This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
head, which has both afeature and a <WH> feature. (1.25) a.
[C,[Pat willread [what<WH> to whom<WH>]]] → IM
b.
[will[C[Pat will read [what<WH> to whom<WH>]]]] → IM
c.
[what<WH> [will[C[Pat will read [what to whom<WH>]]]]] ‘What will Pat read to whom?’
In (1.25a), the Chead will look for a constituent with afeature within its search domain. Once it locates the constituent will, the search will terminate and IM will apply, remerging willin the matrix CP. Since the C head also has a <WH> feature, the head must being another search, this time one that looks for a <WH> counterpart feature. The search will again terminate (successfully) when the first active <WH> is found on the constituent what<WH>. IM will apply once again, this time remerging what<WH> in the matrix CP. Although on the surface these two searches outlined in (1.25) appear to be relatively simple, they are, in fact, tremendously complex. Stroik and Putnam (2007:7) argue: If we look closely at how the derivation proceeds from (1.25b) to (1.25c), we will see some of this complexity. Of note here is the fact that IM can’t apply directly to (1.25). Rather, before IM can apply to (1.25b), the C<WH> head must initiate a search for a constituent with a <WH> feature – that is, a FIND<WH> operation must apply at D. The FIND operation itself is a complex (composite) operation that must LOOK-ATFEATURES of every constituent K in the Derivation, that must DISREGARD allfeatures of K that are not a <WH> feature, and that must GO-TO-NEXT constituent should a <WH> feature not be found on a given K; and of course, all these operations will have to apply to each successive K that is searched. Furthermore, should the search successfully FIND<WH> in D (the what<WH> constituent), then the search ends, but another set of operations must now apply. These operations include a MAKE-COPY operation, an operation that, according to Chomsky (2001), will create a copy without phonetic features8, and REMOVE-ORIGINAL and INSERT-COPY operations, which will remove the constituent located by FIND and replace it with the copy formed by the non-phonetically realized MAKE-COPY opera-
8
This is a rather problematic operation in that it is unclear where this operation could take place.
Chapter 1. Introduction
19
tion. The removed constituent can now undergo IM with D, deriving (1.25c) from (1.25b) after the application of all the intermediate operations defined above.9
Needless to say, building syntactic structure from EM and IM in their current state is extremely costly, especially within processing terms, because of all of the ancillary machinery necessary to operate IM. A derivational analysis according to the Survive Principle simplifies the additional operations necessary to restrain IM. Under the Survive Principle, EM and IM can be reduced to the primitives Merge and Remerge, two operations that are essentially identical in design and purpose. The definition of the function of narrow syntax or the Derivation is thus clarified in a Minimalist Program of syntactic theory adhering to the Survive Principle. The Derivation D according to Survive is thus the iterative mapping of copies of SOs that exist in the Numeration. This notion has far-reaching theoretical effects on the Minimalist Program; no longer are notions of the Copy Theory (cf. Chomsky 1995), look-back and/or look-ahead mechanisms, economy constraints and ontological commitments (e.g., phases) deemed conceptually necessary. All of these aforementioned ‘essential’ components of minimalist syntactic theory are attached to a theory of constituent distal that make use of Internal Merge/Move. The removal of Internal Merge also liberates the theory from these taxing computational constraints. As pointed out by Frampton and Gutmann (2002), look-ahead and lookback mechanisms (i.e., non-cyclic operations) in the grammar impose a massive workload on the CHL. This is clearly an undesirable result in a theory that seeks to minimize to the fullest extent all operations and constraints that function within the grammar. The Survive Principle only employs two primitive operations (Merge and Remerge); hence it brings the theory one step closer to Frampton and Gutmann’s (2002) vision of a crash-proof syntax and how it may potentially operate within the framework of minimalism.10 Second, the Survive Principle also threatens the virtual conceptual necessity of uninterpretable and unvalued features in both the Numeration and in the Derivation. According to the Survive Principle, both Merge and Remerge are not viewed as a ‘checking’ configuration per se, but rather as a fusion process uniting two identical, interpretable features.11 Both Attract and Move models of XP9
Another problem with current minimalism is that in the derivation illustrated in (1.25) the Chead possessing the [Q]-feature that will peer down into its c-command domain to find a relevant probe is previously blind to the wh-items that have existed in the derivation for quite some time. That these wh-items are not found and recognized until the very end of the derivation is quite problematic for a true minimalist analysis of syntax. 10 David Pesetsky (p.c.) points out the application of evaluation processes upon ever iterative step of Merge could also be viewed as a costly processing strain on the grammar. Be that as it may, it is far more in line with minimalist thought to envision of a system with purely local evaluation mechanisms rather than one that employs look-ahead and look-back operations. 11 Similar in structure and purpose to Rizzi’s (1991) Dynamic Agreement.
20
Scrambling and the Survive Principle
displacement rely upon the notion of uninterpretable and/or unvalued features, while Survive only uses interpretable features that are immediately interpretable at the interfaces upon concatenation (e.g., Merge). Under such principles SOs enter the derivation fully interpretable at both the LF and PF interfaces. Third, the Survive Principle addresses Brody’s (1998, 2002) observation that a Minimalist Program of syntax that employs Internal Merge is a mixed theory that is weakly representational in its design. The elimination of Internal Merge in favor of the Survive Principle creates a pure derivational view of minimalism. In conclusion, the Survive Principle amounts to defining the fundamental problem of compositionality. The Survive Principle addresses the prime function of language structure, namely, how can two distinct units ever make up a whole that, in turn, may function as part of another, larger whole (cf. Sigurðsson 2004:65)? In the spirit of Collins (2002), who seeks to eliminate the necessity of labels in a syntactic system, the Survive Principle provides a clear explanation of the purpose of the concatenation of two separate lexical items α and β: The merger of lexical items α and β begins and ends with concatenated features. The operations Merge and Remerge make the newly formed syntactic object {αβ} accessible to interpretation at LF and PF. 1.2.1
Copy Theory and Link!
With the potential conceptual advantages of the Survive version of minimalism in mind, let’s once again return to the sample derivation in (1.17) above. At this juncture, the purpose of this section is to elucidate further on the nature of Remerge and the status of lower ‘copies’ that exist in the derivation for a given SO residing in the Numeration. It is paramount to bear in mind that SOs reside in the Numeration throughout; what appears during the course of the derivation is ‘something else’. What happens immediately upon concatenation – be it an application of Merge or Remerge – is that an identity is formed linking the fused features of head and SO. This newly bound unit is immediately interfaceinterpretable. Call this operation Link!12 (1.26)
12
Link! (Putnam 2006b: 6) If XP is a syntactic object (SO) bearing the interpretable feature α, and a syntactic head Yo enters the derivation bearing the interpretable feature α, XP will remerge into the structural position immediately dominating Yo [SpecYP]13, at which point the matching features will amalgamate and be interpreted at the interfaces.
The primitive function Link! is similar in design and function to Kobele’s (2006) Principle of Immediacy that requires features to be checked as soon as appropriate configurations arise in the narrow syntax. 13 An idea worthy of potential future research is Sigurðsson’s (2004) removal of X’-theory from the minimalist program in favor of a nucleus/edge approach.
Chapter 1. Introduction
21
The primitive function Link! – working in tandem with Merge, Survive and Remerge – maps the interpretable features {α, β , γ} of a SO (W) into the derivation by means of syntactic structure, generated by Merge and Remerge (example taken from Putnam 2006b: 7).14 (1.27) Num = {W(α, β, γ), X(α), Y(β), Z(γ)} a.
Merge {W,X} → WX
b.
Link! {α, α} → α
c.
Survive {W}
d.
Merge {Y, {W, X}} → YWX
e.
Remerge {W, {Y, {W, X}}} → WYWX
f.
Link! {β, β} → β
g.
Survive {W}
h.
Merge {Z, {W, {Y, {W, X}}}} → ZWYWX
j.
Remerge {W, {Z, {W, {Y, {W, X}}}}} → WZWYWX
k.
Link! {γ, γ} → γ
l.
Spell-Out
m.
PF = WXYX
n.
LF = WZWYWZ
The sample derivation in (1.27) is an expanded version of (1.17) with the help of abstract variables and with the addition of Link!15 We no longer have to er14
Due to the dynamic, rapid composition of syntactic structure in the course of real-time construction Remerge and Link! occur simultaneously. Be that as it may, I parse these two operations separate from one another for the sake of clarity. 15 David Pesetsky and Peter Sells (p.c.) have brought to my attention the strikingly familiarity the Survive version of minimalism shares with the HPSG framework (cf. Pollard and Sag 1994) to the extent that heads are familiar with their neighbors thus ‘naturally’ preventing
22
Scrambling and the Survive Principle
roneously speak of copies and chains in Survive; Link! illustrates the true interpretative nature of what remains behind. What resides in these structural positions previously occupied by the object W is the concatenate identity forged through Link! What is phonologically produced at the end of the derivation (cf. (1.27m)) is the highest copy, how this is interpreted throughout the derivation at LF (cf. (1.27n)) are all of the previous applications of Link! A clear advantage to integrating Link! into the Survive view of minimalism is the resultant ability of lower instantiations of Merge and Remerge being fully interpreted at the interfaces. (In current versions of minimalism that maintain either an Attract or Internal Merge/Move approach to constituent displacement, the problem of how lower copies of a movement chain are interpreted remains an unsettled issue – see Section 5.3.) The chief conceptual advantage to a derivational version of syntax with an active lexicon is the elimination of Copy Theory (cf. Chomsky 1995) and the baggage that accompanies this conceptual notion in the form of movement chains. The model of ‘copying’ licensed within Survive-minimalism is much simpler: In such a model, Numeration formation, Merge, Remerge and Link! are all versions of a single copy operation. No other operations are required in the computation of syntactic derivations – no deletion operations and no mechanisms to determine whether or not deletion should take place after a copying operation. Epstein and Seely (2006) postulate the removal of A-chains (and the EPP) from derivational syntactic theory, however according to the Survive Principle; we can forge a bolder claim yet and call for the elimination of all lower copies and chains indicating movement legacies. What remains as a ‘legacy’ or ‘copy’ reflecting the derivational history of a given constituent is the information shipped off to both the LF and PF interfaces via Link! immediately upon the concatenation of two syntactic objects with matching, identical features. The notion of any sort of movement chain is multirepresentational by nature and is purely structural: In other words, chains cannot and do not contribute anything interpretable to the interfaces.16 Chains, be they c-command, m-command or successively cyclic movement history chains, are a theoretical entity with the sole intention to define locality in the narrow syntax. As thoroughly clarified in Stroik’s (2007) work on locality in minimalist syntax, the Survive Principle returns the theory to a strict understanding of locality, namely, extremely local Spec-Head and Head-Comp relations (cf. Fitzpatrick 2002). As a result, all non-local configurations, such as Probe-Goal relations (Agree), are deemed in this system to be too far removed from one look-ahead properties. Although the Survive version of syntax faces the issue of sorting out which head merges first and why, the application of Agree in ‘standard’ minimalism also bears strong affinities to HPSG (the idea of unification found in this framework) and LFG (Bresnan 2001). See also Stroik (2007) for an in depth discussion of these aforementioned issues. 16 See Epstein and Seely (2006:Ch. 2) for a thorough discussion and analysis of how chains are not syntactic objects.
Chapter 1. Introduction
23
another to enact feature-valuation. Furthermore, Probe-Goal relations are predicated on the existence of some sort of look-ahead or look-back mechanism that enables the Goal to somehow stay alive in anticipation of eventual feature-valuation.17 Syntactic operations enacted by the primitives MergeSurvive-Remerge/Link! avoid this confusion and make every cyclic concatenation of syntactic objects in the derivation interpretable at the interfaces. Epstein and Seely (2002, 2006) come close to this sort of reasoning in their ‘level-free’ syntax hypothesis where every iterative application of Merge and Remerge is immediately shipped off to the interfaces – or in their own words, “each rule application is its own ‘self-contained’ Y-model (phasal) derivation” (Epstein and Seely 2006:6). However, their continued reliance on Internal Merge/Move does not fully align their ‘level-free’ syntax with the purely derivational account fleshed out through Survive. We can now revisit and critically revise the notion of Full Interpretation (cf. Chomsky 1993, 1995) to bring it more in line with true minimalist ideology. Recall the earlier description of Full Interpretation (cf. 1.13) repeated for the sake of the reader in (1.28) below. (1.28)
Full Interpretation (cf. Chomsky (1993, 1995))
PF:
Elements that have a uniform, language-independent interpretation in terms of universal phonetics.
LF:
Each object a chain CH (=α1,…,αn): at least with CH a head, an argument, a modifier or an operator-variable construction.
Viewed through the lense of Brody’s criticism of the current mixed state of minimalism and the suggested theoretical adjustments advanced by the Survive Principle, it is now clear Full Interpretation as currently stated in (1.28) above rests on assumptions and theory-internal considerations (i.e., chains) that do not contribute anything interpretable to either the LF or PF interface. Based solely on the grounds of virtual conceptual necessity, Full Interpretation must be considerably reduced to ensure its reliability in a minimalist, derivational theory. In line with the Survive Principle, Full Interpretation can be reduced to the ensurance that each lexical item/syntactic object bearing features that are active in the narrow syntax, must enter into enough concatenative relations (i.e., Link!) to guarantee complete convergence at the interfaces.
17 Another constraint that the Probe-Goal system in its current instantiation must adhere to, but is rarely ever mentioned, is something I label FIND-PHASE EDGE. The Probe is not only responsible for finding a suitable goal with a matching grammatical feature for valuation purpose, but it must also recognize the ontology of strong phases (vP and CP) and find items at the edge of these larger derivational units in order to respect the PIC (cf. Chomsky 2001).
24
(1.29)
Scrambling and the Survive Principle
Revised Full Interpretation
LF & PF:
All syntactic objects bearing syntactic features must enter into an equal number of Link! relations to match the number of features in their inventory.
In a similar fashion to the mainstream version of minimalism, failure to account for an uninterpretable feature in the course of a derivation will, in the words of Chomsky, cause the derivation to ‘converge at gibberish’. Critical to the revision of the lexicon from the status of a mere lexical array containing constituents to an active participant in derivations is the assumption that all features are interpretable i.e., uninterpretable variants do not exist. Stroik and Putnam (2005b) support this claim by noting the inability of neonates to acquire uninterpretable features. The failure to concatenate and subsequently activate each and every necessitated instance of feature-union (Link!) will cause syntactic objects the derivation to crash. Perhaps a better term to introduce at this juncture is the idea that the system does not ‘crash’ per se, but rather stalls. The derivation will not continue further once it has stalled; the only repair mechanism at this point is for the CHL to begin generating the utterance anew. Consider (1.30) and (1.31) below (from Putnam 2006a:118-9). (1.30)
*Scheint [TP Wilhelm das Lied wieder zu singen] seems Wilhelm the song again to sing
(1.31)
Wilhelmi scheint [TP PROi das Lied wieder zu singen] Wilhelm seems the song again to sing ‘Wilhelm seems to be singing that song again’
Infinitival clauses – more particular, the defective To – fail to assign Case to their subjects, in these examples Wilhelm. The raising verb scheinen ‘to seem’ necessitates A-movement of the subject of the infinitival clause Wilhelm in order to properly check the Case feature of this subject. Based on the Case Filter that states that all overt arguments must be Case marked in the syntax, the sentence (1.30) is ungrammatical due to the fact it fails to provide a way and means for the Case feature of the subject Wilhelm to properly be checked. According to the Survive Principle, the failure of the subject Wilhelm to remerge as a local sister to the raising verb scheinen ‘to seem’ will automatically cause the derivation to stall for the reason that an active element of the feature matrix – here Case – on both scheinen and Wilhelm will not participate in Link! offending both LF and PF.18 If the derivation were to continue from this point, 18
It is also conceivable to argue that the derivation will stall at an earlier derivational stage, namely, the failure of the subject to be repelled, i.e. survive, from the embedded TP.
