modules of Checking and q-Theory cannot be strictly complementary. PD constructions in German (and
116

arguably in many other languages as well) are a prime example of doubly q-marked DPs. Lpez Spanish causative example and (less convincingly) his English ECM example also suggest that a revision of Chomskys (1995, 2000, 2001) system is called for. Overgeneration of double q-role assignment can be avoided by requiring that movement be driven exclusively by formal features, never by thematic needs.

3.5

Residual issues

Since random numerations, which can contain elements made up of any combination of features allowed in the language, are a core characteristic of a Minimalist system, the only way to prevent ungrammatical derivations from converging is through restrictions on the structure-building process. Operations like External and Internal Merge must be designed to disallow illicit combinations of feature bundles from being spelled out. As explained in the previous section, the movement restriction given in (56) successfully rules out overgeneration cases like Chomskys (55), my (57), and Lpez (60). A PD cannot accidentally raise to the highest Spec vP projection and receive a proto-agent role. There may, however, be other

overgeneration scenarios that the system developed thus far does not prevent. One such scenario is the following. Say the possessee DP comes with a dative and the possessor DP with an accusative case feature. Nothing prevents the entire possessorpossessee complex from raising into the specifier of the affectee v to check its dative case feature and the possessor DP from entering into a static (non-movement) Agree
117

relation with the higher, accusative-case-checking v. scenario is the ungrammatical example in (61).

A possible result of this

(61)

*Der Einbrecher ruinierte meine Mutter dem Haus.

the burglar ruined my mother (ACC) the house (DAT)

The tree structure in (62) shows this problem-case before v-to-T-to-C movement and raising of the subject from Spec vP via Spec TP to Spec CP (i.e. the movements resulting in VERB-SECOND word order) have taken place.
(62) vP 3 DP1 v' AGENT Der Einbrecher 3 vP v [ACC] DP2 ruinierte AFFECTEE [DAT] 3 v' DP3 POSS D' 3 [ACC] 3 VP v meine D NP g [DAT] Mutter dem Haus V' tv 3 tDP2 V tv

Unless we can straightforwardly rule out this constellation because of a semantic feature clash between the malefactive aspect and the inanimate house, there is a puzzle to be solved here. If it is at all possible to imagine a house with feelings which is negatively affected by the burglars ruining the mother, there must be a structural reason that renders this sentence unacceptable: word order. If the accusative
118

argument is not a pronominal (pronominals occur high in the structure and thus precede full DPs), the dative argument (the PD) must precede the accusative one (the possessee). This is confirmed by the examples in (63). a. ?*Mein Bruder hat das Auto
my

(63)

der Mami

zu Schrott gefahren
driven

brother has the car (ACC) the mom (DAT) to scrap

My brother totaled moms car. b. ?Mein Bruder hat das Auto dem PAPI, nicht der MAMI, zu Schrott gefahren
my brother has the car (ACC) the dad (DAT) not the mom (DAT) to scrap driven

My brother totaled DADs, not MOMs car.

Without special emphasis, (63a) is clearly degraded, and even when dem Papi is contrasted with der Mami, as in (63b), which seems at least better than (a), it is not clear that there is a possessor relation between das Auto and the respective parent. If this sentence is acceptable at all, it is probably a scrambled non-PD (here maleficiary dative) construction, not a PDC. The possessor raising configuration in (62) must then clearly be kept from converging, and since raising of the entire possessorpossessee complex happens for case reasons, the movement restriction in (56) is not sufficient here. Economy considerations might provide at least a temporary solution. One could argue that raising of the bigger DP, the possessee DP, which has the smaller possessor DP in its specifier, is an illegal operation because it would be more economical to just move the smaller DP. In order for this to work, the probe (the affectee v) has to look for any DP that is active, i.e. whose case-feature needs to be

119

valued.38 If given a choice between two equidistant DPs,39 v must attract the smaller one. Since, in (62), it is the bigger DP that gets raised to the specifier of the affectee v, the derivation is deviant. The economy principle of moving a smaller rather than a bigger phrase would then ensure that the PD, not the possessee, ends up with dative case.40 I leave a more thorough investigation of this issue for future research. Another interesting possessor construction to be explored in the system I develop here is shown in (64). Here the possessor is allowed to be in subject position (cf. (57a)). Er hebt die Hand.41

(64)

he (NOM) raises the hand

He is raising his hand. According to Guron (1985), who gives a control-based account of the French PDC, these constructions are limited to expressions of natural gesture (i.e. functional movement of a body part). She analyzes predicates like lever la main raise the hand

38

On this view of case-checking, DPs come with unvalued case-features that need to be filled

in by case-licensing heads.
39

According to asymmetric c-command (Chomsky 1995), neither of the two DPs is closer to

the affectee v.
40

As pointed out to me by Theresa Biberauer (p.c.), the worry here is that we know from

instances of generalized pied-piping (Chomsky 1995) that pied-piping of more material than the syntax actually needs to move routinely takes place.
41

See Wunderlich (2000) for an analysis of this example in the framework of Lexical

Decomposition Grammar and Vergnaud & Zubizarreta (1992) for a discussion of similar examples in English and French. 120

as pronominal verbs or reanalyzed V+NP constructions which are transitive in syntax but intransitive at LF. In effect, the verb and the body part form an idiomatic unit expressing a particular kind of bodily movement.42 As for the PDC-analysis I propose, the question is how the example in (64) is different from the one in (57a), Er bewunderte den Bruder He admired the brother, where no obligatory possessor relation gets established between er and den Bruder. Since the movement restriction given in (56) prevents possessor raising to the specifier of an agentive v, the subject er in (64) cannot have originated in Spec DP of the possessee die Hand. The possessor relation must then be established differently here. This is confirmed by the fact that *sich die Hand heben to raise ones hand is impossible. While the reflexive PD in (65a), a non-natural gesture sentence, is fine, it is unacceptable in (65b). a. Er massiert sich die Fe.

(65)

he massages self (DAT) the feet

He is massaging his feet. b. *Er hebt sich die Hand.

he raises self (DAT) the hand

He is raising his hand.

42

Note that this does not hold for similar constructions where the nominal in possessee-

position is not a body part. The sentence Er griff in die Tasche He reached in the bag/pocket, for example, does not establish a necessary possessor relation between the subject Er and die Tasche the bag/pocket. I thank Jorge Hankamer for bringing up this sentence. 121

This contrast corroborates Gurons reanalysis of V+NP, taking the theme-status away from the NP. If die Hand were the theme argument of heben raise, we would expect it to be able to host a PD in its specifier; and since it is possible to say Er hebt ihr die Hand He raises her hand, perhaps in a context where she is unable to raise her own hand, we know that generally, heben can project an affectee v. One way to make sense of the ungrammaticality in (65b) is then to follow Guron and maintain that the DP die Hand, when used as a body part belonging to the subject er is not a theme argument of heben and also cannot assign a possessee role to the DP in its specifier. Another possible explanation is that, since there cannot be external causation with natural gesture sentences, the subject is automatically the possessor of the body part. While it is possible to ask about the manner of causation given the case of (66) (How did he break his leg?), this makes no sense in the case of (67) (#How did he raise his hand?) Thus, in contexts where there is no distinction between the causing and the resulting event, as in (67), the possessor relation does not need to be expressed through a special construction (namely the PDC). In fact, the difference in grammaticality between the examples in (66) and (67) suggests that this possessor relation really is different in that it is not syntactically encoded. It seems plausible that there is no structural connection (neither via binding nor movement) between the subject and the possessor position. a. *Er brach das Bein.
he broke the leg

(66)

He broke his leg.

122

b. ?Er brach sein Bein.

he broke his leg

He broke his leg. c. Er brach sich das Bein.

he broke self (DAT) the leg

He broke his leg. a. Er hob die Hand.

he raised the hand

(67)

He raised his hand. b. ?Er hob seine Hand.

he raised his hand

He raised his hand. c. *Er hob sich die Hand.

he raised self (DAT) the hand

He raised his hand. The (b)-examples show that the body part in both non-natural-gesture (66) and natural-gesture (67) constructions can marginally be modified with a possessive pronoun. The unmarked versions of these sentences, however, have a PD in the former case (see (66c)) and neither PD nor possessive pronoun in the latter case (see (67a)). In short, I suggest that natural-gesture constructions like (64/(67) cannot and need not be analyzed as PDCs. Other examples of natural-gesture (according to Guron, complex reanalyzed) verbs are die Augen schlieen close ones eyes, mit

123

den Ohren wackeln to wiggle ones ears, and mit der Wimper zucken blink (idiomatic). A type of construction found in English but not in German, which is similar to (64)/(67) in that it potentially has a possessor in subject position, is given in (68). a. The ship tore a sail. b. The car burst a tire. c. The athlete tore a muscle. (Hole to appear-b: 372) Since, in these examples, there is no PD or possessive pronoun, the question is how the possessor relation gets established here. In this case, there is no reason to argue against Spec DP of the possessee being the origin site of the subject because the subjects in (68) are not agents. This means that movement of the possessee from Spec DP to the nominative case position (Spec TP) does not have to proceed via the specifier of the agentive vP and thus does not cause the derivation to crash. In other words, possessor raising into subject position is compatible with the PDC account presented here, as long as the possessor moves directly to Spec TP without an intermediate, purely q-related stop-off point.

(68)

124

3.6

Conclusion

This chapter presents a PDC analysis which is able to account for the fact that German PDs are not only raised, as proposed by Landau (1999), but also obligatorily affected, as proposed for German by e.g. Wegener (1985, 1991), McIntyre (to appear), and Hole (to appear-a, to appear-b) and for French and Spanish by e.g. Guron (1985) and Kempchinsky (1992). The PD plays the role of both possessor and affectee. The framework I propose to allow for double q-role assignment is a dynamic structure-building system generally based on, but in certain points crucially distinct from, Chomskys Minimalist Program (1995, 2000, 2001). As also argued by Lpez (2001), Chomskys strict separation of the modules of Checking and q-Theory, which prevents a moved argument from receiving a q-role, must be modified. Obvious cases of illicit double q-role assignment can be ruled out by the restriction that the movement operation Internal Merge may only be applied if the moving element can check its formal features as a direct result of the movement. Other, less obvious cases may exist and will have to be carefully investigated. Overall, I have shown that the proposed dynamic structure-building approach is a plausible (if not the only) way of making Landaus (1999) convincing possessor raising account compatible with the German facts and thus the more general case of the PDC. The most attractive aspect of the analysis is that it provides a unified way to account for both ordinary affectee datives (non-PDs) and possessor datives. In the case of the former, the affectee v introduces as its specifier an argument that is externally
125

merged. In the case of the latter, the affectee v merges with an argument from inside the syntactic object being built (Internal Merge); the result is possessor raising. The basic idea on which the account is built is that selectional features of heads can in principle be satisfied either by drawing material from the numeration or by way of movement. In a PDC constellation with a complex DP containing two heads that require feature-checking but only one feature-licensing head, the costly solution of moving the non-case-licensed DP to the specifier of an additional case-checking vhead (the affectee v) is the only way to allow for the derivation to converge. It seems worth exploring to what extent this basic idea of external or internal satisfaction of featural requirements (unconstrained by Chomskys strict checking and qcomplementarity) is applicable beyond the PDC and causative contexts. The property of the PDC I explore (and exploit, in fact) in chapter 4 is that it involves two structural positions which potentially span infinitival clause boundaries. Since I have shown that the relation between these positions is subject to the locality requirement of A-movement, and since I follow Wurmbrand (2001) in arguing that only infinitival complements that are bare VPs allow A-movement of an embedded argument into the matrix clause (see chapter 2), the PDC can serve as coherence diagnostic.

126

Possessor Raising as Coherence Diagnostic

This chapter returns to the distinction between COHERENT and NON -C OHERENT infinitive constructions, first discovered by Bech (1955/57) and later reformulated by Wurmbrand (2001) as a more fine-grained set of distinctions between infinitival complements of varying degrees of structural complexity. As summarized in TABLE 1 of chapter 2, repeated below, Wurmbrand distinguishes three subclasses within the class of what are traditionally considered coherence-inducing matrix predicates: F UNCTIONAL R ESTRUCTURING (FR) predicates, LEXICAL R ESTRUCTURING (LR) predicates, and R EDUCED N ON-RESTRUCTURING (RNR) predicates. FR predicates themselves include causative and perception (ACI-introducing) verbs, which make up the subclass of SEMI-FUNCTIONAL RESTRUCTURING (SFR) predicates (not shown as a separate category in the table).

127

TABLE 1
Type FUNCTIONAL RESTRUCTURING (FR) Infinitive = main predicate (vP or VP) Properties, Distribution thematic properties are determined by infinitive possible with: modal, raising,
aspectual, causative, perception, motion verbs

Examples drfen may, gehen go, hren hear, lassen let scheinen seem vergessen forget versuchen try wagen dare beschlieen decide planen plan
versprechen promise

LEXICAL RESTRUCTURING (LR) Infinitival complement = VP REDUCED NON-RESTRUCTURING (RNR) Infinitival complement = vP or TP (FULL) CLAUSAL NON-RESTRUCTURING (NR) Infinitival complement: CP

no embedded (PRO) subject no embedded structural case possible with: control verbs embedded (PRO) subject embedded structural case possible with: control verbs embedded (PRO) subject embedded structural case possible with: control verbs

bedauern regret, behaupten claim vergessen forget

The main difference between the traditional coherence/non-coherence distinction and this new classification is that non-functional coherence-inducing verbs are divided into LR and RNR predicates. In Wurmbrands analysis, the characteristic property of RNR predicates is that they take an infinitival complement consisting of more than a bare VP. In particular, an infinitive heading this type of complement is argued to project a subject-introducing (i.e. transitive/agentive) v-head, which induces opacity in that it does not allow for an argument to move from the embedded infinitival domain into the matrix clause. The diagnostics Wurmbrand appeals to as probes for the presence of an embedded agentive vP are SCRAMBLING (more precisely, nonfocus scrambling) and LONG PASSIVE. Although RNR predicates are larger than

128

VPs in Wurmbrands analysis, they are also no larger than TP. The basis for this conclusion is that they are transparent for movement operations which are blocked by CP, namely pronoun fronting and focus scrambling. In section 4.1, I discuss scrambling and long passive as diagnostics for the presence of vP, and in section 4.2, I propose the PDC as an additional and, in important ways, less problematic probe. Finally, in section 4.3, I turn to AcI-constructions and show how the PDC sheds light on the complement to SFR predicates. As noted in chapter 2, this is the subclass of FR predicates which Wurmbrand leaves partly unanalyzed.

4.1

Scrambling and long passive as diagnostics for infinitival clause size

Wurmbrands (2001) distinction between LR and RNR predicates, the former argued to take a bare VP, the latter a vP (or TP)-complement, is based on two diagnostics: non-focus scrambling and long passive. LR predicates, such as versuchen try and vergessen forget, allow both non-focus scrambling (displacement of an argument which does not bear focus-stress) and long passive, while RNR predicates, such as planen plan and beschlieen decide, allow neither. Both of these processes are taken to be instances of A-movement and therefore subject to the standard locality constraints on A-movement. The presence of a transitive light verb projection (an agentive vP) is known to block A-movement from within its domain. Therefore, the contrasting behavior of LR and RNR predicates can be understood if the former lack

129

an agentive vP (i.e. select a bare VP-complement) and thus permit A-movement from within their complement. RNR predicates, on the other hand, select a complement which subsumes at least an agentive vP-projection, and material in the domain of v is closed off to A-movement. This effect can be obtained by assuming that RNR predicates take either vP or TP-complements. In the following subsections, I scrutinize more closely the reliability of scrambling and long passive as diagnostics for different types of complementation.

4.1.1 Scrambling There are several problems with the use of non-focus scrambling (in Wurmbrands terms, object-shift type movement (p. 274)) as a probe for vP. One is that the assumed distinction between non-focus and focus scrambling is not always clear-cut. It is argued that non-focus scrambling is sensitive to intervening case positions, while focus scrambling is less restricted. (1a), for example, where% stands for the judgment that the scrambled direct object (den Traktor) is only acceptable when it is focused, is taken to show that planen plan selects a vP or TP-complement. This is contrasted with (1b), with the matrix verb versuchen try, which allows any kind of direct object scrambling out of its complement, and is therefore argued to select just a VP-complement. (1) a. % dass Hans den Traktor geplant hat [vP/TP __ zu reparieren].
that Hans the tractor planned has to repair

that Hans planned to repair the tractor.

130

b. weil

Hans den Traktor versucht hat [VP __ zu reparieren].

tractor tried has to repair

because Hans the

because Hans tried to repair the tractor. (Wurmbrand 2001: 270) The problem here is that there are speakers (myself included)1 for whom the contrast between (1a) and (1b) is close to non-existent. Just like (1b), (1a) seems at least marginally acceptable even without special emphasis on den Traktor. As Wurmbrand points out herself, unless the matrix verb is of the b e h a u p t e n / b e d a u e r n (claim/regret)-type, which takes a propositional or factive complement and clearly selects a CP-complement, the judgments on instances of scrambling from an infinitival complement into the matrix clause are very subtle: irrealis infinitives [like the complement of planen] allow focus scrambling and disallow non-focus scrambling for most speakers. However, the two forms of scrambling are not always easy to distinguish and hence speaker variation is not fully systematic (fn. 21). While judgments vary with respect to (1a), there is no question regarding the ungrammaticality of the scrambled direct object in (2a), with the matrix verb bedauern. The contrast between (a) and (b) is considerably stronger than in (1).

The judgments on the data in this chapter are based on my own intuitions and on those of

my informants (two frequently consulted and 8 other, occasionally consulted speakers, mostly from northern Germany). 131

(2)

a. ?* weil den Traktor keiner [CP __ repariert zu haben] bedauert.

because the tractor no-one repaired to have regrets

because no one regrets having repaired the tractor. b. weil den Traktor keiner [VP __ zu reparieren] versucht.
because the tractor no-one to repair tries

because no one tries to repair the tractor. The possibility of focus-related scrambling, which can cross vP/TP-boundaries seems to introduce a confounding factor which may lead speakers to judge examples of scrambling acceptable, regardless of the complexity of material intervening between origin and landing site of the moved phrase. What further complicates Wurmbrands use of scrambling as a diagnostic is that all her examples (as opposed to the examples I provide in (2)) involve object movement from a postverbal infinitival complement. Wurmbrand does not use the term, but these examples are, in fact, instances of the T HIRD CONSTRUCTION (see chapter 2). Given Wurmbrands assumption that examples like (1a-b) are derived by extraposition of the infinitival complement followed by scrambling of an argument from the infinitival domain across the matrix verb into the matrix domain, it may be that speaker judgments are, at least in part, based on the (im)possibility of TP/vP/VPextraposition rather than on the operation of scrambling itself.2

The non-scrambled versions of (1a-b) are completely grammatical, but note that this cannot

be taken as evidence that speakers accept extraposition of complements smaller than CP. Since 132

Aside from these data questions, there is a worry regarding Wurmbrands claim that scrambled elements are (and must in fact be) focused when they cross a vP/TPboundary. The 5th generalization in Grewendorf & Sternefelds (1990) discussion of scrambling is that scrambling cannot apply to focussed phrases (p. 15). Reis & Sternefeld (2004) suggest that Wurmbrands focus-scrambling may be special in that it is a marked type of seemingly unbounded movement known as T-scrambling (Haider & Rosengren 1998), which is associated with a distinct intonation pattern (see footnote 3). This, however, does not appear to be the type of scrambling which leads to the acceptable interpretation of (1a). If, as seems to be the case, the moved direct object in (1a) does not need to bear special focus, the type of scrambling in this example (i.e. DP-scrambling) looks very much like non-focus movement of phrases other than DP. Webelhuth (1987) analyzes DP-scrambling on a par with instances of PP-movement, as given in (3), and von Stechow & Sternefeld (1988) argue that DPscrambling can even be unified with instances of AdvP and AP-movement, as given in (4). Crucially, the moved phrases are not focused here. If they were pronounced with a falling intonation accent (which is typical of focus-marking in German), the examples would be unacceptable.3

restructuring verbs can take either a full or a reduced infinitival complement, the non-scrambled versions of (1a-b) could have full CP-complements.
3

Falling intonation (\) on freiwillig in (4a) and on betrunken in (4b) is to be distinguished

from rising intonation (/) on freiwillig and betrunken followed by a high plateau and then falling 133

(3)

a. weil [PP ohne Liebe]i niemand ti glcklich werden kann.

because without love nobody happy become can

because nobody can find happiness without love. b. weil [PP ber Scrambling]i keiner etwas
because about scrambling

Vernnftiges ti sagen kann.

say can

no-one something reasonable

because nobody has anything reasonable to say about scrambling. (von Stechow & Sternefeld 1988: 465) (4) a. weil [IP [AdvP freiwillig]i [IP dasj [IP niemand ti tj tun wrde]].
because voluntarily that nobody do would

because nobody would do that voluntarily. b. weil [IP [AP betrunken]i [IP niemand ti hineinkommt].
because drunk nobody inside-comes

because nobody would get in drunk. (von Stechow & Sternefeld 1988: 466) Having reason to believe, then, that Wurmbrands direct object scrambling is the same type of operation as scrambling of unfocused PPs, AdvPs, and APs, and knowing that the latter cannot possibly be case-driven, non-focus scrambling in
intonation on niemand (a Hutkontur), associated with T-scrambling (Rosengren & Haider 1998) or, in Brings (1997) terms, a Brckenkonstruktion. (4a) (4b) weil / freiwillig das niemand \ tun wrde weil / betrunken niemand \ hineinkommt.

With this latter type of focus intonation, (4a-b) are acceptable, but that is precisely because in this case, the moved phrase itself is not focused; rather, it has been moved out of the focus domain (see Steube 2001). Crucially, Wurmbrands example given in (1a) is grammatical without this special intonation contour. 134

general can probably not be conceived of as case-driven. This view is consistent with the mainstream take on scrambling in the literature. In their Scrambling and Barriers volume, Grewendorf & Sternefeld (1990) discuss two major proposals as to what kind of movement operation scrambling might be: (i) NP-movement (i.e. an instance of Move a) (see e.g. Fanselow 1990) and (ii) a special kind of Wh-movement (see e.g. Webelhuth 1987). Scrambling shares properties with both types of movement. Like NP-movement, it cannot cross CPboundaries, but like Wh-movement, it seems to target an A-bar position. Both types of proposals posit that the landing site of scrambling is an adjoined position between CP and VP (in Grewendorf & Sternefelds terms, it is adjunction to IP). Thus, regardless of whether scrambling is argued to be more like A-movement or more like A-bar movement, there is general agreement on the landing site being a non-case position.4 The moved argument then must check case before it undergoes scrambling. This clearly speaks against Wurmbrands appeal to scrambling as an operation which is sensitive to intervening case positions.
4

Although Fanselow (1990) argues that scrambling is like A-movement in that it is clause-

bounded and leaves an A-bound trace, the type of NP-movement he identifies with scrambling clearly differs from A-movement in that it does not target a case position. In a more recent account of scrambling as A-movement, Hinterhoelzl & Pili (2003) claim that scrambling targets an A-position (more specifically, scrambled DPs move to the specifiers of heads licensing clitics), but, again, this type of A-position is not a case position. According to Hinterhoelzl & Pili, scrambled DPs move to check a specificity or topicality feature, not a case feature. Even if scrambling is analyzed as an optional EPP-driven object-shift-type movement into the second specifier of the agentive vP (see also subsection 4.3.2), it does not happen for the purpose of case-checking. 135

4.1.2 Long passive Long passive, first discussed by Hhle (1978) and introduced here in chapter 2, is a phenomenon involving passivization of the matrix verb and movement of the embedded internal argument into the matrix clause. As the moved internal argument obligatorily surfaces with nominative case (or, as often stated in the literature, undergoes CASE CONVERSION), it appears to become the subject of the matrix clause. Wurmbrand categorizes predicates of Class 2 (coherence-inducing verbs) which do not pass the long passive test, i.e. have an infinitival complement that is opaque for movement of the direct object of the infinitive into the matrix subject position, as RNR predicates. The verb planen plan, for example, although it does pass traditional coherence tests like pronoun fronting (see (5a)) and is thus commonly associated with coherence, is classified as reduced non-restructuring. This is based on Wurmbrands judgment that the long passive example in (5b) is ungrammatical. In constrast, versuchen try, which passes both the pronoun fronting test (see (6a)) and the long passive test (see (6b)), is classified as an LR predicate. It is important to note that the contrast argued to exist between the (b)-examples in (5) and (6) is far from robust. It seems that, once speakers are sure that they accept (6b), they tend to also accept (5b). (5) a. weil ihn der Hans [ __ zu reparieren] plante.
to repair planned

because it (MASC) the Hans

because Hans planned to repair it.

136

b. * dass der Traktor [ __ zu reparieren] geplant wurde.

that the tractor to repair planned was (PASS)

that they planned to repair the tractor. (Wurmbrand 2001: 267-268) (6) a. weil ihn der Hans [ __ zu reparieren] versuchte.
to repair tried

because it (MASC) the Hans

because Hans tried to repair it. b. dass der Traktor [ __ zu reparieren] versucht wurde.
that the tractor to repair tried was (PASS)

that they tried to repair the tractor. (Wurmbrand 2001: 267-268)

Within Wurmbrands typology assuming that long passive is an instance of casedriven movement triggered by the lack of an embedded case-licensor matrix verbs of the planen-type, which are judged not to allow long passive, must then take a vP or TP-complement, whereas verbs of the versuchen-type, which do allow long passive, take a bare VP-complement. The reason Wurmbrand expects that verbs of the RNRtype, like planen, pass the pronoun fronting test is that fronted pronouns are assumed to target the matrix clause Wackernagel position, which is generally associated with A-bar, i.e. non-case-driven, movement. Intervening A-positions can thus not hinder pronoun fronting. According to Wurmbrand, the long passive test results in (5)-(6) line up with the corresponding scrambling test. As shown in (1), repeated here as (7), she considers
137

scrambling out of the extraposed complement of planen infelicitous unless the scrambled element is focused (see (a) where % indicates that scrambling is ungrammatical when den Traktor is not focused). In contrast, any type of scrambling across the complement boundary of versuchen is possible (see (b)). (7) a. % dass Hans den Traktor geplant hat [ __ zu reparieren].
that Hans the tractor planned has to repair

that Hans planned to repair the tractor. b. weil Hans den Traktor versucht hat [ __ zu reparieren].
tractor tried has to repair

because Hans the

because Hans tried to repair the tractor. (Wurmbrand 2001: 270)

Since one of the generalizations Grewendorf & Sternefeld (1990) list about scrambling is that focused phrases may not be scrambled (see the discussion in subsection 4.2.1), however, it is not clear that the interpretation on which (7a) is grammatical is really focus-scrambling. Hence, if there is a good chance that both the grammatical (7a) and (b) are instances of the same type of movement, the scrambling test really yields no difference between the complexity of the complements to planen and versuchen. Scrambling is not a good way to confirm the long passive data in (4)(5) then.

138

4.1.3 Need for a new diagnostic While I follow Wurmbrand in proposing that infinitival complements with the two different degrees of transparency claimed to be apparent in examples like (5)-(7) should be subdivided into a bare VP-type on the one hand and a bigger, more complex type on the other hand, I also agree with Reis & Sternefeld (2004) that this cannot solely be based on long passive and scrambling. Although there are other authors who consider the possibility of long passive significant enough to be used as coherence diagnostic (see e.g. Bayer & Kornfilt 1989, Wllstein-Leisten 2001, Sabel 2002, and Schmid et al. forthcoming), Reis & Sternefeld are skeptical about relying on long passive data. They follow Hhle (1978) and Kiss (1995) in considering it a highly marked construction which leads to acceptable results with very few verbs only. I agree that judgments on examples with matrix verbs other than versuchen try (and perhaps vergessen forget) vary considerably. In fact, Wllstein-Leisten (2001) presents the results of an empirical study on the acceptability of long passive5 and comes up with a list of matrix verbs allowing long passive which includes a number of verbs that Wurmbrand classifies as RNR predicates, i.e. verbs that
5

According to this study, a number of verbs, besides versuchen (which triggers 100%

acceptance), trigger about 75% acceptance of long passive in the context of a Third Construction. These verbs are ablehen decline, anbieten offer, aufgeben give up, beabsichtigen intend, beschlieen decide, ertragen endure, erwarten expect, geloben vow, genieen savor, glauben believe, hoffen hope, vergessen forget, versumen miss, verweigern refuse, vorschlagen suggest, vorziehen prefer, and wnschen wish. Some of these verbs are also on Wurmbrands long-passive-allowing list of LR predicates, but others are on her long-passive-disallowing RNR list, and again others were not included in Wurmbrands study. 139

Wurmbrand claims do not allow long passive and scrambling (anbieten offer, beschlieen decide, glauben mean, vorschlagen propose, wnschen wish). Furthermore, in a recent empirical study by Schmid et al. (forthcoming), subjects rated examples of long passive as barely acceptable (they gave an average rating of 3.5 with 1 being perfectly grammatical and 5 being completely ungrammatical), significantly less acceptable than examples of other phenomena used to probe for coherence. The average rating for pronoun fronting was 2.85, and that for verbcluster fronting was 3.2. It seems, then, that neither scrambling nor long passive are fully reliable diagnostics for the internal structure of infinitival complements. More specifically, we cannot be sure that an application of scrambling of an embedded argument really means that the infinitival complement lacks a subject and an accusative case position and thus consists of nothing more than a bare VP-projection. As for the possibility of long passive, it does establish that the infinitival complement lacks case-licensing positions, but native speaker judgments are only consistent in the case of very few matrix verbs. In fact, given how unproductive long passive is, it is difficult to eliminate the possibility that it may be a fully lexicalized construction. In the face of this welter of conflicting and unreliable test results, a new diagnostic will be very welcome. With the goal of providing additional evidence for what Wurmbrand sets out to establish, namely that there is a fine-grained phrase-structural

140

distinction within the class of verbs traditionally classified as coherence-inducing, I propose the PDC as an additional diagnostic to probe for vP. The PDC is a syntactic phenomenon which generally leads to more robust results than long passive and scrambling in that it is more productive than the former and more clearly case-driven than the latter. I am not proposing, however, that the PDC can replace scrambling and long passive as diagnostics. Rather, this new tool should be used as a means to verify and refine some of the results of the other tests.