Chapter 1. Introduction
25
the CHL would simply add unto the ungrammaticality of this issue at hand. The failure of the subject to participate in Remerge spells certain doom for this derivation, forcing another attempt by the CHL to correct its previous mistake. In this sense, the derivation and the computational system in general have become ‘crash-proof’ in the sense of Frampton and Gutmann (2002). In sum, with the ousting of traditional Copy Theory in favor of Link! the theory finally has the potential to achieve a fully derivational, ‘crash-proof’ status. 1.3
Survive-Minimalism
The ultimate achievement of the Survive Principle’s effect on and revision of the Minimalist Program in its current state is the massive reduction of theoryinternal constructs and constraints that contribute nothing to the system’s desire to receive Full Interpretation when it presents its final representation – after all subsequent derivational steps are successfully completed – to the interfaces. Ontological notions such as phases, Prolific Domains and even phrase structure rules are deemed to be a burden to the human language faculty. The chief aims of all of these bigger derivational chunks is (i) to prevent the operation known as Internal Merge/Move from participating in constituent distal that is ‘too far away’ and (ii) to prevent Internal Merge/Move from potentially over-generating itself. As pointed out repeatedly by Stroik (1999, 2000, 2007), the Survive Principle reduces our dependency on any sort of economy conditions placed on the CHL due to the fact that the human language faculty is now forced to evaluate all active members in its lexical base upon every application of iterative head-Merge. As pointed out by David Pesetsky (p.c.), the fact that an extensive evaluation process must take place at each occurrence of Merge in the narrow syntax might be perceived as a non-minimal, taxing addition to the theory. I partially agree with Pesetsky’s statement, however, regardless of how costly this ‘extensive evaluation’ system may appear at first blush, it comes with the distinct tactical advantage of eliminating all look-ahead and look-back mechanisms in the grammar. From this point on, all evaluations are extremely local, taking place at the exact moment of the merger of functional heads into the narrow syntax. At this point something can also be said about the notion of successive cyclicity as well and transformational locality conditions. A longheld and strongly supported assumption in generative syntax is the notion of an adherence of grammatical operations and mechanisms to a sense of locality. The common practice of using non-local constructs to account for non-local displacement phenomena such as the wh-movement in (1.32) leads, as noted by Stroik and Putnam (2005b), to the circular argument that we have displacement data because we have displacement operations and we have displacement operations because we have displacement data. The consequence of ‘explaining’ displacement data in terms of displacement (non-local) constructs is that the data are merely replicated and re-labeled
26
Scrambling and the Survive Principle
by the constructs, akin to the way that the that-trace Filter replicated the data it sought to ‘explain’. Furthermore, as I have demonstrated and explained in this brief introduction to the Survive Principle and its implications for the Minimalist Program, non-local accounts of displacement produce uneconomical derivations. For example, if a derivation begins by merging read with what to produce (1.32), the wh-element remains derivationally dormant until some nonlocal [+wh] head affects it. (1.32) [VP read what] Hence, we can’t determine whether this wh-element displaces (as in (1.33a)) or not (as in (1.33b)) until we reach the head of the matrix CP, which has a [+wh] feature. Delaying the displacement in (1.33a) until the derivation reaches the matrix CP, however, forces the derivation to return to the original Merger (in (1.32)) and re-compute the structure. Needless to say, the fact that derivational delay requires eventual derivational re-computation makes such non-local derivations uneconomical and costly from a computational perspective. (1.33) a. b.
What does Pat think that Chris should read? Who thinks that Chris should read what?
The subsequent removal of all non-local operations such as Move, Attract and Agree as well as economy constraints (e.g., Greed, Attract Closest, Phase Impenetrability Condition, etc.) can all be subsumed under the mechanical primitives Merge and Remerge.19 Furthermore, we are not burdening the CHL with the complex ancillary machinery necessary to constraint IM from overgenerating itself (cf. (1.25)). A minimalism composed of other rich ontological commitments (e.g., phases, Prolific Domains, etc.) is also no longer necessary. The notions of phases (vP and CP) and edge-movement are examples of theory-internal constructs set to ensure local operations resulting in harmonious convergence at the interfaces. Grohmann’s notion (following Platzack 1991) of the natural tripartite structure of clauses consisting of three smaller derivational units that he calls Prolific Domains (2000a, 2003 and later subsequent work) is, however, worthy of further consideration. Unlike phases, which are purely structural notions20, Prolific Domains are contextually identified as units responsible for 19 Survive and Link! are also primitive operations, however, Survive technically occurs within the lexicon and not in the narrow syntax. Second, Link! is the functional outcome of Remerge, so in application they are synonymous while remaining distinct in function. 20 Although Chomsky retains that phases are “structurally independent” units at the interfaces, scholars such as Epstein and Seely (2002), Legate (2002), and Boeckx (2003) (among others) point out key flaws in Chomsky’s ascertain. Although it is unclear exactly what Chomsky
Chapter 1. Introduction
27
thematic properties (vP), agreement relations (TP), and discourse-oriented phenomena (CP). As brought forth by Putnam and Stroik (in progress), although ontological commitments needn’t be a part of the narrow syntax anymore, there must exist some sort of spinal structure to which applications of Merge and Remerge can apply. At this point, as a primitive notion, let us maintain that universally three functional heads – namely, C > T > v – make up the backbone of clausal structure. Another appealing attribute of a possible connection between Survive-minimalism and Prolific Domains is Grohmann’s (2000a, 2003, 2006) take on the dynamic processing of these structural units at the interface: upon their completion Prolific Domains are immediately shipped off to the interfaces. In comparison, phases must remain active until the completion of the immediately dominating phases, so in essence not just one but two phases must technically remain active in the human language faculty.21 Although this work will explore the potential of Prolific Domains in more detail due to their contextual, dynamic and interface-interpretable nature, we must also deal with Putnam’s (2006a) criticism of the mapping of middle field phenomena – including middle field Scrambling – in West Germanic directly to natural clauses in a Prolific Domains-framework. According to Abraham (1997), Meinunger (2000) and Putnam (2006a), to name a few, constituents that participate in middle field Scrambling are best interpreted as referential topics, a notion that I also support in this book. This forces a re-evaluation of the middle field (i.e., the Φ-domain), which is believed by Grohmann to be rigorously committed solely to the licensing of agreement relations. We will test if – with minor adjustments – these shortcomings can be overcome and if the means by ‘relatively independent’ in this context, Epstein and Seely (2002:78) raise the following key question: “How can we know that they are relatively independent at the interface if Spell Out applies before the interface is reached, and without access to interface properties?” Epstein and Seely (2002:78) comments further that “…it is a potential architectural paradox to hypothesize that vP and CP are spelled out cyclically, internal to the narrow syntax by virtue of them having the property being later, relatively independent at the interface.” According to Epstein and Seely, another problem with adopting phases is coming to terms with Chomsky’s definition that phases are ‘propositional’ units. There are vPs that are non-propositional, such as who bought what and everyone bought something. As pointed out by Epstein and Seely (2002:78-79), prior to movement these vPs exhibit vacuous quantification; subsequent to movement they are open sentences internal to vP containing free variables. Legate (2002) observes that some of the criterion used by Chomsky to justify his characterization of vP and CP as phases, while excluding V and T, are insufficient. For example, she shows that raising and passive verbs pass three tests that define structural units as phases: 1.) They show instances of reconstruction of wh-phrases at their edge, 2.) They provide a target for QR, and 3.) They license parasitic gaps at their edge. Boeckx (2003) also notes another shortcoming in a phasebased system’s ability to accurately interpret locality in that blind intermediate links (spurious EPP-features) remain in such a system. 21 Putnam (2006a: Ch.3) mentions potential weaknesses that exist in Grohmann’s (2000a, 2003) current formulation of Prolific Domains that prevent them from being immediately shipped off to the external interfaces LF and PF upon their completion. My reformulation of the contextual identity of Prolific Domains in Chapter 4 eliminates this issue.
28
Scrambling and the Survive Principle
notion of contextually-defined larger derivational units known as Prolific Domains can be salvaged in Survive-minimalism. 1.4
Alternative Approaches to Scrambling
There exists a multitude of alternative approaches that attempt to explain the phenomenon of Scrambling within the minimalist framework of syntactic theory. For the sake of time and space, I have sorted competing approaches to a Survive-based explanation to middle field Scrambling into three camps. First, the null hypothesis that must be explored is whether or not any sort of syntactic displacement is involved at all in Scrambling; in other words, the possibility that scrambled items are base-generated in the syntax in their final surface position must be explored. Second, assuming that syntactic movement is involved in Scrambling – either substitution movement of an XP to the specifier of a higher functional head or displacement by XP-adjunction – we must investigate the strengths and weaknesses of claims that interpret middle field Scrambling as movement to Spec,AgrP and those that call on adjunction to Spec,vP. Third, the typological distinction between West Germanic languages that display an underlying word order of SOV (e.g., Dutch and German) are capable of licensing Scrambling due to properties of asymmetry in their clausal structure (cf. Kayne (1994)) whereas West Germanic SVO-languages (e.g., English) are structurally deficient and hence lack the ability to license Scrambling. Although each of these approaches has some merit, all of them are seriously flawed on both conceptual and empirical grounds and are incapable of explaining descriptive characteristics of Scrambling. A base-generated approach to Scrambling such as the one supported by Bošković and Takahashi (1998) requires covert object lowering at LF for θ-role assignment. My analysis developed here is straightforward and does not have recourse to theoretically suspect moves such as LF-lowering that are counter-cyclic in design. Second, movement by means of either substitution or adjunction of a scrambled XP to Spec,AgrP in the middle field fails to capture the lack of an isomorphic A/A’-reflex. Lastly, the assertion that Scrambling is somehow unique to SOV-languages in West Germanic (cf. Haider and Rosengren 2003 among others) is easily refuted by Diesing’s (1997) work on the semantic interpretation of scrambled indefinites in Yiddish. Yiddish, which is synchronically SVO underlylingly, illustrates that middle field Scrambling does occur in Yiddish. 1.4.1
The Null Hypothesis: A Base-Generated Approach
The obvious alternative to an analysis of middle field Scrambling as an operation involving the displacement of syntactic objects from their initially merged position is a base-generated approach. Under such an analysis, items that un-
Chapter 1. Introduction
29
dergo Scrambling are base-generated in the middle field and, as a result, must allow the scrambled object in the middle field to participate in lowering at LF where they will be able to receive and license a θ-role. Positing a syntactic operation (such as Bošković and Takahashi’s (1998) Affect α) that permits LFlowering is problematic for two reasons: First, it unnecessarily complicates the theory, and is to a considerable extent redundant with respect to other locality and economy constraints. The allowance of syntactic objects to move rightward via LF-lowering would force structure in the narrow syntax to be ‘alive’ much longer than is currently assumed in standard minimalist theory. According to the Phase Impenetrability Condition (Chomsky 2001), only constituents at the ‘edge’ (i.e., in a multiple specifier of vP) of a strong phase are accessible for further operations in the higher phase. Therefore LF-lowering of an object in the ‘middle field’ – most likely at the edge of a strong phase (vP) in a phasebased model of minimalism – should not be allowed access to syntactic structures lower than v. Second, allowing lowering operations such as these to exist substantially complicates the computational load placed upon the CHL. This is clearly an unwanted result in a theoretical framework that seeks to be crashproof and minimal in design. In their defense, Bošković and Takahashi (1998) put forward an interesting argument in support of LF-lowering operations based on current minimalist desiderata. Bošković and Takahashi point out that although lowering operations qua Affect α were prohibited in a preminimalist syntactic theory such as Government and Binding (GB), such a restriction on these operations was straightforward. According to the Projection Principle and the θ-Criterion, all θ-roles must be assigned at D-structure. Movement into θ-positions was thus prevented from taking place for a trivial reason: There are no empty θpositions when Move α applies. Movement into a filled θ-position, which would erase the element that previously occupied it, violates recoverability of the deleted object and Projection Principle proper (different elements bear the same θ-role at different levels of representation). With the elimination of internal interfaces (e.g., D-structure and S-structure) in the MP, such a restriction ruling out LF-lowering into θ-positions is no longer a tenable argument. With the disposal of these internal interfaces, Bošković and Takahashi (1998) make the argument that the θ-Criterion and the Projection Principle cannot be checked before LF. Unlike in GB-theory, within the MP all operations must be motivated (Chomsky 1993, 1995). What has lead scholars such as Saito (1989), Bošković and Takahashi (1998) and Bošković (2004), to support a lowering analysis at LF of Scrambling, is that in languages such as Japanese, the scrambled objects must obligatorily reconstruction at LF. This is often referred to as the ‘undoing’ property of Scrambling. As a result, Scrambling is viewed in Japanese as semantically vacuous. Bošković and Takahashi (1998) and Bošković (2004)
30
Scrambling and the Survive Principle
maintain that this ‘undoing’ property of Scrambling shows the way for Scrambling to count as a minimalist operation that adheres to Last Resort. One attractive aspect to this analysis is that in its design it avoids the messy details of the A- and A’-movement characteristics of Scrambling; by relying on lowering at LF rather that displacement within the narrow syntax, it is possible to avoid this distinction. From the standpoint of economy, Bošković and Takahashi (1998) side with Chomsky (1995) and Collins (1995), who argue that derivations should not be compared on the basis of global length or cost. According to them, at a particular point X the decision about whether a particular operation will apply should be made only on the basis of information available at X, which is, of course, a natural result of the MP operating on the Survive Principle. No look-ahead properties are permitted. As a result, a derivation can be ruled out “via the notion of economy only if at a particular point in the derivation an alternative operation is available that is less costly at that point” (Bošković and Takahashi (1998:352)). These analyses maintain that the scrambled constituent in (1.34) sono hono ‘that book’ is base generated in the matrix IP and later lowered to the position marked as t to receive Case and a theta-role. (1.34) [IP Sono hon-o [IP John-ga [[ Mary-ga [ t katta]] to] omotteiru]] that bookACC JohnNOM MaryNOM bought that thinks ‘That book, John thinks that Mary bought.’ In a minimalist syntax such as Survive-minimalism that allows only structurebuilding operations that lack both look-back and look-ahead properties, any syntactic lowering operations such as the one proposed above should, in principle, be impossible not only because they would require the CHL to have a mixed (derivational and representational) design—something that Brody (2002) argues against—but also because they would place an enormous burden on processing. To see this, consider sentence (1.35), in which a base generated constituent daremo-ni ‘everyone’ is scrambled, but there is no place to lower it. (1.35) *[IP Daremo-ni [IP dareka-ga [Mary-ga sonohon-o katta to] omotteiru]] everyone someoneNOM MaryNOM that bookACC bought thinks ‘Everyone, someone thinks that Mary bought that book.’ As pointed out by Stroik (2007:Chapter 4), the problem with (1.35) for the lowering analysis is that this analysis permits us to compute IP structures such as (1.35) and it won’t allow us to rule out (1.35) until we re-compute (look for) all the possible lowerings, hoping to find one that will Case license and theta license the scrambled constituent. Needless to say, all this fruitless computing
Chapter 1. Introduction
31
and re-computing of ill-formed structures is incompatible with O’Grady’s (2005:6) observation that the computational system of CHL should be efficient: it should “minimize the burden on the working memory.” Approaches to middle field Scrambling that assume an LF-lowering operation similar in design to (1.35) above further complicate and burden the working memory. One further reason that LF-lowering is at odds with Survive-minimalism is that it forces us to interpret θ-roles as only being interpreted at LF. If, as predicted by Link!, all syntactic features united by Merge and Remerge and interpreted at the external interfaces, a derivational legacy – a link of information – rests in the syntactic position responsible for licensing thematic interpretations. If this assumption is correct, LF-lowering is not a viable option to license Scrambling due to the lack of a potential landing site in the vP. In our discussion of anaphoric binding in Section 3.2.2, we will see that θ-features can and do have relevance at both LF and PF. Furthermore, middle field Scrambling is not a semantically vacuous movement in West Germanic. This is particularly noticeable in scrambled indefinite DPs (see Heim 1982, Diesing 1992, and Frey 1993). LF-lowering thus enacts operations that are not detectable and are implausible in Surviveminimalist framework.22 1.4.2
Move α: Movement to Spec,FP
By ruling out a base-generated approach to middle field Scrambling, the logical conclusion is, of course, that some sort of constituent displacement is responsible for this phenomenon. One possibility worthy of serious consideration is that Scrambling is driven by substitution movement from its base position in the vP to the specifier of a functional projection in the middle field. In earlier versions of minimalism, both Adger (1994) and Meinunger (1995) independently argue the notion of topichood (for Adger ‘familiarity’) is closely connected with Agr-phrases in the middle field. There are three problems in interpreting middle field Scrambling along these lines. First, if we are to adopt the standard assumption agreement features and discourse features (e.g., [Topic] and [Focus]) are separate unique features, it is not possible to have one functional head responsible for simultaneously checking two separate features. This is a very serious violation of our notion of syntactic concatenation (RemergeLink!) in Survive-minimalism: the creation of a syntactic object qua Link! is only possible between a unique functional head and a lexical item bearing the
22
Miyagawa (2006) takes issue with the conceptual and empirical claims of Bošković and Takahashi (1998) and Bošković (2004) for their claims that Scrambling in Japanese is the result of LF-lowering into θ-positions. Miyagawa provides empirical evidence that the ‘undoing’ does not occur and that the reconstruction effect simply reflects a normal property of A’movements similar to wh-movement in English. Miyagawa further shows that the condition that governs optional Scrambling is Fox’s (2000) Scope Economy.
32
Scrambling and the Survive Principle
same matching feature. Multiple valuation of features with a single application of Link! is strictly prohibited. Another factor which casts serious doubt on movement to Spec,AgrOP to check a Scrambling-feature as well as agreement features is the lack of an isomorphic A-/A’-reflex found in West Germanic middle field Scrambling. For example, middle field Scrambling in West Germanic languages can license parasitic gaps, which is clearly an A’-movement effect (1.36a is a Dutch example and 1.36b is a German example). (1.36) a.
b.
Jan heeft die boekeni zonder ei te bekijken weggelegd. John has the books without to inspect away-put ‘John put away the books without having looked at them.’ …weil er die Zeitungi ohne ei zu lesen weggeworfen hat. …because he the newspaper without to read away-thrown has ‘…because he threw away the newspaper without having read it.’
If middle field Scrambling is triggered by a formal feature, it is erroneous to assume that this feature is somehow connected with agreement features.23 With the eventual removal of Agr-phrases from the MP (ch. Chomsky 1995:Chapter 4), such a theoretical stance is quite antiquated and would require rationale for the re-admittance of Agr-phrases based on the grounds of virtual conceptual necessity. Meinunger (2000) and Jayaseelan (2001) advocate the presence of functional projections in the middle field responsible for licensing discourse properties such as topic and focus. In a version of the MP that makes use of Chomsky’s notion of phases (2000, 2001), one could argue that movement to the edge of a strong phase (i.e., vP or CP) enables the external interfaces to ‘see’ scrambled items and can thus interpret them (see also Parafita (2005) and Putnam (2006a: Chapter 4) who independently support this assumption). In this approach, a language family like West Germanic would license middle field Scrambling based on the necessity to check a [Top] feature, whereas in a language like Hungarian a [Foc] feature would be licensed. (1.37) Ich habe den Manni gestern ti gesehen. I have the man yesterday seen ‘I have seen the man yesterday.’
23 Following recent treatments of agreement features by Bobaljik (2005) and Sigurðsson (2007) it may indeed be the fact that agreement features are purely a grammatical phenomenon licensed in shallow PF.
Chapter 1. Introduction
33
(1.38) Péter valószínüleg [FP A SAROKHÁZAT veszii [VP meg ti]] Peter probably the.house.on.the.cornerACC buys pref ‘It is probably the house on the corner that Peter buys.’ In the German (1.37) and Hungarian (1.38) examples above a definite objects scrambles into the middle field. In German, the definite object den Mann ‘the man’ does not receive prosodic prominence, however, in Hungarian the object a sarokházat ‘the house on the corner’ is focused at the left periphery of VP. In contrast to German, middle field Scrambling in Hungarian is licensed by a [Foc] feature rather than a [Top] feature as illustrated in (1.39). (1.39) a.
b.
German:
[TP [TopP XPi [NegP [vP [VP XPi ]]]]]
Hungarian
[TP [FocP XPi [NegP [vP [VP XPi ]]]]]
Such an approach to middle field Scrambling in West Germanic is on the right track; however, there are a few noticeable shortcomings with this approach. First, a finer distinction of the notion of ‘topic’ is necessary to differentiate Scrambling above the subject into the left-periphery of the clause and Scrambling into the middle field. Following Sportiche (1992), I make the distinction between referential topics licensed by the [+ Ref] feature and discourse topics triggered by the presence of a strong [Top]. Second, it is unclear whether Scrambling is movement by substitution or an instance of XP-adjunction in this model. For example, Lopez (2002) and Newmeyer (2003) question the existence of formal grammatical features and projections for the licensing of semantic and discourse properties of sentences. These approaches admit that the cascade of functional projections at the edge of strong phases represent a singular unminimalist framework for capturing the parametric variation of languages and their left periphery effects. Notwithstanding these aforementioned weaknesses of the claim that middle field Scrambling in West Germanic is driven by some sort of a notion of topichood, I expand and improve upon this idea in this book. Sabel (2005:308) also maintains that Scrambling is a featuredriven operation and identifies the relevant feature simply as [Σ]. Sabel assumes that the Scrambling feature Σ is related to discourse-relevant notions of topic and focus, and thus he labels the (T-level) attracting feature as [T[Σ]/[topic]] and [T[Σ]/[focus]], respectively. The primary constraint on the application of Scrambling, as argued by Sabel, is that it is not string-vacuous, in keeping with a general principle against movements that do not affect the output in any way.24 Although similar in design to Jayaseelan’s (2001) account of Scrambling, Sabel’s approach creates a discourse-level of external interface (his ‘T24
See Chomsky (2001) for a similar proposal with regard to the [EPP] feature on v, which triggers object shift in some Germanic languages.
34
Scrambling and the Survive Principle
level’) that further complicates the CHL and the model of grammar. The approach put forward in this book will show that the only external interfaces necessary to interpret grammatical strings are LF and PF. My last (and somewhat expected) criticism of any system that employs Move/Internal Merge to license middle field Scrambling is the reliance upon movement chains that are not interpretable at the external interfaces. In its current form, it is unclear what a movement legacy would look like that generates both A-movement and A’-movement characteristics. In Survive-minimalism such problems are completely avoided: Every application of Remerge and Link! are instances of feature identity at the interfaces. In his most recent treatment of the properties of topicalization in the left periphery (i.e., CP-layer) of natural clauses, Rizzi (2004) remarks that topicalization appears to be a ‘special’ kind of A’-movement that in fact does permit binding in some instances. What makes middle field Scrambling seemingly incompatible with the traditional A-/A’-movement characteristic distinction is fact that these movement types are not compatible with discourse/semantic features. ‘Movement’ in Survive-minimalism is thus only concerned with feature interpretation. 1.4.3
XP-Adjunction
Another movement-based analysis of middle field Scrambling involves interpreting such displacement as adjunction to vP. In the earliest versions of transformational generative grammar, it was assumed that each grammatical transformation consisted of a structural description (SD) and a structural change (SC), where the SD specifies the domain of the transformation and its analysis into phrases subject to the SC, and the SC is an operation composed of elementary abstract operations, of substitution, adjunction, deletion, etc. (cf. Chomsky 1981:86). The displacement of constituents would then take place by means of either substitution or adjunction followed by the deletion of the moved element. The formalization of the Principles & Parameters (P&P) theory took place in the late 1970s/early 1980s within the Government & Binding (GB) framework brought forth a clear structural distinction between substitution and adjunction. Assuming that both substitution and adjunction are both licit movement rules in the GB-theoretical framework, according to Chomsky (1981:47), the latter always forms a structure introducing either one of the following form options: (1.40) a.
[β α β]
b.
[β β α]
In the representations above (1.40ab), α is adjoined to β by Move α. Dissimilar to phrase structure generated by means of X’-theory, adjunction is a structural-
Chapter 1. Introduction
35
preserving operation that does not involve the further projection of a syntactic head. In this sense, adjuncts cannot head a phrase. The lack of the function to head phrases (among other properties) forced the theoretical notion in GBtheory that adjoined XPs should not technically be included to or excluded from the maximal projections to which they adjoined to during the run of syntactic operations. In this sense, adjoined structures existed; however, their exact nature remained a bit of a mystery. With the advent of the Minimalist Program (cf. Chomsky 1993, 1995), constituent distal was no longer interpreted as operating along the prescribed rubric of Move α at the internal level of S-structure, rather, both the concept of internal interfaces (i.e., D- and S-structure) as well as Move α were eliminated from the theory on grounds of their incompatibility with the notion of virtual conceptual necessity. In the Minimalist Program, displacement operations are constrained by economy conditions such as Greed and Last Resort (cf. (1.14) and (1.15)). The generation of phrase structure is motivated by an iterative application of the extraction of items from the lexical array, the Merge/Adjoin operation, and, if necessary, the copy and eventual deletion of moved items. Within the Minimalist Program, adjunction is interpreted by Chomsky as quite a problematic construction. Consider the following adjunction operation (taken from Chomsky 1995:322). (1.41) L = {γ, {α, K}} Assume that L is formed by adjunction projecting K, so that γ =. However L was formed – be it by either substitution or adjunction – the non-deleted items in (1.40), i.e. L, α, and K, must be visible when presented to the external interfaces LF and PF to research some sort of interpretation to satisfy the then current version of Full Interpretation (cf. (1.13)). The problem with coming to a full understanding of adjunction in the Minimalist Program ca. 1995 is exacerbated when one considers LF roles. Returning again to the adjoined structure L in (1.41), if the adjoined items are non-minimal, we must come to terms with two syntactic objects, i.e. L and K, however only one role at LF.25 Operations such as extraposition, right-node raising, VPadjunction, and Scrambling are what Chomsky calls ‘whatever’ arrangements and the like “differ markedly in character from many other operations of the language faculty, and it may be a mistake to try to integrate them within the same framework of principles (i.e., minimalist inquiry)” (Chomsky 1995:325). A major stumbling block for Chomsky in earlier versions of minimalism regarding the problematic status of adjunction centered on the then perceived 25 Chomsky (1995:323) points out that there is a potential tenable solution to this issue in the case of full reconstruction at LF which eliminates the adjunct entirely. Interestingly enough, Saito (1989) adopts this very same analysis for Scrambling in Japanese.
36
Scrambling and the Survive Principle
inability to conform XP-adjunction to economy constraints (e.g., Last Resort) forcing movement to be driven by feature checking within the N → λ computation. Adjunction was therefore deemed to not highlight ‘minimalist guidelines’ and was once again determined to be driven by linguistic characteristics outside of the narrow syntax. The status of XP-adjunction – especially for movement purposes – is anything but clear in current orthodox minimalism. Putnam (2006a) interjects a new proposal in an attempt to integrate XP-adjunction as a licit movement operation in both the MP and in Universal Grammar (UG): Building on the proposals of Rochemont (1986), Culicover and Rochemont (1991) and Kidwai (2000)26, Putnam argues that the C-I system is the locus of both discursive/pragmatic and logico-semantic interpretation, and both types of interpretations must be licensed by derivations present in the syntactic component of grammar. To accomplish this, Putnam suggests parsing adjoined structures into LF- and PF-adjunction and attempts to show how XP-adjunction potentially conforms to economy and locality restrictions placed upon other movement types in the theory. The prime fault in Putnam’s (2006a) analysis of movement qua XP-adjunction is that it relies on an internal movement mechanism (i.e., Internal Adjoin) and forces a weakly representational interpretation of the theory. Minimalist theory as understood and revised within the Survive framework avoids many of the errors that beset earlier stages of transformational syntactic theory. Of prime importance is the inclusion of the lexicon as an active component of the CHL, which brings forth the ability of the theory to break its dependency from Internal Merge/Move and the concept of feature checking and deletion. Under such assumptions, the traditional distinctions between substitution and adjunction can be subsumed under the primitive operation Link! As a result, XP-adjunction for discourse/semantic properties would be required by the presence of a formal feature (although it remains to be determined if XPadjunction is at all necessary for middle field Scrambling).27 According to Survive-desiderata motivated by Link!, XP-adjunction driven by the +interpretable [+Ref] feature is not motivated by the necessity to ‘check’ a feature, but rather for feature diffusion. To achieve this concatenate relationship of adjunction, I advocate the fusion-like relationship of XPs exhibiting identical features within Rizzi’s (1991) dynamic agreement. The structures in (1.42) below illustrate the difference between Chomsky’s (1995, 2000, 2001) featurechecking and feature-valuation models and Rizzi’s (1991) dynamic agreement. 26 See Kidwai (2000) for an attempt to conform XP-adjunction to minimalist economy conditions placed upon the human language faculty by illustrating the deficiencies in the binding theory in its current formulation. 27 This assumption is anything but novel to the theoretical enterprise of derivational syntax (see Rizzi 1991, Kidwai 2000 and Rubin 2003 among others).
Chapter 1. Introduction
(1.42) a. Dynamic Agreement WP 3 ZPi XP [+INT] 3 Xo YP [+INT] 3 ti Yo
37
b. Feature-checking/Featurevaluation WP 3 ZPi XP [iF] 3 Xo YP 3 [uF] Yo ti
The newly formed syntactic object WP(ZP, XP) in both (1.42a) and (1.42b) display an Extended Sister relationship (in the sense of Grohmann’s (2003) Natural Relations) between ZP and Xo. According to either the featurechecking or feature-valuation model in (1.42b) above, why would an +interpretable feature need to participate in any further operations if it is deemed +interpretable from its conception? The significant difference between the two models lies in the fact that both +interpretable features remain at the end of the derivation to be read at the interfaces and both contribute to the concatenation of syntactic structure (see also Sauerland 1999). The result of this proposal is that the former notion of XP-adjunction can be eliminated from the theory as a potential mechanism in licensing discourse/semantic features that are ‘syntactic’, i.e. those that are interpreted at both the LF and PF interface.28 The prime conceptual advantage of labeling Scrambling as an instance of XPadjunction is often based on its mixed A-A’-movement distinction. Scrambling does exhibit certain A-movement traits such as anaphoric binding. (1.43) …dass Fritz die Gästei gestern einanderi t vorgestellt hat. …that FritzNOM the guestsACC yesterday each introduced has ‘…that Fritz introduced the guests to each other yesterday.’ If XP-adjunction is an A’-position, the only way around this dilemma is to tweak our understanding of adjunction for the purposes of Scrambling as a ‘special’ exception to our standard notion of adjunction (cf. Webelhuth 1989, 1992). Such an adaptation of adjunction is ad hoc and involves some sort of movement chain (although it is not clear how the lower copy of this chain is to be interpreted). As we shall see later in this book (Chapter 3), our current understanding of XP-adjunction in the MP proper will be challenged when we 28
Such a view of eliminating the distinction between specifiers and adjuncts finds its roots in the works of Hoekstra (2001) and Kayne (1994) among others. For Hoekstra (2001), a specifier is an adjunct, which enters into an agreement relationship with the head of the phrase it is adjoined to.
38
Scrambling and the Survive Principle
investigate Fox’s (2003, 2004) treatment of Condition C asymmetries. Based on the argumentation put forth in Stroik and Putnam (2007), I will supply further evidence along with the Scrambling data in this manuscript supporting a much needed re-analysis of XP-adjunction according to Survive-minimalism. Such a re-analysis is crucial in understanding how Survive-minimalism can explain certain epiphenomena found in the Scrambling patterns of West Germanic languages such as the licensing of parasitic gaps. In sum, neither movement by means of substitution to the specifier of a functional projection nor XP-adjunction proved to provide adequate explanation of middle field Scrambling in West Germanic. Both forms of constituent displacement fail on both conceptual and empirical grounds. Here lies the crux of the dilemma in understanding Scrambling as a syntactic operation: If middle field Scrambling is not base-generated in the middle field, yet does not undergo movement to a Spec,FP or a higher adjoined position, how can it possibly be ‘syntactic’ in design? The theory of ‘movement’ advanced in Survive-minimalism has a straightforward answer to these complicated issues that beset our current understanding of the MP. In the end, any discussion about adjunction in a labelfree syntax (and Survive-minimalism is a label-free syntactic system) is irrelevant. Given that the syntax has only two primitive operations at its disposal – Merge and Remerge – there is simply no reason to continue to make use of A/A’- and Spec/non-Spec-distinctions. If one considers the information contained within a lexical item that are communicated to the external interfaces qua Link!, we can see that structural notions such as adjunction are not interpretable, whereas features are. Therefore the above distinctions have no room in a syntax that begins in lexical items and ends with concatenated features. In such a system as this, it is simply not clear where and how adjunction could arise. In the following section I will continue the argument against the interpretation of purely structural considerations in the syntax in my treatment of the OV vs. VO head-parameter distinction in West Germanic and its (dis)connection with middle field Scrambling. 1.4.4
OV vs. VO: Head-Parameter as a Licenser of Scrambling
Another view in the literature interprets middle field Scrambling as an operation available only to languages that have an underlying deep structure of OV (i.e., head-final) (cf. Haider and Rosengren 2003, Richards 2005 among others) with regard to the ordering of V and its complement (Comp). As an example of this approach, Haider and Rosengren (2003:207) purport that “the linear aspect of the head-complement relation, that is, VO versus OV, is determined by the parametric direction of structural licensing and a universal restriction against merging to the right.” The formulations in (1.44) illustrate the structural difference between the two projection options:
Chapter 1. Introduction
39
(1.44) a.) [XP3 [XP2 [XP1 V ]]] b.) [Vi [XP3 [ei [XP2 [ei XP1 ]]]
head-final head-initial
According to Haider and Rosengren, these formulations (1.44) are meant to show that (1.44b) is much more complex than (1.44a). In (1.44a) the head and each of its intermediate projects are licensed to the left. Consequently, all XPs are with the identification domain of the head. This domain may be extended by additional projections adjoined to the projection. Haider and Rosenberg refer to the minimal, non-extended part of the V-projection as the “minimal project complex,” or MAC. Thus, the MAC is the V-projection that minimally dominates the three argument positions. The distinction between (1.44a) and (1.44b) is represented in the following diagrams labeled (1.45a) and (1.45b) respectively: (1.45) a.