4.2

Possessor datives: A more reliable probe

Chapter 3 has established that the German possessor dative construction (PDC) involves case-driven movement. It follows and this is the property of the PDC which makes it relevant for diagnosing coherence that the PDC is subject to the locality restrictions typical of A-movement: an argument may not raise beyond the edge of a subject-containing phrase (namely a CP, a TP, an agentive vP, and a complex DP) to a higher A-position. PD-raising is then restricted to apply within the minimal agentive vP which contains it. More precisely, the affectee v-head which licenses the dative case on the PD must be on the same side of the agentive vPboundary as the origin site of the PD (the specifier of the possessee nominal). Unlike an embedded direct object, which can check accusative case with the agentive v-head at a distance, i.e. via a static Agree relation, a PD must move to the specifier of its case-licensor (see chapter 3). Thus, given that possessor raising is obligatory and
141

strictly local, the possibility of a PD moving from an embedded infinitival complement into the domain of the matrix verb is expected to depend on the complexity of the infinitival complement. If the complement is bigger than a bare VP, PD-movement into the matrix clause is expected to be blocked. If, on the other hand, the complement lacks an agentive vP-projection, the PD may raise beyond the infinitival domain boundary. Given the latter scenario, depending on whether the affectee vP is projected by the infinitive or the matrix verb (I assume that both options exist)6, the PD will remain in the infinitival complement or move into the matrix clause (henceforth, I will refer to the latter option as long PD-movement). The goal of the following subsections is to lay out in detail how the PDC (analyzed as discussed in chapter 3 and briefly described above) interacts with the different types of (non-)restructuring contexts covered in Wurmbrands (2001) typology of infinitival clause size. The prediction is that a PD which originates in an infinitival complement can only undergo long movement into the matrix clause in the context of a restructuring (FR or LR) predicate. In other words, we expect that the agentive vP-

I assume that, in principle, a reduced infinitival complement (which lacks an agentive vP-

projection) can project an affectee vP. Thus, affectee vPs do not always have to be dominated by an agentive vP. In a simple unaccusative PDC like weil dem Jungen (DAT) der Stein ( ACC ) auf den Kopf fiel because the rock fell on the boys head, for example, the affectee vP which licenses the PD, dem Jungen, does not cooccur with an agentive vP. What is crucial here is simply that, when an infinitival complement which lacks an agentive vP does not have the ability to check dative case, nothing prevents an embedded PD from raising into the higher clause to find a licensor. 142

boundary intervening between the PD origin site and the matrix clause domain in the context of a non-restructuring (RNR or NR) predicate is opaque for PD-raising.

4.2.1 NR predicates A full NR predicate, which takes an infinitival CP-complement with a null C (see (9)) is expected to be just as incompatible with long PD-movement as a matrix predicate which takes a finite embedded clause (see (8)). In neither case should a PD be able to move into the matrix clause. As the (b)-examples of the following data show, this is indeed the case. (8) a. Tim hat bedauert [CP dass er der Nachbarin das Auto waschen musste].
Tim has regretted that he the neighbor (DAT) the car wash must (PAST)

Tim regretted that he had to wash the neighbors car. b. *Tim hat der Nachbarin bedauert [CP dass er das Auto waschen musste]. (9) a. Tim hat behauptet [CP der Nachbarin schon den Hof gefegt zu haben].
Tim has claimed the neighbor (DAT) already the yard swept to have

Tim claimed to have already swept the neighbors yard. b. *Tim hat der Nachbarin behauptet [CP schon den Hof gefegt zu haben]. Notice that the infinitival complement in (9) is postverbal. (9b) with the PD der Nachbarin as part of the matrix clause domain is then an ungrammatical instance of the Third Construction. Since the (un)grammaticality of all the examples that follow does not hinge on the fact that they are Third Constructions, I will continue to use
143

examples with postverbal complements.7 Given Zwarts (1997) view, which is not crucial for the main point here but which I propose as a reasonable way to deal with the Third Construction, postverbal infinitival complements are in their base-position (i.e. not literally extraposed) and, depending on their complexity, may allow movement of an embedded argument into the matrix clause.

4.2.2 RNR and LR predicates Moving on to the PDC in the context of RNR predicates, the prediction is again that long PD-movement is ungrammatical. This is confirmed by (10b).

In many cases, the infinitival complement is more natural in postverbal position.

Furthermore, if the complement is preverbal (or, in traditional terms, intraposed), PD-movement is string-vacuous and therefore does not reveal whether the complement boundary has been crossed or not. PD-movement can be made visible if the matrix clause subject is a quantificational pronominal (see (i)-(iii) below), but PD-movement to the left of such a subject necessarily involves scrambling which is precisely what the PDC-test is meant to supplement. (i) *Dass der Nachbarin
that

alle den Hof gefegt zu haben behaupten, ist doch wohl klar.
the yard swept to have claim is but clear

the neighbor (DAT, FEM) all

Its obvious that everyone claims to have swept the neighbors yard. (ii) ?Dass dem Tim wirklich keiner das Fahrrad zu reparieren plant, glaube ich nicht.
that the Tim (DAT) really no-one the bike to repair plans believe I not

I dont believe that theres really no one who plans on reparing Tims bike. (iii) Dass seiner Schwester dafr jemand das Radio heile zu machen versucht, finde ich lustig.
that his sister (DAT) there-for somebody the radio intact to make tries find I funny

I think its funny that somebody tries to fix his sisters radio instead.

While the judgments on the high PD-position in (i) and (iii) are clear and identical to the judgments on long PD-movement in the corresponding Third Construction context (see (9b) and (11b)), the high PDposition in (ii), where the matrix verb is planen, is just as hard to judge as Wurmbrands example of direct object scrambling in (7a). 144

(10)

a. Die Nachbarin hatte geplant [vP/TP dem Tim dafr das Fahrrad reparieren
the neighbor had planned
the Tim (DAT) there-for the bike

repair

zu lassen].
to let/have

The neighbor had planned to get Tims bike repaired in return. b. *Die Nachbarin hatte dem Tim geplant [vP/TP dafr das Fahrrad reparieren zu lassen]. The ungrammaticality of (10b) is exactly in line with Wurmbrands test results. Recall that planen passes the traditional coherence tests, for example pronoun fronting, but not Wurmbrands long passive and non-focus scrambling diagnostics. Since I am not sure I agree with Wurmbrands judgments on the relevant data (see (5)-(7) above), however, the PDC test in (10) helps solidify the RNR-status of planen. In contrast to RNR predicates like planen, LR predicates, which take a bare VPcomplement, are expected to allow long PD-movement. The data in (11) show that this prediction is borne out as well. (11) a. Tim htte machen].
make

lieber versucht [vP/CP seiner Schwester das Radio heile zu

his sister (DAT) the radio intact to

Tim would-have rather tried

Tim would have rather tried to fix his sisters radio. b. Tim htte lieber seiner Schwester versucht [VP das Radio heile zu machen.]

145

The reason the infinitival domain boundary in (11a) is marked as vP/CP is that all LR predicates, though they may just take a bare VP, also have the option of taking a full clausal complement. Since there is no evidence of restructuring here, the complement could be of either the restructuring or the full clausal type. The accusative case on das Radio could be licensed by either the matrix or an embedded agentive v-head (the former via static Agree), so there need not be an agentive vP in the complement. Given the word order in (11a), the complement does, however, need to include an affectee vP because the inherent dative case on the embedded PD seiner Schwester cannot be checked via static Agree. Recall that the affectee vP needs an argument in its specifier to assign its q-role to. It could be argued that the high position of the PD in (8)-(11) is a result of complement-internal possessor raising followed by scrambling and thus non-casedriven movement into the matrix clause. The PDC would then not be a diagnostic for complement size at all because the origin and landing site of the PD would be on the same side of the complement boundary. Rather, the judgments would reflect the (im)possibility of scrambling out of the complement. The reason I believe that the PDC-examples in (8)-(11) and also in (12)-(13) below cannot be treated on a par with instances of scrambling is that the judgments on examples with irrealis matrix verbs like planen are much more clear-cut in the context of a PDC than in contexts of regular object movement (see, again, Wurmbrands example in (7b)). Making the traditional assumption that scrambling is adjunction to IP (TP) and therefore not
146

blocked by intervening case-positions, there is no explanation for the strong contrast between the (a) and (b)-examples in (10)-(11) and between most of the examples in (12) and (13) below. Scrambling across any complement boundary smaller than CP should be equally acceptable. The fact that long PD-movement in (10b) and in all of the examples in (12) is clearly degraded suggests that we really are not dealing with scrambling here. PD-movement is subject to a stricter locality condition and therefore serves as a better diagnostic for complement complexity than direct object scrambling. Another argument against treating PD-movement on a par with

scrambling is that an indefinite DP is not expected to be allowed to scramble (see Bring 2001), yet the indefinite PD in (11) is grammatical as part of the matrix clause. (11) Tim htte machen].
make

lieber einem hbschen Mdchen versucht [VP das Radio heile zu

tried the radio intact to

Tim would-have rather a beautiful girl (DAT)

Tim would have rather tried to fix a pretty girls radio. If not for case-reasons, an indefinite DP should not move out of its DATO > ACCO position inside VP. The position of the PD here then cannot be the result of scrambling into an adjoined position.8

What exactly it is that rules out scrambling of the PD into the matrix clause after case-

checking in (10b) and (12) is not clear to me. These data could be taken as evidence for scrambling being adjunction to vP, not TP. If scrambling is in fact restricted to applying within vP, however, we 147

According to Wurmbrand, the matrix verbs in (12) are RNR predicates and the matrix verbs in (13) are LR predicates. The PDC-test results corroborate this. The star on the PD in the higher position in the examples in (12) indicates that the PD may not occur in the domain of RNR matrix verbs. In (13), on the other hand, where we have LR matrix verbs, which can be used with either a full or reduced infinitival complement, both the lower PD-position within the infinitival complement and the higher PD-position within the matrix clause are (at least marginally) acceptable. (12) a. Der Frisr hat (*der Kundin) einfach beschlossen [(der Kundin) die
decided (the client (DAT)) the

the hair-dresser has (the client (DAT) simply

Haare rot zu frben].

hair red to dye

The hair-dresser simply decided to dye the clients hair red. b. Der Lehrer hat (*dem Schler) entschieden [(dem Schler) das Leben
the teacher has (the student (DAT)) chosen (the student (DAT)) the life

zur Hlle zu machen].

to-the hell to make

The teacher chose to make the students life hell.

are left with the unexplained judgments on scrambling examples like (7a). Scrambling of the embedded direct object into the matrix clause should not be possible at all, unless the scrambled element is part of a Brckenkonstruktion with a Hutkontur intonation pattern (see subsection 4.1.1, footnote 3). 148

c. Herr Mller hat (*seinem Sohn) erwogen [(seinem Sohn) das

Mr. Miller has (his son (DAT)) contemplated (his son (DAT)) the

Taschengeld zu streichen].
pocket-money to cross (out)

Mr. Miller contemplated taking away his sons allowance. d. Mein Mann hat (*meinen Eltern) geplant [(meinen Eltern) den
my husband has (my parents (DAT)) planned (my parents (DAT)) the

Videorekorder zu reparieren].
video-recorder to repair

My husband planned to repair my parents video recorder. (13) a. Ich habe (meiner kleinen Nichte) versucht [(meiner kleinen Nichte) die
I have (my little niece (DAT)) tried (my little niece (DAT)) the

Zhne zu putzen].
teeth to clean

I tried to brush my little nieces teeth. b. Der Mann hat (seiner Frau) tatschlich vergessen [ (seiner Frau) den
the husband has (his wife (DAT)) actually forgotten (his wife (DAT)) the

Rcken zu massieren].
back to massage

The husband actually forgot to massage his wifes back. c. Der Papa hat (der Kleinen)
the dad

gerade erst begonnen [(der Kleinen)

begun

die

has (the little one (DAT)) just now

(the little one (DAT)) the

Haare zu streicheln].
hair to stroke

The dad has just now begun to stroke the little ones hair.

149

d. Ich habe (?der Professorin) irgendwie versumt [(der Professorin) die

I have (the professor (DAT)) somehow missed (the professor (DAT))) the

Hand zu geben].
hand to give

I somehow missed the opportunity to shake the professors hand. e. Mein Bruder hat tatschlich (?dem Rockstar) gewagt [(dem Rockstar)
my brother has actually (the rock star (DAT)) dared (the rock star) (DAT))

die Gitarre zu stimmen].

the guitar to tune

My brother actually had the guts to tune the rock stars guitar.

Long PD-movement then provides new evidence for Wurmbrands distinction between RNR and LR predicates. RNR predicates like planen and beschlieen take at least a vP -complement which prevents movement of an embedded argument beyond the infinitival domain boundary, while LR predicates like versuchen and beginnen have the option of taking a transparent VP-complement. Before we turn to the interaction of the PDC with FR predicates in subsection 4.2.3, I would like to return to the list of verbs that Wllstein-Leisten (2001) classifies differently than Wurmbrand. As discussed in subsection 4.1.3, this list includes the following verbs: anbieten offer, beschlieen decide, glauben believe, vorschlagen propose, and wnschen wish. The PDC diagnostic supports Wurmbrands test results in that beschlieen (see (12b)) and wnschen (see (14b)) take an infinitival complement which is not transparent to possessor raising, but it supports Wllstein-Leistens test

150

results in that the complement to glauben is transparent (see (14a)). The high position of the PD is clearly more readily acceptable in (14a) than in (b). 9 (14) a. Der Student hat (?dem Professor) geglaubt [(dem Professor) die Tasche
the student has (the professor (DAT)) believed (the professor (DAT)) the bag

tragen zu mssen].
carry to must

The student believed he had to carry the professors bag. b. Die Mutter hatte (*ihrem Sohn) gewnscht [(ihrem Sohn) den Teddy
the mother had (her son (DAT)) wished (her son (DAT)) the teddy

heile machen zu knnen].

intact make to can

The mother had wished she could fix her sons teddy. The fact that the high position of the PD in (14a) is judged to be at least marginally acceptable also sheds light on Reis & Sternefelds (2004) observation that glauben, while allowing object scrambling in the context of a Third Construction, as shown in Bechs (1955/57) example in (15a), 10 does not pass the long passive test shown in (15b).

Note that (17a) is best when there is an intonation break between Professor and geglaubt. Just as in the PDC example in (14a), Bechs original (1955/57) object scrambling example

10

is most readily acceptable when there is an intonation break between the dislocated nominal and glauben. 151

(15)

a. dass sie eine Absicht glaubten [ __ verbergen zu knnen, die

that they an intention believed hide

so

to be-able-to which so

zutage lag].
disclosed lay

that they believed to be able to hide an intention which was so clearly visible. (Bech 1955/57: 66) b. * weil der Wagen [ __ zu reparieren] geglaubt wurde.
because the car to repair believed was

because they believed that the car was being repaired. (Reis & Sternefeld 2004: 473) In the case of conflicting test results like this, the PDC can be used as a tie-breaker, so to speak. The possibility of long PD-movement confirms the scrambling/Third Construction test result in (15a) and thus helps classify glauben as an LR predicate, taking a bare VP-complement. Correspondingly, the PDC test result in (14a) casts further doubt on the reliability of the long passive test. The major limitation of the PDC as restructuring diagnostic is that PDCs are incompatible with verbs which select both an infinitival complement and a dative argument. As briefly touched upon in chapter 3 (see footnote 30), a PD may not originate as the possessor of a core dative argument. In (16), helfen help and gratulieren congratulate, for example, select a dative-marked nominal as their internal argument, der Tochter the daughter in (a) and dem Vater the father in (b). In this case, the respective non-core datives, unseren Nachbarn our neighbors and
152

seiner Freundin his girlfriend cannot be interpreted as PDs, in fact, they cannot be accommodated at all. (16) a. *Ich habe unseren Nachbarn der Tochter
I have our neighbors (DAT)

geholfen.

the daughter (DAT) helped

I helped our neighbors daughter (for their benefit). b. *Er hat seiner Freundin dem Vater
he has his girlfriend (DAT)

gratuliert.

the father (DAT) congratulated

He congratulated his girlfriends father (for her benefit).

The PDC can then not be used as a diagnostic to confirm the restructuring status of the dative-taking control verbs empfehlen recommend and versprechen promise in examples (17a-b). (17) a. Man hat ihr
one

empfohlen [das Auto zu verkaufen].

to sell

has her (DAT) recommended the car

They recommended that she sell the car. b. Sie hat ihrem Mann versprochen [ das Sofa reinigen zu lassen].
the couch clean to have

she has her husband (DAT) promised

She promised her husband to have the couch cleaned. Since empfehlen and versprechen both take a core dative argument, the addition of a PD would be ungrammatical, and in the context of these verbs, a single dative nominal in the matrix clause, despite its cooccurrence with a possible possessee in the infinitival domain, is not normally interpreted as a PD. The dative nominals ihr in
153

(17a) and ihrem Mann in (17b) are automatically interpreted as arguments selected by the verb and are thus grammatical regardless of whether or not they are interpreted as possessor of the respective embedded direct object.11 Other examples of dativetaking verbs, which the PDC diagnostic cannot properly classify, are anbieten offer and vorschlagen propose. Recall that these verbs did pass Wllstein-Leistens but not Wurmbrands long passive test. Whether they are potentially VP-taking LR predicates or vP/TP-taking RNR verbs thus remains to be established.

4.2.3 FR predicates Finally, turning to FR predicates, in the context of a modal or raising verb (see (18)), which can take either a VP or vP-complement, but which itself is a non-q-assigning functional head, there appears to be no way to test whether a PD can move from the domain of the infinitive into the domain of the finite verb because there really is just one lexical domain (see Wurmbrands chapter 3). Assuming Wurmbrands

organization of clause structure, a modal or raising verb construction with a transitive

11

Note that, unlike empfehlen and versprechen, the verbs in (14), glauben and wnschen, are

in principle compatible with a non-core dative argument in the matrix clause domain, and this is so despite the fact that these verbs can take a core dative argument, as in Der Student hat seinem Professor (DAT) geglaubt The student believed his professor and Die Mutter hat ihrem Sohn ( DAT) nur das Beste gewnscht The mother wished her son only the best. The reason is that these verbs never take both a core dative argument and an infinitival complement. The datives here are then not automatically interpreted as core, non-possessor arguments. The purpose of the intonation break between the PD and the verb in (14a) (see footnote 11) is probably to facilitate the interpretation of the dative as non-core, i.e. as an element not selected by the verb in the same way as a regular dative argument is selected. 154

or unergative infinitive consists of a lexical VP hosting the infinitive, an agentive vP introducing the subject, and layers of functional structure which host the modal/raising verb and case-license the subject. If such a construction includes a PD, it is necessarily the infinitive (the lexical V) which projects the PD-licensing affectee vP. The PD then raises from the specifier of the possessee within the VP to the specifier of the affectee vP, and there is no chance of this affectee vP being outside the domain of the infinitive.12 The examples in (18) show the PDC in the context of the modal mssen must (a) and the raising verb scheinen seem (b). (18) a. dass der Student [vP der Professorin
that the student

das Auto waschen] musste.

wash must (PAST)

the professor (DAT, FEM) the car

that the student had to wash the professors car for her. b. dass der Professor [vP der Studentin
that the professor

die Hausaufgaben zu
to

the student (DAT, FEM) the homework

machen] schien.
make seemed

that the professor seemed to do the students homework for her. As Wurmbrand puts it, the mono-clausality of FR constructions necessarily follows from the architecture of the clause. PD-raising has but one option of applying. As discussed in the following section, AcI-introducing semi-functional restructuring
12

On the view that modals and raising (FR) verbs are V-heads which project a defective vP

and select a vP-complement, the verbal shell of an FR predicate could in principle include an affectee vP. Whether this is indeed possible remains unclear both as a theoretical and an empirical matter. Since word order would not be affected, there may not be a way to probe for this structure. 155

(SFR) predicates behave very differently from modal and raising verbs in this respect. We will see why classifying SFR predicates as a special type of FR predicate is not an adequate way of capturing the restructuring status of AcIs.

4.3

Possessor datives in the context of AcIs

Picking up the task of going beyond the traditional coherence/non-coherence distinction at the point where Wurmbrand (2001) left off, namely where the focus is on the internal structure of Accusativus cum Infinitivo (AcI), SFR-type complements, the property of AcIs I aim to shed light on here is that AcI-infinitives are clearly able to project an opacity-inducing, agentive vP. Recall from chapter 2 that, although it is well-known that AcIs can be headed by a transitive infinitive and thus have an agentive subject, AcI-constructions are commonly described as obligatorily coherent and therefore non-satzwertig (but see Grewendorf 1987, 1988 for a critique of this view). According to Bech (1955/57), verbs governing the 1st or 3rd status (i.e. a zuless infinitive or a past participle) are always coherently constructed. Furthermore, the fact that AcI-complements can neither occur postverbally nor constitute a tense or negation domain separate from the matrix clause makes them seem very much dependent on, and thus in a sense incomplete without, the matrix clause. This does not mean, however, that AcIs cannot be complete semantically. Their ability to host an agentive subject that is referentially distinct from the matrix subject is a

156

property shared by finite complement clauses, and there is no question that finite embedded clauses are fully opacity-inducing. My assumptions concerning the syntactic correlate of agentivity and thus the phrasestructural distinction between transitive and unergative AcI-infinitives on the one hand and passive and unaccusative AcI-infinitives on the other hand will be laid out in detail in chapter 5. In order to show how the PDC interacts with AcI-

constructions, however, I already anticipate these assumptions here. The following subsections will establish that, just like infinitive constructions of the LR-type, AcIconstructions differ with respect to their compatibility with long PD-movement. If the AcI-infinitive projects a proto-agent-introducing vP, PD-movement is restricted to applying within the AcI.

4.3.1 Transitive and unergative versus unaccusative AcIs Given that the PDC is subject to the locality restrictions of A-movement PDs may not move across subject-containing and adjunct boundaries possessor raising is only expected to be able to span an AcI-boundary if the AcI is not closed off by an agentive vP. The success of long PD-movement then indicates that the AcI is a bare VP, while the failure of long PD-movement indicates that the AcI is more complex. Since the impossibility of the infinitival complement occurring postverbally and the lack of an independent tense and negation domain speak against the presence of a TP and thus also a CP-projection, it is in fact safe to assume that vP is the highest level of
157

complexity an AcI can have. An example of a PDC in the context of a transitive AcI is given in (19). The infinitive, massieren massage, has an agentive AcI-subject, Willi, here. (19) a. Maja lsst Willi ihrer Freundin die Fe massieren.
massage

Maja lets/has Willi (ACC) her friend (DAT, FEM) the feet

Maja lets/has Willi massage her friends feet. b. *Maja lsst ihrer Freundin Willi die Fe massieren. The PDC-test results are as expected. The fact that the PD must stay inside the AcI, to the right of the AcI-subject, confirms that this AcI is closed off by an agentive vPboundary. An example of an unaccusative AcI is given in (20). Here, the AcIinfinitive, fallen fall, has a non-agentive AcI-subject, den Stein the rock. (20) a. Der kleine Junge lsst den Stein
the little boy lets

seinem Freund auf den Kopf fallen.

on the head fall

the rock (ACC) his friend (DAT)

The little boy lets the rock fall on his friends head. b. Der kleine Junge lsst seinem Freund den Stein auf den Kopf fallen. Again, the results are as expected. The fact that the PD can be on either side of the AcI-subject confirms that unaccusative AcIs are not closed off by an agentive vPboundary which could prevent long PD-movement. These data suggest, then, that the subject of an unaccusative AcI-infinitive (a proto-patient) induces transparency, while the subject of a transitive AcI-infinitive (a proto-agent) induces opacity and must
158

therefore be associated with an agentive vP-layer. Thus, depending on the agentivity of the AcI-subject, AcI-introducing verbs could be grouped with either coherenceinducing/restructuring or non-coherence-inducing/non-restructuring predicates. Like the complement of FR predicates, AcIs are syntactically inseparable from the matrix clause (regardless of whether they are bare VPs or agentive vPs), but, like the complement of RNR predicates, AcIs can host a second subject. In fact, since this second subject can be referentially distinct from that of the matrix predicate, agentive AcIs even share properties with complements of full NR predicates taking a finite complement.

4.3.2 The position of affectee vPs The reason I do not indicate the boundaries of the AcI in examples (19)-(20) is that the structural position of the affectee vP, the landing site of the PD, as well as the position of the AcI-subject are not immediately obvious (at least not in (20)). In the transitive AcI in (19), where the AcI-subject is clearly the external (proto-agent) argument of the infinitive, the structural positions are straightforward. As shown in (21)13, the AcI (the circled part of the structure) includes both an agentive vP and an

13

In the structures in both (21) and (22), I simplified the matrix clause by making vP left-

headed, and I am abstracting away from verb-second-related movements (the TP and CP-projections are not shown). According to Vikner (1995), all verbal heads and T are phrase-final in German, and the verb moves via T to C (the verb-second position). The subject then moves via Spec TP to Spec CP (see Haider 1993 for a different view). Note that on Zwarts (1997) view, which I adopt to make sense of the Third Construction, verb phrases are head-initial, and SOV order is derived by movement of 159

affectee vP. The AcI-subject is in the specifier of the former, and the PD in the specifier of the latter projection. (21)

vP 2 DP v' Maja 2 (agentive) v VP lsst 2 vP V 2 tv DP v' Willi 2 vP v (agentive) 2 DP v' ihrer Freundin2 (PD) VP v (affectee) 2 DP V 14 2 massieren tDP D' (PD) 2 D NP die Fe

Maja lets Willi massage her friends feet.

In contrast, when it comes to the unaccusative AcI in (20b) (diagrammed in (22)), we are dealing with a different instantiation of lassen, one which selects a complement without (verbal) functional structure, certainly without an agentive vP and, as I

object and clausal complements to the left of the verb. In my tree structures, I use a mix of the standard head-final and Zwarts head-initial verbal structure to present the familiar picture of German being an underlying SOV language, while also saving space, i.e. getting the word order right without having to show the matrix verb-second domain.
14

Whether or not non-finite Vs in German move to the light verb (v) layers they project is not

a clear-cut issue. See Hankamer & Lee-Schoenfeld forthcoming for an analysis of VP-fronting which provides an argument against nonfinite V-to-v movement. 160

assume here, also without an affectee vP.15 The affectee vP must then be part of the matrix verb shell. Notice that word order does not distinguish between an

unaccusative AcI-complement that includes both the AcI-subject and the PD and one that only includes the AcI-subject. The unaccusative AcI-subject, den Stein the rock in (22), originates inside the embedded lexical VP and is thus necessarily merged lower than the landing site of the PD (the specifier of the affectee vP), whether it be projected by the AcI-infinitive or the matrix verb.16,17
15 16

But see footnote 6, for the possibility that the affectee vP may also be part of the AcI. In order to have the affectee vP in (22) be part of the matrix clause, I must assume that the

AcI-introducing verb lassen is introduced into the derivation as a lexical V-head which projects and raises to an agentive v, not as an aspectual, semi-functional v-head, as Wurmbrand (2001) proposes. I know of no evidence against this.
17

It should be noted that, if the linear order of den Stein and auf den Kopf in (20b) reflects

their hierarchical order, as shown in (22), den Stein would be closer to the affectee v than the possessor (PD) in the specifier of den Kopf. Instead of targeting the PD, seinem Freund, to raise and check case with, the dative-case licensing affectee v would then target den Stein. This would lead to the wrong outcome, of course, and the derivation would, in fact, crash as desired den Stein would cause a defective intervention effect with respect to case-licensing of the possessor, and the inanimate Stein is incompatible with the affectee role. The question is, however, how to guarantee that the right derivation converges. The solution is to reverse the hierarchical order of the DP den Stein and the PP auf den Kopf in (22). This is reasonable given that the DP is the obligatory internal argument of the embedded verb fallen and the PP, while also an argument, is optional. The DP should then be introduced into the derivation by first Merge with V, and the PP by second Merge. In this reversed configuration, the PD and den Stein are equidistant with respect to the affectee v (because neither commands the other and asymmetric command is the standard measure of relative prominence), and, given the semantic mismatch between the affectee role and den Stein, only raising of the PD allows the derivation to converge. Linearization must then be responsible for deriving the correct surface order of den Stein and auf den Kopf. If the DP is indefinite (e.g. einen Stein), the hierarchical order of PP > DP is in fact also a possible linear order. 161

(22)
vP 2 DP v' Der kleine Junge 2 (agentive) v vP lsst 2 DP v' seinem Freund 2 (PD) VP v (affectee) 2 tv V VP tv 2 DP V' den Stein 2 PP V 2 fallen P DP auf 2 tDP D' (PD) 2 D NP den Kopf The little boy lets the rock fall on his friends head.

It may not be a meaningful question to ask whether the PD seinem Freund his friend should be an argument of the AcI-infinitive fallen fall or the matrix verb lassen let, allow, have. The questions raised here about compositionality and argument structure are very subtle. If the affectee vP is part of the infinitival verb shell, it is the caused event, namely the falling of the rock, that affects the friend. If, on the other hand, the affectee vP is part of the matrix verb shell, it is the causing event, namely the boy letting something happen, that has consequences for the friend. These two interpretations are very hard (perhaps impossible) to distinguish. It is for these kinds of reasons, I believe, that so many investigators have had the intuition that the two verbs, fallen and lassen, form such a close unit that their respective argument structures blend. Haider (1993), for example, has proposed that verbs in coherent
162

constructions are base-generated as a verbal complex with a complex projection base and fused argument-structures. In line with Wurmbrands idea of restructuring infinitives being bare VPs, I suggest here that the extreme transparency of an unaccusative AcI is reflected structurally in that the AcI has no case-licensing heads at all, not even an inherent dative-case-checking affectee v, but the derivation of the sentence in (22) would converge and yield a syntactic object with virtually the same interpretation if the AcI did include the affectee vP. Perhaps a more meaningful question and one which permits a reasonably clear-cut answer, is how to account for the word order in (20a), where den Stein precedes seinem Freund. This order could be the result of either scrambling in the traditional sense, i.e. non-case-driven movement of den Stein into a TP-adjoined position since the matrix subject and the finite verb move into the CP (verb-second)-domain, they will still precede den Stein or a case-driven type of movement triggered by an optional EPP feature on the agentive v. In the latter case, den Stein would be raised into the second specifier position of the matrix v, which would be the equivalent of object-shift. Returning to the main issue at hand here, what is most noteworthy about the structures in (21) and (22) is that AcI-introducing verbs like lassen do not always have the same complement type. Just as there are two different options of infinitival

complementation in the case of LR verbs, AcI-introducing verbs come with the selectional need for either a bigger, clausal (vP) complement or a smaller, purely verbal (VP) complement.

163

4.3.3 From the PDC to binding to phases At this point, it has been established that a PD which originates in an AcI can only precede the AcI-subject when the AcI is not closed off by an opacity-inducing, agentive vP-layer. Since chapter 5 will show that agentive vPs also turn out to provide a domain in which syntactically-bound pronominals are free, i.e. do not incur a Condition B violation, despite having an antecedent in the same sentence, I ultimately propose to connect the observed opacity/non-coherence effects with Chomskys (2000, 2001) notion of phase (see Part III of the dissertation). Intuitively, what makes transitive AcIs phase-like is their semantic completeness. Clearly, AcIs can be much more satzwertig than their syntactic dependence on the matrix verb seems to suggest. In sum, this chapter has shown that the use of the PDC as a coherence/restructuring diagnostic, in particular as an addition to Wurmbrands use of scrambling and long passive, provides a more solid foundation for the existence of the LR-RNR subcategorization of the class of coherent verbs. The PDC also helps establish that AcIintroducing verbs are unlike any other type of (non-)restructuring predicate in that the complement of a transitive AcI not only includes an opacity-inducing vP but also hosts a subject that is referentially distinct from the matrix subject. The following chapter discusses reflexive and pronominal binding as another way to show that not all AcIs can simply be labeled subclausal in the sense of nicht satzwertig. AcIs
164

including an agentive subject are opaque for pronominal binding, while those with an unaccusative AcI-subject are not. The binding data will furthermore suggest that opacity-inducing domains are not necessarily headed by an agentive v. Other categories, which have in common that they are potentially phase-defining, namely D and P, can also head domains within which a syntactically bound pronominal is free.

165

PART III COHERENCE BY PHASE

166

A Phase-Based Binding Account of (Non-) Complementarity in German

Having established now that the POSSESSOR DATIVE CONSTRUCTION (PDC), analyzed as an instance of A-movement (see chapters 3 and 4), serves as a diagnostic for infinitival clause size, I now turn to the interaction of reduced infinitive constructions with binding. Since reflexive binding is generally assumed to be clause-bounded, it is commonly used as a coherence test. Von Stechow & Sternefeld (1988), for example, present the observations in (1) as evidence that ACIS construct coherently. (1) Der Knigi lsst die Leutej fr sichi/j arbeiten.
the king lets the people for self work

The king has the people work for him. (corresponds to subscript i) The king lets the people work on their own. (corresponds to subscript j) (von Stechow & Sternefeld 1988: 409) Parts I and II of this dissertation have already made it clear that AcIs cannot simply be classified as coherent. Here, I will further strengthen this finding by showing that AcIs which are closed off by an agentive vP are opacity-inducing with respect to pronominal binding, and that the long-distance binding possibility for the reflexive in (1) does not actually indicate that AcI-introducing verbs are obligatorily coherenceinducing.

167

In this chapter, I revisit some of the key literature on binding in AcI-constructions and present a new PHASE-based analysis of reflexive and pronominal binding which also accounts for binding facts in non-AcI contexts. In chapter 6, I focus specifically on binding as a way to probe for the complexity of AcI-complements and extend the notion phase to the PDC and other coherence diagnostics.

5.1

Introduction

Patterns of complementarity are found in many areas of linguistic inquiry. When the environments in which two forms occur never overlap, we say that the forms stand in complementary distribution. The main body of research on syntactic

complementarity comes from the study of anaphora, more specifically, the binding behavior of reflexive and non-reflexive pronouns in various phrase-structural contexts (see e.g. Safir 2004). My contribution here to this body of research is the

investigation of anaphoric relations in contexts where complementarity seems to break down. Since the empirical evidence comes from German, which does not have logophorically used reflexives, the cases discussed here have the useful theoretical property that they cannot be dismissed as being beyond the scope of a theory of binding. Furthermore, assuming a configurational theory of binding, I argue that the distribution of reflexive and non-reflexive pronouns provides crucial evidence regarding the clause structure of reduced infinitive constructions.

168

In most syntactic environments, the distribution of reflexives and pronominals is complementary. Use of one of the forms obviates the other form. However, certain contexts are known to allow the two forms to overlap. German AcI-constructions provide one such context. Although binding in AcI-constructions has been

extensively discussed in the literature (see e.g. Reis 1976, Grewendorf 1983, Pustejovsky 1984, Haider 1985, Primus 1989, and Frey 1993), the full extent of noncomplementarity in AcIs, and other configurations that involve embedding of a subclausal constituent, has not been given a satisfactory account.

5.1.1 AcIs: A context for (non-)complementarity AcI-constructions are found as complements of permissive or causative lassen let, allow, have and perception verbs like hren hear, sehen see, fhlen feel, and spren sense (members of Wurmbrands (2001) SFR-predicates). An example is given in (2). As indicated by my initials (L.-S.), the coreference relations given here are based on my judgments.1

In order to verify my native speaker intuitions, I conducted an informal study in the form of

two questionnaires (see Appendix A and B). For each questionnaire a number of native speakers of German (mostly from northern Germany) were presented with sentences containing AcI-constructions and asked to rate coreference relations between anaphoric elements and their antecedents on a scale from 1 (perfectly grammatical) to 5 (absolutely ungrammatical). The results generally match up with my judgments. Thus, when I refer to my judgments, I am generalizing over myself and my informants. When I present other authors examples and add my initials to theirs, I share their judgments. There may, however, be additional binding possibilities for me that are not indicated. In 169

(2)

Martini hrt [AcI den Mannj ber sichi/j/ihni/*j reden].

Martin hears the man (ACC) about self/him talk (INF)

(L.-S.)

Martin hears the man talk about himself/him. While the pronominal behaves as expected in that it can only refer to the higher, matrix subject Martin or some other male person mentioned in previous discourse, the reflexive sich can either refer to the lower subject den Mann or engage in seemingly long-distance binding with the matrix subject. We have a case of noncomplementarity. The domain in which the reflexive must be bound is bigger than the domain in which the pronominal must be free. The nominal in the embedded clause, den Mann, is marked with accusative case. Here, it plays the agent role with respect to the infinitive reden, and I will be calling it the AcI-subject.2 In the matrix clause, Martin is the nominative-marked subject of hren and checks case with the matrix clause T-head. If the infinitive were also embedded in a TP, the AcI-subject would check case with the lower T. Since there is no TP in an AcI, however (see chapters 2 and 4, as well as Haider 1993), it must check case with a verbal projection in the matrix clause. The standard assumption is that the accusative case on AcIsubjects results from a checking relation with the matrix clause verb (hren in (2)).

section 5.3, I will explicitly take issue with Haider (1985), Reis (1973, 1976), and Freys (1993) empirical generalizations.
2

As in previous chapters, my use of AcI-subject is to be distinguished from subject. A

working definition of subject will be developed in sections 5.2 and 5.3. AcI-subject (in my use of the term) stands for the topmost accusative-marked nominal in the AcI, regardless of its semantic role. 170

As for the unexpected long-distance binding ability of the reflexive in examples like (2), Reis (1973, 1976) and Haider (1985) claim that the reflexive pronoun sich can be bound across the intervening AcI-subject here because it is embedded in a certain type of PP. Reis characterizes PPs out of which this apparent long-distance binding of sich is possible as less obligatory and gives examples such as those shown in (3a-b). Haider specifies that these PPs have to be non-arguments and backs this up with examples like (4). Frey (1993) argues that another factor which comes into play here is the unaccusative-unergative distinction. If the AcI-infinitive is unaccusative, the AcI-subject does not interfere with the binding relation between the matrix subject and sich. (3) a. Hansi lsst [AcI die Mnnerj [PP ber sichi/ber ihni] herfallen].
Hans lets the men upon self/upon him fall

(R., L.-S.)

Hans lets the men attack him.

b. Hansi lsst [AcI die Verantwortungj [PP auf sichi/auf ihni] zukommen]. (R., L.-S.)
Hans lets the responsibility on self/ on him to-come

Hans lets the responsibility come to him. (fig. Hans wants to cross that bridge when he gets to it.) (Reis 1973: 522)

(4)

Eri lie [AcI die Leutej [PP fr sichi/j/ihni] Schnaps besorgen].

he let the people for self/him liquor get

(H., L.-S.)

He had (allowed) the people (to) get liquor for themselves/him. (Haider 1985: 244)

171

5.1.2 Scope of the chapter After a detailed discussion of the accounts by Reis (1973, 1976), Haider (1985), and Frey (1993) in section 5.2, I will argue that, for some speakers, the binding of sich is in fact even less restricted than the data in (3)-(4) suggest. This is illustrated by the following examples in (5a-b). Here the reflexive is not embedded in a less obligatory or non-argument PP. Rather, it appears to be an argument of the AcIinfinitive and thus a coargument of the AcI-subject. Yet, it is allowed to be bound by the matrix subject and, crucially, across its own subject. (5) a. Die Mutteri lsst [AcI die Kleinej
the mother lets

sichi/j/ihri/*j die Schokolade in den Mund

(L.-S.)

the little girl (ACC) self/her (DAT) the chocolate (ACC) in the mouth

stecken].
stick

The mother lets the little girl stick the chocolate in its/her mouth.

b. Die Spieleri hren [AcI die Fansj

the players hear

sich?i/j/siei/*j

anfeuern].

(L.-S.)

the fans (ACC) self/them (ACC) on-cheer

The players hear the fans cheer them on. Since the AcIs in (5a-b) can each be considered a COMPLETE FUNCTIONAL COMPLEX (CFC),3 and since, in both cases, the reflexive can have an antecedent outside of this CFC, my conjecture is that German sich, while preferably bound within the minimal CFC, has the possibility of being anteceded by any subject within the minimal TP
3

A Complete Functional Complex (CFC) is a maximal projection which contains a lexical

head and in which all the grammatical functions compatible with that head are assigned (from Chomsky (1986)). 172

containing it. As exemplified by the strong contrast in (6) the reflexive is clearly worse in (a) than in (b) it is a well-established fact that sich cannot be bound across a CP or TP-boundary. I claim here that this upper-limit boundary may in fact be the only insurmountable domain-limitation for the reflexive.4 (6) a. Die Spieleri hren, [CP dass [TP die Fans sich*i anfeuern]].
the players hear that the fans self on-cheer

(L.-S.)

The players hear that the fans cheer them on. b. Die Spieleri hren [AcI die Fans sich?i anfeuern].
the players hear the fans self on-cheer (L.-S.)

The players hear the fans cheer them on. As mentioned above, German reflexives (unlike English anaphors) do not have a logophoric or emphatic use. The apparent long-distance binding ability of the reflexive thus cannot be dismissed as going beyond the scope of a theory of binding. While the English reflexive in (7a) is grammatical because, in Reinhart & Reulands (1993) terms, it does not reflexive-mark the predicate and is thus exempt from Condition A of the Binding Theory, the German equivalent in (7b) is ruled out. (7) a. This letter was addressed only to himself. (adapted from Reinhart & Reuland 1993: 672) b. *Dieser Brief war nur an sich adressiert.
this
4

(L.-S.)

letter was only to self addressed

As pointed out by Harbert & Srivastav (1988), Gurtu (1985) makes a similar claim for

Hindi, and Kluender (1986), for German. 173

The data then force a three-way distinction among relevant cases. (6a) is an example of Case 1: reflexive and antecedent are separated by a CP-boundary. This type of long-distance binding is clearly ungrammatical. There is no variability in judgments. Haiders Schnaps-example given in (4) is a proto-type of Case 2: the reflexive is embedded in a PP, and there is no CP or TP-boundary intervening between sich and its matrix-clause antecedent. Again, though more subtle than in Case 1, the judgments are relatively clear: long-distance binding between the reflexive and the matrix subject is generally allowed. Finally, Case 3 is exemplified by (5a-b) and (6b): the reflexive is in a non-PP AcI-argument position (dative or accusativemarked). Here, the judgments are variable. Many speakers categorically rule out long-distance binding between sich and the matrix subject; others allow it at least marginally. The goal of the analysis I present is to account for all three cases, in particular 2 and 3, and the distinctions among them. As for pronominal binding, I propose that a pronominal can be bound in the same domain in which a reflexive must be bound (i.e. TP), as long as it is separated from its antecedent by an external argument, more specifically an agentive subject.5 This

As in previous chapters, I am using agentive in the sense of proto-agent (see Dowty

1991). 174

claim is supported by the type of binding contrast illustrated in (8a-b). The examples are adapted from Frey (1993) and will be discussed further in sections 5.2 and 5.3.6 (8) a. Der Knigi lsst [AcI den Gefangenenj vor sichi/vor ihmi niederknien].
the king lets the prisoner before self/before him down-kneel

(L.-S.)

The king has the prisoner kneel down before him.

b. Hansi lsst [AcI den Stein sichi/ihm*i auf den Kopf fallen].
Hans lets the rock self/him on the head fall

(L.-S.)

Hans lets the rock fall on his head. For now, I will use these examples to provide a descriptive preview of the account I propose for pronominals in AcI-constructions. In (8a), where the AcI-subject, den Gefangenen, is agentive, the pronominal ihm is grammatical when referring to the matrix subject der Knig. The reflexive and the pronominal thus overlap. In (8b), however, where the AcI-subject, den Stein, is non-agentive, ihm is impossible. In section 5.3, I propose that the existence of a domain which allows the pronominal to be free and thus grammatical falls out from the syntactic correlate of agentivity, a central notion in all three parts of this dissertation. At that point, I also propose to extend the analysis from AcIs to the nominal domain. Finally, in section 5.4, I reinterpret the results in terms of a phase-based understanding of binding domains.

The bracketing in example (8b) abstracts away from the possibility that both the AcI-subject

and the anaphoric element might have raised out of the AcI into the matrix clause. As discussed further in footnote 20, this is a harmless expositional simplification with respect to the material of this chapter. 175

The analysis will unify the binding conditions for AcI and DP-constructions and account for the three cases of long-distance reflexive binding ability mentioned above.

5.2

Previous accounts

5.2.1 Reis 1973, 1976 and Haider 1985 According to Reis and Haider, the coreference relation represented by the i-index in examples (5a-b) and (6b) (Case 3) should be ungrammatical because their accounts only allow the reflexive sich to be bound across the intervening AcI-subject when it is embedded in a PP. Examples of such long-distance reflexive binding were given in (3)-(4) above and are repeated here in (9)-(10). (9) a. Hansi lsst [AcI die Mnnerj [PP ber sichi/ber ihni] herfallen].
Hans lets the men upon self/upon him fall

(R., L.-S.)

Hans lets the men attack him. b. Hansi lsst die [AcI Verantwortung [PP auf sichi/auf ihni] zukommen]. (R., L.-S.)
Hans lets the responsibility on self/on him to-come

Hans lets the responsibility come to him. (fig. Hans wants to cross that bridge when he gets to it.) (Reis 1973: 522)

(10)

Eri lie [AcI die Leutej [PP fr sichi/j/ihni] Schnaps besorgen].

he let the people for self/him liquor get

(H., L.-S.)

He had (allowed) the people (to) get liquor for themselves/him. (Haider 1985: 244)
176

Haiders (1985) analysis is partially successful when applied to examples like (10). He proposes the following reflexive binding rule (his (51b), p. 243). (11) A reflexive pronoun has as its antecedent the external argument of its governor7. In AcI-constructions where the reflexive is governed by the infinitive this rule predicts that the reflexive must be bound by the external argument of the infinitive, i.e. the AcI-subject. However, in the case of AcI-constructions like (10), where the reflexive is governed by P, which does not have an external argument, the rule, as stated above, does not apply and thus wrongly predicts that the reflexive is ungrammatical. Haider therefore adds another clause to his rule. If the governor of the reflexive does not have an external argument, i.e., as Haider puts it, if the reflexive is embedded in a non-argument PP, the reflexive must find an antecedent elsewhere in the sentence (TP). In (10), there are two choices: The reflexive can be linked either with the external argument of the AcI-infinitive, die Leute, or with the external argument of the matrix verb, er; hence the ambiguity.

By its governor, Haider (1985) means the element that is responsible for its case valuation.

According to Haiders Realisationsprinzip, the governor assigns a case index to its argument. If this case index is structural (NOM or ACC ), case must be valued externally, and that means that it is not valued through the verb which assigns the index (p. 232). In an AcI, the AcI-subject then gets its case index from the infinitive, but its case is valued through the matrix verb. 177

Haider states that the converse of the reflexive binding rule in (11) holds for pronominals: a personal pronoun may not be anteceded by an argument of its governor (p. 244). With respect to the pronominal, however, Haider does not add a second clause. If the governor of the pronominal does not have an external argument, the converse of (11) is vacuously satisfied and allows for the pronominal to be bound by any other nominal. In the case of (10), this wrongly predicts that the pronominal can have the AcI-subject as its antecedent. Even if it were plural (sie them), the pronominal could not be bound by die Leute. Thus, Haiders (1985) system is successful in accounting for the binding ambiguity regarding reflexives, but it fails to sufficiently constrain the binding possibilities of the pronominal. Furthermore, it is not obvious how Haider would treat examples like (9a-b), taken from Reis 1973. The AcI predicates are the idiomatic expressions ber jemanden herfallen to attack (lit. to fall over) someone and auf jemanden zukommen to approach (lit. to come toward) someone. Since the PPs are clearly complements of the verbs, they should be arguments of the AcI-infinitive and have the AcI-subject as their external argument. Although Haiders system, according to which PP-

embedded reflexives are governed by P, presumably regardless of whether the PP is an argument or adjunct (see footnote 6), makes the right prediction for the reflexive here, the only cases he discusses involve adjunct-PPs. Examples with argument-PPs like (9a-b) certainly seem to be at odds with Haiders claim that binding ambiguity can only result if the reflexive is part of a non-argument.
178

Reis (1973) remains vague regarding the issue of whether the PPs in her examples are arguments or adjuncts. She simply states that the constraint which disallows reflexives to be bound across an intervening deep structure subject can be escaped by prepositional (and therefore less obligatory?) object NPs, although a good deal of lexical variation is to be observed (p. 522).8 In her later work (1976), Reis actually categorizes the PPs in (11) and (12) as obligatory. This then clearly goes against restricting long-distance binding to PP-adjunct-embedded reflexives.

5.2.2 Frey 1993 An alternative account, which is endorsed in Haiders later work (1993) and goes beyond the adjunct versus argument status of the constituent containing the anaphoric element, has been proposed by Frey. He gives the following binding conditions for reflexives and pronominals (his (28)-(29), ch. 7).

As discussed by Harbert & Srivastav (1988), the availability of the higher subject as an

antecedent for an embedded reflexive is equally disputed in Hindi. While some speakers allow longdistance binding when the reflexive is part of an argument, most speakers only allow it when the reflexive is part of an adjunct. 179

(12)

Principle (A)9: (i) An anaphor must be bound in its local domain K, when K contains a SUBJECT distinct from the anaphor, otherwise, (ii) it must be bound by a SUBJECT in the minimal CFC that contains the anaphor and a domain-closing element. Principle (B): A pronominal must be free in the minimal CFC containing the pronominal, its governor, and a SUBJECT. (Frey 1993: 120)

If all AcI-constructions were CFCs containing a SUBJECT, only reflexives in AcIsubject position would be allowed to have an antecedent outside the AcI. Reflexives that are object-DPs or embedded in a PP would never be expected to be bound by the matrix clause subject. In order to account for sentences that exhibit precisely this
9

In order to understand Freys (1993) binding conditions as translated in (12), the following

definitions must be kept in mind. Since some of these definitions are not entirely clear to me, I am providing both Freys exact words in German (pp. 119-120, (23-24), (26-27)) and literal English translations.
(i) Ein CFC ist die minimale abschlieende Kategorie, in der smtliche Elemente, die durch einen gegebenen lexikalischen Kopf eine Projektionslizenz erhalten, realisiert sind. A CFC is the minimal domain-closing category in which all the elements that get a license to project from a given lexical head are realized.

(ii) Die lokale Domne fr einen Ausdruck A ist der minimale CFC, der smtliche Lizensierer von A enthlt. The local domain for an expression A is the minimal CFC that contains all licensors of A. (iii) Domnen abschlieende Elemente sind: I und Elemente, welche die [Spec, DP]-Position realisieren. Domain-closing elements are: I and those elements that realize the [Spec, DP]-position. (iv) Unter dem Begriff SUBJEKT werden jene syntaktischen Reprsentanten einer externen Argumentstelle, welche in eine Bindungsbeziehung mit einer DP eintreten knnen, und Domnen abschlieende Elemente zusammengefasst. The cover term SUBJECT stands for domain-closing elements and those syntactic representatives of an external position which can enter into a binding relationship with a DP.

180

thus far unpredicted binding possibility in his system, Frey appeals to the distinction between unaccusative and unergative verbs. Assuming that auxiliary choice and passive verb forms are direct indications of argument structure, Frey claims that all verbs selecting sein be (as opposed to haben have) as their perfect auxiliary are unaccusative and thus do not have an external argument. For unaccusative AcI-verbs, Freys binding conditions make the following predictions. An AcI with an unaccusative infinitive is a SUBJECT-less local domain within which requirement (i) of Principle (A), as stated above, cannot be met. As a consequence of requirement (ii), the entire sentence (TP) is the relevant CFC, and the matrix clause subject must bind the reflexive. Once the domain is extended to include the whole sentence, the pronominal cannot be free and is thus predicted to be ungrammatical. In discussing example (13) (his (10b), ch. 7), Frey refers to Haiders (1987) and Grewendorfs (1988) notion of argument structure unification. According to this notion, the whole argument structure of an unaccusative AcI-infinitive blends with that of the matrix verb. The result is a coherent, reduced structure with only one complex predicate (see chapter 2 for an overview of various approaches to coherent infinitive constructions).

181

(13)

Der Knigi lsst den Gefangenenj vor sichi/*vor ihmi niederknien.

the king lets the prisoner before self/before him down-kneel

(F.)

10

The king has the prisoner kneel down before him. (Frey 1993: 116) Combining Freys binding conditions and his assumptions about unaccusativity with Haiders and Grewendorfs notion of argument structure unification, the account of (13) is as follows. Since it is a sein-selecting verb, the AcI-infinitive niederknien is unaccusative, and den Gefangenen is not realized as its external argument. The AcI thus does not have a SUBJECT. The reflexive must find its antecedent in the next higher CFC that does contain a SUBJECT, namely the matrix clause. Since the argument structures of lassen and the unaccusative infinitive blend, the AcI is not considered a sentential complement of lassen, and there is no intervening subject to worry about. The only available SUBJECT-antecedent is der Knig. Within this blended, coherent domain, the pronominal cannot be free, and, as a consequence, the interpretation indicated in (13) is unavailable. The following sentences are two more of Freys unaccusative AcI examples (his (10a) and (10c), ch. 7). Again, it is claimed that the reflexive is allowed to be, and in fact,

10

Note that I do not agree with Freys judgment here. As discussed in subsection 5.2.3, the

majority of my informants find the indicated binding possibility for the pronominal acceptable. Reis (1976), who also uses this example (her (49g)), confirms that the pronominal is at least marginally acceptable. 182

must be bound by the matrix subject because the AcI does not contain a SUBJECT and is not even a domain that can be considered separate from the matrix clause. (14) a. Hansi lsst sichi/*ihmi ein Buch
Hans lets

von Maria geben.

give

(F., L.-S.)

self/him (DAT) a book (ACC) by Maria

Hans has Maria give him a book. b. Hansi lsst sichi/*ihmi einen Stein
Hans lets self/him (DAT) a

auf den Kopf fallen.

head fall

(F., L.-S.)

rock (ACC) on the

Hans lets a rock fall on his head. (Frey 1993: 116) In both (14a) and (b), the anaphoric element precedes the DP that I have been calling the AcI-subject (namely the topmost accusative-marked nominal in the AcI). In (14a), we have an example of a passive lassen-construction (equivalent to the causative faire par infinitive in French), in which the word order of the anaphoric element and the AcI-subject is fixed. In (14b), which is the example I adapted in (8b), the word order can be changed, so that the AcI-subject precedes the anaphoric element: Hansi lsst den Stein sichi/*ihmi auf den Kopf fallen.11 Even when den Stein intervenes, the reflexive can and must be bound by the matrix subject Hans. This is consistent with Freys judgment on (13).

11

In order for the accusative object Stein rock to be allowed to precede the dative reflexive

pronoun sich, it must be definite (den Stein the rock) (see Bring 2001). 183

When it comes to AcIs with transitive infinitives, Frey contends that the argument structures of the AcI-infinitive and the matrix verb do not blend because both the matrix verb and the infinitive have an external argument. Freys Principle (A) predicts that, since the AcI contains a SUBJECT, the reflexive must be bound within its local domain and cannot have the matrix subject as its antecedent. In other words, reflexive binding across the intervening AcI-subject should not be possible. Rather, a pronominal must be used in order to refer to the matrix subject. According to Freys Principle (B), the pronominal is grammatical because it is free in its minimal SUBJECT-containing CFC. Examples (15a-c) (his (11), ch. 7) illustrate that these predictions are consistent with Freys grammaticality judgments. (15) a. Hansi lsst [AcI mich *sichi/ihmi ein Buch geben].
Hans lets me self/him a book give

(F.)

Hans has me give him a book. b. Der Knigi lsst [AcI den Gefangenen *sichi/ihmi huldigen].
the king lets the prisoner self/him honor (F.)

The king has the prisoner honor him. c. Hansi hrt [AcI den Professor *mit sichi/mit ihmi sprechen].
Hans hears the professor with self/with him speak (F.)

Hans hears the professor speak with him. (Frey 1993: 116)

184

As for examples like (4), repeated here as (16), with a transitive AcI-infinitive where the anaphoric element is embedded in what Haider (1985) calls a non-argument PP, Freys system is similar to Haiders. (16) Eri lie [AcI die Leutej [PP fr sichi/j/ihni] Schnaps besorgen].
he let the people for self/him liquor get

(H., L.-S.)

He had (allowed) the people (to) get liquor for themselves/him. (Haider 1985: 244) If the preposition lacks an external argument in Freys terms, if the external argument of the preposition is identified with the event q-role of the sentence (p.127) there is no SUBJECT in the local domain of the reflexive, and according to part (ii) of Freys Principle (A), the reflexive must be bound by a SUBJECT in the minimal CFC that contains a domain closing element. The extended binding domain is then the whole sentence. This is illustrated in (17) (Freys (55a), ch. 7). (17) Hansi lsst [AcI ihnj [PP bei sichi/j] arbeiten].
Hans lets him at self work

(F., L.-S.)

Hansi lets himj work at hisi/j place. (Frey 1993: 128) The reflexive sich can be bound by either Hans or ihn. As in Haiders system, the ambiguity results from the fact that the reflexive binding condition (Freys Principle (A) given in (12)) consists of a two-part requirement. Since the minimal CFC, the PP bei sich, does not contain an external argument, it is SUBJECT-less and therefore
185

does not fulfill part (i). Part (ii) demands that the binding domain now be extended, crucially not to the next bigger CFC that contains an external argument (the AcIsubject) but to the minimal CFC that contains a domain-closing element, namely T. With the entire sentence as its extended binding domain, the reflexive then has two possible antecedents, the AcI-subject, ihn, and the matrix subject, Hans. Unlike Haider (1985), Frey also has an account for PP-embedded pronominals (i.e. non-reflexive pronouns). Freys Principle (B) differs from his Principle (A) in that it consists of only one step. For pronominals, there is no local domain defined as a CFC that contains all the licensors of the anaphoric element. The only domain that counts for a pronominal is the minimal CFC that contains a SUBJECT. If the infinitive is transitive or unergative, the CFC is the AcI, with the AcI-subject as the SUBJECT. In this domain, the pronominal can be free and therefore grammatical. Interestingly, all of Freys adjunct-PP examples are locatives introduced by prepositions like neben next to and bei at (someones place). There is no question about the adjunct status of these PPs since they are clearly optional and unpredictable (in the sense that their content is not selected by the verb).12 Frey gives examples like (18a-b) (his (11c) and (55b), ch. 7), where the binding behavior of a reflexive embedded in an adjunct-PP (neben sich next to self) is contrasted with that of a reflexive embedded in an argument-PP (mit sich with self).
12

This does not hold for the context of verbs like setzen, stellen, and legen (all put), where

locative PPs, like neben next to, are in fact selected. 186

(18)

a. Hansi hrt [AcI den Professor [PPadj neben sichi] sprechen].

Hans hears the professor next to self speak

(F., L.-S.)

Hans hears the professor speak next to him. b. Hansi hrt [AcI den Professor [PParg*mit sichi/mit ihmi] sprechen].
Hans hears the professor with self/with him speak (F.)

Hans hears the professor speak with him. (Frey 1993: 116) As also claimed by Haider (1985), the reflexive can only be bound by the matrix subject, across the intervening AcI-subject, if it is embedded in an adjunct-PP. Frey is more explicit than Haider in that he distinguishes between adjunct-PPs and argument-PPs by assuming that the former constitute a CFC with an external argument that is identified with the event q-role of the sentence,13 whereas argumentPPs are not CFCs at all because they do not contain all the licensors of the anaphoric element. The licensor of an anaphoric element embedded in an argument-PP is the verb, not the preposition. Although Freys system is an improvement over Haiders, there are still problems. First, though the distinction between adjuncts and arguments is notoriously fuzzy, both Haider and Frey categorically classify the fr (for)-PP in (16) as an adjunct. Intuitively, however, this fr-PP is much more closely connected to the verb besorgen
13

The event q-role, the referential argument of the sentence, is not licensed by the adjunct.