VP = MAC (1.44a) 3 XP3 VP’ 3 XP2 V’ 3 Vo XP1
(1.45) b.
VP = MAC (1.44b) 3 V’ XP3 3 Vi VP 3 XP2 V’ 3 XP1 ei (Haider and Rosengren 2003:208-9)
Two of the three arguments in (1.45b), namely XP2 and XP1, are within the directionality domain of V, but one is to the left, in the Spec position of VP. Haider and Rosengren elucidate and simultaneously summarize this crucial difference between SOV and SVO base word orders in the following statement (from Haider and Rosengren 2003:209): In the head-initial structure, the lowest head position cannot canonically license more than a single complement, namely its complement XP1. Further merging in the directional domain of the head, that is, to the right, would require right adjunction and is excluded by the UG constraint against left-associative structures.
40
Scrambling and the Survive Principle
According to such a framework, Scrambling should not occur in SVOlanguages and could potentially boil down to a language needing to structurally conform to Kayne’s Linear Correspondence Axiom (1994) (i.e., all heads and specifiers existing in an asymmetrical c-commanding relationship to one another) prior to Spell-Out. This would (once again) regulate Scrambling to linearization principles at PF external to the narrow syntax. This treatment of Scrambling sits on shaky ground, for the only evidence that one would need to disprove it would be an SVO-language that exhibits Scrambling. Incidentally, one needn’t look outside of the family of West Germanic languages to find such a candidate in Yiddish. Yiddish – a language that was historically an SOV-language – is an example of a synchronic West Germanic language that is underlyingly SVO and licenses Scrambling. Diesing (1997) points out a simple fact in her research on Yiddish Scrambling patterns: in ditransitives, the expected OV order (both objects preceding the verb) is not possible. There is only one available open position immediately preceding the verb, and it is exclusively reserved for contrastively focused items. This is not at all what one should expect in a language with an underlying OV order (data from Diesing 1997:392). (1.46) a.
b.
(1.47) a.
b.
*Nekhtn hot Maks nit dem yingl dos bukh gegebn. yesterday has Max not the boy the book given Intended: ‘Max did not give the book to the boy yesterday.’ *Nekhtn hot Maks nit ken yingl ken bukh gegebn. yesterday has Max not no boy no book given Intended: ‘Max gave a boy a book yesterday.’ Nekhtn hot Maks dem Jingl dos bukh nit gegebn. yesterday has Max the boy the book not given ‘Max didn’t give the boy the book yesterday.’ Nekhtn hot Maks ken yingl ken bukh nit gegebn. yesterday has Maks no boy no book not given ‘Max didn’t give a boy a book yesterday.’
The contrast between (1.46) and (1.47) would not be at all expected if the base order of Yiddish was OV; the impossibility of placing both objects of a double object verb in the immediate preverbal position effectively rules out the hypothesis of the OV-order in the base. As pointed out to me by Molly Diesing (p.c.), the status of Yiddish as an SVO-language also explains why the follow-
Chapter 1. Introduction
41
ing strings are marked in Yiddish (data from Diesing 1997:391; contrastively focused): (1.48) a.
b.
?Maks hot nekhtn a bukh geleyent. Max has yesterday a book read ‘Max read a book yesterday.’ ?Maks hot nekhtn dos bukh geleynt. Max has yesterday the book read ‘Max red the book yesterday.’
If the underlying order of Yiddish was OV, we should not expect these forms to be marked, but rather they should represent the underlying order of objects. One would not expect that contrastive stress in the immediate preverbal position is required as the data in (1.48) suggests. Even the limited corpus of Yiddish data in this section makes a strong case that languages that are underlyingly SVO (such as Yiddish and Russian) can and do license middle field Scrambling. Yiddish remains consonant with other West Germanic languages in that referentiality is the triggering feature motivating this movement type. Agreeing in principle with Yang (1999), I interpret Merge to be a syntactic, concatenative operation that is uninvolved with the linearization of the items that are united together as one syntactic object through the iterative application of this operation. The primitive operation Merge is therefore not concerned with the linearization of the constituents united as one syntactic object under its power. Based on the Yiddish data above it is also unclear what Haider and Rosengren’s (2003) MAC is supposed to explain or achieve. Such a proposal is purely structural and forces the interfaces to somehow interpret syntactic structure. An easier, more straightforward approach to understanding middle field Scrambling would be one that makes use of grammatical features interpreted at LF and PF. Agreeing in principle with movement analyses (while simultaneously opposing their reliance upon Move/Internal Merge) that call upon discourse/semantic features to license Scrambling (Section 1.4.2), I make the claim in this book that middle field Scrambling in West Germanic is driven by the necessity to license [referentiality] in the narrow syntax to ensure derivational success. 1.5
Scrambling and the Survive Principle
As seen throughout this introduction, our derivational model of minimalist syntax – Survive-minimalism – takes a radically new look at the relationship between lexical items and syntactic derivations. Stroik (2007:39-40) summa-
42
Scrambling and the Survive Principle
rizes the purpose and advantage that a program of derivational syntax adhering to the Survive Principle possesses: By locating lexical items, and their features, already within the interfaces prior to participating in any derivations, we are forced to recognize that derivations are not, as Chomsky (2001, 2004) suggests, involved in checking the legibility of lexical features. Lexical features are inherently legible at the interfaces, so this legibility has no need to be checked. What syntactic operations derivationally check are not lexical features, but concatenation relations. That is, derivations are concerned with the legibility of larger-than-lexical-item syntactic objects (SOs), not the legibility of the atomic features. It is the SOs that are interpreted at the interfaces. Any operations that might obfuscate, or improperly delay, the concatenation relations of SOs will prevent the SOs from being interpreted legitimately; hence, feature deletion operations and copy deletion operations must be rejected because they disrupt the interpretation of SOs formed in a derivation, and feature segregation operations must be rejected because they build SOs that are not legible at the interfaces until they are re-built by covert syntactic operations.
The purpose of this manuscript is two-fold: First, in this book I seek to analyze middle field Scrambling in West Germanic languages within a minimalist framework of syntactic derivation that adheres to Stroik’s Survive Principle. In this endeavor we will discover if Survive-minimalism is a valid, reliable framework of syntactic analysis and to what extent this system presents us with superior analytic tools to better understand the responsibility of the syntax in generating and evaluating grammatical strings involving middle field Scrambling. Secondly, in this analysis of Scrambling within Survive-minimalism I hope to enhance and improve upon our knowledge of middle field Scrambling and its properties. Some of the key questions I hope to address in the duration of this work are as follows: (Q1) (Q2) (Q3) (Q4) (Q5)
Is middle field Scrambling in West Germanic the result of truly optional reordering operations? Is middle field Scrambling feature-driven, and is it A-movement, A’movement, both or neither? What is/are the relevant landing site(s)? Does middle field Scrambling (sometimes/always/never) have a logicosemantic effect? What are the consequences for typology and learnability; that is, does having the identified processes correlate with certain morphological properties of language?
These questions listed above represent some of the fundamental issues that have beset analyses of middle field Scrambling in the Principles & Parameters tradition. Although I will elaborate in much more detail on each of these questions and surrounding issues that accompany them, a simple derivation of my
Chapter 1. Introduction
43
analysis of middle field Scrambling championed in this work will demonstrate my line of thinking. Consider the following sentence. (1.49) Ich habe den Manni gestern ti gesehen. I have the manACC yesterday seen ‘I saw the man yesterday.’ In example (1.49) the direct object den Mann ‘the man’ has scrambled from its base position in the vP. The motivation for this displacement is the necessity to license the [+ Ref] feature in the middle field. The portion of the derivational history of (1.50) pertinent to the scrambling of the object den Mann ‘the man’ is provided below. (1.50) a.
Merge {den Mann, gesehen} → den Mann gesehen
b.
Survive {den Mann}
c.
Adjoin {gestern, {den Mann, gesehen}} → gestern den Mann gesehen
d.
Merge {REF, {gestern, {den Mann, gesehen}}} → REF gestern den Mann gesehen
e.
Remerge {den Mann, {REF, {gestern, {den Mann, gesehen}}}} → den Mann REF gestern den Mann gesehen
Middle field Scrambling is clearly not an optional operation in this system; this ‘movement’ is driven by the necessity to license [referentiality] in the middle field (ad Q1). This displacement type is thus feature-driven, but importantly, it is not A-movement or A’-movement. As a matter of fact, such a distinction can be henceforth eliminated: In Survive-minimalism the interpretation of syntactic objects and the interpretation of ‘lower copies’ rests on the properties of individual features (ad Q2). Regarding the relevant landing site in the middle field, the object den Mann ‘the man’ must enter into an immediately local union (Spec-Head) with the functional head bearing an identical [+ Ref] feature in the middle field (i.e., above vP) (ad Q3). The scrambling of the object den Mann ‘the man’ does have a discourse/semantic reading accompanying this displacement. The definite DP is interpreted as a thematic (discourse old) entity. Building upon the proposals of Rochemont (1986) and Culicover and Rochemont (1991), we illustrate here that pragmatic/semantic readings can be licensed within the narrow syntax (ad Q4). Regarding the learnability and typology of middle field Scrambling within this system, I will postpone this discus-
44
Scrambling and the Survive Principle
sion for later consideration (in Chapter 3) (ad Q5). As we can see, my analysis of middle field Scrambling within Survive-minimalism is straightforward and has the distinct advantage of being able to eliminate much of the unnecessary hardware and constraints that traditionally are found in analyses of Scrambling. Example (1.49) and its abbreviated derivational history (1.50) offers an obvious simplification of some of the more difficult issues that we will encounter in this book. Nonetheless, the framework provided above represents the core analysis supported here and will be elaborated on further in the coming chapters. 1.6
Scope and Content of this Book
In this book I will make the case for the existence of the discourse/semantic feature [+ Ref] in the narrow syntax responsible for the licensing of middle field Scrambling in West Germanic languages. I will demonstrate how a version of the MP that adheres to Stroik’s Survive Principle improves upon our descriptive and explanatory power to understand the responsibility of the narrow syntax in generating and evaluating completed representations. As my analysis will show, the syntax does not need to call upon XP-adjunction, the archaic A/A’-movement dichotomy, the head-parameter order of languages (e.g., OV/VO), or other extraneous structural constraints that play no role in interface interpretation. The distinct advantage to my approach is that it reunites middle field Scrambling with other empirical data argued to traditionally represent displacement phenomena. The organization of the remainder of the book is as follows. In Chapter 2, I lay out the myriad of grammatical properties involved in middle field Scrambling as far as West Germanic languages are concerned. Following Putnam (2006a), I provide an in-depth analysis of Scrambling according to traditional A/A’-movement characteristics to illustrate the complex nature of this movement type within this archaic classification system. I then investigate the involvement of established non-syntactic features such as pragmatic, prosodic and scopal features. In the end, I make a case for the reality and necessity of the referential feature [+ Ref] as the central motivating feature for middle field Scrambling in syntactic computation. In Chapter 3, I unite my theory of middle field Scrambling driven by a [+ Ref] with Survive-minimalism. At the beginning of this chapter I discuss the conceptual reality of recognizing [referentiality] as a syntactic feature based on empirical data from typological diverse languages (including Germanic). I provide both empirical and conceptual reasons to eliminate the view of treating middle field Scrambling as a form of XP-adjunction (contra Webelhuth 1989, 1992; Kidwai 2000, Haider 2002 and Putnam 2006a). This chapter also addresses some of the potentially problematic issues that arise from such an analysis and attempts to provide clear explanations when possible. As another in-
Chapter 1. Introduction
45
stance of argumentation in favor of my analysis, I provide a novel solution based on the placement of [+ Ref] feature in Pennsylvania German and German clauses to explain how the former language has *lost* its ability to license middle field Scrambling. Chapter 4 concerns itself with the phonological aspects of Scrambling. Here I once again defend my analysis of middle field Scrambling driven by referentiality [+ Ref] against purely prosodic accounts. Based on Putnam (2006a) and follow-up work by Grohmann and Putnam (in press), I investigate the syntax-phonology interface with respect to dynamic stress assignment. By upholding my referentiality analysis, I am able to keep the narrow syntax pure from phonological features (i.e., AntiFocus [AF]). Lastly, Chapter 5 caps off this work with a reformulation of the properties of the middle field, directions and suggestions for future research, and a novel sketch of how my analysis of middle field Scrambling within Survive-minimalism may be adapted and applied to typologically distinct languages and language families that also license Scrambling.
CHAPTER 2 PROPERTIES OF SCRAMBLING As mentioned in the introductory chapter, Scrambling embodies a complex set of grammatical properties. In the extensive literature on Scrambling from a syntactic perspective, scholars have attempted to classify this phenomenon as either movement for agreement (A-), e.g., Case, phi-features, etc., or nonagreement (A’-) purposes, e.g., wh-movement, topicalization. The purpose of this chapter is to offer a representative sample of the characteristics that define Scrambling in West Germanic. Creating a panoramic profile of Scrambling exposes potential triggers for this movement type. At the same time, data also function to eliminate other grammatical relations and features from playing a significant role in determining licit scrambling configurations. In this chapter we shall seek exactly what makes Scrambling in West Germanic tick. Here I will support the hypothesis that the discourse/semantic notion of referentiality is the most viable candidate to license middle field Scrambling in West Germanic. Rather than approaching this issue from the traditional A-/A’-characteristic distinctions exhibited by Scrambling in West Germanic, this book analyzes Scrambling according to the following categories: •
Scrambling as NP-movement
•
Scrambling as wh-movement
•
Scrambling as Topicalization
If Scrambling is purely an instance of NP-movement, we should expect universal A-movement characteristics across the board for constituents that scramble. Likewise, if Scrambling is either wh-movement or topicalization, A’movement distinctions should abound. The rationale behind further parsing the traditional A’-categories of wh-movement and topicalization is motivated on the following grounds: First, wh-items represent new, thematic pragmatic information (i.e., Focus) at the level of discourse, whereas items marked as topics function as previously established, rhematic material. Second, as will be discussed in this chapter, Scrambling in West Germanic is clausal-bound, whereas wh-movement can be long-distance, i.e., spanning across multiple clauses. From the standpoint of a ‘traditional’ syntactic analysis of Scrambling, it is highly desirable that Scrambling would resemble all (or at least the majority) of one of these three categories established above. Although Scrambling does
Chapter 2. Properties of Scrambling
47
display certain characteristics that resemble NP-movement, wh-movement and topicalization, the data in this chapter suggest Scrambling does not fully align itself with any of these three camps. This, of course, forces us to look at other grammatical features that have been regarded to exist outside the realm of traditional syntax, e.g., referentiality, scope (pure LF-considerations), pragmatic/discourse configurationality and prosody. Maintaining a modular view of language composition by the faculty of human language, these conventionally non-syntactic attributes of Scrambling are commonly interpreted as phenomena that are licensed at the interfaces, i.e., outside of the syntax. Within these categories we see the greatest degree of consistency with Scrambling in West Germanic. From the standpoint of discourse configurationality, scrambled XPs in the West Germanic middle field are regarded as presuppositions. As shown by Frey (1993) and Lechner (1997), scrambled XPs also create scope ambiguities. It is important to bear in mind that I am only concerned with uncovering the grammatical relations that regulate Scrambling at this stage. The discussion as to how and if Scrambling can be licensed within the framework of syntax will be taken up in Chapter 3. Another aspect of Scrambling that must also be addressed is the structural and selectional requirements that this movement types seems to obey. As noted by Ross (1967) and Wexler & Culicover (1980) among others, it was originally observed that Scrambling could not take place from a unit that has already moved from its base position (e.g., Freezing effects). As far as the selectional properties of verbs are concerned, Webelhuth (1992) provides evidence substantiating the claim that only verbs that can c-select a complementizer phrase (CP) or a determiner phrase (DP) license Scrambling. Psych verbs such as ablehnen ‘to refuse’ and sich freuen ‘to be happy’ also prohibit Scrambling. 2.1
Strong vs. Weak Scrambling
Assuming the base order of DP arguments in the predicate (vP) adhere to the order Subject >> Indirect Object (IO) >> Direct Object (DO), we can observe a contrast between the permitted middle field order of arguments in Dutch and German. Unlike German, in Dutch the relative order of DP arguments must not be permuted in the middle field. This holds for subject-object order as well as for the order of the indirect object before the direct object (data from Haider 2005a:247). (2.1)
[CP dat (er) [TP iemandi krakersk [vP ti tk oppakt.]]] that (there) someone squatters arrests ‘that someone arrests squatters’
48
Scrambling and the Survive Principle
(2.2)
*[CP dat (er) [TP krakersk iemandi [vP ti tk oppakt.]]] that (there) squatters someone arrests ‘that someone arrests squatters’
(2.3)
[CP dat [TP iki de jongenj het/een boekk [vP ti tj tk gegeven heb]]] that I the boys the/a book given have ‘that I gave the boys the/a book.’
(2.4)
*[CP dat [TP iki het/een boekk de jongen tj [vP ti tj tk gegeven heb]]] that I the/a book the boys given have ‘that I gave the boys a book.’
In this regard, in Dutch (and Afrikaans) the base order of DP arguments in the vP must be maintained in the middle field with few exceptions.1 The scrambled word order of arguments in Dutch also functions as a prime example of the notion of cyclic derivation in the syntax. c d (2.5)
[CP … [TP α
β
e γ
c d [vP α
β
e
γ]]]
The diagram in (2.5) above symbolizes the cyclic derivation exhibited by Dutch middle field Scrambling. The ‘optimal’ linearization of arguments in the middle in German also obeys this principle; however, unlike Dutch, German allows Crossing and Nesting Paths (cf. Pesetsky 1982). (2.6) Nesting Paths
Crossing Paths
1
Zwart (1993:132) points out that for many speakers there is one context in which the order of arguments in a double object construction is not entirely rigid. With ditransitive verbs containing a particle, the objects can occur in both orders according to these speakers. (i)
…dat Jan Marie het boek terug gegeven heeft that JanNOM MarieACC the bookACC back given has ‘…that Jan gave Mary the book back.’
(ii)
…dat Jan het boek Marie terug gegeven heeft.
In addition, DO >> IO orders also seem to improve by stressing (via prosody) the IO.
Chapter 2. Properties of Scrambling
49
With both Nesting and Crossing Paths, the starting position of one chain ccommands the other. Middle field Scrambling in German is iterative. Examples (2.7) and (2.8) below demonstrate legal Nesting and Crossing Paths. (2.7)
[CP dass [IP das Buch2 dem Mann1 [vP die Frau
t1 t2 gegeben hat]]]
that the bookACC the manDAT the womanNOM given has ‘…that the woman gave the book to the man.’
(2.8)
[CP dass [IP dem Mann1 das Buch2 [vP die Frau
t1 t2 gegeben hat]]]
that the manDAT the bookACC the womanNOM given has ‘…that the woman gave the book to the man.’ In the German examples (2.9) – (2.16) below I show that the argument order in the German middle field needn’t strictly adhere to the base-generated order in the vP. (2.9)
[CP dass [TP jemandi die schlafenden Pennerk [vP gestern ti tk aufgeweckt hat]]] that someone the sleeping bums yesterday up-woke has ‘that someone woke the sleeping bums yesterday’
(2.10) [CP dass [TP die schlafenden Pennerk jemandi [vP gestern ti tk aufgeweckt hat]]] that the sleeping bums someone yesterday up-woke has ‘that someone woke the sleeping bums yesterday’ (2.11) [CP dass [TP jemandi den Pennernk ein paar Kippenj [vP ti tk tj gegeben hat]]] that someoneNOM the bumsDAT a few cigarettesACC given has ‘that someone gave a few cigarettes to the bums’ (2.12) [CP dass [TP jemandi ein paar Kippenj den Pennernk [vP ti tk tj gegeben hat]]] (2.13) [CP dass [TP den Pennernk jemandi ein paar Kippenj [vP ti tk tj gegeben hat]]]
50
Scrambling and the Survive Principle
(2.14) [CP dass [TP den Pennernk ein paar Kippenj jemandi [vP ti tk tj gegeben hat]]] (2.15) [CP dass [TP ein paar Kippenj jemandi den Pennernk [vP ti tk tj gegeben hat]]] (2.16) [CP dass [TP ein paar Kippenj den Pennernk jemandi [vP ti tk tj gegeben hat]]] Concerning the examples involving the ditransitive predicate, the example that maintains the constituent order found in the vP (2.11) is the unmarked, default order in the middle field (TP-layer). As mentioned in the introduction to this book, although the other permuted orders are acceptable, they are slightly marked. The crossing of Scrambling patterns below the subject is possible as indicated by (2.12); however, the construction is marked and only possible if the lower indirect object receives prosodic prominence. Scrambling both objects above the subject is possible; however, both constructions are highly marked. Scrambling above the subject as attested in (2.13) and (2.16) is acceptable with the highest object receiving secondary prosodic stress. These data bring two important facets of middle field Scrambling in West Germanic to the forefront. First, the term ‘Scrambling’ in West Germanic actually embodies two separate movement types – one that allows object order alterations in the middle field (Strong Scrambling) and those that must obligatorily follow the base-generated order of arguments in the vP (Weak Scrambling). The latter is commonly realized as the apparently ‘optional’ ordering of adverbs and direct objects: DO/ADV – ADV/DO. Although this may appear at first sight to be structurally optional, discourse considerations demonstrate the function of objects remaining in the vP and those that scramble into the middle field are pragmatically distinct from one another. German is the only continental West Germanic language that exhibits both strong and weak versions of middle field Scrambling, the other West Germanic languages (e.g., Afrikaans, Dutch, Frisian) only license Weak Scrambling. Weak Scrambling is quite similar to Object Shift in Scandinavian in that the order of arguments in the vP must be maintained in the middle field. A clear distinction needs to be drawn between Weak Scrambling in West Germanic on the one hand and Object Shift in Scandinavian. Whereas Object Shift displays ubiquitous A-movement characteristics, Weak Scrambling in West Germanic languages such as Dutch maintains the pragmatic distinctions briefly mentioned above as well as the mixed A/A’-movement characteristics also attributed to Strong Scrambling. I will continue to refer to examples of Scrambling in West Germanic throughout this work; however, the dominant focus of this work will concern itself with Strong Scrambling and permuted argument order in the West Germanic middle field
Chapter 2. Properties of Scrambling
51
and the problems that this latter type of Scrambling presents for syntactic theory. Secondly, thus far it is apparent from the data presented that non-syntactic linguistic phenomena (e.g., pragmatics and prosody) also have a say in determining the grammaticality of scrambled items in the middle field. This, of course, brings us back to one of the central issues discussed in this book: Can Scrambling be licensed purely by means of syntactic constraints and structure? Assuming a derivational approach under which syntax operates as a modular component of human grammar, it is difficult to envision a ‘purely syntactic’ account of Scrambling at this stage. The interaction of non-syntactic elements such as discourse configurationality and prosody has lead to explanations of Scrambling from a representational standpoint (most notably, Optimality Theory) focusing solely on well-formedness constraints imposed upon completed output structures (cf. Büring 2001a, 2001b, Choi 1999 among others)). On the contrary, others have maintained a derivational account to Scrambling and have accounted for these non-syntactic facets of Scrambling by means of introducing formal grammatical features to exist in the syntax and/or suggesting that prosodic structure assignment is accounted for prior to Spell-Out (Abraham and Molnárfi 2002, 2003; Feng 1995, 2003; Zubizaretta 1998 among others). 2.2
Selectional Properties of Verbs
Restrictions placed on Scrambling patterns in West Germanic are not only structural, but also appear as a result of the argument structure selected by a particular verb. Selectional properties of verbs play a decisive role in determining whether a given verb possesses the argument structure necessary to license Scrambling. Consider the following examples from Webelhuth (1992:103-4). (2.17) Ich glaube [das] I believe that ‘I believe that.’ (2.18) Ich glaube [dass Martin krank ist] I believe that Martin sick is ‘I believe that Martin is sick.’ (2.19) [Dass Martin krank ist]i glaube ich ti. that Martin sick is, believe I ‘I believe that Martin is sick.’ Only those verbs that c-select DP in addition to CP allow their complement clause to topicalize as illustrated in examples (2.19), (2.22) and (2.25).
52
Scrambling and the Survive Principle
(2.20) Ich weiss [das] I know that ‘I know that.’ (2.21) Ich weiss [dass Christoph den Spitznamen “Kartoffel” hat] I know that Christoph the nickname “potato” has ‘I know that Christoph has the nickname “potato”.’ (2.22) [Dass Christoph den Spitznamen “Kartoffel” hat]i weiss ich ti. that Christoph the nickname “potato” has know I ‘I know that Christoph has the nickname “potato”.’ (2.23) Ich verstehe [das] I understand that ‘I understand that.’ (2.24) Ich verstehe [dass Leo heute arbeiten muss] I understand that Leo today work must ‘I understand that Leo has to work today.’ (2.25) [Dass Leo heute arbeiten muss] verstehe ich. that Leo today work must understand I ‘I understand that Leo has to work today.’ For verbs that allow their complements to be realized by a sentence (CP) rather than a DP, Scrambling is not permitted. Psych verbs such as sich freuen ‘be happy,’ sich ärgern ‘be angry’ belong to this class. (2.26) *Ich freue mich [das] I am-happy Refl that ‘I am happy that.’ (2.27) Ich freue mich [dass Leo ein neues Mikrophon gekauft hat] I am-happy Refl that Leo a new microphone bought has ‘I am happy that Leo bought a new microphone.’ ich mich ti. (2.28) *[Dass Leo ein neues Mikrophon gekauft hat]i freue that Leo a new microphone bought has am-happy I Refl ‘I am happy that Leo bought a new microphone.’ None of the complement sentences can be topicalized because they cannot be linked to a DP-trace (or lower copy of the transposed DP), since the verb does not c-select a thematic DP. (2.29) Ich kann nicht glauben [dass Martin krank ist]
Chapter 2. Properties of Scrambling
53
I can not believe that Martin sick is ‘I can’t believe that Martin is sick.’ (2.30) Ich kann mich nicht freuen [dass Martin krank ist] I can Refl not be-happy that Martin sick is ‘I can’t be happy that Martin is sick.’ Middle field Scrambling of these CP complements is also disallowed by these psych-verbs. (2.31) weil ich [CP dass Martin krank ist]i nicht glauben kann ti. because I that Martin sick is not believe can ‘I cannot believe that Martin is sick.’ (2.32) *weil ich [CP dass Martin krank ist]i mich nicht freuen kann ti. because I that Martin sick is Refl not be-happy can ‘I cannot be happy that Martin is sick.’ This suggests that the triggering function licensing Scrambling above the subject (i.e., topicalization) targets argument structures rather than entire clausal units, or at the very least phrasal units licensed by V. As previously stated, this implies that verbs that can take either a DP or CP as their complements permit Scrambling of these structures into the middle field as well as above the subject. These data insinuate that Scrambling targets arguments in the VP; however, the fact that not only DPs but also but also clausal units (CP) are able to be scrambled presents serious difficulties for any account of Scrambling that depends upon the presence of strong agreement features on Do. Moreover, as will be illustrated in the following section, any formal grammatical feature/function triggering Scrambling (if such a feature exists) may indeed fall outside the scope of traditional morphosyntactic features. 2.3
Scrambling is not NP-movement
If Scrambling were to be classified as NP-movement, we would expect it to display movement types that reflect traditional A-movement characteristics. Although Scrambling does embody some of these distinctions, such as anaphoric binding and the absence of crossover effects, it appears that Scrambling effects show a much stronger tendency to be defined under A’-movement reflexes, i.e., reconstruction effects and the ability to license parasitic gaps. Further evidence against the notion that Scrambling is NP-movement is the fact that robust morphological Case is not compulsory for a language that licenses Scrambling. If Scrambling is not driven by the necessity to license Case and
54
Scrambling and the Survive Principle
does not universally display A-movement characteristics, it stands to reason that Scrambling is not synonymous with NP-movement. 2.3.1
Weak Crossover Effects (WCO)
WCOs occur when an item is displaced to the left of an anaphor with which it is co-indexed. Wh-items (A’-movement) are the quintessential example of a displacement type exhibiting WCO effects in English, as demonstrated below in (2.33). (2.33) * Whoi does hisi mother love ti? As far as WCOs are concerned, Scrambling in German does not give rise to (clear) effects of this phenomenon. In this regard, Scrambling in German is classified as showing A-movement properties. The generalization that changes in the word order of arguments do not lead to WCO effects is observed by many authors (cf. e.g., Webelhuth 1989, Frank, Lee and Rambow 1996). (2.34) Weni hat ti seine Mutteri immer ti geküßt? whomACC has his motherNOM always kissed “Whom does his mother always kiss?” Unlike the English example in (2.33), the German sentence in (2.34) does not produce WCO effects. The wh-item wen ‘whom’ moves to the left of the DP with which it is co-referenced without producing a WCO effect. 2.3.2
Anaphoric Binding
Moved constituents that have the ability to enter into new binding relationships with anaphors undergo A-movement. Therefore, if Scrambling in West Germanic is for agreement purposes, we should expect the potential of new binding relations existing between scrambled DPs and anaphors. This is where the trouble starts. Perhaps one of the most problematic issues connected with middle field Scrambling in German is the fact dative objects that scramble out of the vP cannot enter into new anaphoric binding relationships, whereas accusative objects can (cf. Grewendorf 1984, 1988; Müller 2001 among others) (data taken from Putnam 2005:287). (2.35) …dass Fritz die Gästei gestern einanderi t vorgestellt hat. …that FritzNOM the guestsACC yesterday each introduced has ‘…that Fritz introduced the guests to each other yesterday.’
Chapter 2. Properties of Scrambling
55
(2.36) *…dass Fritz den Gästeni gestern einanderi t vorgestellt hat. …that FritzNOM the guestsDAT yesterday each other introduced has ‘…that Fritz introduced one another to the guests yesterday.’ der Mariam sichh/*m t auf dem Photo zeigte (2.37) …weil der Hansh …since the JohnNOM the MaryDAT self in the picture showed ‘…since John showed Mary himself/*herself in the picture.’ Example (2.35) demonstrates that the accusative object die Gäste ‘the guests’ can and does create a new binding relation with the anaphor einander ‘each other.’ Assuming that the predicate adverbial adjunct gestern ‘yesterday’ is adjoined to vP, it is evident that some form of displacement outside of the vP has occurred. Conversely, the dative objects in (2.36) and (2.37) above evidence that the indirect objects den Gästen ‘the guests’ and der Maria ‘Mary’ cannot exist in a binding relation with the reflexive pronoun sich ‘self’. The exact Case of sich ‘self’ in German is difficult to determine, provided that both the dative and accusative reflexives share the same morpheme with no distinctions. These results are unexpected and puzzling, especially given the fact that scrambled datives can exist in a co-referential binding relationship with full DPs. (2.38) …dass der Jörg jedem [seinen Vater] gestern [t gezeigt hat]. …that the JörgNOM everyoneDAT his fatherACC yesterday shown has ‘…that Jörg has shown his(x) father to everyone(x) yesterday. Although scrambling the dative indirect object over the direct object can create a well-formed operator variable in German (2.38), it cannot create the anaphor reciprocal binding relation in (2.36) above. Based on the data above, it appears that middle field Scrambling in German displays both A- and A’characteristics with regard to its ability to create new anaphoric binding relations. 2.3.3
Reconstruction Effects
Reconstruction for binding effects is regarded as a property of items that have undergone A’-movement. Once again we obtain a complex picture of the behavior of Scrambling with respect to reconstruction. Certain cases of Scrambling give rise to reconstruction effects (example (2.39) is from Frank, Lee and Rambow (1996:77) and (2.40) from Haeberli (2002:119)).