(Frey 1993, p. 45). Consequently, this external argument does not count as a SUBJECT, and the binding domain for the reflexive must be extended beyond the PP to include a possible antecedent. 187

get/buy than the neben (next to)-PP in (18b) is to the verb sprechen speak. This intuition is confirmed in that besorgen just like English get or buy, is often complemented by a benefactive, i.e. the person for whom something is gotten or bought. The benefactive can either be expressed as a fr-PP or a dative DP. (Sie besorgten ihm (DAT) Bier. They got him some beer.) Since the PP-versions of applicatives (like benefactives and passive by-phrases) are considered systematically optional arguments,14 it seems that the fr-PP in (16) could be either an adjunct (licensed by the preposition) or an argument (licensed by the verb). Again, Reis (1973), who acknowledges these problems, notes that a good deal of lexical variation is to be observed (p. 522). The second problem concerns Freys unaccusativity analysis, as explained above. It unexpectedly rules out the pronominal in Reis (1973) examples given in (9a-b), repeated here as (19a-b). More specifically, the problem is that auxiliary choice does not seem to correlate with the binding possibilities in the way that Freys account suggests. (19) a. Hansi lsst [AcI die Mnnerj [PP ber sichi/ber ihni] herfallen].
Hans lets the men upon self/upon him fall

(R., L.-S.)

Hans lets the men attack him.

14

Reis (1976) calls fr-PPs semi-obligatorische Verbergnzungen (p. 28). 188

b. Hansi lsst [AcI die Verantwortung [PP auf sichi/auf ihni] zukommen]. (R., L.-S.)
Hans lets the responsibility on self/on him to-come

Hans lets the responsibility come to him. (fig. Hans wants to cross that bridge when he gets to it.) (Reis 1973: 522)

The infinitives ber jemanden herfallen attack someone and auf jemanden zukommen come toward/approach someone select sein as their perfect auxiliary. According to Frey, they should thus be unaccusative, and consequently be subject to argument structure unification. The pronominal is predicted to have no chance of being free. The fact that it can be free here suggests that, in Freys terms, argument structures have not blended and that the nominals die Mnner men and d i e Verantwortung responsibility have external argument status. This in turn means that the infinitives cannot be unaccusative despite the fact that they select sein. Thus, neither Haider nor Frey offer a convincing account for Reis well-established data in (19a-b). One of the points I make in the following subsection is that pronominal binding facts are a better probe for unaccusativity than auxiliary choice (see also LeeSchoenfeld (to appear)).

5.2.3 Reconsidering the facts As the preceding discussion reveals, neither of the previous analyses presented here is able to account for all of the commonly agreed upon binding facts in examples involving AcI-constructions. Furthermore, the empirical generalizations presented in
189

Reis, (1973, 1976), Haider (1985, 1993), and Frey (1993) are in need of sharpening and revision. In particular, according to their work, there is no reflexive binding ambiguity when the reflexive is an argument (or, for Reis, a non-prepositional object NP) of the AcI-infinitive; sich has to be bound by the external argument of the AcI. Thus, Case 3, namely coreference of a reflexive in argument (or non-PP) position with a DP beyond the intervening embedded subject is strictly ruled out. Section 5.4 aims to give an account of binding in AcI-constructions that explains both the facts reported in the literature thus far (Case 2) and the judgments at issue here (Case 3). There needs to be an explanation, then, for (5a-b), repeated here as (20a-b), where sich is a non-PP argument of the AcI-infinitive and can be bound by either the AcI or the matrix subject. (20) a. Die Mutteri lsst [AcI die Kleinej
the mother lets

sichi/j/ihri/*j die Schokolade in den Mund

(L.-S.)

the little girl (ACC) self/her (DAT) the chocolate (ACC) in the mouth

stecken].
stick

The mother lets the little girl stick the chocolate in its/her mouth.

b. Die Spieleri hren [AcI die Fansj

the players hear

sich?i/j/siei/*j

anfeuern].

(L.-S.)

the fans (ACC) self/ them (ACC) on-cheer

The players hear the fans cheer them on. In (20a), sich is a (non-core) possessor dative (PD), and in (b), it is a direct accusative-marked object. Although the more local binding possibility is generally preferred for the reflexive (especially in (20b)), binding across the intervening
190

embedded subject is not impossible. As for the pronominals (ihr and sie), there are three options: They can be bound by the matrix subject, refer to a DP from previous discourse, or be used deictically. Intuitively, it is precisely this ambiguity involving the pronominal that allows the reflexive to provide an alternative way of referring to the matrix subject. For many of the consulted informants, it is most natural to think of the pronominal in (20a) as referring to a third person not mentioned in the sentence. At the same time, although they find it possible, they do not feel completely comfortable with the reflexive referring to the matrix subject either. There does not seem to be a perfect way of establishing this binding relation. When it comes to an accusative-marked reflexive as in (20b), the long-distance binding possibility is even harder to get. Descriptively speaking, it seems that acceptability of the reflexive decreases, the more clear-cut the argument status of the phrase containing it is. Still, if the speaker puts him or herself in the position of die Spieler the players, and the context of the described situation is clear, the reflexive cannot be ruled out it is at least possible and in this way contrasts sharply with (6a) in which the reflexive is within a finite CP and binding is utterly impossible. Again, since German reflexives cannot be used logophorically and thus be exempt from the binding conditions, the judgments in (20a-b) are a significant empirical finding which a binding account must at least attempt to explain (section 5.4 presents the analysis ultimately proposed here).

191

As shown by the example pairs in (21) and (22), what rules out coreference of the reflexive with the matrix subject in Haider and Reis data is probably not the given syntactic configuration. The (a)-examples are taken from Haider and Reis and represent their judgments of the binding facts. The (b)-examples are slightly altered versions of the respective counterparts in (a) and show that the binding possibilities change, or are less constrained than they seem to be, depending on lexical choices and possibly perspective. Crucially, in all of the following examples, the anaphoric element is a dative-marked (non-PP) nominal. (21) a. Eri lie [AcI die Leutej sichj/*i/ihmi/*j Schnaps besorgen].
he let the people self/him (DAT) liquor get

(H.)

He had (allowed) the people (to) get liquor for him/themselves. (Haider 1985: 244) b. Eri lie [AcI den Oberj sichj/?i/ihmi/*j einen Schnaps bringen].15
he let the server self/him (DAT) a liquor bring (L.-S.)

He had (allowed) the server (to) bring him a shot of liquor/bring liquor for himself. (22) a. Hinzi lsst [AcI Kunzj *sichi/ihmi
Hinz lets Kunz

den Brief geben].

(R.)

self/ him (DAT) the letter give

Hinz has Kunz give him the letter. (Reis 1973: 522)

15

Coreference between den Ober and sich is pragmatically very odd here. With the addition

of the emphatic element selbst self, however, the coreference relation becomes acceptable and is therefore marked as grammatical here. 192

b. Hinzi lsst [AcI Kunzj sich?i/ihmi den Brief auf den Schreibtisch legen].
Hinz lets Kunz self/him (DAT) the letter on the desk put

(L.-S.)

Hinz has (allows) Kunz (to) put the letter on his desk.

Without making a significant phrase-structural difference, the slightly altered context in the (b)-examples clearly facilitates the binding possibility between reflexive and matrix subject. The same kind of observation can be made about Freys (non-PP) examples in (15a-b). The reflexive can reach beyond the embedded subject if the context is slightly changed. As for pronominal binding, a system like Freys, in which all sein-selecting verbs are considered unaccusative, has no way of accounting for the binding facts in (23). (23) Der Knigi lsst [AcI den Gefangenenj vor sichi/j/ihmi niederknien].
the king lets the prisoner before self/him down-kneel

(L.-S.)

The king has the prisoner kneel down before him. Despite the fact that niederknien is a sein-verb, many of my consultants and I find that both reflexive and pronominal are perfectly acceptable when coreferring with the matrix subject. Unlike the pronominal, the reflexive also has the possibility of being bound by the closer, embedded subject. Given the strong tendency of sich to be bound locally, this is unsurprising. Pragmatically, of course, the coindexation of sich and den Gefangenen is only possible if the non-linguistic context provides a mirror.16 In sum, based on the judgments I report on here, the binding relations in (23) are no
16

This corresponds to a Proxy-reading of the reflexive, as discussed in Safir (2004). 193

different from examples with transitive or unergative AcIs. There is then no motivation for argument structure unification and thus no evidence of a coherent construction. The judgments do not seem to vary when it comes to examples like (14a) (the causative passive example) and (14b), repeated below as (24a-b).17 Here, an account based on unaccusativity and blended argument structure becomes much more attractive (though, as we will see, complex verb formation and argument structure blending are not ultimately devices that will be needed for the proposal I develop). (24) a. Hansi lsst sichi/*ihmi ein Buch
Hans lets

von Maria geben.

Maria give

(F., L.-S.)

self/him (DAT) a book (ACC) by

Hans has Maria give him a book. b. Hansi lsst sichi/*ihmi einen Stein auf den Kopf fallen.
Hans lets self/him (DAT) a rock (ACC) on the head fall (F., L.-S.)

Hans lets a rock fall on his head. (Frey 1993: 116) In these cases, where the pronominal is clearly ungrammatical and the reflexive only has one binding possibility, it is reasonable to attribute the transparency of the AcI to
17

Note that, the AcI-introducing verb lassen has many different shades of meaning going

beyond permissive and causative (see Hhle 1978 for a discussion of 8 types of lassen). I follow Reis (1976), however, in assuming that there is no evidence for structural distinctions based solely on these different shades of meaning. In section 5.3, I propose a structural distinction between transitive/ unergative and unaccusative AcIs, which reflects at least some of the semantic differences between the many types of lassen. 194

the unaccusativity of the infinitive. This means, however, that niederknien kneel down in (23), which does not enter into a coherent (in the sense of blended) construction with lassen, must not be an unaccusative verb, despite the fact that it selects sein. I conclude that auxiliary selection is not a reliable diagnostic for unaccusativity. Rather, as suggested by the contrast between (23) and (24b), it is the agentivity (i.e. the properties associated with proto-agent arguments) of the AcIsubject which classifies an intransitive AcI-infinitive as unaccusative or unergative. While den Gefangenen the prisoner in (23) is agentive and seems to be able to turn the AcI into an opaque domain within which the pronominal can be free, einen Stein a rock in (24b) and (25a) is non-agentive and does not constitute a domain boundary relevant for the pronominal. Some minimal-pair-like examples that support this contrast in agentivity and its correlation with the binding facts are shown in (25)-(28).18 When the AcI-subject is non-agentive, as in the (a)-examples, coreference between the matrix subject and the pronominal is generally impossible. When the AcI-subject is agentive, as in the (b)examples, on the other hand, it is suddenly much easier to accept the pronominal as coreferring with the matrix subject. While the reflexive is always preferred, the pronominal in the (b)-examples is at least marginally acceptable and consistently

18

The focus is on the acceptability of the pronominal here. As for the reflexive, it can be

coindexed with both the matrix and the AcI-subject where pragmatically possible. In (28b), for example, sich can be bound by the AcI-subject die Polizisten if it takes on its reciprocal function. 195

better than in the (a)-examples. To show that it really is agentivity19, not auxiliary selection, that makes the difference here, I chose sein-verbs for all of the following AcI-infinitives.20 (25) a. Der kleine Jungei lsst [AcIden Stein sichi/ihm*i auf den Kopf fallen]. (L.-S.)
the little boy lets the rock self/him on the head fall

The little boy lets the rock fall on his head. b. Die Gromutteri lsst [AcIden Wellensittich sichi/ihr?i auf den Kopf fliegen].
the grandmother lets the parakeet self/her on the head fly

(L.-S.)

The grandmother lets the parakeet fly onto her head.

19

In (25)-(28), all the (b)-examples have proto-agent AcI-subjects, which are literally

agents. Recall from chapter 2, however, that I am using the term agentive subject to refer to any subject with proto-agent properties in the sense of Dowty (1991), i.e. subjects that are agents, experiencers, causers, etc. The AcI-subjects in (i)-(iii), for examples, are not agents but (according to my judgments) still induce opacity in that they are enough of a separation between matrix clause antecedent and pronominal. (i) Hansi lie die Krankheit sich?i/ihni ohne jede Gegenwehr tten. Hans let the disease kill him without any resistance. (ii) Hansi lie die Kugel sich?i/ihni nicht treffen. Hans did not let the bullet hit him. (iii) Der Arzti sah das Medikament sich?i/ihm?i helfen. The doctor saw the medication help him.
20

(L.-S.) (L.-S.) (L.-S.)

Again, the bracketing in the examples with a possessor dative is an expositional

simplification in that it abstracts away from the discussion in chapter 4. As explained there and also in chapter 6, a PD-licensing affectee vP in a construction with an unaccusative AcI is probably projected by the matrix verb, and the non-agentive AcI-subject, which originates in Spec VP of the AcI, may precede the PD as a result of scrambling. Crucially, the pronominal here cannot be bound by the matrix subject, whether the AcI-subject precedes it or follows it. 196

(26)

a. Willii lsst [AcIdie Chance sichi/ihm*i nicht durch die Finger gleiten]. (L.-S.)
Willi lets the chance self/him not through the fingers slide

Willi doesnt let the chance slip through his fingers. b. Majai lsst [AcIdie Katze sichi/ihr?i nicht ins
Maja lets the cat self/her not

Haus kommen].

(L.-S.)

into-the house come

Maja doesnt let the cat enter her house. (27) a. Der Bauarbeiteri sieht [AcIdas Gerst auf sichi/ihn?*i niederstrzen]. (L.-S.)
the scaffolding on self/him down-crash

the construction worker sees

The construction worker sees the scaffolding crash down on him. b. Der Knigi lsst [AcIden Gefangenen vor sichi/ihmi niederknien].
the king lets the prisoner before self/him down-kneel (L.-S.)

The king lets the prisoner kneel down before him. (28) a. Brittai lie [AcIden Ball auf sichi/sie?*i zurollen].
Britta let the ball on self/her to-roll

(L.-S.)

Britta let the ball roll toward her. b. Die Demonstrantini sah [AcIdie Polizisten auf sichi/siei zukommen].
the demonstrator (FEM) saw the policemen on self/her to-come (L.-S.)

The demonstrator saw the policemen come toward her. It is evident that the pronominal generally becomes more acceptable when embedded in a PP. While the pronominal possessor datives are clearly ungrammatical in (25a) and (26a) and noticeably better in the corresponding (b)-examples, the PP-embedded pronominals in (27) and (28) get better judgments overall, in both the (a) and (b)examples. As I explain in section 5.4, PP-embeddedness can render a syntactically
197

bound pronominal acceptable, regardless of whether or not the AcI-subject is agentive. It would be possible, in principle, to build reference to agentivity directly into the definition of binding domains. It should be clear from the previous chapters of this dissertation, however, that I explore a different path. It has been established that the crucial property of a non-coherent (opaque) AcI is the presence of an agentive AcIsubject. Recent work, in the Minimalist Program and elsewhere, has explored the syntactic correlate of expression of an agentive (in traditional terms external) argument. This line of thought opens up the possibility of defining binding domains in phrase-structural terms and thereby capturing the agentivity effects just documented indirectly. I argue, beginning in section 5.3, that this framework yields a relatively straightforward account of the facts. More specifically, I will show that it is the presence or absence of the syntactic head introducing the agentive AcI-subject which determines the binding possibilities for the pronominal. As explained in chapters 3 and 4, this head is v, the instantiation of Burzios Generalization, functioning as both assigner of the external argument (proto-agent) role and licensor of accusative case.

5.3

Binding domains and agentivity

To reiterate the basic intuition that drives my reanalysis of the binding facts in German AcI-constructions, I propose that the reflexive binding possibilities need to
198

be less restricted than usually assumed and that the distribution of reflexives and pronominals partially overlaps when the AcI-subject is agentive. More specifically, for some speakers, the only binding restriction on the reflexive seems to be that it be bound within the minimal TP containing it. As for the pronominal, it can be bound within the same TP as the reflexive, as long as it is contained in a smaller domain within which it can be free. After a discussion of the syntactic correlate of agentivity in 5.3.1, the analysis will be extended to instances of binding in the nominal domain in 5.3.2.

5.3.1 A phrase-structural analysis of unaccusativity Instead of relying on a concept like Freys (1993) argument structure unification (either base-generation of a complex verbal head or a derivational process of restructuring), which is difficult to reconcile with recent assumptions concerning the building blocks and operations available to narrow syntax (see Chomsky 1995, 2000, 2001), I appeal to a theory of phrase structure which allows for a clear-cut and purely syntactic distinction between passive and unaccusative verbs on the one hand and transitive and unergative verbs on the other. Within this framework, the

configurations of these types of verbs are more radically different than in the classic Principles & Parameters system. This more radical difference is desirable because it offers a solution to the following puzzle. If, as traditionally assumed, the argument of an unaccusative verb, though base-generated as the verbs internal argument, moves

199

to check case in external argument position, it ends up in the same position as the argument of an unergative verb and the higher argument of a transitive verb. Thus, structurally speaking, both types of verbs have an external argument. Why then should the binding behavior of their arguments (or more specifically, the transparency/opacity-effects they induce) be different? Recall that, within the framework of the phrase structure theory assumed here (see Hale & Keyser 1993, Chomsky 1995, and Kratzer 1996), transitive and unergative verb structures are syntactically complex in that they subsume both a lexical VP and an agentive vP-projection, as shown in (29). The verb originates as the V-head and raises to little v. 21 Each of the verbal heads assigns a q-role to the argument with which it merges. The verbs internal argument then is the complement (or specifier) of V, while the external argument is in Spec vP. Crucial to the analysis to be developed here, an agent role can only be assigned by v. (29)

vP 2 DP v' SUBJ 2 VP v 2 (DP) V (OBJ) | unerg./(tr.)

21

At least in the case of finite verbs, V-heads in German are generally assumed to move to

the v-heads they project. 200

In contrast to this double-layer configuration of unergative and transitive verbs, the characteristic structure of passive and unaccusative verbs is less elaborate. It lacks vP and thus consists of the lexical VP only. No agent role is assigned. This is shown in (30). (30)

VP 2 DP V SUBJ unacc.

It follows that only subjects of transitive and unergative, not passive and unaccusative verbs can receive an agent role. I assume that either one of the verbal structures in (29) and (30) can occur as complement of AcI-matrix verbs. The tree structures in (31a-b)22 show both the unergative and the unaccusative type of AcI-configuration embedded under lassen let, allow, have. The former corresponds to the (b)-example and the latter to the (a)-example of the sentence pair in (27).

22

As in previous chapters, to save space, my tree structures include a mix of left and right-

headed verb phrases. Abstracting away from verb-second related movement allows me to leave out the CP-layer of the diagrammed sentences. On the view that the underlying order (at least in Dutch) is SVO (see Zwart 1997), the left-headed agentive vP in the matrix clause of my structures is in fact standard. 201

(31)

a.

TP 2 DPi T' Der Knig 2 T vP 2 ti v' 2 v VP lsst 2 vP V 2 tv DPj v' den Gefangenen 2 AGENT VP v | V' 2 PP V vor niedersichi/j/ihmi/*j knien

b.

TP 2

DPi T' Der 2 Bauarbeiter T vP 2 ti v' 2 v VP lsst 2 VP V 2 tv DP V' das Gerst 2 PP V auf niederstrzen 23 sichi/ihn*i

23

I give a categorical *-judgment here, although there is an interpretation of this sentence

(especially when the matrix verb is sieht sees) which allows the pronominal, at least marginally (see (27a)). This has to do with either perspective or the finding discussed in subsection 5.4.4-5 that certain PPs can constitute their own binding domains. 202

Case-checking and q-assignment proceed as follows. In both (31a) and (b), the matrix subject checks nominative case with T, while the AcI-subject checks accusative case with the matrix v. The matrix v assigns an external argument role to the matrix subject, which then moves to Spec TP, leaving a trace in Spec vP. In the unergative AcI-structure in (a), the AcI-subject is assigned an agent role by the embedded v, whereas there is no embedded v to assign an agent role to the AcIsubject in the unaccusative AcI-structure in (b). This latter structure crucially lacks a vP-projection. These two types of AcI-structures lead to different binding possibilities. The extra structural layer of unergative and transitive AcIs, resulting from the presence of vP (see (31a)), provides a domain in which the pronominal can be free. The assumption that v is needed to assign a proto-agent role to the AcI-subject explains the intuition that the pronominal is grammatical as long as it is separated from its antecedent by an agentive subject. The crucial binding domain for the pronominal should thus be the minimal agentive vP containing it. The reflexive, on the other hand, can transcend agentive vP boundaries. While it is preferably bound within the same domain in which the pronominal must be free, some speakers find it grammatical, as long as it has a binder within its minimal TP. The preliminary binding conditions for AcI-

203

constructions are thus given in (32). I assume provisionally that subject refers to DPs in Spec vP.24 A formal definition is offered in the next subsection. (32) a. A reflexive g must be bound by a subject within the minimal TP containing g.

b. A pronominal d must be free within the minimal agentive vP containing d. In section 5.4, I will propose a way to eliminate the disjunction (TP versus vP) implicit in (32). For now, consider how this formulation will do its work. In (31a), the pronominal ihm can refer to the matrix subject der Knig because it is free in the minimal vP containing it. The AcI-subject den Gefangenen is the intervening agentive subject. If the AcI-infinitive does not include a vP, as is the case in (31b), the minimal vP containing the pronominal is the matrix vP, which also contains the trace of the matrix subject antecedent. Within this higher vP, the pronominal cannot be free and is thus ruled out. As for the reflexive, sich has two binding possibilities in (31a). Both den Gefangenen and der Knig are within the minimal TP that contains sich and can thus antecede it. (Again, the prisoner would have to be in front of a mirror). In (31b), only the matrix subject der Bauarbeiter is a possible antecedent. Pragmatically, only very few verbs allow for non-agentive subjects to antecede a reflexive. An example is in sich zusammensacken to (internally) collapse as in Er sah den Kucheni in sichi zusammensacken He saw the cake collapse. In general, we

24

I focus solely on subject-oriented anaphora here. For cases of reflexives with object

antecedents, see e.g. Grewendorf (1985). 204

only get non-complementarity of the reflexive and the pronominal if the phrase structure of the AcI corresponds to (31a), i.e. the transitive/unergative type. A consequence of the proposed analysis is that pronominal binding facts may, but auxiliary choice cannot, be taken as a reliable probe for unaccusativity. The fact that motion verbs, for example, select sein be as their perfect auxiliary does not necessarily mean that they are unaccusative. Agentivity and thus the presence of an agentive subject is only one of the factors that influence auxiliary selection. Although agentivity (or control) are cross-linguistically associated with
HAVE-selection,

there

are other factors like change of location, which, combined with agentivity, may lead to BE-selection (see Soraces (2000) Auxiliary Selection Hierarchy). The point is that unaccusativity dynamically depends on the subtleties of sentence interpretation rather than being fixed once and for all for a given lexical item.25

5.3.2 An extension to the nominal domain In the previous subsection, it has been established that AcI-constructions call for binding conditions which have TP as the relevant domain for the reflexive and the minimal agentive vP as the relevant domain for the pronominal. Since the former domain is bigger than the latter, we get non-complementarity. Before proposing a way to unify the reflexive and pronominal domains in section 5.4, I focus here on

25

See Lee-Schoenfeld to appear for an in-depth discussion of the non-correlation of

agentivity and auxiliary choice. 205

determining what impact the findings thus far have on the study of German anaphora in a broader sense. That is, does the current analysis extend to non-AcI contexts? This subsection explores two such contexts: complex DPs with a possessor in Spec DP and plain (i.e. possessor-less) DPs. Consider first the complex DP example in (33). (33) Martini hrt nicht gern
Martin hears not

[DPThorstensj Geschichten ber sichi/j/ihni/*j]. (L.-S.)

stories about self/him

with-pleasure Thorstens

Martin doesnt like to hear Thorstens stories about himself/him. The binding conditions as stated in (32) do not predict the coindexation of the pronominal ihn with the matrix subject to be possible because there is no embedded agentive vP within which the pronominal is free. Here, ihn is embedded in a PP that is part of a complex DP. There is a remarkable parallel between the binding facts in this complex DP example and those in unergative and transitive AcI-constructions. Compare (33) with (34). (34) Martini hrt [AcIThorstenj ber sichi/j/ihni/*j reden].
Martin hears Thorsten about self/him talk

(L.-S.)

Martin hears Thorsten talk about himself/him. The reflexive sich is ambiguous in that it can be bound either by the subject/possessor in the embedded domain or, within TP, by the matrix subject. The pronominal, on the other hand, can only be bound by the matrix subject or refer to someone mentioned in
206

previous discourse. The fact that the pronominal in (33) is at all grammatical suggests that DP, just like vP, constitutes a sufficient domain within which the pronominal can be free. The possessor in Spec DP (see (33)) is the equivalent of the agentive subject in Spec vP (see (34)). Both intervene between the pronominal and its matrix clause antecedent. As for the reflexive binding facts, TP is still the only upper boundary needed. If the possessor in Spec DP is considered a subject, the binding condition for the reflexive, as stated in (32a), still holds. The definition of subject, however, must be broadened now. It needs to include DPs in the specifier of both agentive vPs and DPs. I propose to adopt Safirs (2004) definition (his (21), ch. 5), which I first give in its original wording and then paraphrase in Minimalist terms, in (35). Furthermore, the pronominal binding condition must now allow for a pronominal to be free within the minimal DP containing it. The revision (adding another disjunction to be eliminated shortly) is given in (36). (35) Safir (2004): The syntactic subject of a, a a lexical category, is any b such that b is the SPEC of an extended projection of a, and b is in an A-chain relation with SPEC-a.

207

Adaptation: A subject is any DP-specifier of (the extended projection of) a q-assigning head.26,27 (36)

A pronominal d must be free within the minimal agentive vP or DP containing d.

My adaptation of Safirs definition of syntactic subject includes the specifier of the immediate or extended projection of any lexical category. With respect to the AcIconstructions in (31a-b), the definition may be applied as follows. Both the AcIintroducing matrix verb and the AcI-infinitive are Vs and thus q-assigning lexical categories. Since TP is an extended projection of the matrix V, and the respective matrix subjects (der Knig in (31a) and Hans in (b)) are in Spec TP, these nominals are correctly defined as subjects. The DP den Gefangenen in the embedded Spec vP in (a) also qualifies as subject because Spec vP is the extended projection of the
26

This reformulation of Safirs definition of subject is based on the bare phrase structure

definition of complement and specifier, according to which complements are phrases that are incorporated by first merge to a given head and all subsequent merges produce specifiers.
27

According to this definition of subject, possessor datives, which originate in the specifier

of the possessed nominal and raise to the specifier of a dative-case-licensing vP (see chapter 3), are subjects as well. Although it has been established in the literature that accusative-marked nominals are better reflexive binders than dative nominals, it is not impossible for datives to antecede sich (see e.g. Grewendorf 1985). In connection with the reflexive binding condition stated in (32a), the subject definition in (35) roughly makes the right predictions, even when it comes to object-oriented reflexive binding. Whether or not it can be maintained that possessor datives are subjects in the more general (non-binding-related) sense, however, is a different issue. As discussed in chapter 6, possessor datives do not have the same opacity-inducing quality as proto-agent subjects (introduced by agentive v-heads). 208

lexical V of the AcI. As for the complex DP example in (33), N is a lexical category with DP as its extended projection. Hence, the possessor in Spec DP, Thorsten, is a subject in the relevant sense. In sum, Safirs definition and my adaptation make the right predictions for the reflexive binding facts in constructions with both AcIs and complex DPs. All the possible binders of sich in examples like (31a-b) and (33) are subjects as defined in (35). A question that arises from this discussion is whether the disjunction vP or DP in (36) can be eliminated by referring instead to a subject-containing category (which would be natural). At first sight, data like (37) seem to suggest that the answer is no. It appears that even DPs without a possessor in Spec DP constitute a domain for pronominals. (37) Eri sah [AcI das Boot neben sichi/ihmi untergehen].
he saw the boat next to self/him under-go

(L.-S.)

He saw the boat next to him sink. According to (32b), a pronominal should not be grammatical in an unaccusative AcI. The lack of an agentive vP and thus an intervening agentive subject between the pronominal and its antecedent is predicted to rule out the pronominal. The wellformedness of (37) could be explained by arguing that das Boot neben sich/ihm has the structure [DP [PP ]]. As in the complex DP example in (33), the anaphoric element is then embedded in a PP that is part of a DP. Within the domain of the DP, the pronominal is free. If the [DP [PP ]]-analysis is forced, i.e. if the possibility of
209

interpreting the DP and the PP as separate arguments of the infinitive is excluded, the pronominal seems even more natural. This is shown in (38). (38) Eri sah [AcI [DP das Boot [PP neben sichi/ihmi]], aber nicht [DP das Boot [PP hinter
he saw the boat next-to self/him but not the boat behind (L.-S.)

sichi/ihmi]] untergehen].
self/him under-go

He saw the boat next to him, but not the boat behind him sink. Examples like (37) and (38) would then suggest that, even without a subject, a DP behaves like an agentive vP in that it provides enough intervening structure between a pronominal and its antecedent to license the pronominal. Note, however, that the difference between the complex DP example in (33) and the plain DP examples in (37)-(38) is that the pronominal in the former is embedded in what looks like an argument PP, whereas the pronominal in the latter is embedded in an adjunct PP. The fact that the pronominal is free in (37)-(38) may thus be a consequence of the adjunct PP-boundary which intervenes between the matrix clause antecedent and the embedded pronominal, and not a consequence of the plain DP-boundary. Subsection 5.4.4 confirms that pronominals embedded in adjunct PPs are free regardless of what type of DP the PP may be contained in. Hence, there is no reason to believe that plain DPs are equally opacity-inducing as complex DPs. Just as verbal binding domains need to be closed off by an agentive subject, nominal binding domains seem to require a subject (a possessor in Spec DP) in order to be opacity-inducing. In the next section, I will address what agentive vPs and complex DPs have in common and
210

sketch a proposal of how to eliminate both the vP or DP-disjunction in the pronominal binding condition and the need for distinct domains for reflexive and pronominal (TP vs. vP/DP).