56
Scrambling and the Survive Principle
jeder ti besucht hat. (2.39) …dass seinen Vateri …that his fatherACC everyoneNOM visited has ‘..that everyone has visited his father.’ (2.40) …dass sichi Hans ti einen Brief geschrieben hat. …that himselfDAT HansNOM a letterACC written has ‘…that John wrote himself a letter.’ In both examples provided above, it appears that scrambled objects must undergo reconstruction, which would favor an analysis of reconstruction in West Germanic along the lines of an A’-movement property. In (2.39), the scrambled direct object is bound by the quantifier subject jeder ‘everyone.’ In similar fashion, in (2.40) the scrambled reflexive indirect object has to undergo reconstruction to be properly bound by the subject Hans. The picture becomes more complex when we consider Scrambling of direct objects over indirect objects. As a matter of fact, we get counter-evidence from the effects that we saw in (2.39) and (2.40) (data taken from Moltmann (1991:123)). (2.41) * …dass der Jörg seinen Vateri jedem ti gezeigt hat. fatherACC everyoneDAT shown has … that the JörgNOM his ‘…that Jörg showed his father to everyone.’ (2.42) * Ich habe die Bilder von einanderi den Männern ti gezeigt. I have the picturesACC of each otherDAT the menDAT shown ‘ I have shown the pictures of the men to each other.’ Basically, the examples in (2.41) and (2.42) show us that a change in the order of the objects disrupts the binding relationships. In (2.41), the possessive pronoun heading the direct object cannot be bound by the quantificational indirect object. The anaphor embedded within the direct object in (2.42) faces the same fate as the possessive pronoun in (2.42). These examples suggest that these scrambled objects cannot be reconstructed, therefore Scrambling may not involve A’-movement with regards to reconstruction effects. At this juncture we are forced to draw a similar conclusion regarding Scrambling and reconstruction effects as we did with Scrambling and its ability (or lack thereof) to create new anaphoric binding relations: Neither can be classified as clearly displaying either A- or A’-movement characteristics.2
2
Under the assumptions laid out in Survive-minimalism, all ‘reconstructions’ involve noncovert concatenations. In other words, nothing is actually ‘reconstructed’. The reconstructions that emerge in (2.39) and (2.40) exist because the dislocated, i.e. remerged, DPs had previous instances of Merge that are interpretable at LF.
Chapter 2. Properties of Scrambling
2.3.4
57
Parasitic Gaps
Parasitic gaps, i.e. the licensing of an empty category by the presence of an A’trace, is designated as an A’-movement phenomenon. This property of objects in Dutch and German is illustrated in (2.43) and (2.44) with parasitic gaps represented by the empty position e.3 (2.43) Jan heeft die boekeni zonder ei te bekijken weggelegd. John has the books without to inspect away-put ‘John put away the books without having looked at them.’ ohne ei zu lesen weggeworfen hat. (2.44) …weil er die Zeitungi …because he the newspaper without to read away-thrown has ‘…because he threw away the newspaper without having read it.’ Although parasitic gaps are canonically regarded as A’-movement phenomena, Webelhuth (1989) and Neeleman (1994:403) argue that the licenser of a parasitic gap in Dutch can also bind an anaphor, which is a property of Amovement, as illustrated in (2.45) and (2.46) below (Data from Bennis and Hoekstra (1985) and Parker (1999)).4 (2.45)
...dat Jan de rivaleni namens elkaar zonder ei aan te kijken feliciteert that Jan the rivals on behalf of each other without at to look congratulates ‘…that John congratulates the rivals on behalf of each other without looking.’
(2.46)
Ik ben mijn oomi [na ti jaren niet gezien te hebben] gisteren weer tegen gekomen. I am my uncle after year not seen have yesterday again against come ‘Yesterday I met my uncle again after year of not having seen (him).’
The main argument put forth by Neeleman and Webelhuth above delivers the false prediction that licensing conditions for parasitic gaps are identical for English, German, and Dutch. Dutch appears to manifest a peculiar type of con3
These examples are taken from Bennis and Hoekstra (1984:68) and Haeberli (2002:110) respectively. 4 Neeleman (1994) observes two additional differences between Dutch and English with respect to the licensing of parasitic gaps. First, the subject of a passive sentence can license a parasitic gap in Dutch. Second, parasitic gaps can even be licensed in nominalized constituents.
58
Scrambling and the Survive Principle
struction that appears to contain parasitic gaps where no licensing gaps are present. If the parasitic gap in (2.46) refers to mijn oom ‘my uncle’, it would have to be bound by an element in an A-position. Disagreeing with Neeleman and Webelhuth, I posit that classifying parasitic gaps as an A-movement phenomenon is problematic. First, it is an attempt to create undesirable language-specific rules. Why does it appear crosslinguistically that parasitic gaps exhibit A’-movement characteristics for every language expect German and Dutch? Second, and most importantly, are the Dutch examples above authentic parasitic gaps? Based on the research of Parker (1999) and Kidwai (2000) – which will be presented and explained in full in Chapter 3 – I show how parasitic gaps in West Germanic are universally A’constructions. In this very brief overview of the A/A’-characteristics demonstrated by West Germanic Scrambling, the only consistency present is inconsistency. These data confirm Webelhuth’s Paradox that Scrambling universally does not exhibit neither A- nor A’-characteristics across the board. 2.3.5 Case Requirements Do Not License Scrambling
If Scrambling were to be classified as A-movement, we would expect that Case requirements would force XPs into the middle field. The conflicting nature of Scrambling with regards to A/A’-movement characteristics as illustrated in the previous chapter makes it extremely difficult, if not impossible, to assign this displacement phenomenon as an operation driven by agreement features. The cross-linguistic argument stemming from research connecting robust Case morphology and word order freedom is that languages with extensive Case-marking tend to also have greater word order freedom than languages without. One of the most transparent shortcomings of the hypothesis that morphological Case drives Scrambling is in the errant assumption that morphological Case richness can be reduced to the simple presence or absence of a marker on a given form. For example, consider the following data from German. (2.47)
…dass Fritz Maria ein Buch gegeben hat ...that FritzNOM MariaDAT aACC book given has ‘…that Fritz gave Maria a book.’
(2.48)
…dass Fritz ein Buchi Maria ti gegeben hat ...that FritzNOM aACC book MariaDAT given has ‘…that Fritz gave Maria the book.’
In example (2.48), the direct object ein Buch ‘a book’ is able to scramble across the indirect object Maria even though no overt morphological Case
Chapter 2. Properties of Scrambling
59
markings are present on either constituent. The data provided in the remainder of this section supports the theory that non-morphological notions (e.g., referential, possibly prosodic features, as well as other structural notions) motivate Scrambling rather than robust morphological Case. This sample of languages and the degree to which they allow Scrambling in connection with their morphological Case inventory, although not exclusive, is taken to be representative for the West Germanic language family on the whole. 2.3.5.1 Swiss German Standard German expresses four morphological Case distinctions, illustrated in (2.49). (2.49) Standard German Case System Singular
Plural
der Mann
die Männer
des Mann-es
der Männer
Dative
dem Mann
den Männer-n
Accusative
den Mann
die Männer
Nominative Genitive
As has been established earlier in this work, German allows the Scrambling of one object across another. Barring minor exceptions, most modern continental German dialects allow Scrambling patterns similar to the standard language (Russ et al. 1989). Grisons Swiss German, a dialect exhibiting the same fourCase distinction as Standard German, however, does not permit objects to scramble over one another, although the richness of morphological Case in this dialect rivals that of Standard German. The Swiss German data in (2.50) (from Kiparsky 1997:14) shows that scrambling the direct object over the indirect object is prohibited. (2.50) a.
b.
…und den het dr dogdor S. e m Bbuur de KB erklärt. and then has the doctor S.NOM the farmerDAT the AIACC explained ‘…and then Dr. S. explained the artificial insemination to the farmer.’ *…und den het der dogdor S. de KB1 em Bbuur t1 erklärt.
60
Scrambling and the Survive Principle
The fact that only a weak correlation, if any, at all exists between word order freedom and morphological Case is borne out in the Swiss German data. 2.3.5.2 Pennsylvania German Particular variants of Pennsylvania German, a West Germanic language originating in Southeastern Pennsylvania in the 18th century by means of the amalgamation of various Alemannic, Swabian and Palatinate dialectal variants, continue to exhibit a three-Case distinction in its nominal paradigm.5 Non-plain Pennsylvania German (NPG) maintains this three-Case distinction, whereas Plain Pennsylvania German (PPG), i.e., the vernacular of the Amish, has undergone paradigmatic leveling in the last three generations and now display nominal inflections similar to American English (i.e., a two-Case distinction marking nominative and objective Case). Since the 1870s, many scholars and native speakers with a passion for Pennsylvania German have composed formal grammars and primers of the language. Unfortunately, none of the more detailed accounts of the language (Frey 1945, Buffington and Barba 1954, Haag 1982) give mention to ‘scrambling’ or any sort of word-order permutations when dealing with possible word order variations. NPG, with an active three-Case distinction, does not license Scrambling as it is available in German (focused items indicated by capital letters), rather strict adherence to the base ordering of arguments (e.g., SUBJ >> IO >> DO) is required. (2.51) a.
b.
5
Ich hab der Bu der Balle gewwe. I have the boy the ball given ‘I gave the boy the ball.’ * Ich hab der Balle der Bu gewwe. I have the ball the boy given Intended: ‘I gave the boy the ball.’
Regional differences, which are still evident are now overshadowed by a single sociolinguistic distinction, that of membership in conservative Anabaptist sects. Those speakers of Pennsylvania German who are Anabaptist sectarians are thus often referred to as ‘plain’ in the literature; while those who are not, ‘non-plain’ (Louden 1989, Huffines 1987, Keiser 1999). This social dichotomy is giving rise to the evolution of two distinct variants of Pennsylvania German, namely Plain Pennsylvania German (PPG) and Non-plain Pennsylvania German (NPG). Only recently have studies focused on differences existing between these variants of Pennsylvania German.
Chapter 2. Properties of Scrambling
61
c.
Ich hab en der Balle gewwe. I have him the ball given ‘I gave him the ball.’
d.
Ich hab der Balle gewwe zu der BU (un net zu's Meedel). I have the ball given to the boy (and not to the girl) ‘I gave the ball to the BOY.’
e.
* Ich hab der BU der Balle gewwe. I have the boy the ball given Intended: ‘I gave the ball to the BOY.’
f.
Ich hab der Bu en gewwe. I have the boy him given I gave it to the boy.
g.
Ich hab en gewwe zu der BU. I have him given to the boy I gave it to the BOY.
h.
Ich hab en en gewwe. I have it him given I gave it to him.
(Mark Louden, p.c.)
Note that if the direct object were another gender, for example, neuter as with ‘s Buch “the book,” the order of the object would be the same: IO >> DO. (2.52) Ich hab en’s gewwe. I have him it given I have him it given Thus far, the underlying word order for NPG appears to be quite rigid, similar to English. As expected from a non-Scrambling language, no ‘German-style’ Scrambling is permitted, i.e., no orders with DO >> IO, either with both pronouns and DO pronoun >> IO full noun.6 (2.53) * Ich hab der Balle en gewwe. I have the ball him given Intended: ‚I gave the ball to him.’
6
Such constructions are not only ungrammatical in English and Pennsylvania German, but also in a number of other Germanic languages (e.g., Icelandic). This is due to a connection between definiteness/animacy and scrambling/object shift.
62
Scrambling and the Survive Principle
Provided that Pennsylvania German finds its linguistic home in the eastern Palatinate region of Germany (the Vorderpfalz to be exact), the next obvious question would be whether Scrambling is permitted in the current regiolects from which Pennsylvania hails. Contemporary residents of the region and speakers of the dialect (Michel Werner, Jörg Meindl, p.c.) report that Scrambling is still permitted. So when did Pennsylvania German lose the ability to license Scrambling? Was the option ever available to Pennsylvania German? After all, in the developmental stages of Pennsylvania German, it is very well possible that syntactic leveling took place to some extent. Once again we encounter a dilemma that has attracted little attention in the literature. Not surprising, the data indicate that this absence of Scrambling predates the loss of the dative. Below are sample from literary Pennsylvania German works from 19th century authors John Oberholtzer (1809-1895) and Edward H. Rauch (1820-1902):7 (2.54) Oberholtzer: …dann hot [mir ein Mann es] ksawt, der mich noch nie net beloge hot then had [meDAT a manNOM itACC] said, whoRP meACC yet not lied has “…then a man told it to me, who has never lied to me.” In (2.54) the IO and DO pronouns are not scrambled; however, the full DP subject follows the IO; i.e., the IO has scrambled above the subject DP. This is no longer possible in modern Pennsylvania German. (2.55) Rauch: a.
Ich will aa nau mentione, as wenn du ebber weescht, as gerne Grumbeere ferkawfe will fer a Faddel Dollar die Buschel, as ich reddi bin, [‘m sie] abzukawfe un ‘m Note gewwe fer’s Cash I want to also mention now, that if you know someone who wants to sell potatoes for a quarter-dollar a bushel, that I’m ready to buy them from him and give him my note for the cash
b.
Gell, Pit, du waerscht nau net bees, wann ich [der’s] sawg… Right, Pit, you wouldn’t be angry now, if I told it to you…
7
I am deeply indebted to Mark Louden (p.c.) for bringing these data to my attention.
Chapter 2. Properties of Scrambling
63
In both (2.54) and (2.55) above the linear order from pronouns is IO >> DO in both instances, the same as in modern PG. It appears then that syntactic changes/alterations have preceded the loss of the dative case in West Germanic languages. We will discuss the effects that such a claim would have on the loss of Scrambling in the conclusion of this work. Furthermore, Pennsylvania German appears to be typologically consistent in its development. This provides additional evidence why recent scholarship (Keel 1994, Fuller 1997, Putnam 2002, 2003) interprets the Case morphology attrition that has taken place in the last three generations in PPG to be independent of external influences (i.e., American English). 2.3.5.3 Amana German Another German-American dialect community that contributes interesting data to this discussion is the vernacular spoken in and around Iowa County, Iowa known as Amana German. In 1843, 600 German Inspirationists located in Central Hesse under the leadership of Christian Metz established a pietist settlement near Buffalo, New York. In 1854 these pietists settled in villages (originally six but later expanded to seven) along a track of the Iowa River approximately 20 miles west of Iowa City. The linguistic diversity within the individual villages themselves accurately reflected the ‘melting pot’ dialect known as Amana German. Prior to the Change of 1932, each village operated independently, with its own Council of Elders exercising final authority in temporal and spiritual affairs, hence independent village identity and settlement history helped shape and maintain this intra-dialectal variations of Kolonie-Deutsch, e.g., Amana German. The Change of 1932 ushered in many sociolinguistic changes to Amana German. Perhaps the most striking were the ever-increasing usage of English and the gradual erosion of the most significant intra-dialectal differences among the individual villages. The emergence of a conglomerate ‘Amana German’ by those who continued to speak Kolonie-Deutsch in the Colonies can be seen as a quasi-defense mechanism to unite this unique German heritage shared by those who still spoke German in the villages. Amana German has also participated in morphological Case attrition (Rettig 1976), however unlike Plain Pennsylvania German (PPG), the former dialect retained the dative and ousted the accusative rather than the pattern exhibited by Pennsylvania German. Regardless, native speakers of Amana German today exhibit a two-Case distinction (i.e., nominative and non-nominative (dative)) with regards to definite articles while maintaining a three-Case distinction in the pronominal system. Even with a weaker correlation between the-
64
Scrambling and the Survive Principle
matic roles and overt Case markings, Scrambling in Amana German is still a viable construction (Lawrence Rettig, p.c.). (2.56) a.