5.4

Taking the analysis to the next phase

At this point, it has been established that a German reflexive which is not embedded in a PP is most readily acceptable when bound locally (i.e. within the minimal agentive vP containing it), but that, for at least some speakers, it can also reach beyond an embedded vP-boundary and engage in apparent long-distance binding. Furthermore, both agentive vPs and complex DPs have been shown to be relevant binding domains for the pronominal within TP (Case 1). The goals of this final section are (i) to account for the parallel between the binding facts in transitive/unergative AcI and complex DP-constructions and thus, more generally, between vP and DP as binding domains and (ii) to simultaneously explain the widely attested default case of complementarity between reflexive and pronominal, while still allowing for the (at times marginal) long-distance binding ability of the reflexive, which leads to non-complementarity (Cases 2 and 3).

5.4.1 Unifying vP and DP One way to shed light on the parallel between vP and DP as pronominal binding domains is to appeal to Chomskys (2000, 2001) notion of phase.
211

Within a Minimalist framework, the derivation of a sentence proceeds in phases. A phase is a subpart of a derivation whose internal content is closed off to grammatical interactions with external elements. The only parts of a phase which can interact with external elements (introduced later in the computational sequence) are those at its edge. The edge of a phase in turn consists of its defining head, the specifiers of that head and any adjoined material. Among the grammatical operations which cannot penetrate phase boundaries are feature-checking operations; it follows that higher heads can establish such relations with material inside a lower phase only if that material is at the phase-edge. Once a phase is completed, its contents get evaluated at LF (and also at PF, though that will be of less concern here). Any unchecked features, or material that is otherwise dependent on phase-external elements, causes the derivation to crash. One of the category-types that define phases is CP. As shown, for example, by the successive-cyclic nature of Wh-movement, raising out of a CP entails stopping off in Spec CP (McCloskey 2002). With respect to the current analysis, it is crucial that vP and DP have also been argued to constitute phases. Chomsky (2000) (building on work by Fox (2000) and Nissenbaum (2000)) makes the argument for vP; McCloskey (2000) and Svenonius (2004) for DP. In both cases, extraction seems to proceed via the respective specifier position. If vP and DP indeed have this commonality, the binding condition for the pronominal does not need to be

212

stated as a disjunction (see (35)). It can simply require that the pronominal be free in the smallest phase that contains it.28 While allowing for a unified, more general pronominal binding condition, the proposal that vP and DP (in addition to CP) define phases has interesting and potentially problematic consequences for the reflexive binding condition. Since the internal contents of a phase are not accessible to higher elements, the binding requirements of a phase-internal reflexive cannot be satisfied by a phase-external antecedent. While the subject in Spec vP of transitive or unergative AcI-

constructions, for example, is an accessible binder for a phase-internal sich, the matrix subject (being external to the embedded vP-phase) is not. The question is then how to explain the possibility (for some speakers) of long-distance binding for sich in AcI-examples like (31a) and complex DP-examples like (33). What allows sich to be bound by the matrix subject? The following subsections offer a solution based partly on Safirs (2004) proposal of drawing a parallel between German sich and reflexive clitics in Romance, which are known to be able to raise from inside VP to the inflectional layer.

28

Since we have come across at least two different kinds of vP agentive and affectee it

remains to be seen whether all kinds of vP have the opacity-inducing quality that is characteristic of phases. This issue will be addressed in chapter 6. 213

5.4.2 Reflexive raising In The Syntax of Anaphora, Safir (2004) develops a theory in which the distribution of anaphors and pronominals is determined by universal principles applying to select the best available form-to-interpretation match (ch. 3, p. 1). As there can only be one best available form, the distribution is complementary. Non-complementarity can only exist if there is a difference in interpretation between the use of the reflexive and the use of the pronominal. Since the data at issue here show a large area of noncomplementarity which does not coincide with clear interpretive differences, the empirical generalizations seem incompatible with Safirs theory. In his discussion of German AcI-constructions, however, Safir does point out a characteristic of the reflexive pronoun sich that provides the means for reconciling his theoretical claims and the empirical observations here. His discussion is based on the mainstream view that sich can be bound in a clause higher than its thematic assignment (p. 19, ch. 5) when it is embedded in a PP. Safir gives one of Reis (1976) examples, here shown in (39). (39) Hansi lsst [AcIdie Mdigkeit [PPber sichi] kommen].
Hans lets the tiredness over self come

(R., L.-S.)

Hans lets tiredness overcome him. (Safir 2004: (28a), ch. 5, p. 19)

214

To account for the larger domain of the reflexive, Safir compares German sich to the French reflexive clitic se.29 He assumes that covert clitic movement from prepositional object position is possible in German for sich and that certain causative constructions permit the domain of covert clitic movement to pass a specified subject (p. 19, ch. 5). In other words, the reflexive comes to be in the same domain as its antecedent by covert movement from the AcI into the higher clause.30 Safirs covert clitic movement, i.e. the ability of the reflexive to raise covertly, justified by the fact that the Romance reflexive has the ability to do so overtly, constitutes the starting point for the account of reflexive raising I develop in the following subsections. If sich is able to covertly move up and adjoin to the edge of its vP or DP-phase, even if it is not introduced as the object of a preposition, it becomes accessible to elements merged later in the derivation. It can then find the matrix subject as its antecedent, while still being pronounced as part of the embedded

29

See also Pica 1987. Although I follow Safir in rejecting Picas account of head movement,

it should be noted that Safirs proposal of covert clitic movement for German sich relies on the essence of Picas analysis, uniting subject-orientation and movement domains.
30

This explains the grammaticality of the reflexive, but it does not provide an answer for why

the pronominal can be considered grammatical as well (see (i)). Reis (1973) puts a question mark next to the pronominal, and I agree that ihn him is at least marginally acceptable.
(i)

Hansi lsst [AcIdie Mdigkeit [PPber sichi/?ber ihni] kommen].

Hans lets the tiredness over self/over him come

(R., L.-S.)

Hans lets tiredness overcome him. (Reis 1973: 522)

Safir does not discuss the pronominal in this context. It is unclear why, in his terms, the most dependent available form, namely the reflexive sich, does not obviate the pronominal ihn. 215

domain. Before we get to the details of the account, however, I will lay out some basic assumptions concerning movement and the nature of reflexive binding. Assuming the COPY T HEORY OF MOVEMENT, according to which movement consists of the operations copy, merge, and delete, the first two steps of covert movement proceed just as they would in overt movement. The only difference lies in the deletion-part of the two types of movement. Instead of deleting (or not pronouncing) all but the highest copy of the moved element, covert movement only leaves the lowest copy to be pronounced (see e.g. Bos&kovic@ 2001, Bobaljik 2002, and Reintges, LeSourd & Chung (to appear)). On this view, covert reflexive raising then happens in narrow syntax. Since the goal here is to explain a case of non-complementarity, it is crucial that the pronominal be unable to undergo this covert raising process. But this seems reasonable given that reflexives are generally grammatically active in ways that pronominals are not. Reflexives, for example, must be syntactically bound, while pronominals can refer to an antecedent mentioned in previous discourse or may not have a linguistic antecedent at all. As has been established by Kayne (1975), the binding behavior of French clitics involves a type of agreement relation that is closely related to the system of A-movement. Both clitics and A-moved phrases can engage in apparently unbounded grammatical interactions by means of establishing successive-cyclic local relations. Cross-linguistically, reflexives in particular have special morphosyntactic properties that often implicate A-movement (Burzio 1986). Although it is unlikely that the feature-checking operation driving A-movement and
216

the mechanism responsible for establishing anaphoric binding relations are the same thing, it is reasonable to assume that the two are subject to the same basic locality restriction. Within the theoretical landscape of Minimalism, where derivations are subdivided into phases, such a locality restriction is necessarily encoded featurally. Features are either actively involved in narrow syntax and must be checked before LF, or they persist throughout narrow syntax and get evaluated at LF, i.e. once the syntactic object which has been built up to that point gets transferred to the LF interface.31 In the case of binding, I assume that evaluation of the relevant features happens at LF, upon completion of each phase (see also Baltin 2003). The reason is that, unlike A-movement, binding appears not to be driven by f and case-feature checking. Both reflexives and pronominals show person and number agreement32 with the nominal they refer to, but case-checking of an anaphoric element and its antecedent involves two completely separate checking relations. Binder and bindee check case independently of each other with two different case-licensing heads. The local nature of reflexive binding must then be the result of a feature or property which

31

Note that evaluation at LF does not mean that there is a separate derivational LF level

which has access to the entire syntactic object and where this object can be further manipulated. Rather, it means that the feature combinations created by narrow syntax are interpreted at the LF interface. If there are any narrow-syntactic featural requirements which are not satisfied at the interface, the derivation cannot be properly interpreted. I propose here that Condition A does not involve such uninterpretable narrow syntactic features.
32

While the 3rd person reflexive pronoun is the invariable sich-form, the 1st and 2nd person

forms vary. They are homophonous (or, on another view, identical) with the corresponding nonreflexive pronouns. 217

does not get eliminated in narrow syntax but demands that, at the LF interface, reflexive and antecedent be part of the same portion of the syntactic object (phase), being evaluated at that point (Condition A). The opposite requirement (Condition B) holds for the pronominal: its antecedent may not be part of the same phase. The reason we get cases of non-complementarity is precisely that reflexives, but not pronominals, have the ability to undergo raising of a type similar to the classic Amovements. Note that the inability of the reflexive to be bound across a TP or CP-boundary (Case 1), i.e. the fact that cases of non-complementarity do not extend beyond tensed clauses, also falls out from the parallel we are drawing between covert reflexive raising in German and its overt counterpart in Romance. It is a matter of fact that overt reflexive clitics target a position no higher than the inflectional layer (i.e. not the edge of a CP-phase). In Reinhart & Reulands (1993) terms, reflexives form a special kind of predicate, so that (loosely speaking) reflexive pronouns most naturally occur in the verbal domain (including functional vP and inflectional projections, but not CP). Put another way, the movement operations which reflexives undergo are Amovements, solely within the inflectional layer (again, see Kayne (1975)). In the general case of local reflexive binding, then, where the reflexive is not at the phase-edge, the only possible antecedent is the closest subject. When it comes to long-distance reflexive binding, however, there are two ways (corresponding to Case

218

2 and Case 3) for the reflexive to be a part of the next higher phase. Case 2, which exists for all speakers of German, will be addressed in subsection 5.4.4. Case 3, which is at issue here and only available to some speakers, involves covert reflexive raising. This next subsection discusses the details of the narrow-syntactic

mechanisms which I propose lead to Case 3.

5.4.3 Binding by phase Consider again the AcI-binding scenario in (40). (40) Die Spieleri hren [AcIdie Fansj sich?i/j anfeuern].
the players hear the fans self on-cheer

(L.-S.)

The players hear the fans cheer them/themselves on. The reflexive binding possibility that is most readily available to all speakers (although pragmatically dispreferred) is marked with the j-index. In this case, no binding-related operation takes place in narrow syntax. The feature bundle

representing sich probably includes a feature that identifies it as a reflexive, but this is not a feature which triggers an Agree or movement relation. It is presumably an interpretable feature which persists throughout narrow syntax and comes to be significant at LF. It is not clear whether this interpretable feature is connected to the LF evaluation process we associate with Condition A, but it seems safe to assume that Condition A (whatever it is exactly) comes into the picture once the narrow syntax has completed the embedded vP (the AcI)-phase (again, see Baltin 2003). The AcI
219

passes the evaluation here because the phase contains a binder (die Fansj) for the reflexive. Since only the material at the edge of the AcI-vP, namely the AcI-subject in Spec vP is part of the next higher phase and thus still accessible to the rest of the derivation, sich, which is trapped within VP, cannot have a matrix-clause antecedent. The other (more marginal) reflexive binding possibility in (40), namely Case 3, is marked with the i-index. Here, a binding relation is established between the reflexive and the matrix subject (die Spieleri) although sich is the internal argument of the AcIinfinitive (within the embedded VP). To be accessible to the matrix subject at the point when the matrix clause portion of the derivation gets transferred to the LF interface, sich must have raised to Spec vP, the edge of the AcI. Since the material at the edge of the embedded vP and the contents of the matrix vP undergo the LF interpretation process as part of the same phase, Condition A is satisfied. The crucial question is what triggers reflexive raising. Since reflexive pronouns in many languages have special morphology that distinguishes them from their non-reflexive counterparts (in German, this distinction exists for all 3rd person forms), let us follow up on my speculative suggestion above and postulate that all reflexive pronouns (cross-linguistically and cross-dialectally) bear an interpretable reflexive feature. This may then be the same feature that Condition A recognizes when it checks whether reflexive and antecedent are in the same phase at LF. For speakers who allow the long-distance binding scenario in (40)

220

and thus have covert reflexive raising in their grammar, there must be an additional reflexive feature that is uninterpretable and can, when paired with the EPP feature,33 trigger movement. Since, in the case of AcIs,34 reflexive raising targets the edge of the embedded Spec vP, the uninterpretable reflexive feature and the EPP must be on v, rendering it an active probe. Whether the reflexive feature also needs to be uninterpretable on the head of the reflexive DP depends on ones assumptions concerning the exact mechanism of the Agree-relation. It is not entirely clear that both probe and goal must be active. If, in the case at hand, the goal does need to be active and thus bear an uninterpretable feature, this could be a second kind of reflexive feature, which is only part of some speakers grammar. This issue, however, is not central to the proposed analysis. What is crucial is the uninterpretable reflexive feature paired with the EPP on v because it is this feature combination which seeks out the reflexive DP and moves it to Spec vP. (Recall that we are dealing with covert movement here. The copy of sich which gets pronounced is the lower (in-situ) copy.) More generally, it is then the presence of an uninterpretable feature on the embedded v which, by widening the reflexive range, enables sich to be in the same domain as its distant binder. This scenario is very familiar from apparently unbounded syntactic
33 34

The EPP feature demands that the head bearing this feature acquire an extra specifier. We should keep in mind that German also allows long-distance binding across a DP-phase

boundary. This means that the raising mechanism described here must also hold for covert movement of sich to the edge of DP. 221

phenomena like raising and Wh-movement. In order for the moving element to interact with and reach its ultimate destination, there must be stop-off points at intermediate phase-edges, and the movements from phase-edge to phase-edge must be featurally-driven. This leads us to the next question: If only some speakers have the mechanism justdescribed, needed for reflexive raising, in their grammar, what is it that makes longdistance reflexive binding possible for all speakers (i.e. categorically) when sich is embedded in (certain) PPs (Case 2)? An answer to this question is offered in subsection 5.4.4. Before we add PPs to the picture, however, I will restate the binding conditions (cf. (32) and (36)) in phase-based terms. Both pronominal and reflexive now have the phase as their relevant binding domain. (41) a. A reflexive must be bound within the minimal phase containing it.35,36 b. A pronominal must be free within the minimal phase containing it. The conditions stated as such account for the general case of reflexive-pronominal complementarity. The particular case of non-complementarity seen in (40) is a result of reflexive raising, i.e. the ability of the reflexive, but not the pronominal, to covertly
35

As will become clear in subsections 5.4.4 -4.5, this condition has to be refined: the reflexive

must be bound within the minimal phase which contains it and in which its binding requirements can in principle be met (Chomsky 1986).
36

I am leaving out the restriction that the reflexive must be bound by a subject here. All the

possible binders in the type of AcI-constructions we have been examining here are subjects by the definition given in (35) anyway. 222

raise to the edge of its phase. Crucially, being located at the phase-edge means being part of the next higher phase (see Chomsky 2001). More specifically, given two phase-defining heads H1 and H2 (assuming no other phase-defining head intervenes between them):
ext. arg. H1 edge H2

Material at the edge of H2 can be bound by elements in the domain of H1 or by the external argument introduced by H1. Put another way, edge material of a phasedefining head belongs in a certain sense to both phases, the higher and lower. Thus, if a reflexive in an embedded vP or DP moves to the edge of its phase, it is contained in the matrix vP. Within this higher phase, it can find the matrix subject in the higher Spec vP as its antecedent. A pronominal, on the other hand, does not have the ability to raise and is thus always trapped within the lower phase.

5.4.4 Why PP-embedded reflexives are different While long-distance binding between sich and the matrix subject in examples like (40) is only a possibility for some speakers (Case 3), all native speakers of German seem to agree that the PP-embedded sich in examples like (42) can have either the AcI or the matrix subject as its antecedent (Case 2).

223

(42)

Hansi lsst [AcIseinen Kollegenj [PPbei sichi/j/ihmi] arbeiten].

Hans lets his colleague at self/him work

(L.-S.)

Hans lets his colleague work at his place. The reflexive is embedded in a clearly optional and unpredictable bei-PP here. Assuming that non-selected PPs, with a P that assigns its own internal q-role independently from the verb,37,38 are not complements of V but adjoined higher (see Grewendorf 1983), it seems reasonable to postulate v' as the relevant adjunction-site. We then get the following constellation.
(42')
TP 2 DPi Hans T' 2 T vP 2 ti v' 2 v VP lsst 2 vP V 2 tv DPj v' seinen Kollegen2 PP v' bei sichi/j/ihmi 2 VP v tv arbeiten

37

See Hestviks (1991) distinction between three types of PPs: (i) PPs which assign an

independent q-role to the prepositional object (these PPs are CFCs, regardless of whether they are complements or adjuncts), (ii) PPs which assign a q-role to the prepositional object through the verb (these PPs are not CFCs), and (iii) PPs which are not involved in q-role assignment at all, i.e. where the prepositional object gets its q-role directly from the verb (these PPs are not CFCs).
38

A recent account of long-distance binding out of PPs which appeals to the meaningfulness

of P is Gunkel 2003. The more concrete the meaning of the preposition, the more likely it is that the PP in an AcI-construction can undergo (long) reflexivization (p.127). 224

Being adjoined to v', sich is part of the edge of the embedded vP-phase and is thus contained in the next higher, the matrix phase, which contains both seinen Kollegen and Hans. Hence, both the matrix and the AcI-subject are in positions that allow them to antecede the reflexive. Crucially, no reflexive raising to the vP-phase-edge is involved. From the high PP-adjunction site, sich can have access to both subjects without having to move. This then explains why both speakers who have covert reflexive raising in their grammar and speakers who do not, accept the reflexive binding ambiguity here. However, while the PP-adjunction story accounts for the binding behavior of the reflexive, it runs the risk of compromising what has thus far been said about the pronominal. If the adjoined PP in (42') is part of the higher phase, how can ihm (with the given coindexation) be free? The only way to prevent the matrix subject Hans and the pronominal ihm from being in the same minimal phase is to claim that, besides CP, vP, and DP, PP can be a phase as well. This does not affect the reflexive binding behavior because, treating PPs as phases in the sense of CFCs, PPs are inherently subjectless and thus cannot possibly fulfill the binding requirements of the reflexive. Following Chomsky (1986), the reflexive binding domain must then be extended to the next higher subject-containing domain (phase in my terms).39 This

39

As in the case of a vP-phase, the way a reflexive embedded in an inherently subjectless PP-

phase gains access to the next higher binding domain must be by raising to the phase-edge. Regardless 225

may seem like a step backward in that it rephrases part of Freys (1993) CFC-based binding conditions in terms of phases, but the merit of replacing CFCs with phases is that the latter are independently needed to account for island phenomena, ordering, and more generally, reduction of search space (Chomsky 2000), while the notion of CFC is of use only in the domain of binding. As further discussed in the next subsection, if it is independently plausible to assume that PP-adjuncts are phases, both the reflexive and the pronominal binding facts in examples like (42) are accounted for when the PP is adjoined high and thus part of the matrix vP-phase.40

5.4.5 PP-phases In order to account for the pronominal binding possibility in (42), namely between the matrix subject and a PP-embedded pronominal at the AcI-phase-edge, as well as instances of pronominals which can be bound by the matrix subject when embedded in nothing but a PP, we are now asking the question whether, in addition to agentive vPs and complex DPs, certain PPs constitute pronominal binding domains as well. If
of whether or not the raising mechanism is the same within PPs as within vPs, a reflexive must be able to find an antecedent introduced later in the derivation, thus reaching beyond the PP-phase.
40

Note that (42) does not show the full paradigm of the pronominal binding possibilities. No

index is given for the relation between the AcI-subject seinen Kollegen and ihm. Since I argue that the phase-edge-adjoined PP here is a phase, this binding relation is predicted to be a possibility; and it is, in fact, a possibility, as, for example, in Er lsst seine Kollegini bei ihri (zu Hause) arbeiten He lets his colleague work at her place. The reason ihm does not seem to be able to refer to seinen Kollegen in (42) is that both Hans and seinen Kollegen are masculine singular and thus agree with the pronominal ihm. When there is this choice of antecedents, the pronominal is naturally used to refer to the higher antecedent, Hans. 226

they do, and if the goal is to maintain phase-based binding domains, at least certain PPs must be added to the types of constituents that potentially qualify as phases. As it turns out, Baltin (1982) has argued that extraction from PP proceeds by way of its specifier. Assuming that extraction via the specifier of a phrase is a viable diagnostic for phasehood, PPs may indeed be a reasonable addition to the class of phase-like constituents. Furthermore, letting phases do the work of the binding-specific construct CFC, I follow Hestvik (1991) in assuming that CFC/phasehood is characterized by independent and phrase-internally complete q-role assignment, i.e. phrase self-sufficiency and saturation. More specifically, I assume that all PPs, argument or adjunct, whose prepositional object receives its q-role from P (without involvement the verb) are phases. Pronominal binding then serves as a phasehood diagnostic. The following are both AcI and non-AcI examples which suggest that a pronominal can be free inside a PP that is not contained in a vP or DP-phase. Since the use of untergehen in (43a) triggers an unaccusative AcI-structure here, there is no embedded vP-projection. Similarly, there are no embedded vP or DP-boundaries which could explain the acceptability of the pronominals in (43b) and (c). The PPs alone must be the relevant binding domains for the pronominals here. (43) a. Welches Boot lie eri [PP-phase neben sichi/ihmi] untergehen?
which boat let he next-to self/him under-go

(L.-S.)

Which boat did he let sink next to him?

227

b. Eri sah [PP-phase direkt neben sichi/ihmi] eine Schlange auf dem Boden. (L.-S.)
he saw directly next-to self/him a snake on the ground

He saw a snake on the ground directly next to him. c. Eri setzte den groen Teddybren [PP-phase neben sichi/ihn?i].41
he put/sat the big teddy-bear next-to self/him (L.-S.)

He sat the big teddy bear next to him. These data confirm that both adjunct PPs (43a-b) and argument PPs (43c) can be phases.42 Crucially, in both cases, the P assigns its own (independent) q-role to the prepositional object. If all three PPs in (43) constitute pronominal binding domains and thus phases, another question worth asking is which sub-sentential phrases do not qualify as such domains. As for bare VPs (in my system, the immediate and only projection of passive and unaccusative verbs), section 5.3 has already established that they do not
41

The facts here seem to be the opposite in English, at least for those speakers who share the

judgments in (i). (i) Shei put the book next to heri/??herselfi.

The current analysis has no explanation for this contrast between German and English. The English facts in binding contexts like these are notoriously unclear, however. Some speakers find pronominal and reflexive equally acceptable here.
42

While the pronominal in (43c) is only marginally acceptable, it is clearly better than the

pronominals in the q-dependent PPs auf and fr in (44b-c). The reason that ihn in (42c) is marginal could be that the object of the preposition here gets its q-role from both the P and the verb setzen. The verbs setzen, stellen, legen (all put) select a directional PP. Thus, while the exact locative relation between the direct object and the object of the preposition is specified by P, the directional nature of this relation is specified by the verb (cf. footnote 32). 228

provide enough (or not the right type of) structure for a syntactically bound pronominal to be free. In line with Chomsky (2000, 2001), the data presented here thus suggest that the projection of passive and unaccusative verbs (in Chomskys system a defective vP) is not an opacity-inducing (strong) phase.43 This is confirmed by simplex clauses with a transitive verb and a pronominal in direct object position, as shown in (44a). The VP-internal pronominal cannot be coreferent with the VPexternal subject. It is also clear that APs and argument-PPs in simplex clauses like (44b) and (c) are not pronominal binding domains. In neither case can the

pronominal be syntactically bound. Note that the P fr in (44c) does not assign a qrole to the prepositional object. In other words, the P is semantically contentless, and the q-role is supplied directly by the inherently reflexive verb sich interessieren. (44) a. weil der Manni [VP sichi/ihn*i kennt].
self/him knows (L.-S.)

because the man

because the man knows himself. b. Die Elterni sind [AP stolz auf sichi/sie*i].
the parents are proud on self/them (L.-S.)

The parents are proud of themselves. c. Die Fraui interessiert sich nur [PP(non-phase) fr sichi (selbst)/sie*i].
the woman interests self only for self (emphatic)/her (L.-S.)

The woman is only interested in herself.

43

This is contra Legate (2003), who argues that all vPs as well as bare VPs constitute phases.

Her claim is based on reconstruction, quantifier raising, and parasitic gaps in English. 229

Non-phase argument-PPs in unaccusative AcIs are expected to show the same pronominal binding pattern as (44c). Since there is no vP-layer, and since the PP is qdependent on the verb, the pronominal cannot be free. As illustrated by the (a)examples of (27) and (28), repeated here as (45a) and (b), the facts tend to support this expectation.44,45 (45) a. Der Bauarbeiteri niederstrzen].
down-crash

sieht [VP das Gerst [PP(non-phase) auf sichi/ihn?*i]

the scaffolding on self/him (L.-S.)

the construction-worker sees

The construction worker sees the scaffolding crash down on him.

44

The fact that speakers may not completely rule out the pronominals in (45a-b) could have to

do with the possibility of interpreting the PPs as non-selected and thus as phases. The verbs would just be strzen and rollen, and the PPs, consisting of the pre- and post-positions aufnieder and aufzu, would be optional specifications of direction. As explained in subsection 5.4.4, the reflexive would still be able to find the matrix subject as its binder because PPs are inherently subjectless (see Hestvik (1991)) and therefore allow the reflexive to look beyond a PP-phase.
45

In contrast to (45), Reis (1973) example given in (19a), repeated here as (i), allows for a

syntactically bound pronominal because, in my terms, die Mnner is an agentive subject in Spec vP and therefore makes the AcI the relevant binding domain for the pronominal. As for (19b), repeated here as (ii), either speakers interpret die Verantwortung as an agent or causer, so that it is a proto-agent and turns the AcI into a vP-phase, or they analyze the aufzu-PP as an adjoined PP-phase (see footnote 43). Either way, ihn is embedded inside a phase and thus not at risk of violating Condition B. The question of why speakers who do not have covert reflexive raising would get long-distance binding of sich here, will be addressed in connection with semi-obligatory PPs below. (i) Hansi lsst [vP die Mnnerj [PP(non-phase) ber sichi/ber ihni] herfallen]. Hans lets the men attack him. (ii) Hansi lsst [vP die Verantwortung [PP(non-phase) auf sichi/auf ihni] zukommen]. Hans lets the responsibility come to him. (fig. Hans wants to cross that bridge when he gets to it.) (Reis 1973: 522) 230
(R., L.-S.)

(R., L.-S.)

b. Brittai lie [VPden Ball [PP(non-phase)auf sichi/sie?*i] zurollen].

Britta let the ball on self/her to-roll

(L.-S.)

Britta let the ball roll toward her. To sum up, subsections 5.4.4-4.5 have dealt with the following Case 2 binding scenarios: TABLE 3
EXAMPLE TYPE OF ACI POSITION OF ANAPHORIC
ELEMENT

(42) unergative (vP) PP-adjunct (phase)

(43) unaccusative (VP) PP-adjunct/argument (phase)

(45) unaccusative (VP) PP-argument (nonphase) ?*

REFLEXIVE PRONOMINAL

In all three sets of examples, the argument/adjunct status of the PP is relatively clear. The assumption that PP-adjuncts are adjoined to v' (i.e. the AcI-phase-edge) accounts for the invariable grammaticality of the reflexive, while the claim that q-independent PPs are phases explains the acceptability of the syntactically bound pronominal. To reiterate, a reflexive embedded in an inherently subjectless binding domain (phase) has relatively easy access, so to speak, to the next higher domain which contains a potential antecedent. The proposed analysis of PPs then allows for a way to incorporate Reis (1973, 1976), Haiders (1985), and Freys (1993) intuition concerning the long-distance binding ability of PP-embedded reflexives into the broader framework of the current proposal.
231

As discussed by Reis (1976), there are other instances of Case 2 that are less clearcut. Reis presents a number of examples in which both pronominal and reflexive can be anteceded by the matrix subject although they are embedded in a PP that does not straightforwardly qualify as a phase (Reis calls these PPs semi-obligatory). One of my examples that falls into this category is (2), repeated here as (46). (46) Martini hrt [AcI den Mannj [PP ber sichi/j/ihni/*j reden]].
Martin hears the man about self/him talk

(L.-S.)