Ich geb dr Rosie di Blume. INOM give the RosieDAT the flowerACC ‘I give the flower to Rosie.’
b.
Ich geb sie dr Rosie. INOM give themACC the RosieDAT ‘I give them to Rosie.’
c.
Ich geb re di Blume. INOM give herDAT the flowerACC ‘I give her the flower.’
d.
Ich geb s re.8 INOM give itACC herDAT ‘I give it to her.’
e.
Di Blume geb ich dr Rosie. the flowerACC give INOM the RosieDAT ‘The flower, I give to Rosie.’
f.
Dr Rosie geb ich di Blume. the RosieDAT give INOM the flowerACC ‘To Rosie, I give the flower.’
Amana German – a German-American speech enclave that has undergone Case syncretism in a similar fashion to that found in the history of English – maintains ‘German-style’ Scrambling, whereas Non-plain Pennsylvania German with its 3-Case distinction with regards to definite articles has espoused the construction altogether. Scrambling above the subject into the CP-layer is also possible as well as split-NPs in Amana German whereas both are ungrammatical in Pennsylvania German (Lawrence Rettig, p.c.). (2.57) a.
Dr Max gibt dr Maria des Buch. the MaxNOM gives the MariaDAT the bookACC
b.
Dr Maria gibt dr Max des Buch. the MariaDAT gives the MaxNOM the bookACC
8
Amana German
This order is permitted only with the s [Standard German 'es'] pronoun (Lawrence Rettig, p.c.).
Chapter 2. Properties of Scrambling
c.
Des Buch gibt dr Max dr Maria. the bookACC gives the MaxNOM the MariaDAT ‘Max gives the book to Maria.’
d.
* Der Maria gibt der Max des Buch. Pennsylvania the MariaDAT gives the MaxNOM the bookACC German
e.
* Des Buch gibt der Max der Maria. the bookACC gives the MaxNOM the MariaDAT ‘Max gives the book to Maria.’
65
This distinction between Amana German and Pennsylvania German, two German-American heritage dialects with similar morphological Case inventories, is not due to the status of robustness of their respective morphological Case inventories. If this were the case, we should expect similar, if not identical behavior of these two dialects with regards to the degree to which they license Scrambling both below and above the subject. This is not evidenced in the data, thus strengthening the claim that Scrambling is not influenced by Case requirements. 2.3.5.4 Dutch Similar to Amana German, Dutch displays a two-Case distinction in its nominal inflection. However, unlike Amana German, the legal word order possibilities for scrambled DPs below the subject are much more restricted. Thus, as (2.58) shows (based on data from Neeleman 1994:416), arguments generally have to occur in the order Subj >> IO >> DO and Dutch therefore only the Scrambling of single objects over adverbials. (2.58) a.
...dat Jan de mannen deze film toont. that JanNOM de mannenACC the filmACC shows ‘…that Jan shows the film to the men.’
b.
*…dat Jan deze film de mannen toont.
c.
* ...dat deze film Jan de mannen toont.
d.
* ...dat deze film de mannen Jan toont.
e.
* ...dat de mannen Jan deze film toont.
f.
* ...dat de mannen deze film Jan toont.
66
Scrambling and the Survive Principle
These distinctions stand in stark contrast to German, which allows all possible linear orders in (2.58) above. The examples from other synchronic West Germanic languages support the hypothesis, however, that the overt presence of robust morphological Case distinctions is not an obligatory requirement for whether a given language licenses Scrambling or not. In addition, prosodic stress on the indirect object in the DO >> IO word order in Dutch improves the grammaticality of the statement. Two facts can be gleaned from this: First, something other than Case requirements are driving Scrambling in Dutch. Second, prosody plays a role in repairing/improving the grammaticality of DO >> IO word orders in Dutch. The application of the recovery capabilities of prosodic stress can be expanded to include all West Germanic languages. 2.3.5.5 Afrikaans To offer a death knell to the alleged rigid requirement that only if rich morphological Case is present then Scrambling is a possible licit construction, I would like to draw upon data from Afrikaans. Unlike NPG, Afrikaans exhibits a depleted inflectional inventory, and in many instances no distinctions can be recognized (e.g., the inflectional paradigm for 3rd person plural pronouns is completely absence in modern Afrkiaans). Regardless of the absence of morphological Case in many instances, Afrikaans licenses Scrambling similar to Dutch. Notice that if the object DP remains vP in situ, it receives prosodic stress (indicated by boldface type). Definite arguments that do not scramble outside of the vP in modern spoken Afrikaans are also marked with the morphological marker vir ‘for’. Molnárfi and Abraham (2002, 2003) purport that the presence of this shibboleth morphology indicates a rhematic interpretation for the object phrase that it heads (data from Molnárfi 2003:1153): (2.59) …dat ek die meisie [VP gister gesoen het] …that I the girl yesterday kissed have gesoen het] (2.60) ....dat ek haar [VP gister …that I her yesterday kissed have ‘…that I kissed the girl yesterday.’ (2.61) Ek het [VP gister (vir) die MEISIE gesoen] I have yesterday (for) the girl kissed (2.62) Ek het [VP gister *(vir) HAAR gesoen] I have yesterday (for) her kissed ‘I kissed the girl yesterday.’ The adverb gistern ‘yesterday’ is adjoined to vP. This clearly illustrates that the object in both (2.59) and (2.60) have moved outside of their base-generated
Chapter 2. Properties of Scrambling
67
positions. It is still unclear at this point what kind of a feature, if any, would motivate this movement. The non-subcategorized adverb gistern ‘yesterday’ in the Afrikaans examples provide solid evidence that Scrambling is not driven by agreement features. The placement of non-subcategorized adverbials in modern NPG is predominantly extraposed, as is seen in (2.61) and (2.62), which makes the application of the litmus test for movement out of the vP utilized for the Afrikaans data is opaque in Pennsylvania German. Notice again, however, the prohibition on the Scrambling of pronouns in the inner field. (2.63) …as ich's Meedel [VP gebusst hab geschder] …that I the girl kissed have yesterday (2.64) ...as ich sie [VP gebusst hab geschder] …that I her kissed have yesterday ‘…that I kissed her yesterday.’ (2.65) Ich hab's Meedel [VP gebusst geschder] I have the girl kissed yesterday (2.66) Ich hab sie [VP gebusst geschder] I have her kissed yesterday ‘I kissed her yesterday.’ Concerning the position of adverbs, Mark Louden (p.c.) contends that PG historically patterned itself after Standard German in this regard. Whether or not the shift of non-subcategorized adverbials in modern NPG from adjunction to vP (as in modern German) to the final position in the VP (similar to modern English) can be attributed to language contact with American English is a complex assertion to defend or contend against. If external influence is the principle impetus for the change, it is not totally off base to assume that NPG assimilated over time not to the morphological Case inflections of American English, but rather to the general discourse and prosodic structure of American English. Once the ‘German-style’ discourse archetype had been removed, the need for multiple morphological Case distinctions became superfluous, and eventually eroded over time (e.g., as in PPG). If this hypothesis is correct, it may indeed be the case (pun intended) that paradigmatic leveling of Case distinctions in Pennsylvania German – as well as other German-American dialects – is influenced by the linguistic encroachment of American English. It is noteworthy to mention that the birth and following transition of Afrikaans from a creole to a pidgin entails that the morphological inflections were never as strong as they were in other West Germanic languages (cf. Ponelis 1993). The fact that Afrikaans still exhibits discourse and prosodic requirements at the phrasal level similar to other West Germanic languages is the reason that
68
Scrambling and the Survive Principle
Scrambling, even to a limited extent, is still possible, whereas it is no longer legal in Pennsylvania German. In sum, any account of Scrambling as Amovement (i.e., motivated by either Case or any other agreement features) cannot be upheld. 2.4
Scrambling is not wh-movement
If Scrambling does not exhibit A-movement characteristics, the next question to ask concerns A’-movement traits. Although the A-movement phenomena from the previous section, e.g., anaphoric binding of scrambled accusatives and WCOs, provide evidence that middle field Scrambling is not purely A’movement either, Webelhuth (1992) elucidates the close relationship between wh-items and scrambled XPs. Based on the work of Saito (1989, 1992), Webelhuth (1992) and Frank, Lee, and Rambow (1996) viewing movement types in the narrow syntax of only consisting of the A-/A’-distinction are too restrictive in scope and do not reveal all possible movement options. Displacement can also be classified as being either movement to enter into an operatorvariable relationship (operator) or movement for other purposes (nonoperator). Given this added categorical distinction within A’-movement types, we have the opportunity to see if Scrambling aligns itself more closely with operator (wh-movement) or non-operator (topicalization) A’-movement. The chart below (2.67) (from Webelhuth 1992:180) surveys two kinds of movement chains: Operator-Chains (O-Chains) and Agreement-Chains (AChains). (2.67) Properties of Operator and Agreement Chains Property
O-Chains
A-Chains
Moves DP
+
+
Moves PP
+
-
Movers lack Case
+
-
Licenses parasitic gaps
+
-
Strands prepositions
+
-
As a means of comparison I will juxtapose Scrambling in German as representing the archetype O-Chain and Scandinavian (in the examples below Swedish is the representative language) Object Shift corresponding to A-Chain movement (the comparative data comes from Webelhuth 1992:180-188).
Chapter 2. Properties of Scrambling
2.4.1
69
Scrambling and Object Shift Move DP
(2.68) weil Hans [DP das Buch]i wahrscheinlich ti gelesen hat. because HansNOM the bookACC probably read has ‘because Hans probably has read the book.’ (2.69) Han köptei [DP den]j inte ti tj he bought it not ‘He didn’t buy it.’ As predicted, both the German sentence involving Scrambling (2.68) and the Scandinavian example with Object Shift (2.69) allow DP-movement. This is not surprising due to the fact that both Operator and Agreement chains can be headed by DPs. 2.4.2
Scrambling Moves PP
(2.70) weil er [PP mit ihr]i nicht ti tanzen wollte. because he with her not dance wanted ‘Because he didn’t want to dance with her.’ (2.71) *Jag tror [PP på det]i inte ti. I believe in it not ‘I don’t believe in it.’ Here we see a grammatical difference between the scrambled example (2.70) and the one displaying Object Shift (2.71): Whereas the former operation can license the movement of either a DP or a PP, the latter only can allow the movement of DPs into the middle field. 2.4.3
Scrambled XPs lack Case
To further support the data and assumption that Scrambling does not involve Case requirements, the DP den Jungen ‘the boy’ can remain in situ, i.e., does not have to move, or can scramble outside of the verb phrase without creating an ungrammatical utterance. This shows that Case, however it may be assigned in the system, does not trigger Scrambling. (2.72) weil [DP den Jungen]i [vP niemand ti gesehen hat] because the boyACC no one seen has ‘because no one saw the boy.’ (2.73) weil [vP niemand [DP den Jungen] gesehen hat] because no one the boyACC seen has ‘because no one saw the boy.’
70
Scrambling and the Survive Principle
Although subtle dialectal differences exist with respect to Object Shift in the Scandinavian languages, one element that remains constant among all Scandinavian languages is that all DPs that undergo Object Shift must land within the c-commanding domain of their Case-assigning verb (cf. Holmberg’s Generalization (1986)). The following sentences show that Object Shift can only apply in main clauses that also allow the Case-assigning verb to raise. (2.74) Han [C’ köptei [DP den]j inte ti tj]] he bought it not ‘He didn’t buy it.’ (2.75) *att han [DP den]i inte [köpte ti] that he it not bought ‘that he didn’t buy it.’ The presence of the complementizer att ‘that’ prevents movement of the verb köpte ‘bought’ to a higher position (Co) than the shifted pronoun den ‘it’. This confirms Holmberg’s Generalization (1986) that DPs that undergo Object Shift cannot move outside of the c-commanding domain of their Caseassigning verb.9 2.4.4
Scrambling Licenses Parasitic Gaps
As noted in the previous section discussing Scrambling’s relationship with NPmovement, scrambled elements in German can license parasitic gaps. As expected, items that have undergone Object Shift cannot, given their status as AChain movement. (2.76) ? weil er [DP den Artikel]i [ohne PRO vorher e zu lesen] ti ablegte because heNOM the articleACC without before to read filed ‘because he filed the article without first having read (it).’ (2.77) *Jag kastade [DP den]i inte ti [innan jag hade läst e] INOM threw itACC not before I had read ‘I didn’t through it away before I had read (it)’10 Once again in example (2.77) we see that Object Shift behaves like DPmovement, whereas Scrambling (2.76) exhibits O-Chain characteristics. 9
Although Icelandic permits Object Shift in subordinate clauses, it still must adhere to Holmberg’s Generalization. The reason this is permissible in Icelandic is due to the fact that the underlying word order of Icelandic is SVO rather than SOV. In other words, the Case-assigning verb that raises does not move all the way to the CP-layer. 10 This sentence is from Holmberg (1986:173).
Chapter 2. Properties of Scrambling
2.4.5
71
Scrambling Can Strand Prepositions
As a final test to confirm that Scrambling bears strong semblance to Operator Chains rather than Agreement Chains, data show that Scrambling can strand a preposition whereas Object Shift cannot.11 (2.78) weil dai niemand [PP ti gegen] war because that nobody against was ‘because nobody was against that.’ (2.79) *Jag tror deti inte [PP på ti]12 I believe it not in ‘I don’t believe in it.’ The following data prove Webelhuth’s hypothesis that Scrambling in German correlates with operator-driven A’-movement. On the other hand, Object Shift patterns itself according to typological movement distinctions akin to Agreement Chains. This captures a key difference between Scrambling and Object Shift and shows how the former shares many characteristics with Operator Chains. 2.4.6
Scrambling in West Germanic is Clause-bound
Unlike wh-movement in West Germanic that can cross-clausal boundaries, Scrambling patterns in German are strictly clause-bound. Consider the data below. kennt. (2.80) dass den Leoi jeder ti That the LeoACC everyoneNOM knows ‘that everyone knows Leo.’ kennt]. (2.81) *weil den Leoi ich glaube [dass jeder ti Because the LeoACC INOM believe that everyoneNOM knows ‘because the Leo I believe that everyone knows.’ In sentence (2.80), the scrambled object Leo targets an intra-clausal landing site below the complementizer dass ‘that’. In the second example it is evidenced that extraction of the object Leo across the clausal boundary produces an ungrammatical result. This characteristic; namely, that Scrambling is strictly clausal-bound, is unique to West Germanic and is not shared by other languages licensing this movement type (such as Japanese). On the other hand, 11 In DP-movement, preposition stranding is possible under very special (often lexically governed) circumstances (cf. Hornstein and Weinberg 1981). 12 From Holmberg (1986:165).
72
Scrambling and the Survive Principle
wh-movement phenomena in German allow an unbounded wh-phrase to move far from its base position but restrict Scrambling to within the clause (data from Kidwai 2000:34-35). (2.82) *dass niemand [VP Puddingi sagt [CP ti dass sie ti mag]] that no one pudding says that she likes ‘That nobody says that she likes pudding.’ (2.83) [CP Wasi sagt niemand [CP ti dass die ti mag]] what says nobody that she likes ‘What does nobody say that she likes?’ Interestingly, languages such as Russian and Hindi-Urdu exhibit the exact opposite constraints than German with regards to wh-movement and Scrambling: In contrast to German, in both Russian and Hindi-Urdu wh-movement is clausal bound whereas Scrambling may undergo long-distance movement. Agreeing with Kidwai (2000), whatever grammatical feature triggers Scrambling in German (as opposed to Russian and Hindi-Urdu) cannot be substitution into Spec,CP, because German does not use this position as an ‘escape hatch’. This is a decisive difference between Scrambling and wh-movement in German. As evidenced from the data above, Scrambling and wh-movement do not target the same landing sites. As an example of a language that licenses long-distance Scrambling, consider Japanese. German and Japanese are similar in the fact that they allow more than one DP to scramble from its base position; however, German differs significantly from Japanese in two respects: First, while Japanese scrambled DPs exhibit A-scrambling effects, German DPs that have undergone Scrambling show a mixture of both traditional A- and A’-movement properties. Second, as noted above, German Scrambling is always clausal bound, whereas Japanese allows long-distance Scrambling (illustrated in (2.84) and (2.85) below (taken from Grewendorf and Sabel 1999:11). (2.84) Japanese: [TP sono hon-o1 [TP John-ga [VP Bill-ni [CP Mary-ga t1 motteriu to] itta]]] (koto) Mary.NOM have C said fact that book.ACC John.NOM Bill.DAT ‘That book, John said to Bill that Mary has.’