Martin hears the man talk about himself/him. Although the embedded agentive vP ensures that the i-index of the pronominal is accounted for, regardless of whether the PP is a phase or not, the reflexive is only expected to be judged grammatical by all speakers if it is part of a PP that is adjoined to the AcI-phase-edge. Since it is not obvious that PPs which are potentially selected by the verb should be allowed to occupy a VP-external position, we are faced with a puzzle. So far, I have assumed that only q-independent PPs can occur outside the thematic selection domain of the verb. The fact that the ber-PP in (46) seems to marginally license a syntactically bound pronominal in simplex clauses like Eri spricht nur [ber ?ihni selbst] He only talks about himself, when the emphatic element selbst self is added, suggests that this PP may be treated like the qindependent PPs in (42)-(43). This would explain, then, why speakers who do not have covert reflexive raising, can get the long-distance binding option here: sich has a non-movement way of reaching the embedded phase-edge. Since, besides the
232

notoriously fuzzy argument-adjunct distinction, various non-syntactic factors46 may have an impact here, however, there could well be other reasons for the acceptability of Case 2-binding examples like (46). Abstracting away from Case 2-examples involving semi-obligatory PPs, the system I propose has clear-cut ways to account for both Case 2 and Case 3 binding possibilities. The crucial mechanism needed to explain Case 3 is reflexive raising, a type of movement which is independently needed for many apparently unbounded phenomena (e.g. raising and Wh-movement). Assuming that syntactic derivations proceed by phase, overt or covert movement to phase-edges is necessarily an integral part of the system. Since transitive AcIs containing a sequence of two nonprepositional arguments (often both accusative-marked) are marginal to begin with, even without the involvement of anaphora, it is not surprising that, depending on
46

To address just a few of the non-syntactic factors at play here, Reis data suggest that, even

for speakers who categorically rule out Case 3, there are exceptional contexts that make long-distance binding across an embedded external argument less of a violation. In transitive AcIs, for example, where it is pragmatically unlikely that the anaphoric element corefers with the AcI-subject, use of the reflexive to refer to the matrix subject is perfectly acceptable despite the intervening external argument (e.g. Hansi lsst den Mannj auf sichi/auf ihn*i eiferschtig werden. Hans lets the man get jealous of him. (Reis 1976: 31)). If, on the other hand, ambiguity could arise, the pronominal is judged to be the more acceptable way of referring to the matrix subject (e.g. Hansi lsst den Vaterj ruhig stolz auf sich*i/auf ihni sein. Hans has no problem letting the father be proud of him. (Reis 1976: 31)). The judgments Reis documents implicate that, although the pronominal should be a possibility in these contexts, the chance of using sich obviates the pronominal. Other non-syntactic factors, such as avoidance of two homophonous reflexive pronouns in the same sentence, and involvement of idiomatic expressions which often make the reflexive sound better than the pronominal, complicate the facts even further. 233

exposure, speakers may or may not make covert reflexive raising part of their grammar. As for Case 2, the proposed analysis accurately captures the invariable judgments involving PPs that are either clearly arguments or clearly adjuncts. Again, no new technology needs to be introduced: the postulation of both high PP-adjunction and PP-phases is in line with previously made claims or uncontroversial assumptions.

5.5

Conclusion

The proposals I have made in this chapter have consequences for two areas of grammar: binding and the phrase-structure of reduced infinitive constructions (in particular AcIs). The main empirical contribution made to the study of anaphora is the finding that there are several contexts (more than previously recorded) in which the reflexive and the pronominal are not in complementary distribution. The reflexive whether part of a PP or not can be bound across an agentive AcI-subject or a possessor in a complex DP, and the pronominal may be grammatical even if its antecedent is only a vP, DP, or PP-boundary away. The non-complementarity results from the fact that the reflexive is more versatile in finding a binder than the pronominal in being free. This can be formalized configurationally, without

reference to argument structure and the binding-specific construct CFC. Chomskys (2000, 2001) notion of derivation by phase offers an explanation for both the versatility of the reflexive and the easily gained freedom of the pronominal. As long as the reflexive is at an embedded phase-edge, it is accessible to the contents of
234

both the embedded and the next higher phase. Its binding requirement can thus be satisfied within a widened range. While both reflexives and pronominals can reach the phase-edge via PP-adjunction, only reflexives are endowed with the ability to covertly raise. This, we have assumed, is a reflection of the larger pattern that reflexive pronouns are morphosyntactically active in a way that ordinary pronouns are not. Given that it is reasonable to treat not only vP and DP but also PP as potentially phase-defining categories, and given that Condition A and B apply upon completion of each phase, the proposed analysis is largely successful at covering the facts. Both the majority of the previously recorded and the newly discovered empirical observations are accounted for. As for the potentially problematic reflexive binding possibilities in what Reis (1976) calls semi-obligatory PPs, it is important to keep in mind the marginal status of transitive AcI-constructions in general. The variability in judgments on binding, probably influenced by various non-syntactic factors, is then not surprising.

235

Binding, Possessor Datives, and Phasehood

The phase-based analysis of binding phenomena I propose in chapter 5 is to a large extent grounded in the finding that both agentive vPs and complex DPs constitute binding domains. I have shown, for example, that a pronominal can be bound by the matrix subject of an AC I-construction, as long as pronominal and antecedent are separated by an agentive vP-boundary. In Part II of the dissertation, I established that the same two phrase-types agentive vPs and complex DPs (with a genitive DP embedded in a larger DP) are also relevant domains in POSSESSOR DATIVE CONSTRUCTIONS (PDCs). A PD may not raise beyond an agentive vP-boundary and, apparently, it is also blocked by a double-layer DP-boundary. The possibility of unifying the categories which define the relevant domains for both binding and the PDC then certainly looks attractive. If reference to the notion P HASE can be substituted for vP and DP as binding domains, it should also be extendable to the locality restriction on possessor raising. Whether reference to phasehood actually turns out to be a simplification, however, depends on the extent to which phases and the binding/locality domains identified here overlap. One of the central questions this chapter addresses is what reference to phasehood adds over and above what is given by the notion intervening subject (i.e. Chomskys (1973) SPECIFIED SUBJECT CONDITION).

236

Before tackling this broader question, however, I aim to integrate into a single proposal the two strands of analysis I have developed in earlier chapters to account for the PDC and binding in a range of syntactic contexts. This will involve, first, undoing an expository simplification I made in chapter 5 regarding the position of affectee vPs in unaccusative AcIs, and then, assessing what commonalities hold of the PDC and the patterns of pronominal binding. Hence, section 6.1 is structured as follows. I start by providing an overview of the crucial pronominal binding facts from chapter 5. Then, I point to the parallel between binding and the PDC by reviewing some of the relevant PDC-test results from chapter 4 and examining them alongside the pronominal binding facts. Finally, section 6.2, which is more

speculative, is devoted to the broader question raised above, addressing issues of opacity, the notion subject, and phasehood. The main focus will be on the types of phrases which binding and the PDC diagnose as opacity-inducing, and then on how the common notion of opacity which emerges from that discussion can best be understood.

6.1

Parallels between binding and possessor datives as probes for transparency

Chapter 5 presents binding, in particular the binding possibilities of AcI-embedded pronominals, as another COHERENCE/RESTRUCTURING diagnostic. AcIs which are closed off by an agentive subject provide a legitimate domain in which a syntactically
237

bound pronominal is free. Thus, if an AcI-embedded pronominal bound by the matrix subject is grammatical, the AcI must consist of an agentive vP. The following data (re-)present the crucial evidence. In (1), the AcI-infinitives are transitive (a) and unergative (b-c), whereas in (2), they are unaccusative (a-b) and passive (c).1, 2 (1) a. Der Professori [vP ti lsst [vP die Studenten ihni zu Hause anrufen]].
the professor lets the students him at-home call

The professor lets the students call him at home. b. Siei [vP ti lsst [vP ihn mit ihri tanzen]].
she lets him with her dance

She lets him dance with her. c. Die Gromutteri [vP ti lsst [vP den Wellensittich ihr?i
the grandmother lets the parakeet

auf den Kopf fliegen]].

head fly

her (DAT) on the

The grandmother lets (allows) the parakeet fly onto her head.

Assuming that the PP in (1b) has argument status and is q-dependent on the verb (the fact that *Siei tanzt mit ihri selbst is ungrammatical suggests that the mit-PP is not a separate binding domain), it must be the embedded agentive vP, intervening between

The bracketing in (1)-(3) corresponds to the tree structures in chapters 4 and 5, where I show

the matrix v as left-headed. On Zwarts (1997) view, this is in fact the correct structure. On the traditional view of Germanic verb-second structures (see e.g. Vikner 1995), all verbal projections are head-final, and verb-second is derived by V-to-I-to-C movement. The subject is assumed to move via Spec TP to Spec CP.
2

In (1c) and (2b) (as in the PDC-examples in chapters 3 and 4), PD and possessee (i.e. origin

and landing site of the PD) are in bold face. In (3), (9) and (10), the origin site of the PD is marked with a trace. 238

the subject trace and the pronominal, which makes it possible for the pronominal to be bound by the matrix subject in all three examples. The infinitive in (1c), fliegen, is a directed motion verb, which, I assume, projects either an unergative or an unaccusative structure, depending on the agentivity of the subject. The subject of fliegen is agentive here and, as predicted, the pronominal PD ihr is at least marginally acceptable. It is noticeably better than the pronominal in (2a), where fliegen has a non-agentive subject and is used in an unaccusative structure. Like the mit-PP in (1b), the auf-PP in (2a) is assumed to have q-dependent argument status. The AcIinfinitive fallen fall in (2b) is another motion verb, but there is less variation with respect to its use. Its subject is normally non-agentive, forcing an unaccusative structure. Finally, in (2c), we have a passive AcI (similar to the French faire-par)construction. In none of these examples is the pronominal acceptable. (2) a. Die Gromutteri [vP ti lsst den Ballon ihr?*i [VP direkt aufs Auto fliegen]].
the grandmother lets the balloon her (DAT) directly on-the car fly

The grandmother lets (has) the balloon fly directly onto her car. b. Der kleine Jungei [vP ti lsst den Stein ihm*i [VP auf den Kopf fallen]].
the little boy lets the rock him (DAT) on the head fall

The little boy lets the rock fall on his head. c. Hansi [vP ti lsst ihm*i [VP ein Buch von Maria geben]].
Hand lets him (DAT) a book by Maria give

Hans has a book given to him by Maria.

239

I take the ungrammaticality of the pronominals here to indicate that these AcIs, which lack an agentive subject, are bare VPs, even in the case of (2a-b), which involve PDCs. Recall from chapter 4 that the affectee vP which licenses the PD is probably not part of the AcI. I assume that the affectee vP is projected by the matrix verb, and that the AcI-subject (den Ballon in (2a) and den Stein in (2b)) can precede the PD by scrambling from the AcI-VP into the matrix clause. The bracketing in (2c) shows the dative complement of geben as AcI-external as well. Unlike the PDs in (2a-b), however, the core dative in (c) does not originate in the AcI. Importantly, whether the verbal projections which license the dative pronominals in (2a-c) are part of the respective AcIs or not (I suspect that they are not), there is no agentive vP-boundary intervening between the ungrammatical pronominals and their binders. As for the subtleties of the judgments provided here, the possibility of scrambling is certainly a confounding factor which contributes to the complexity of the task speakers face in having to judge these examples. Overall, the results of the binding test in (1)-(2) correlate accurately with the possibility of long PD-movement. As already shown in (2a-b) and confirmed below in (3b), an unaccusative AcI allows a PD to move from its possessor position within the possessee into the matrix clause. Since the non-agentive AcI-subject, which originates inside the embedded VP, has the option of scrambling in front of the PD (see (3b')), the PD may occur on either side of the AcI-subject. The transitive AcI shown in (3a), on the other hand, only allows the PD to occur to the right of the AcI240

subject, indicating that PD-movement may not cross the AcI-boundary. It is evident, then, that the presence of the AcI-subject in the specifier of an agentive vP induces opacity. (3)
a. Maja [vP lsst (*ihrer Freundin) [vP Willi (ihrer Freundin) t die Fe massieren]]
Maja lets (her friend (DAT, FEM)) Willi (her friend (DAT, FEM)) the feet massage

Maja lets Willi massage her friends feet.

b. Der kleine Junge [vP lsst seinem Freund [VP den Stein auf t den Kopf fallen]].
the little boy lets his friend (DAT) the rock on the head fall

b'. Der kleine Junge [vP lsst den Stein seinem Freund [VP auf t den Kopf fallen]].
the little boy lets the rock his friend (DAT) on the head fall

The little boy lets the rock fall on his friends head.

A similar parallel between pronominal binding and the PDC concerns the opacity of complex DPs. It has been established that a complex DP, with a possessor in Spec DP, provides a domain in which a syntactically bound pronominal can be free. The relevant example from chapter 5 is repeated here in (4). (4) Thorsteni hrt nicht gern
Thorsten hears not

[DP Martins Geschichten ber ihni].

about him

with-pleasure Martins stories

Thorsten doesnt like to hear Martins stories about him. Again, given that the ber-PP is a q-dependent argument of the noun, it must be the DP-boundary which is responsible for the possibility of pronominal binding here. In

241

contrast to this complex DP, the plain DPs in (5) seem not to induce opacity. The pronominals are unacceptable with the given coindexation. (5) a. Die Kinderi haben nur [DP Interesse an ihnen*i].
the children have only interest at them

The children only have interest in themselves. b. Die Fraui schreibt [DP ein Buch ber sie*i].
the woman writes a book about her

The woman is writing a book about herself. Apparent counterexamples, like those in (6), can be explained by assuming that the specifiers of these seemingly plain DPs are actually occupied by a PRO-subject which is referentially distinct from the pronouns antecedent (see e.g. Chomsky 1986). (6) a. Mariai liest [DP PROj einen Bericht ber siei].
Maria reads a report about her

Maria is reading a report about her. b. Thorsteni hrt nicht gern [DP PROj Geschichten ber ihni].
Thorsten hears not with-pleasure stories about him

Thorsten doesnt like to hear stories about him. It is implied in these examples that the report and the story are written and told by somebody other than the respective subjects, Maria and Thorsten. It is reasonable to assume, then, that these seemingly plain DPs are in fact complex.

242

As for the plain DPs presented in chapter 5, repeated here in (7), we now know that the reason the pronominals are acceptable is that they are embedded in adjunct PPs, which constitute their own binding domains. (7)
Eri sah [AcI [DP das Boot [PP neben sichi/ihmi]], aber nicht [DP das Boot [PP hinter
he saw self/him the boat under-go next-to self/him but not the boat behind

sichi/ihmi]] untergehen]. He saw the boat next to him, but not the boat behind him sink.

This provides reason to believe that plain DPs are transparent for binding relations in the same way that bare (agentless) VPs are. At first sight, the complex DP data in the context of PDCs, presented in chapter 3, seem to line up with the binding facts. Complex DPs which contain a genitive DP embedded in a larger DP appear to be opacity-inducing domains. The preferred interpretation of examples like (8), for instance, is that the PD Lena is the possessor (here, owner or caretaker) of the larger (containing) DP, not of the embedded DP. (8) a. Tim pflegte Lena [DP das Fohlen [DP der Stute]] gesund.
the foal the mare (GEN) healthy

Tim treated Lena (DAT)

Tim cured the mares foal which belongs to Lena. b. Tim pflegte Lena [DP die Mutter [DP des Fohlens]] gesund.
the mother the foal (GEN) healthy

Tim treated Lena (DAT)

Tim cured the foals mother which belongs to Lena.

243

Although it is clear that both foal and mother somehow belong to Lena, it is primarily the foal that is interpreted as being Lenas in (8a), and the mother in (b). Thus, given the structure in (9) (see Longobardi 2001), it seems that the PD must originate in the specifier of the structurally higher accusative-marked nominal. As discussed in chapter 3, this falls out from the locality restrictions which characterize A-movement. If the PD originated in the specifier of the embedded, genitive-marked DP, it would have to move across the specifier position of the larger DP and thus across an intervening A-position. The circled specifier positions in (9) indicate that, given my analysis of the facts in (8) (which I return to in subsection 6.2.4), only the higher, not the lower Spec DP is a possible origin site for a PD. (9)

DP [ACC] 3 PD D' 3 D FP das 3 F PossP | 3 N DP [GEN] Poss' Fohlen 1 3 *PD D' Poss NP 1 tN 6 D NP tN der Stute

I have been using the term complex DP to refer to nominals which induce opacity both for pronominal binding and the PDC. The nature of the complexity, however, is not quite the same. While a DP that is opaque for pronominal binding is complex in the sense of having a filled specifier, a DP that is opaque for PD-movement is
244

complex in the sense that there is a DP-boundary intervening between the origin site of the PD and its landing site. In both cases, there is an A-position intervening between the relevant two points of interaction (i.e. between binder and bindee and between landing site and origin site, respectively). In the binding context, this Aposition is filled; in the PDC context, it is not. Whether both types of opacity are subsumed under the notion phase is one of the questions I speculatively address in the following section.

6.2

The phase as the key to opacity effects

Building on the notion BOUNDING NODE (Chomsky 1973, 1977) and the core idea of the B ARRIERS-framework (Chomsky 1986), Chomsky 2000 proposes that phases, defined by C and v, impose a strict locality on syntactic computation. Like the nodes formerly identified as bounding nodes and barriers, phase-heads are argued to trap elements inside their complement. As explained in chapter 5, once a phase is built, its contents are transferred to the (LF and PF) interfaces and are therefore inaccessible to elements introduced later in the derivation. Only material at the phase-edge is still accessible. Like a domain closed off by a bounding/barrier node, a phase then forces extraction via its specifier (i.e. guaranteeing SUCCESSIVE CYCLIC derivations), and like a COMPLETE FUNCTIONAL COMPLEX (Chomsky 1986), a phase is a domain in which all the selectional requirements of the head of the domain are satisfied. In Derivation by Phase, Chomsky (2001) makes a distinction between STRONG and
245

W EAK phases, arguing that not all types of v-heads have the strong-phase-defining property of inducing opacity. He also alludes to the possibility that C and v are not the only phase-heads; D and P may be phase-defining as well. Exactly what types of v, D, and P qualify as (strong) phases3 is the focus of much recent work (see e.g. Legate 2003, Abels 2003, and Svenonius 2004). In this section, I contribute to this line of work by tentatively proposing a definition of phase based on the phrase-types the PDC, binding, and other locality-driven phenomena identify as opacity-inducing.

6.2.1 Agentive versus verbalizer vPs As confirmed by the data in section 6.1, passive and unaccusative AcIs are transparent for both PD-movement and pronominal binding. Following Hale & Keyser (1993), Chomsky (1995), Kratzer (1996), and Wurmbrand (2001), I have been assuming that the difference in transparency between transitive and unergative verbal shells on the one hand and passive and unaccusative verbal shells on the other hand is reflected in their internal structural complexity. While verbal shells of the former type consist of both a lexical VP projection and a functional vP-layer, verbal shells of the latter type consist of a lexical VP only. According to Chomskys (2000, 2001) later work, however, a verbalizer vP-layer is a part of all verbal shells, including those projected by passive and unaccusative verbs. Unlike proto-agent-introducing
3

I use the term phase to refer to what Chomsky labels strong phase. 246

agentive vPs, these verbalizer vPs are considered to be defective in that they neither introduce an external argument nor have accusative-case-licensing ability (cf. Burzios Generalization). On this view, the unaccusative AcI in (10) has the structure given in (10'). (10)

Der kleine Junge [vP lsst seinem Freund [vP-def [VP den Stein auf t den Kopf fallen]]].
the little boy lets his friend (DAT) the rock on the head fall

The little boy lets the rock fall on his friends head.

(10')

vP 3 DP v' Der kleine Junge 3 (agentive) v vP lsst 3 DP v' seinem Freund 3 (PD) VP v (affectee) 3 tV vP V 3 tV VP v (defective/verbalizer) 3 DP V' den Stein 3 PP V 3 fallen P DP auf 3 tPD D' 3 D NP den Kopf

As explained in chapter 4 and reiterated in the previous section of this chapter, I assume that the PD-licensing affectee vP is projected by the matrix verb. The PD,

247

which originates in the specifier of the possessed nominal (here den Kopf), then raises from the AcI into the matrix clause. The word order would remain unchanged, and the derivation would converge, yielding virtually the same interpretation, if the affectee vP were projected by the AcI-infinitive (here fallen). The assumption that it is not, however, is more in line with Wurmbrand 2001 in that the characteristic property of reduced infinitival complements is their lack of case positions. Given the structure in (10'), then, the verbalizer vP projected by the unaccusative AcI-infinitive fallen intervenes between the possessor position within the possessee and the affectee vP. Yet, possessor raising is not blocked.4 The structure in (11') illustrates that defective verbalizer vPs are also transparent for pronominal binding. The pronominal is degraded here because the verbalizer vP does not define a domain in which a syntactically-bound pronominal is free.5 (11) Der Bauarbeiteri [vP ti lsst [vP-def [VP das Gerst
the construction-worker lets

auf ihn?*i niederstrzen]]].

down-crash

the scaffolding on him

The construction worker lets (has) the scaffolding crash down on him.

See chapter 4, footnote 18, for a discussion of why the affectee v does not target the closer

DP, den Stein, instead of the PD.

5

As discussed in chapter 5, subsections 5.4.4-5, the reason the pronominal in (2a) is not

completely ruled out may be due to the possibility of analyzing aufzu as a pre- (and post-) position which is not necessarily selected by the verb fliegen. If it is an adjunct PP, it could constitute its own binding domain and therefore allow the pronominal to be free. 248

(11')

TP 3 DPi T' Der Bauarbeiter 3 T vP 3 tDPi v' (agentive) v lsst

3
VP 3 vP 3 V tV v (defective/verbalizer)

VP 3 DP V' das Gerst 3 PP V auf ihn?*i niederstrzen

The lack of opacity induced by the defective verbalizer vPs in (10) and (11) is corroborated by evidence from long-distance case and agreement marking in impersonal, expletive-es constructions. The defective vP projected by the passivized verb in the impersonal passive construction in (12), for example, does not interfere with the static Agree relation between the nominative case-licensor T and the VPinternal subject. (12) [TP Es wurden damals [vP-def [VP viele Professoren
it were (PL) then

eingestellt]]].

many professors (NOM, PL) hired

Many professors were hired then. Similarly, there is no interference with case and agreement marking in the unaccusative expletive constructions in (13).

249

(13)

a. [TP Es ist

noch nie [vP-def [VP ein Meister

never

vom

Himmel gefallen]]].
fallen

it is (SG) yet

a master (NOM, SG) from-the sky

It has not yet happened that a master fell from the sky. (fig. It takes practice to become good at something.) b. [TP Es haben aber schon [vP-def [VP viele
it have (PL) but already

ihr Glck versucht]]].

tried

many (NOM, PL) their luck

But many have already tried their luck.

Possessor raising, pronominal binding, and case/agreement marking in impersonal passive and unaccusative constructions all converge, then, on the lack of opacity induced by the defective outer layer of passive and unaccusative verbal shells. Furthermore, extending the notion defective verbalizer vP to the complements of Wurmbrands (2001) restructuring predicates, any vP-layer projected by a restructuring infinitive must be transparent for the long object movement operations discussed in chapter 2. Relevant examples are shown in (14). The movement operation in the (a)-example is an instance of long passive, and the (b)-example is an instance of long object movement out of the complement of an unaccusative restructuring predicate.6 (14) a. [TP Die Traktoren wurden [vP-def [VP __ zu reparieren]] versucht].
to repair tried

the tractors (NOM, PL) were (PL)

They tried to repair the tractor.

As in the tree structures in (10') and (11'), no CP-layer is shown here. 250

b. [TP Der Brief

ist ihm [vP-def [VP __ zu entziffern]] gelungen].

to decipher succeeded

the letter (NOM, SG) is him

He managed to decipher the letter. Unlike in (12)-(13), the Agree relation between T and the respective nominal is coupled with movement here. Crucially, in neither case does the highest layer of the embedded verbal shell make the argument of the infinitive inaccessible to an Agree relation with the matrix T. Given that the characteristic property of a phase-defining head is precisely that its complement is inaccessible to elements introduced later (higher) in the derivation, the defective verbalizer vPs projected by passive/ unaccusative verbs and restructuring infinitives cannot be phases. Thus, they are unlike agentive vPs, which clearly are phases.

6.2.2

Affectee vPs

The possessor raising analysis proposed in chapter 3 and illustrated again in (10') of subsection 6.2.1 introduces a third type of vP: the PD-licensing affectee vP projected by verbs which are compatible with affected dative-marked possessors. The obvious question, then, is whether affectee vPs are of the phase-defining, agentive type or the defective, verbalizer type. Making the standard assumption that the accusative case of AcI-subjects comes from the matrix verb, any layers of structure intervening between the matrix agentive v and the AcI-subject must be transparent for casechecking. Since a phase-boundary would induce opacity and since the affectee vP in the unaccusative AcI-construction in (10') necessarily intervenes between the matrix v
251

and the VP-internal AcI-subject, affectee vPs must not be phases. Rather, they pattern with verbalizer vPs (projected by passive/unaccusative verbs) in that they lack the phase-defining opacity property of agentive vPs. Based on the fact that the crucial opacity-inducing property of vPs and DPs with respect to pronominal binding is the presence of a subject in their specifier, a phase could simply be defined as any phrase closed off by a subject (cf. Chomskys (1973) Specified Subject Condition). Given the definition of subject in chapter 5 (see (35)), a phase would then be defined by any q-assigning head with a filled specifier. This understanding of phasehood works for the contrast between agentive and defective verbalizer vPs, but it also predicts that affectee vPs are phases. As just discussed, this cannot be right. Appealing to the notion agentive (proto-agent) subject instead of just filled specifier might take care of excluding affectee vPs, but it introduces new problems. The possessor in complex DPs is not usually a protoagent, and PPs, which can also be opacity-inducing, in all probability lack subjects altogether (Hestvik 1991, building on earlier, unpublished work by Bresnan). An even stronger argument against the Specified Subject-definition of phase is that it excludes CPs, which are phases by all other criteria, and in everyones system. In (15), I propose a characterization of phase which makes the correct distinction and does not appeal to the notion Specified Subject. (15) Phases are defined by those verbal shells whose highest projection has a filled specifier.
252

The highest projection of a passive/unaccusative verbal shell is a defective verbalizer vP. Since this vP has an empty specifier, passive/unaccusative verbal shells are not phases. Affectee vPs do have a filled specifier the position is filled either via Internal Merge by a case-seeking PD or via External Merge by a non-PD (i.e. an ethical, estimative, or ficiary dative) but they are never the highest projection in the verbal shell. If the verbal shell of an affectee vP is headed by a transitive/unergative verb, the affectee vP is embedded under an agentive vP. In this case, since the specifier of the agentive vP is occupied, the verbal shell is a phase. If the verbal shell of an affectee vP is headed by a passive/unaccusative verb, on the other hand, the affectee vP is embedded under a defective verbalizer vP. In that case, the specifier of the highest verbal projection is empty, and the verbal shell is not a phase.

6.2.3

DPs, PPs, and CPs

Extending the definition of phase given in (15) to the non-verbal domain, i.e. DPs, PPs, and CPs, we have four potentially phase-defining heads: v, D, P, C. In order to make the definition work for P and C, the requirement of the highest projection having a filled specifier needs to be modified. The idea is that a filled specifier is only a phase-defining property if the position can be filled in virtue of pure selectional requirements, not if it can only be filled in virtue of the semantically vacuous EPP feature. In other words, since C and P never select more than one argument (merged as a complement) and only attract an argument into their specifier
253

if they have the EPP property, a filled specifier is not a phase-defining property for these heads. As for vP and DP, however, a filled specifier is a necessary condition because both v and D can potentially select two arguments, one merged as a complement (by first Merge) and the other merged as a specifier (by second Merge). Given this distinction between the nature of P and C on the one hand and v and D on the other hand, I propose the characterization of phase in (18), now going beyond the verbal domain. (16) A phrase of type a is topmost of it is not itself the complement of a phrase of type a. (17) A phrase of type a is saturated if it has the maximum number of arguments that lexical items of type a can in principle take. (18) Characterization of phase: A phrase of type a, with a being v, D, P, or C, which is saturated and topmost is a phase. Together with the definition of saturated in (17), this characterization of phase ensures that v and D-heads, which can in principle take two arguments, are not phasedefining when they do not have an argument in their specifier. This means that defective verbalizer vPs, projected by passive/unaccusative verbs, are not phases. P and C-heads, which maximally take one semantically selected argument, are only identified as non-phase-defining if they take no semantically selected argument at all. P and C-heads then do not need a specifier to be saturated. The requirement that a be
254

topmost (see (16)) ensures that affectee vPs, despite being saturated, do not qualify as phases. An affectee vP is the complement of a phrase headed by type a, namely another v, because it is necessarily embedded under either an agentive v or a defective verbalizer v. The understanding of phasehood expressed by the characterization in (18) then accounts for the well-established opacity effects induced by CP as well as most of the opacity/non-coherence effects presented thus far in Parts II and III of this dissertation. The local nature of derivations proceeding phase by phase essentially subsumes the locality restriction associated with A-movement. Since (18) identifies agentive vPs as phases, it explains the opacity of agentive vPs with respect to possessor raising and pronominal binding, and since it also identifies PPs as phases, it explains the opacity of q-independent PPs with respect to pronominal binding. The fact that q-dependent PPs with a P which is semantically contentless and thus lacks the ability to q-mark its complement do not induce opacity falls out from the requirement that a phasedefining phrase be saturated (see (17)). Since a contentless P takes no semantically selected argument at all, it is not saturated and thus not phase-defining. As for DPs, more needs to be said. On the one hand, the characterization of phase in (18) correctly predicts that complex DPs with a filled possessor position should induce opacity. Since a D with both a complement (an internal argument) and an (external) argument in its specifier is not defective, it qualifies as phase-defining.
255

This accounts for the fact that complex DPs, but not plain DPs, are opaque with respect to pronominal binding. On the other hand, (18) does not predict that complex DPs with a genitive-marked nominal embedded in a larger nominal do not seem to allow possessor raising out of the embedded DP, only out of the larger, containing DP. If Landaus (1999) analysis of the relevant data (see (8)) presented in chapter 3 is correct, the larger DP, embedding a genitive DP, has the opacity-inducing quality of a phase despite not having a (semantically) filled specifier. This clearly does not fall out from the understanding of phasehood this section has developed.

6.2.4 Double-layer DPs revisited My conclusions regarding the opacity of DPs which embed a genitive DP (henceforth double-layer DPs) have been based almost entirely on the data in (19) and (20) (first introduced in chapter 3). Notice that all these examples involve relational nouns. (19) a. Tim pflegte Lena [DP das Fohlen [DP der Stute]]
the foal

gesund.

Tim treated Lena (DAT)

the mare (GEN) healthy

Tim cured the mares foal which belongs to Lena. b. Tim pflegte Lena [DP die Mutter [DP des Fohlens]] gesund.
the mother the foal (GEN) healthy

Tim treated Lena (DAT)

Tim cured the foals mother which belongs to Lena. (20) a. Dann stecke ich mir
then stick I

einen Ring auf [einen Finger [der linken Hand]].

ring on a finger the left hand (GEN)

me (DAT) a

Then I put a ring on a finger of my left hand.

256

b. Mir

fiel der Hammer auf [die Spitze [des linken Zeigefingers]].

on the tip the left index-finger (GEN)

me (DAT) fell the hammer

The hammer fell on the tip of my left index finger. In this subsection, I revisit these facts and return to the question addressed in chapter 3, which is whether the specifier of the embedded genitive DP in the structure in (9), repeated below in (21), is a possible PD-origin site. (21)

DP [ACC] 3 D' 3 D FP das 3 F PossP | 3 N DP [GEN] Poss' Fohlen 1 3 ?PD D' Poss NP 1 tN 6 D NP tN der Stute

While it is true that the interpretation of a possessor relation between the dative Lena and the embedded DP in (19) is dispreferred, there is no detectable preference for a possessor relation with the larger, rather than the embedded, DP in (20). In (a), there is virtually no difference in interpretation between a syntactic configuration with the DP einen Finger versus the DP der Hand being the origin site of the PD mir. The same holds for (b). Logically, the PD could have originated in the specifier of either die Spitze or des linken Zeigefingers. Thus, when the two nouns of the double-layer DP are relational, depending on the lexical items involved, it seems that either there is
257

no preferred possessor interpretation, or there is a slight preference for the referent of the larger DP belonging to the referent of the PD, perhaps because the raised N of that DP is linearly closer to the PD. Since the nominals in a double-layer DP consisting of relational nouns will either necessarily have the same possessor, as in the body-part examples in (20) if the tip of the finger is mine, the finger must be, too or at least tend to have the same possessor, as in the kinship examples in (19), these types of examples are not the best test cases. Considering the double-layer DPs in (22), which involve nouns that are not inherently related, it is clear that there must be a way for the PD to originate in the specifier of the embedded DP. (22) a. Tim pflegte Lena [DP das Pferd [DP t einer Freundin]]
Tim treated Lena (DAT) the horse a friend (GEN, FEM)

gesund.
healthy

Tim cured the horse of one of Lenas friends for her (Lena). b. Wir haben unserer besten Kundin [DPden Computer [DP t einer
we have our best client (DAT, FEM)) the computer an

Bekannten]]

repariert.

acquaintance (GEN) fixed

We fixed, for our best client, the computer of an acquaintance of hers. c. Der Grtner wusch der Frau
the

[DP das Auto [DP t des Sohns]].7

the son (GEN)

gardener washed the woman (DAT) the car

The gardener washed the womans sons car.

7

In colloquial German, one would say das Auto vom Sohn the car of the son since the use of

the genitive is rather formal. 258

While the long-distance possessor relation between the dative nominal and the embedded DP is most efficiently expressed by a possessive pronoun (see chapter 3, subsection 3.3.3) in that case, the dative is a beneficiary non-PD the use of a PD to express this relation must be a legitimate alternative. PD-movement from the embedded Spec DP of a double-layer DP is then not at odds with the binding facts in (4)-(6). Like the binding facts, which showed that a syntactically bound pronominal can be free in a DP closed off by a possessor in Spec DP, but not in a plain DP, the new PDC facts presented here are accounted for by the characterization of phase given in (18). Unless its specifier is filled (by virtue of semantic selection), a DP is not a phase and therefore does not block PD-raising across its left edge to the specifier of the affectee vP. This means that again, given Longobardis (2001) DP structure in (21) only DPs with a SAXON GENITIVE, which occupies Spec DP, the topmost position in the nominal shell, not those with only a genitive in the specifier of the lower PossP, constitute a phase. DPs can then be treated on a par with vPs, needing two arguments in order to be semantically saturated. A potential argument against drawing this parallel between vP and DP is that, unlike a verbs external argument, which is generally assumed to be introduced as the specifier of v, a nouns possessor is often argued to be introduced as either the specifier of D or the specifier of N. In the case of deverbal nouns, for example,
259

which stand in an obvious q-relation to both their internal and their external argument, the external argument is commonly assumed to originate in Spec NP. Since a DP with a Saxon genitive in Spec NP (and thus an unfilled Spec DP) is not defined as a phase by (18), a syntactically bound pronominal embedded in such a DP should not be free. In German, this is counter to fact. As shown by the acceptability of the pronominals in (23a-b), the possessor of both the noun Geschichten stories and the noun Angst fear renders the nominal domain opaque for pronominal binding. And this is so despite the fact that Geschichten (a story-type noun8) and its possessor could stand in a number of semantic relations to each other the possessor could be the writer, the owner, the experiencer, or the main character while Angst (a fear-type noun) and its possessor can only stand in one relation to each other: sensation-experiencer. (23) a. Thorsteni hrt nicht gern
Thorsten hears not

[DP Martins Geschichten ber ihni].

about him

with-pleasure Martins stories

Thorsten doesnt like to hear Martins stories about him. b. Martini ist erstaunt ber
Martin is surprised about

[DP Thorstens Angst vor ihmi].

Thorstens fear of him

Martin is surprised about Thorstens fear of him. I assume that both the PPs here are q-dependent and thus do not qualify as phases. The crucial opacity-inducing domain must then be that of the noun, Geschichten in
8

See Sturgeon (2003) for a discussion of binding in the context of story versus fear-type

nouns in Czech. 260

(a) and Angst in (b). This leads to at least two possible conclusions. Either the possessor of Angst sits in Spec NP, and even plain DPs (with an unfilled topmost specifier) can be phases, or, in the presence of a Saxon genitive, Spec DP is always filled, whether via direct External Merge or Internal Merge (i.e. movement from Spec NP to Spec DP). Since the former of these conclusions is incompatible with the understanding of phasehood developed here, and I know of no evidence against the latter, I tentatively assume the latter. A final concern to be addressed here is the fact that a syntactically bound pronominal can apparently be free even in the absence of a Saxon genitive. As just discussed with respect to the PDC, a double-layer DP structure with a genitive DP embedded in a larger DP (see (21)), does not qualify as a phase. The binding possibility between the matrix subject and the pronominal in (24) is surprising then. (24)
Thorsteni hrt nicht gern
Thorsten hears not

[DP die Geschichten [PossP seiner Schwester ber ihni]].

his sister (GEN) about him

with-pleasure the stories

Thorsten doesnt like to hear his sisters stories about him.

A possible solution to this puzzle is that PossP, which, like vP and DP, can take two semantically selected arguments, may be a phase-defining category.9,10 It certainly
9

Note that the N Geschichten head-moves to F via the Poss-head (see (21)). This means that,

if PossP were a phase, N could still move out because, like its specifier, the head of a phase-defining category is part of the phase-edge.
10

PossP could, in fact, be equivalent to nP. See Grimshaws (1990, 2000) proposal of

extending the clausal architecture of the verbal domain to the nominal domain. 261

fulfills the criterion of being semantically saturated. Importantly, the proposal that PossP is a phase is not at odds with the ability of a PD in the specifier of a genitive DP to move out of the nominal domain to check case with an affectee v. Since the embedded genitive DP of a double-layer DP structure is the specifier of Poss, it is at the edge of PossP and thus, if PossP were a phase, it would be at the edge of this phase and therefore accessible to elements in the next higher phase.

6.2.5 Beyond coherence: A unified account In conclusion, there is reason to believe that the various opacity effects (i.e. effects of Satzwertigkeit) which I bring together in the form of the three strands of analysis corresponding to the three parts of this dissertation can indeed be unified by the notion phase. The notion intervening subject alone makes sense of the opacity induced by agentive vPs and complex DPs, but it falls short of accounting for the transparency induced by affectee vPs, and it has no explanatory value with respect to inherently subjectless PPs and CPs. In contrast, the characterization of phase I offer here potentially lays the groundwork not only for a unified analysis of coherence/noncoherence phenomena in German but also for locality constraints cross-linguistically. Building on, and in support of, Wurmbrands (2001) fine-grained typology of infinitival clause size, this analysis accounts for various control verb phenomena and the often misanalyzed hybrid case of AcI-constructions. Having developed a more nuanced understanding of Bechs (1955/57) original binary distinction between

262

coherent and non-coherent infinitive constructions, I hope to have taken issues of Satzwertigkeit to a new level, certainly beyond coherence.

263

Appendix A: Questionnaire Results for Pronominal Binding in AcI-constructions

The following shows the pronominal binding results of the two questionnaires I designed to elicit grammaticality judgments on binding in AcI-constructions. This was an informal study used only to back up my own judgments as a native speaker. Some of the judgments may have been influenced by factors I could not control for, but there are nonetheless clear patterns which generally support my empirical generalizations. I do not include the individual results of each participant here (see Appendix B for a complete list of sentences and results), but the two questionnaires were completed by eight and ten speakers, respectively, mostly from northern Germany. In most cases, the sentences were read to the participants. First they heard a given sentence with the reflexive, then the same sentence with the pronominal. After each version of the sentence, they indicated which of the given antecedents they felt the anaphoric element referred to by picking a number from 1 to 5: 1 = immediately comes to mind 2 = fine with appropriate context but isnt the first thing that comes to mind 3 = possible but awkward 4 = sounds pretty much wrong 5 = absolutely impossible
264

To the right of each sentence I show the average of the evaluation numbers the participants gave for the pronominal in coreference with the matrix subject (both in bold face). The results for the reflexive and the binding possibilities with respect to other antecedents are not shown here (again, see Appendix B). The sentences are ordered according to the acceptability of the pronominal, going from least to most acceptable. Since it is generally the case that the sentences with less acceptable pronominals also have less agentive AcI-subjects (underlined), and the sentences with more acceptable pronominals have more agentive AcI-subjects,11 the results of this questionnaire support my claim that the AcI-binding facts correlate with the agentivity of the AcI-subject. Der kleine Junge lsst den Stein sich/ihm auf den Kopf fallen. The little boy let the rock fall on his head.
(2)

(1)

4.8

Willi lsst die Chance sich/ihm nicht durch die Finger gleiten. Willi doesnt let the chance slip through his fingers.

4.4

(3)

Britta lie den Ball auf sich/sie zurollen. Britta let the ball roll toward her.

4.3

11

Obvious exceptions to this generalization are the AcI-subjects die Mdigkeit the tiredness

in (5) and die Verantwortung the responsibility in (7), which are non-agentive and should therefore make the respective pronominals less acceptable. I suspect that either the PPs ber over and aufzu on to can be interpreted as not selected by the verb kommen come, so that the PPs themselves constitute a binding domain for the pronominal, or the verbs kommen and zukommen auf come toward caused participants to somehow personify the non-agentive AcI-subjects. 265

(4)

Der Bauarbeiter sah das Gerst auf sich/ihn niederstrzen. The construction worker saw the scaffolding crash down on him.

3.7

(5)

Maja lie die Mdigkeit ber sich/sie kommen. Maja let tiredness overcome her.

3.6

(6)

Der Polizist lie den leblosen Krper auf sich/ihn fallen. The policeman let the lifeless body fall on him.

3.6

(7)

Die jungen Eltern lieen die Verantwortung auf sich/sie zukommen. The young parents let the responsibility come toward them.

3.4

(8)

Andrea lsst die Katze sich/ihr nicht ins Haus kommen. Andrea doesnt let the cat come into her house.

3.3

(9)

James Bond lie die junge Frau auf sich/ihn fallen. James Bond let the young woman fall on him.

3.2

(10)

James Bond lie die junge Frau sich/ihm in die Arme fallen. James Bond let the young woman fall into his arms.

3.2

(11)

Die Gromutter lsst den Wellensittich sich/ihr auf den Kopf fliegen. The grandmother lets the parakeet fly onto her head.

(12)

Die Demonstrantin lie den Polizisten auf sich/sie zukommen. The demonstrator let the policeman come toward her.

2.5

(13)

Der Fugnger sah die Radfahrer auf sich/ihn zurasen. The pedestrian saw the bicyclists race toward him.

2.4

(14)

Der Knig lie den Gefangenen vor sich/ihm niederknien. The king let the prisoner kneel down in front of him.
266

2.3

Appendix B: Questionnaires for Reflexive and Pronominal Binding in AcI-constructions, PPs, and complex DPs

The following shows the complete list of sentences and results of the two questionnaires I used to elicit grammaticality judgments on pronominal and reflexive binding. The anaphoric element is embedded in either just an AcI, a PP within an AcI, or a complex DP. Unlike Appendix A, which focuses solely on the average score of the judgments concerning the pronominal binding possibilities, the data here also include the individual results for each participant with respect to both pronominal and reflexive binding. The first survey was completed by eight speakers and the second by ten speakers, mostly from northern Germany. In most cases, the sentences were read to the participants. First they heard the sentence with the reflexive, then with the pronominal. After each version of the sentence, they indicated which of the given antecedents they felt the anaphoric element referred to by picking a number from 1 to 5: 1 = immediately comes to mind 2 = fine with appropriate context but isnt the first thing that comes to mind 3 = possible but awkward 4 = sounds pretty much wrong 5 = absolutely impossible

267

The 1-5 evaluation results are shown with respect to the different antecedents, first for the use of the reflexive, then for the use of the pronominal. The average score for each binding possibility is the underlined number following the different antecedents. The individual results for each speaker are listed following the speakers initials. The data within each of the questionnaires are organized by binding context. The first set of sentences (I) embed the anaphoric element in just an AcI, the second set (II) in a q-dependent PP within the AcI, the third set (III) in a q-independent PP within the AcI, and the fourth set (IV) in a complex DP.

Fragebogen zur Bindungstheorie (#1) Bitte lesen Sie die folgenden Stze und entscheiden Sie, worauf sich das Reflexivpronomen (sich) bzw. das Personalpronomen (ihn/ihm/sie/ihr) bezieht. Sie werden oft Mehrdeutigkeiten feststellen. Benutzen Sie die vorgegebene Bewertungsskala (1-5). Tragen Sie also in jedes Kstchen eine Zahl von 1 bis 5 ein. Dieselbe Zahl kann mehr als einmal pro Satz benutzt werden. Bitte nehmen Sie sich Zeit und fgen Sie, falls erforderlich, einen erklrenden Kommentar hinzu. Bewertungsskala: 1 Ja, kommt mir sofort in den Sinn. 2 Ja, mit entsprechendem Kontext; kommt mir aber nicht als erstes in den Sinn. 3 Ja, vielleicht, aber kommt mir komisch vor. 4 Kaum mglich; hrt sich falsch an. 5 Nein, eindeutig falsch.

I.
(1) Hans lsst sich/ihm einen Stein auf den Kopf fallen sich bezieht sich auf: Hans 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihm bezieht sich auf: Hans 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 Hans lsst den Stein sich/ihm auf den Kopf fallen. sich bezieht sich auf: Hans 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihm bezieht sich auf: Hans 4.9 V.S.: 4, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1

(2)

268

(3)

Die Mutter lsst das Kind sich/ihr die Schokolade in den Mund stecken. sich bezieht sich auf: Mutter 2.9 V.S.: 2, E.S.: 2, F.S.: 2, S.F.: 1, S.K.: 4, M.O.: 4, K.S.: 5, I.S.: 3 Kind 1.1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 2, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihr bezieht sich auf: Mutter 1.3 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 3, K.S.: 1, I.S.: 1 andere Person 2 V.S.: 2, E.S.: 2, F.S.: 2, S.F.: 5, S.K.: 2, M.O.: 1, K.S.: 1, I.S.: 1 Hans lsst sich/ihm ein Buch von Maria geben. sich bezieht sich auf: Hans 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihm bezieht sich auf: Hans 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 Hans lsst mich sich/ihm ein Buch geben. sich bezieht sich auf: Hans 3.1 V.S.: 1, E.S.: 1, F.S.: 4, S.F.: 1, S.K.: 5, M.O.: 3, K.S.: 5, I.S.: 5 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihm bezieht sich auf: Hans 1.6 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 5, S.K.: 1, M.O.: 2, K.S.: 1, I.S.: 1 andere Person 1.5 V.S.: 1, E.S.: 2, F.S.: 2, S.F.: 1, S.K.: 2, M.O.: 1, K.S.: 1, I.S.: 2 Der Knig lsst den Gefangenen sich/ihn anschauen. sich bezieht sich auf: Knig 2.5 V.S.: 2, E.S.: 2, F.S.: 2, S.F.: 1, S.K.: 3, M.O.: 1, K.S.: 5, I.S.: 4 Gefangener 1.3 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 2, S.K.: 1, M.O.: 2, K.S.: 1, I.S.: 1 andere Person V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihn bezieht sich auf: Knig 1.3 V.S.: 1, E.S.: 2, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 2, K.S.: 1, I.S.: 1 Gefangener 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 1.5 V.S.: 1, E.S.: 1, F.S.: 2, S.F.: 2, S.K.: 2, M.O.: 1, K.S.: 1, I.S.: 2

(4)

(5)

(6)

II.
(7) Martin hrt Thorsten ber sich/ihn reden. sich bezieht sich auf: Martin 2.3 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 2, S.K.: 2, M.O.: 2, K.S.: 5, I.S.: 4 Thorsten 1.1 V.S.: 2, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihn bezieht sich auf: Martin 1.4 V.S.: 2, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 2, M.O.: 2, K.S.: 1, I.S.: 1 Thorsten 4.5 V.S.: 3, E.S.: 4, F.S.: 5, S.F.: 5, S.K.: 4, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 1.5 V.S.: 1, E.S.: 2, F.S.: 2, S.F.: 2, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 2 Hans hrt den Professor mit sich/ihm sprechen. sich bezieht sich auf: Hans 3 V.S.: 2, E.S.: 1, F.S.: 2, S.F.: 5, S.K.: 2, M.O.: 3, K.S.: 5, I.S.: 4 Professor 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihm bezieht sich auf: Hans 1.5 V.S.: 1, E.S.: 2, F.S.: 1, S.F.: 1, S.K.: 3, M.O.: 2, K.S.: 1, I.S.: 1 Professor 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 1.4 V.S.: 1, E.S.: 1, F.S.: 2, S.F.: 2, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 2

(8)

269

(9)

Der Knig lsst den Gefangenen vor sich/ihm niederknien. sich bezieht sich auf: Knig 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 Gefangener 3.8 V.S.: 3, E.S.: 5*, F.S.: 5*, S.F.: 5, S.K.: 4, M.O.: 2, K.S.: 5, I.S.: 3 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 *Kommentar: aber mglich, wenn vorm Spiegel ihm bezieht sich auf: Knig 2.3 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 4, M.O.: 2, K.S.: 4, I.S.: 4 Gefangener 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 1.4 V.S.: 1, E.S.: 2, F.S.: 2, S.F.: 2, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1

III.
(10) Der Professor lsst den Assistenten fr sich/ihn arbeiten. sich bezieht sich auf: Professor 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 Assistent 2 V.S.: 1, E.S.: 1, F.S.: 2, S.F.: 2, S.K.: 2, M.O.: 2, K.S.: 5, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihn bezieht sich auf: Professor 2.5 V.S.: 1, E.S.: 1, F.S.: 3, S.F.: 1, S.K.: 2, M.O.: 2, K.S.: 5, I.S.: 5 Assistent 4.8 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 3, I.S.: 5 andere Person 1.5 V.S.: 2, E.S.: 2, F.S.: 1, S.F.: 2, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 2

IV.
(11) Martin hrt nicht gern Thorstens Geschichten ber sich/ihn. sich bezieht sich auf: Martin 2.3 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 2, S.K.: 3, M.O.: 1, K.S.: 5, I.S.: 4 Thorsten 1.1 V.S.: 1, E.S.: 2, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihn bezieht sich auf: Martin 1.9 V.S.: 2, E.S.: 3, F.S.: 1, S.F.: 1, S.K.: 4, M.O.: 2, K.S.: 1, I.S.: 1 Thorsten 3.6 V.S.: 3, E.S.: 1, F.S.: 3, S.F.: 5, S.K.: 4, M.O.: 3, K.S.: 5, I.S.: 5 andere Person 1.5 V.S.: 1, E.S.: 2, F.S.: 2, S.F.: 2, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 2 Jane ist erstaunt ber Danielas Wut auf sich/sie. sich bezieht sich auf: Jane 3.3 V.S.: 2, E.S.: 4, F.S.: 2, S.F.: 2, S.K.: 4, M.O.: 3, K.S.: 5, I.S.: 4 Daniela 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 sie bezieht sich auf: Jane 1.1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 2, K.S.: 1, I.S.: 1 Daniela 4.9 V.S.: 4, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 2.4 V.S.: 2, E.S.: 2, F.S.: 2, S.F.: 2, S.K.: 2, M.O.: 1, K.S.: 3, I.S.: 5 Martin ist entsetzt ber Thorstens Meinung von sich/ihm. sich bezieht sich auf: Martin 3.4 V.S.: 2, E.S.: 2, F.S.: 4, S.F.: 2, S.K.: 4, M.O.: 3, K.S.: 5, I.S.: 5 Thorsten 1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 1, K.S.: 1, I.S.: 1 andere Person 5 V.S.: 5, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 ihm bezieht sich auf: Martin 1.1 V.S.: 1, E.S.: 1, F.S.: 1, S.F.: 1, S.K.: 1, M.O.: 2, K.S.: 1, I.S.: 1 Thorsten 4.9 V.S.: 4, E.S.: 5, F.S.: 5, S.F.: 5, S.K.: 5, M.O.: 5, K.S.: 5, I.S.: 5 andere Person 1.8 V.S.: 2, E.S.: 2, F.S.: 2, S.F.: 2, S.K.: 2, M.O.: 1, K.S.: 1, I.S.: 2

(12)

(13)

270

Fragebogen: Bindung in AcI-Konstruktionen (#2) Bitte lesen Sie die folgenden Stze und entscheiden Sie, worauf sich das Reflexivpronomen (sich) bzw. das Personalpronomen (ihn/ihm/sie/ihr) bezieht. Sie werden sowohl Mehrdeutigkeiten als auch unmgliche Kombinationen vorfinden. Benutzen Sie die vorgegebene Bewertungsskala und tragen Sie in jedes Kstchen Ihr Urteil, also eine Zahl von 1 bis 5, ein. Dieselbe Zahl kann mehr als einmal pro Satz benutzt werden. Wenn Sie meinen den Kontext genauer erklren zu mssen, fgen Sie bitte einen Kommentar hinzu. Vielen Dank frs Mitmachen! Bewertungsskala: 1 Ja, kommt mir sofort in den Sinn. 2 Ja, mit entsprechendem Kontext; kommt mir aber nicht als erstes in den Sinn. 3 Ja, vielleicht, aber kommt mir komisch vor. 4 Kaum mglich; hrt sich falsch an. 5 Nein, eindeutig falsch.

I.
(15) Der kleine Junge lsst den Stein sich/ihm auf den Kopf fallen. sich bezieht sich auf: Der kleine Junge 1.3 E.S.: 2, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 2, M.: 1, F.S.: 1 ihm bezieht sich auf: Der kleine Junge 4.7 E.S.: 5, V.S.: 4, A.M.: 5, F.S.: 4, S.F.: 5, S.K. 5, P.R.: 4, R.S.: 5, M.: 5, F.S.: 5 andere Person 1.3 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 2, R.S.: 2, M.: 1, F.S.: 1 Der kleine Junge lsst sich/ihm den Stein auf den Kopf fallen. sich bezieht sich auf: Der kleine Junge 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 ihm bezieht sich auf: Der kleine Junge 5 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Person 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 Die Gromutter lsst den Wellensittich sich/ihr auf den Kopf fliegen. sich bezieht sich auf: Die Gromutter 1.4 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 2, M.: 1, F.S.: 3 ihr bezieht sich auf: Die Gromutter 3 E.S.: 3, V.S.: 2, A.M.: 2, F.S.: 4, S.F.: 4, S.K. 5, P.R.: 2, R.S.: 4, M.: 1, F.S.: 3 andere Person 1.2 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 2, M.: 1, F.S.: 1 Willi lsst die Chance sich/ihm nicht durch die Finger gleiten. sich bezieht sich auf: Willi 1.2 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 2, M.: 1, F.S.: 1 ihm bezieht sich auf: Willi 4.5 E.S.: 5, V.S.: 4, A.M.: 4, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 4, R.S.: 5, M.: 3, F.S.: 4 andere Person 3.3 E.S.: 4, V.S.: 5, A.M.: 5, F.S.: 2, S.F.: 2, S.K. 1, P.R.: 4, R.S.: 2, M.: 4, F.S.: 4 Willi lsst sich/ihm die Chance nicht durch die Finger gleiten. sich bezieht sich auf: Willi 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 ihm bezieht sich auf: Willi 5 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 4, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Person 3.2 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 1, S.F.: 2, S.K. 2, P.R.: 4, R.S.: 2, M.: 2, F.S.: 4

(16)

(17)

(18)

(19)

271

(20)

Andrea lsst die Katze sich/ihr nicht ins Haus kommen. sich bezieht sich auf: Andrea 1.7 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 5, S.K. 1, P.R.: 1, R.S.: 2, M.: 2, F.S.: 2 ihr bezieht sich auf: Andrea 3.3 E.S.: 3, V.S.: 2, A.M.: 3, F.S.: 3, S.F.: 5, S.K. 5, P.R.: 2, R.S.: 5, M.: 3, F.S.: 2 andere Person 2.2 E.S.: 2, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 5, S.K. 2, P.R.: 3, R.S.: 2, M.: 4, F.S.: 1 Andrea lsst sich/ihr die Katze nicht ins Haus kommen. sich bezieht sich auf: Andrea 1.3 E.S.: 2, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 2, R.S.: 1, M.: 1, F.S.: 1 ihr bezieht sich auf: Andrea 4.8 E.S.: 5, V.S.: 4, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Person 1.6 E.S.: 4, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 3, S.K. 1, P.R.: 2, R.S.: 1, M.: 1, F.S.: 1 Der Vater lsst den Jungen sich/ihm Zigaretten besorgen. sich bezieht sich auf: Der Vater 2.7 E.S.: 2, V.S.: 2, A.M.: 2, F.S.: 4, S.F.: 3, S.K. 3, P.R.: 2, R.S.: 5, M.: 1, F.S.: 3 den Jungen 1.2 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 2, R.S.: 1, M.: 1, F.S.: 1 ihm bezieht sich auf: Der Vater 1.5 E.S.: 2, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 3, P.R.: 1, R.S.: 2, M.: 1, F.S.: 1 den Jungen 4.7 E.S.: 4, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 3, F.S.: 5 andere Person 1.4 E.S.: 1, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 3, R.S.: 1, M.: 1, F.S.: 2 Der Vater lsst sich/ihm den Jungen Zigaretten besorgen. sich bezieht sich auf: Der Vater 3.9 E.S.: 5, V.S.: 4, A.M.: 4, F.S.: 4, S.F.: 2, S.K. 3, P.R.: 3, R.S.: 5, M.: 5, F.S.: 4 den Jungen 3.5 E.S.: 1, V.S.: 5, A.M.: 5, F.S.: 2, S.F.: 2, S.K. 3, P.R.: 5, R.S.: 5, M.: 2, F.S.: 5 ihm bezieht sich auf: Der Vater 4.8 E.S.: 3, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 den Jungen 5 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Person 2.9 E.S.: 3, V.S.: 3, A.M.: 4, F.S.: 2, S.F.: 3, S.K. 3, P.R.: 3, R.S.: 2, M.: 5, F.S.: 1 Die Mutter lsst die Kleine sich/ihr die Schokolade in den Mund stecken. sich bezieht sich auf: Die Mutter 3.1 E.S.: 2, V.S.: 2, A.M.: 2, F.S.: 3, S.F.: 3, S.K. 3, P.R.: 3, R.S.: 5, M.: 5, F.S.: 3 die Kleine 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 ihr bezieht sich auf: Die Mutter 1.6 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 3, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 5, F.S.: 1 die Kleine 4.9 E.S.: 4, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Person 1.7 E.S.: 1, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 3, S.K. 2, P.R.: 2, R.S.: 2, M.: 1, F.S.: 2 Die Mutter lsst sich/ihr die Kleine die Schokolade in den Mund stecken. sich bezieht sich auf: Die Mutter 4.4 E.S.: 4, V.S.: 4, A.M.: 5, F.S.: 4, S.F.: 3, S.K. 5, P.R.: 4, R.S.: 5, M.: 5, F.S.: 5 die Kleine 3.3 E.S.: 1, V.S.: 4, A.M.: 5, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 ihr bezieht sich auf: Die Mutter 4.9 E.S.: 4, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 die Kleine 5 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Person 2.9 E.S.: 3, V.S.: 4, A.M.: 4, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 3, R.S.: 5, M.: 5, F.S.: 1 Maja lsst Willi sich/ihr eine Geschichte erzhlen. sich bezieht sich auf: Maja 2.7 E.S.: 3, V.S.: 1, A.M.: 1, F.S.: 2, S.F.: 1, S.K. 5, P.R.: 1, R.S.: 5, M.: 5, F.S.: 4 Willi 2 E.S.: 2, V.S.: 2, A.M.: 2, F.S.: 1, S.F.: 3, S.K. 3, P.R.: 3, R.S.: 1, M.: 1, F.S.: 2 ihr bezieht sich auf: Maja 1.4 E.S.: 2, V.S.: 1, A.M.: 1, F.S.: 2, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 2, M.: 1, F.S.: 1 andere Person 1.2 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 2, P.R.: 1, R.S.: 1, M.: 1, F.S.: 2

(9)

(10)

(11)

(12)

(13)

(14)

272

(15)

Maja lsst sich/ihr Willi eine Geschichte erzhlen. sich bezieht sich auf: Maja 3.8 E.S.: 4, V.S.: 3, A.M.: 5, F.S.: 3, S.F.: 2, S.K. 5, P.R.: 4, R.S.: 5, M.: 2, F.S.: 5 Willi 3.6 E.S.: 2, V.S.: 5, A.M.: 5, F.S.: 2, S.F.: 3, S.K. 3, P.R.: 3, R.S.: 5, M.: 5, F.S.: 5 ihr bezieht sich auf: Maja 4.5 E.S.: 4, V.S.: 5, A.M.: 5, F.S.: 4, S.F.: 4, S.K. 3, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Person 2.1 E.S.: 1, V.S.: 3, A.M.: 4, F.S.: 2, S.F.: 3, S.K. 1, P.R.: 2, R.S.: 2, M.: 2, F.S.: 1 Die Spieler hren die Fans sich/sie anfeuern. sich bezieht sich auf: Die Spieler 2.9 E.S.: 3, V.S.: 1, A.M.: 1, F.S.: 2, S.F.: 3, S.K. 5, P.R.: 4, R.S.: 5, M.: 2, F.S.: 3 die Fans 1.8 E.S.: 2, V.S.: 2, A.M.: 2, F.S.: 3, S.F.: 2, S.K. 3, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 sie bezieht sich auf: Die Spieler 1.4 E.S.: 1, V.S.: 1, A.M.: 2, F.S.: 2, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 2 die Fans 4.9 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 4, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Leute 1.8 E.S.: 2, V.S.: 2, A.M.: 1, F.S.: 2, S.F.: 3, S.K. 2, P.R.: 2, R.S.: 1, M.: 1, F.S.: 2 Die Spieler hren sich/sie die Fans anfeuern. sich bezieht sich auf: Die Spieler 4.4 E.S.: 4, V.S.: 5, A.M.: 5, F.S.: 2, S.F.: 3, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 die Fans E.S.: 2, V.S.: 5, A.M.: 5, F.S.: 4, S.F.: 4, S.K. 3, P.R.: 5, R.S.: 5, M.: 2, F.S.: 5 sie bezieht sich auf: Die Spieler 5 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 3, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 die Fans 4.8 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 3, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Leute 1.1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 Maja lie die Mdigkeit sich/sie berkommen. sich bezieht sich auf: Maja 3.1 E.S.: 3, V.S.: 1, A.M.: 1, F.S.: 5, S.F.: 4, S.K. 3, P.R.: 3, R.S.: 5, M.: 5, F.S.: 1 sie bezieht sich auf: Maja 2.7 E.S.: 4, V.S.: 4, A.M.: 4, F.S.: 3, S.F.: 2, S.K. 3, P.R.: 3, R.S.: 1, M.: 1, F.S.: 2 andere Person 3.5 E.S.: 3, V.S.: 4, A.M.: 4, F.S.: 2, S.F.: 3, S.K. 3, P.R.: 5, R.S.: 5, M.: 3 F.S.: 3 Maja lie sich/sie die Mdigkeit berkommen. sich bezieht sich auf: Maja 3.6 E.S.: 4, V.S.: 5, A.M.: 4, F.S.: 1, S.F.: 3, S.K. 5, P.R.: 3, R.S.: 5, M.: 2, F.S.: 4 sie bezieht sich auf: Maja 4.7 E.S.: 4, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 4, R.S.: 5, M.: 4, F.S.: 5 andere Person 3.5 E.S.: 4, V.S.: 5, A.M.: 5, F.S.: 2, S.F.: 3, S.K. 3, P.R.: 3, R.S.: 2, M.: 5, F.S.: 3 James Bond lie die junge Frau sich/ihm in die Arme fallen. sich bezieht sich auf: James Bond 1.9 E.S.: 1, V.S.: 2, A.M.: 2, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 5, M.: 2, F.S.: 3 ihm bezieht sich auf: James Bond 3.2 E.S.: 5, V.S.: 2, A.M.: 3, F.S.: 4, S.F.: 3, S.K. 5, P.R.: 2, R.S.: 1, M.: 5, F.S.: 2 andere Person 1.2 E.S.: 1, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 2

(16)

(17)

(18)

(19)

(20)

II.
(21) Maja lie die Mdigkeit ber sich/sie kommen. sich bezieht sich auf: Maja 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 sie bezieht sich auf: Maja 3.6 E.S.: 2, V.S.: 3, A.M.: 5, F.S.: 4, S.F.: 4, S.K. 5, P.R.: 5, R.S.: 4, M.: 2, F.S.: 2 andere Person 2.9 E.S.: 2, V.S.: 4, A.M.: 5, F.S.: 1, S.F.: 3, S.K. 3, P.R.: 5, R.S.: 1, M.: 3, F.S.: 2

273

(22)

Die jungen Eltern lieen die Verantwortung auf sich/sie zukommen. sich bezieht sich auf: Die jungen Eltern 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 sie bezieht sich auf: Die jungen Eltern 3.4 E.S.: 2, V.S.: 2, A.M.: 4, F.S.: 4, S.F.: 3, S.K. 2, P.R.: 4, R.S.: 5, M.: 5, F.S.: 3 andere Person 1.4 E.S.: 2, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 2 Der Bauarbeiter sah das Gerst auf sich/ihn niederstrzen. sich bezieht sich auf: Der Bauarbeiter 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 ihn bezieht sich auf: Der Bauarbeiter 3.7 E.S.: 3, V.S.: 4, A.M.: 4, F.S.: 3, S.F.: 3, S.K. 5, P.R.: 5, R.S.: 5, M.: 2, F.S.: 3 andere Person 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 Der Fugnger sah die Radfahrer auf sich/ihn zurasen. sich bezieht sich auf: Der Fugnger 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 die Radfahrer 3.1 E.S.: 2, V.S.: 3, A.M.: 2, F.S.: 1, S.F.: 3, S.K. 5, P.R.: 5, R.S.: 4, M.: 2, F.S.: 4 ihn bezieht sich auf: Der Fugnger 2.4 E.S.: 2, V.S.: 1, A.M.: 2, F.S.: 2, S.F.: 3, S.K. 5, P.R.: 2, R.S.: 3, M.: 2, F.S.: 2 andere Person 1.1 E.S.: 1, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 Der Knig lie den Gefangenen vor sich/ihm niederknien. sich bezieht sich auf: Der Knig 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 den Gefangenen 4.1 E.S.: 5*, V.S.: 3, A.M.: 2, F.S.: 5*, S. F.: 5, S.K. 5, P.R.: 5*, R.S.: 4, M.: 5, F.S.: 2 *nur wenn vorm Spiegel ihm bezieht sich auf: Der Knig 2.3 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 4, S.F.: 3, S.K. 3, P.R.: 2, R.S.: 5, M.: 1, F.S.: 2 den Gefangenen 5 E.S.: 5, V.S.: 5, A.M.: 5, F.S.: 5, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 5 andere Person 1.1 E.S.: 1, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 Britta lie den Ball auf sich/sie zurollen. sich bezieht sich auf: Britta 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 sie bezieht sich auf: Britta 4.3 E.S.: 3, V.S.: 4, A.M.: 4, F.S.: 4, S.F.: 5, S.K. 5, P.R.: 5, R.S.: 5, M.: 5, F.S.: 3 andere Person 1.2 E.S.: 1, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 Die Demonstrantin lie den Polizisten auf sich/sie zukommen. sich bezieht sich auf: Die Demonstrantin 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 sie bezieht sich auf: Die Demonstrantin 2.5 E.S.: 2, V.S.: 1, A.M.: 1, F.S.: 3, S.F.: 3, S.K. 5, P.R.: 3, R.S.: 3, M.: 3, F.S.: 1 andere Person 1.4 E.S.: 2, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 2, R.S.: 1, M.: 1, F.S.: 1 Der Polizist lie den leblosen Krper auf sich/ihn fallen. sich bezieht sich auf: Der Polizist 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 ihn bezieht sich auf: Der Polizist 3.6 E.S.: 4, V.S.: 4, A.M.: 4, F.S.: 4, S.F.: 3, S.K. 5, P.R.: 2, R.S.: 5, M.: 3, F.S.: 2 andere Person 1.1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 2, R.S.: 1, M.: 1, F.S.: 1

(23)

(24)

(25)

(26)

(27)

(28)

274

(29)

James Bond lie die junge Frau auf sich/ihn fallen. sich bezieht sich auf: James Bond 1.1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 2, M.: 1, F.S.: 1 ihn bezieht sich auf: James Bond 3.2 E.S.: 3, V.S.: 3, A.M.: 4, F.S.: 4, S.F.: 3, S.K. 5, P.R.: 2, R.S.: 4, M.: 1, F.S.: 3 andere Person 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1

III.
(30) Der Pessimist fhlte den Himmel ber sich/ihm einstrzen. sich bezieht sich auf: Der Pessimist 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 ihm bezieht sich auf: Der Pessimist 3 E.S.: 2, V.S.: 3, A.M.: 3, F.S.: 3, S.F.: 2, S.K. -, P.R.: 4, R.S.: 3, M.: 4, F.S.: 3 andere Person 2.4 E.S.: 3, V.S.: 3, A.M.: 3, F.S.: 1, S.F.: 5, S.K. -, P.R.: 3, R.S.: 1, M.: 2, F.S.: 1 Aber er lie den Mann neben sich/ihm untergehen. sich bezieht sich auf: er 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 ihm bezieht sich auf: er 2.1 E.S.: 3, V.S.: 2, A.M.: 3, F.S.: 3, S.F.: 3, S.K. 1, P.R.: 1, R.S.: 1, M.: 2, F.S.: 2 andere Person 1.5 E.S.: 1, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 2, P.R.: 2, R.S.: 3, M.: 1, F.S.: 1 Meine Mitbewohnerin lie das Essen bei sich/ihr unterm Bett verschimmeln. sich bezieht sich auf: Mitbewohnerin 1 E.S.: 1, V.S.: 1, A.M.: 1, F.S.: 1, S.F.: 1, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 1 ihr bezieht sich auf: Mitbewohnerin 2.4 E.S.: 3, V.S.: 2, A.M.: 2, F.S.: 2, S.F.: 2, S.K. 2, P.R.: 3, R.S.: 1, M.: 5, F.S.: 2 andere Person 1.4 E.S.: 2, V.S.: 2, A.M.: 1, F.S.: 1, S.F.: 2, S.K. 1, P.R.: 1, R.S.: 1, M.: 1, F.S.: 2

(31)

(32)

275

Bibliography
Abels, Klaus. 2003. Successive cyclicity, anti-locality, and adposition stranding. Ph.D. dissertation. University of Connecticut, Storrs. Aissen, Judith. 1974. The syntax of causative constructions. Ph.D. dissertation. Harvard University. (1978. New York: Garland.) Aissen, Judith & Perlmutter, David. 1976. Clause reduction in Spanish. In H. Thompson, K. Whistler, V. Edge, J. Jaeger, R. Javkin, M. Petruck, C. Smeall, & R. Van Valin (eds.) Proceedings of BLS 2, 1-30. (1983. Studies in Relational Grammar 1. Chicago: The University of Chicago Press, 360-403.) Allen, Barbara, Donald Frantz, Donna Gardiner & David Perlmutter. 1990. Verb agreement, possessor ascension, and multistratal representation in Southern Tiwa. In P. Postal & B. Joseph (eds.) Studies in Relational Grammar 3. The University of Chicago Press, 321-383. Anagnostopoulou, Elena. 2002. Two clauses of double object verb. Ms. University of Crete. Baker, Mark. 1988. Incorporation. Chicago: The University of Chicago Press. Baker, Mark. To appear. Applicatives. Syntax Companion (SynCom). Oxford: Blackwell. Baltin, Mark. 1982. A landing site theory of movement rules. LI 13 (1). 1-38. Baltin, Mark. 2003. The interaction of ellipsis and binding: implications for the sequencing of Principle A. NLLT 21. 215-246. Bayer, Josef & Jaklin Kornfilt. 1990. Reconstruction effects in German. In E. Engdahl et al. (eds.) Parametric variation in Germanic and Romance. 21-42. Bayer, Josef, Tanja Schmid & Markus Bader. To appear. Clause union and clausal position. In M. den Dikken & C. Tortora (eds.) The function of function words and functional categories. Amsterdam: John Benjamins. Bech, Gunnar. 1955/57. Studien ber das deutsche Verbum infinitivum. Kopenhagen. 2nd edition 1983. Tbingen: Niemeyer.
276

Besten, Hans den & Jean Rutten. 1989. On verb raising, extraposition and free word order in Dutch. In D. Jaspers et al. (eds.) Sentential complementation and the lexicon. Studies in honor of Wim de Geest. Dordrecht: Foris, 41-56. Besten, Hans den & Jerold Edmonson. 1983. The verbal complex in Continental West Germanic. In W. Abraham (ed.) On the formal syntax of the Westgermania. Amsterdam: Benjamins, 155-216. Borer, Hagit & Josef Grodzinsky. 1986. Syntactic cliticization and lexical cliticization: The case of Hebrew dative clitics. In H. Borer (ed.) Syntax and Semantics 19. New York Academic Press, 175-217 Bobaljik, Jonathan. 2002. A-chains at the PF-interface: Copies and covert movement. NLLT 20. 197-267. Bos&kovic@, Z&eljko. 2001. On the nature of the syntax-phonology interface: Cliticization and related phenomena. Amsterdam: Elsevier. Bring, Daniel. 1997. The meaning of topic and focus. The 59th Street bridge accent. London & New York: Routledge. Bring, Daniel. 2001. Lets phrase it! In Mller, G. & W. Sternefeld (eds.) Competition in syntax. Studies in Generative Grammar 49. Berlin & New York: de Gruyter. 69-105. Burzio, Luigi. 1986. Italian Syntax: A Government-Binding Approach. Dordrecht: Reidel. Chomsky, Noam. 1973. Conditions on transformations. In S. Anderson & P. Kiparsky (eds.) A festschrift for Moris Halle. New York: Holt, Rinehard and Winston. Chomsky, Noam. 1977. On wh-movement. In P. Cullicover, T. Wasow, & A. Akmajian (eds.) Formal syntax. New York: Academic Press, 71-132. Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In R. Martin, D. Michaels, & J. Uriagereka (eds.) Step by step: Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, MA: MIT Press, 89-156.
277

Chomsky, Noam. 2001. Derivation by phase. In M. Kenstowicz (ed.) Ken Hale, A life in language. Cambridge, MA: MIT Press, 1-51. Chung, Sandra. 2003. Restructuring and verb-initial order in Chamorro. Syntax 7, 199-233. Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67, 547-619. Drosdowski, Gnther & Werner Scholze-Stubenrecht. 1992. Duden: Redewendungen und sprichwrtliche Redensarten (Band 11). Mannheim: Dudenverlag, Evers, Arnold. 1975. The transformational cycle in Dutch and German. Ph.D. dissertation, University of Utrecht, The Netherlands. Fanselow, Gisbert. 1989. Coherent infinitives in German. In C. Bhatt, E. Lbel, & C. Schmidt (eds.) Syntactic phrase structure phenomena in noun phrases and sentences. Amsterdam: Benjamins, 1-16. Fanselow, Gisbert. 1990. Scrambling as NP-movement. In G. Grewendorf & W. Sternefeld (eds.) Scrambling and barriers. Amsterdam: Benjamins, 113-140. Fox, Danny. 2000. Economy and semantic interpretation. Cambridge, MA: MIT Press. Frey, Werner. 1993. Syntaktische Bedingungen fr die semantische Interpretation. Studia Grammatika 35. Berlin: Akademie Verlag. Gallmann, Peter. 1992. Dativanhebung? Groninger Arbeiten zur Germanistischen Linguistik 35, 92-122. Geilfu, Jochen. 1991. Scrambling und Pseudoscrambling. Arbeitspapier Nr. 11 des SFB 340. Universitt Stuttgart/Tbingen. Gibson, Jeanne & Eduardo Raposo. 1986. Clause union, the stratal uniqueness law and the chmeur relation. Natural Language and Linguistic Theory 4, 295331.

278

Grewendorf, Gnther & Wolfgang Sternefeld. 1990. Scrambling theories. In G. Grewendorf & W. Sternefeld (eds.) Scrambling and barriers. Linguistik Aktuell 5. Amsterdam: Benjamins, 3-37. Grewendorf, Gnther. 1988. Aspekte der deutschen Syntax. Tbingen: Narr. Grewendorf, Gnther. 1987. Kohrenz und Restrukturierung. Zu verbalen Komplexen im Deutschen. In B. Asbach-Schnitker & J. Roggenhofer (eds.) Neuere Forschungen zur Wortbildung und Historiographie der Linguistik. Tbingen: Narr, 123-144. Grewendorf, Gnther. 1985. Anaphern bei Objekt-Koreferenz im Deutschen: Ein Problem fr die Rektions-Bindungs Theorie. In W. Abraham (ed.) Erklrende Syntax des Deutschen. Tbingen: Narr, 137-171. Grewendorf, Gnther & Joachim Sabel. 1994. Long Scrambling and incorporation. LI 25, 263-308. Grimshaw, Jane. 1990. Argument structure. Cambridge, MA: MIT Press. Grimshaw, Jane. 2000. Locality and extended projection. In P. Coopmans, M. Everaert & J. Grimshaw (eds.) Lexical specification and insertion. Amsterdam: Benjamins, 115-133. Guron, Jacqueline. 1985. Inalienable possession: PRO inclusion and lexical chains. In H.-G. Obenauer, J.-Y. Pollock & J. Guron (eds.) Grammatical representation. Dordrecht: Foris, 43-86. Gunkel, Lutz. 2003. Reflexivierung in AcI-Konstructionen. In L. Gunkel, G. Mller, & G. Zifonun (eds.) Arbeiten zur Reflexivierung. Tgingen: Niemeyer, 115-133. Gurtu, M. 1985. Anaphoric relations in Hindi and English. PhD Dissertation. Central Institute of English and Foreign Languages, Hyderabad. Haegeman, Liliane & Henk van Riemsdijk. 1986. Verb projection raising, scope, and the typology of rules affecting verbs. LI 17, 417-466. Haider, Hubert. 1985. ber sein oder nicht sein: Zur Grammatik des Pronomens sich. In W. Abraham (ed.) Erklrende Syntax des Deutschen. Tbingen: Narr, 223-254.

279

Haider, Hubert. 1987. Deutsche Syntax Generativ. Parameter der deutschen Syntax. Ms. University of Vienna. Haider, Hubert. 1993. Deutsche Syntax Generativ. Tbingen: Narr. Haider, Hubert & Inger Rosengren. 1998. Scrambling. Sprache und Pragmatik 49, 1-104. Hankamer, Jorge & Vera Lee-Schoenfeld. 2005. What moves in German VPtopicalization? LSA Annual Meeting, Oakland, CA. Hale, Kenneth & Samuel Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In K. Hale & S. Keyser (eds.) View from Building 20, 53-109. Harbert, Wayne & Veneeta Srivastav. 1988. A complement/adjunct binding asymmetry in Hindi and other Languages. Cornell Working Papers in Linguistics 8, 79-105. Hestvik, Arild. 1991. Subjectless binding domains. NLLT 9, 455-496. Hinterhlzl, Robert & Diana Pili. 2003. Argument shift phenomena across language types. Ms. Humbold University, Berlin and University of Siena. Hole, Daniel. To appear. Reconciling possessor datives and beneficiary datives Toward a unified account of dative binding in German. In C. Maienburg & C. Wllstein-Leisten (eds.) Event arguments in syntax, semantics, and discourse. Tbingen: Niemeyer. Hole, Daniel. 2005. Extra argumentality A binding account of possessor raising in German, English, and Mandarin. In J.-Y. Kim, B. H. Partee, & W. A. Lander (eds.) Possessives and beyond: Semantics and syntax. GLSA Publications, Amherst, MA. Hornstein, Norbert. 1999. Movement and Control. LI 30, 69-96. Hhle, Tilman. 1978. Lexikalistische Syntax. Die Aktiv-Passiv-Relation und andere Infinitivphrasen im Deutschen. Niemeyer, Tbingen. Huang, James. 1982. Logical relations in Chinese and the theory of grammar. Ph.D. dissertation. MIT, Cambridge, MA.

280

Isac&enko, Alexander. 1965. Das syntaktische Verhltnis der Bezeichnungen von Krperteilen im Deutschen. Studia Grammatica 5, 7-27. Kayne, Richard. 1975. French syntax: The transformational cycle. Cambridge, MA: MIT Press. Kayne, Richard. 1993. Toward a modular theory of auxiliary selection. Studia Linguistica 47, 3-31. Keach, Camillia and Michael Rochemont. 1992. On the syntax of possessor raising in Swahili. Studies in African Linguistics 23, 81-106. Kempchinsky, Paula. 1992. The Spanish possessive dative construction: q-role assignment and proper government. In P. Hirschbuhler (ed.) Linguistic Symposium on Romance Languages 20, 135-149. Kiss, Tibor. 1995. Infinite Komplementation. Tbingen: Niemeyer. Kiss, Tibor. 2001. Anaphora and Exemptness. A comparative treatment of anaphoric binding in German and English. In D. Flickinger & A. Kathol (eds.) The Proceedings of the 7th International Conference on Head-Driven Phrase Structure Grammar. CSLI Publications 2001, 182-197. Kluender, R. 1986. Stzchen: German small clauses as Ss. In S. Berman, J.-W. Choe, & J. McDonough (eds.) Proceedings of NELS 16. Somerville, MA: Cascadilla Press, 274-292. Kratzer, Angelika. 1996. Severing the external argument from its verb. In J. Rooryck & L. Zaring (eds.) Phrase structure and the lexicon. Dordrecht: Kluwer, 109-137. Kubo, Miori. 1990. Japanese Passives. Ms. MIT, Cambridge, MA. Landau, Idan. 1999. Possessor raising and the structure of VP. Lingua 107, 1-37. Lee-Schoenfeld, Vera. 2003. Agentivity vs. auxiliary choice: evidence from pronominal binding in German AcI-constructions. Talk given at the Workshop on Cross-linguistic Auxiliary Selection, UC Davis. May 2003. (To appear in conference volume. Amsterdam: Benjamins.) Legate, Julie. 2003. Some interface properties of the phase. LI 34, 506-516.
281

Longobardi, Giuseppe. 1985. Connectedness, scope and c-command. LI 16, 163192. Longobardi, Giuseppe. 1996. The syntax of N-raising: A minimalist theory. OTS Working Papers-TL-96-005, Utrecht. Longobardi, Giuseppe. 2001. The structure of DPs: Some principles, parameters and problems. In M. Baltin & C. Collins (eds.) The handbook of contemporary syntactic theory. Blackwell Publishing, 562-603. Lpez, Luis. 2001. On the (non)complementarity of q-theory and checking theory, LI 32, 694-716. Maldonado, Ricardo. 2002. Objective and subjective datives. Cognitive Linguistics 13, 1-66. McCloskey, James. 2000. Quantifier float and Wh-movement in an Irish English. LI 31, 57-84. McCloskey, James. 2002. Resumption, successive cyclicity, and the locality of operations. In S. Epstein & D. Seeley (eds.) Derivation and explanation. Blackwell Publishers, 184226. McFadden, Thomas. 2004. The position of morphological case in the derivation. Ph.D. dissertation. University of Pennsylvania. McIntyre, Andrew. To appear. The interpretation of German datives and English have. In W. Abraham, D. Hole, & A. Meinunger (eds.) Dative and similar cases (working title). Amsterdam: Benjamins. Miyagawa, Shigeru. 1987. Restructuring in Japanese. In T. Imai and M. Saito (eds.) Issues in Japanese linguistics. Dordrecht: Foris, 273-300. Miyagawa, Shigeru & Takae Tsujigka. 2004. Argument structure and ditransitive verbs in Japanese. Journal of East Asian Linguistics 13, 1-38. Moore, John. Reduced construction in Spanish. 1991. Ph.D. dissertation. University of California, Santa Cruz. Mller, Gereon. 1996. A constraint on remnant movement. NLLT 14, 355-407.

282

Nissenbaum, Jon. 2000. Investigations of covert phrase movement. PhD dissertation. MIT, Cambridge, MA. Payne, Doris & Immanuel Barshi. 1999. External possession: what, where, how, and why. In D. Payne & I. Barshi (eds.) External possession. Amsterdam: Benjamins. Pereltsvaig, Asya. 2003. Syntax of denominal and ditransitive verbs reconsidered. Ms. University of Sheffield, United Kingdom. Pica, Pierre. 1987. On the nature of the reflexivization cycle. In J. McDonough & B. Plunkett (eds.) Proceedings of NELS 17. Somerville, MA: Cascadilla Press, 483-499. Primus, Beatrice. 1989. Parameter der Herrschaft: Reflexivpronomina im Deutschen. Zeitschrift fr Sprachwissenschaft 8, 53-88. Pustejovsky, James. 1984. Opacity and the accessibility of subject in German A.c.I.constructions. In C. Jones & P. Sells (eds.) Proceedings of NELS 14. Somerville, MA: Cascadilla Press, 360-376. Pylkknen, Liina. 2002. Introducing Arguments. Ph.D. dissertation. MIT, Cambridge, MA. Reinhart, Tanya & Eric Reuland. 1993. Reflexivity. LI 24, 657-720. Reintges, Chris, Philip LeSourd & Sandra Chung. To appear. Movement, Whagreement, and apparent Wh-in-situ. In L. Cheng & N. Cover (eds.) WhMovement on the Move. Cambridge, MA: MIT Press. Reis, Marga. 1973. Is there a rule of subject-to-object raising in German? CLS 9, 519-529. Reis, Marga. 1976. Reflexivierung in deutschen A.c.I.-Konstruktionen. Ein transformations-grammatisches Dilemma. Papiere zur Linguistik 9, 5-82. Reis, Marga & Wolfgang Sternefeld. 2004. Review of Wurmbrand, Susanne: Infinitives. Restructuring and clause structure. Linguistics 42, 469-508. Rizzi, Luigi. 1978. A restructuring rule in Italian. In S. Keyser (ed.) Recent transformational studies in European languages. Cambridge, MA: MIT Press, 113-158.
283

Rizzi, Luigi. Forthcoming. On the form of chains: Criterial positions and ECP effects. Ms. University of Siena. Rosengren, Inger. 1992. Zum Problem der kohrenten Verben im Deutschen. In P. Suchsland (ed.) Biologische und soziale Grundlagen der Sprachen. Tbingen: Niemeyer, 265-295. Ross, John. 1967. Constraints on variables in syntax. Ph.D. dissertation. MIT, Cambridge, MA. Rutten, Jean. 1991. Infinitival complements and auxiliaries. Ph.D. dissertation. University of Amsterdam. Sabel, Joachim. 1994. Restrukturierung und Lokalitt: Universelle Beschrnkungen fr Wortstellungsvariationen. Ph.D. dissertation. University of Frankfurt. Sabel, Joachim. 2002. Das deutsche Verbum infinitum. Deutsche Sprache 29, 148175. Safir, Ken. 2004. The Syntax of anaphora. Oxford University Press. Santorini, Beatrice & Anthony Kroch. 1990. Remnant extraposition in German. Ms. University of Pennsylvania, Philadelphia. Schmid, Tanja, Markus Bader, & Josef Bayer. Forthcoming. Coherence An experimental approach. Ms. University of Konstanz. Sorace, Antonella. 2000. Gradients in auxiliary selection with intransitive verbs. Language 76, 859-890. Shibatani, Masayoshi. 1994. An integrational approach to possessor raising, ethical datives and adversative passives. Proceedings of BLS 20, 461-486. Stechow, Arnim von & Wolfgang Sternefeld. 1988. Bausteine syntaktischen Wissens: Ein Lehrbuch der modernen generativen Grammatik. Opladen: Westdeutscher Verlag. Steube, Anita. 2001. Grammatik und Pragmatik von Hutkonturen. Linguistische Arbeitsberichte 77, 7-29.

284

Sturgeon, Anne. 2003. Two-tiered approach to binding domain formation: Evidence from Czech. In B. Partee, W. Brown, J.-Y. Kim, & R. Rothstein (eds.) Proceedings of the 11th Annual Formal Approaches to Slavic Linguistics. Ann Arbor, Mi: Michigan Slavic Publications, 495-514. Svenonius, Peter. 2004. On the edge. In D. Adger, C. de Cat & G. Tsoulas (eds.) Peripheries. Dordrecht: Kluwer, 259-287. Szabolcsi, Anna. 1981. The possessive construction in Hungarian: A configurational category in a non-configurational language. Acta Linguistica Academiae Scientiarum Hungaricae 31. Szabolcsi, Anna. 1983. The possessor that ran away from home. The Linguistic Review 3, 89-102. Ura, Hiroyuki. 1996. Multiple feature checking: A theory of grammatical function splitting. Ph.D. dissertation. MIT, Cambridge, MA. Vergnaud J.-R. & M.L. Zubizarreta. 1992. The definite determiner and the inalienable constructions in French and in English. LI 23, 595-652. Vikner, Sten. 1995. Verb movement and expletive subjects in the Germanic languages. New York & Oxford: Oxford University Press. Webelhuth, Gert. 1987. A universal theory of scrambling. In V. Rosn (ed.) Papers from the Tenth Scandinavian Conference on Lingusitics 2. University of Bergen. Wegener, Heide. 1985. Der Dativ im heutigen Deutsch. Tbingen: Narr. Wegener, Heide. 1991. Der Dativ Ein struktureller Kasus? In G. Fanselow & S. Felix (eds.) Strukturen und Merkmale syntaktischer Kategorien. Tbingen: Narr, 70-103. Wunderlich, Dieter. 1996. Dem Freund die Hand auf die Schulter legen. In G. Harras and M. Bierwisch (eds.) Wenn die Semantik arbeitet. Tbingen: Niemeyer, 331-360. Wunderlich, Dieter. 2000. Predicate composition and argument extension as general options A study on the interface of semantic and conceptual structure. In B. Stiebels and D. Wunderlich (eds.) Lexicon in Focus. Berlin: Akademie Verlag, 247-270.
285

Wllstein-Leisten, Angelika. 2001. Die Syntax der dritten Konstruktion. Tbingen: Stauffenburg. Wurmbrand, Susi. 2001. Infinitives: Restructuring and clause structure. Berlin & New York: Mouton de Gruyter. Wurmbrand, Susi. 2004. Licensing case. Ms. University of Connecticut. Zwart, Jan-Wouter. 1997. Morphosyntax of verb movement. Dordrecht: Kluwer.

286