Chapter 2. Properties of Scrambling
73
(2.85) German: *[CP dass [TP dieses Buch1 [TP Hans [VP dem Studenten gesagt hat [CP dass Maria t1 besitzt]]]]] that this book.ACC Hans.NOM the student.DAT told has that MaryNOM owns ‘Hans told the student that Mary owns this book.’ The German example (2.85) is a direct translation of the previous Japanese sentence (2.84) exhibiting intra-clausal Scrambling. Perhaps even more interesting is the fact that Japanese intra-clausal Scrambling appears to display Amovement characteristics whereas German middle field Scrambling has been show to display a mixed collection of A- and A’-movement. On the whole, Scrambling in German exhibits many traits that are similar to Operator Chains. However, in the end it is clear that wh-movement and Scrambling do not target the same landing sites and are thus not driven by the same mechanisms in the grammar. 2.5
Scrambling is not Topicalization
Lastly, it is also important to investigate Scrambling’s relationship to nonoperator A’-movement chains, e.g., topicalization. Here we can observe a stark contrast between the behavior of definite and indefinite objects that scramble into the middle field. Discussing definite objects first, Lenerz (1977, 2002) and more recently Meinunger (2000) conclude that definite objects that leave the vP are discourse topics to the extent that they represent old discourse information. Topics are traditionally classified as being non-operator movement and non-quantificational. Discourse requirements have the potential of being neither interpreted at either the phonological (PF) or logical (LF) interface, but rather licensed by a separate modular unit outside of the narrow syntax (Maria Carmen Parafita and Sara Rosen, p.c.). Abraham and Molnárfi (2002:2) state: “Other than movement for derivational purposes (such as tensing and Casemarking), scrambling in German is optional, partly semantically motivated and, since principally refocusing, context-linking.” The optionality attributed to Scrambling, therefore, has nothing to do with morphosyntactic features, but rather the presence or absence of discourse-oriented or semantic items in the computation. Lenerz (2002) also advocates an approach in which the linear order of arguments (and adverbials) in German is relatively free regarding its morphosyntactic categories, which of course, underlies certain restrictions, namely, these do not apply to the so-called unmarked order for arguments and adverbials. One of the contemporary approaches (which is the approach taken in this book) in light of such an approach is to take the unmarked word order as basic and derive all other orders from it by Scrambling, whatever its specific characteristics may be. For example, thus far we have hypothesized that the
74
Scrambling and the Survive Principle
dominant but not exclusive underlying base order of arguments with the vP are AG >> GODAT >> THACC, although it is possible that the order of internal arguments may be able to be altered (cf. McFadden 2004, Grohmann 2003:127). Lenerz views the observable restrictions obtained for particular linear orders as constraints upon movement operations, i.e., Scrambling. Below (2.86) is a representation of Lenerz’s restrictions for direct objects (DO) and indirect objects (IO) in German: (2.86) a. [+ def IO] > [+ def DO]: “unmarked order,” regardless of focus position (cf. (112a), (113a), (114a)). b. [+ def DO] > [IO]F : scrambling of [+ def, - F] is ok (cf. (112b)). c. * [+ def DO]F > IO = Don’t scramble focus! (cf. (113)) d. * [- def DO] > [IO]F = Don’t scramble (existential) indefinites! (cf.(114)).
Examples (2.87-2.89) illustrate Lenerz’s proposed restrictions in (86): (2.87)
Wem hast du das Buch gegeben? ‘Whom did you give the book to?’
Q: IO
a. Ich habe [dem/einem StuDENten]F das/ein Buch gegeben. I have the/a student the/a book given [+ def. IO]F > [+ def. DO] (‘unmarked order’) b. Ich habe das Buch [dem/einem StuDENten]F gegeben. I have the book the/a student given [+ def. DO] > [+ def. IO]F (scrambled [+ def DO, -F] is o.k.) ‘I gave the book to the student.’ (2.88)
Was hast du dem Studenten gegeben? ‘What did you give to the student?’
Q: DO
a. Ich habe dem Studenten [das BUCH]F gegeben. I have the student the book given [+ def. IO] > [+ def. DO]F (‘unmarked order’) b. */? Ich habe [das BUCH]F dem Studenten gegeben. I have the book the student given (* scrambled focus) ‘I gave the student the book.’
Chapter 2. Properties of Scrambling
(2.89)
Wem hast du ein Buch gegeben? ‘Whom did you give a book?’
75
Q: IO
a. Ich habe [dem/einem StuDENten]F ein Buch gegeben. I have the/a student a book given [+ def. IO]F > [- def. DO] (‘unmarked order’) b. * Ich habe ein Buch [dem StuDENten]F gegeben. I have a book the student given * [- def. DO] > [+ def. IO]F (* scrambled indefinite NP) c. * Ich habe ein Buch [einem StuDENten]F gegeben. I have a book a student given * [- def. DO] > [-def IO]F (* scrambled indefinite NP) ‘I gave a book to the student.’13 As recognized in the data above, a distinction in behavior exists between constituents that are marked for focus (F-marked), e.g., as new information receiving prosodic stress, and those that are not, as well as definites and indefinites. The Scrambling effects of definite objects in German are potentially the result of encoding the notions Topic and Focus in the Computational System.14 At its base, Focus is clearly not only a phonological phenomenon in German. Topics are also discourse related, but can be represented in syntactic structure. Abraham (2003) points out, however, that much typological divergence exists in language families regarding their individual realization of what Focus is. In Molnárfi (2002) and Abraham and Molnárfi (2002), the typological classifica13
Lenerz (2002:180-181) provides the following explanation to his data sets: As the standard examples (2.87) – (2.89) show, IO > DO is assumed to be the unmarked order for most verbs taking two objects; here, no specific restrictions apply: every distribution of definite or indefinite article NP and focus is possible for that order, cf. (2.87a), (2.88a), (2.89a). Scrambling a definite DO to the left of a focused IO gives a possible order, too, cf. (2.87b). If, however, the scrambled DO is a focused NP, it must be scrambled in front of an IO, cf. (2.88b). Thus, (2.86c) “Don’t scramble focus!” is a crucial restriction on scrambling in German. It may be accounted for by the interaction of focus placement, focus projection and, possibly, the proper assignment of prosodic features…Thus, this restriction may find a plausible functional explanation ensuring the proper interpretation of a sentence with regards to backgroundfocus structure and the formal means for its expression, i.e., linear order and prosodic prominence. 14 This is, of course, the weaker version of how discourse-oriented features are encoded into the Computational system. Banerji (2003) supports the stronger hypothesis that Focus and Topic are already present in the lexicon when selection for the numeration occurs. This will be discussed in Chapter 3 in connection with the possibility of understanding XP-adjunction as a formal movement type in universal grammar.
76
Scrambling and the Survive Principle
tion of ‘anti-focus’ was extended to Germanic, whereas languages such as Hungarian retain their distinction as ‘focus’ languages. Abraham and Molnárfi (2002) forward the notion that the pragmatic characteristics of theme (informationally old) and rheme (informationally new) can be linked to Universal Grammar (UG). Their research centers on the premise that Scrambling is trigged by discourse-functional requirements. “As such it is violable, thus explaining the optionality of word order variation in a natural way. The proposed destressing mechanism allows a new typological classification of discourse configurational languages, by distinguishing between discourse prominence by focus licensing of the Hungarian type and discourse prominence by antifocus licensing of the West Germanic type. The two processes are independent of one another and that the assignment of verbal morphology and legitimizing overt verbal positions are not to be handled under one single derivational process. What is generally held to be case driven movement and to mistakenly indicate the (assumedly universal) existence of AgrDO, AgrIO positions in German and other West Germanic is fundamentally discourse driven movement, triggered by a formal destressing rule for discourse themas” (Abraham and Molnárfi 2002:3-5, emphasis mine). From the perspective of discourse requirements, some studies loosely associate middle field Scrambling as topicalization due to the fact that scrambled XPs are familiar (i.e., thematic) in a pragmatic sense. A finer distinction of the notion topichood will clear up this confusion. There are two different explications of the notion ‘sentence topic’ that have been around in the theory (cf. Reinhart 1981, 1995). According to the familiarity concept, topics are expressions that denote given, known entities (e.g. Kuno 1972, Krifka 1992, Haftka 1995, Jäger 1996, Rizzi 1997). The other explication, the aboutness concept, takes as central that a topic whose referent the sentence is about (e.g. Reinhart 1981, 1995, Molnár 1991, 1998). Some authors characterize topics by both properties. Frey (2004:165) argues that the pure aboutness concept of topichood without the familiarity does not constitute ‘true’ syntactic topics. First, anaphoric phrases do not have to appear in the topic position. (2.90) Gestern hat Paul [eine nette Frau]1 kennengelernt. Er wird hoffentlich [die Dame]1 wiedersehen. yesterday has Paul a nice woman made the acquaintance of. He will hopefully the woman again-see ‘Paul met a nice woman yesterday. He will hopefully see the woman again.’ Example (2.90) shows that familiarity is not a sufficient condition for being a topic in German. According to Frey and Pittner (1998), Pittner (1999) and Frey (2004), they are in the ‘middle field’ in German:
Chapter 2. Properties of Scrambling
77
In the middle field of the German clause, directly above the base position of sentential adverbials (SADVs), there is a designated position for topics: all topic phrases occurring in the middle field, and only these, occur in this position. (Frey 2004:157)
This hypothesis put forward in this study agrees with Frey that topics in German appear above SADVs and that they are adjoined to TP. The fundamental difference between the approach adopted here and Frey’s statement is that Frey still advocates the use of an additional projection above TP (AgrSP), a projection shown by Zwart (1997) to be conceptually unnecessary to the theory. Additional evidence in support of topics only existing above the subject in the CPlayer German is also observable in the fact that XPs above SADVs can be coindexed with cataphoric pronouns whereas constituents in the traditional ‘middle field’ cannot. According to Reinhart (1981, 1995) and Kuno (1972) a cataphoric pronoun has to relate to a topic. The following data show that an expression which is co-referential with a cataphoric pronoun and which occurs in the ‘middle field’ has to be in front of a SADV, thereby confirming Frey’s (2004) hypothesis. (2.91) Weil er1 gut trainiert hat, wird Leo1 wahrscheinlich morgen spielen. because he good trained has will Leo probably tomorrow play (2.92) *Weil er1 gut trainiert hat, wird wahrscheinlich Leo1 morgen spielen. because he good trained has, will probably Leo tomorrow play ‘Because he trained well, Paul will probably play tomorrow.’ (2.93) Sein1 Vater wird dem Hans1 glücklicherweise bei dem Vorhaben helfen. with the event help his fatherNOM will the HansDAT fortunately (2.94) *Sein1 Vater wird glücklicherweise dem Hans1 bei dem Vorhaben helfen. his fatherNOM will fortunately the HansDAT with the event help ‘His Father will fortunately help Hans with the event.’ If the SADVs wahrscheinlich ‘probably’ and glücklicherweise ‘fortunately’ are adjoined to TP, Leo in (2.91) and Hans in (2.93) reside in a landing site above the subject within the CP-layer. The subjects in (2.92) and (2.94) remain within the middle field and are therefore not as readily linked to external discourse, although they represent information that is old to the discourse. What this shows is that XPs above the subject – those above SADVs – are contrastive topics that are accessible to discourse-linking with external discourse notions (D-linking) whereas XPs in the ‘true’ middle field only denote familiar/old information.
78
Scrambling and the Survive Principle
A further distinction between middle field Scrambling and topicalization can be observed in remnant movement patterns. Remnant movement is movement of an XP β from which extraction of α has taken place, cf. (2.95). (2.95) [β … ti … ]2 … [ …α1 … [ … t2 …]] From the data we have seen thus far we can confirm that discourse topics target a landing site above the subject, i.e., not in the middle field. Remnant movement, i.e., extraction of an XP from a unit that has already itself undergone movement from its base position, must not have both items that resided in the original structure occupying the same field after substitution (movement) has taken place. (2.96) [PRO t1 zu kaufen]2 habe ich das Brot1 der Abby versprochen t2. to buy have INOM the breadACC the AbbyDAT promised ‘I promised Abby to buy bread.’ (2.97) *Gestern habe ich [PRO t1 zu kaufen]2 das Brot1 der Abby t2 versprochen. yesterday have INOM to buy the breadACC the AbbyDAT promised ‘Yesterday I promised Abby to buy bread.’ In (2.96) the VP [PRO ti zu kaufen] ‘to buy’ has moved above the subject and has been topicalized. The DP das Brot ‘the bread’; however, has moved to the middle field via Scrambling. In (2.97) we observe that the VP [PRO ti zu kaufen] ‘to buy’ has undergone Scrambling as well as the DP das Brot ‘the bread’. This results in an ungrammatical utterance. From these data we can draw two preliminary conclusions: First, aboutness topics do not exist in the German middle field; the final landing site for contrastive, aboutness topics in German resides above the subject in the CP-layer. Scrambling and topicalization therefore do not target the same landing sites and are independent operations driven by independent grammatical operations. Second, it appears that the potential exists for a constraint regulating a minimal distance of traversal of XPs from their base position. In sentence (2.97), movement of both elements of previously unified phrase structure into the middle field results in an ungrammatical sentence. This latter concept; namely, a lower-bound restriction of XPmovement, will be taken up in Chapter 3. 2.5.1
Indefinites and the Mapping Hypothesis
Definite objects and indefinite objects that scramble into the middle field possess unique licensing strategies. As seen throughout this chapter regarding data with definite objects, their ability to move appears to be ‘optional’ from a mor-
Chapter 2. Properties of Scrambling
79
phosyntactic standpoint. Definites leave their base position if they are interpreted as discourse-old (thematic) material. Indefinites can only scramble if they are interpreted as generic, i.e., specific, rather than existential (data from Webelhuth 1992:198). (2.98) *weil er ein Buchi wohl [VP ti gelesen hat] because heNOM a bookACC probably read has ‘because he probably has read a book.’ (2.99) weil er ein Buchi wohl [VP ti lesen würde] because heNOM a bookACC probably read would ‘because he probably would read a book.’ Sentence (2.98) is ungrammatical; however, if we change the tense and mood in (2.99) so that the indefinite object that scrambles into the middle field can receive a generic/specific interpretation, the sentence is well-formed. The theoretical claim that indefinites move out of the VP to check features with an existential operator-head (∃) corresponds nicely to the work of Diesing (1992, 1997) who assumes that indefinites that move outside of the vP do so to enter into an operator-variable relationship with the operator [Gen_x]. Most scholars argue that existential indefinites do not leave the VP (Diesing 1992, Abraham 2003, Molnárfi 2002, 2003, Abraham and Molnárfi 2003, Grewendorf and Sabel 1999 among others). Such an assumption is supported by the fact that existential indefinites cannot scramble above predicate-level adverbs, i.e. adverbials adjoined to the vP. Take, for example, the sentences presented in (2.100) and (2.101) (from Diesing 1992:70-71). (2.100) a.
b. (2.101) a.
b.
…weil Elly immer Lieder singt. since Elly always songs sings ‘…since Elly always sings songs’ ALWAYSt [time(t)] ∃x song(x) & sing(Elly,x) …weil Elly Lieder immer singt. since Elly songs always sings ‘…since, (generally) if it’s a song, Elly will sing it. ALWAYSx [song(x)] sing(Elly,x)
Sentential adverbs such as immer ‘always’ and negation serve as ‘signposts’, and thus we see that the indefinite object Lieder ‘songs’ in (2.101a) remains in situ. By not Scrambling the object out of the VP, the object receives an existential reading (marked by the existential marker: ∃). Only when a generic in-
80
Scrambling and the Survive Principle
terpretation of the object is possible (or preferred) can the indefinite move out of the vP. Thus we can observe the pattern that unscrambled indefinites remaining in the vP receive an existential interpretation, whereas generic indefinites can move into the middle field (i.e., TP-layer). Diesing captures this generalization in her definition of the Mapping Hypothesis. (2.102) Mapping Hypothesis (tree splitting) TP 3 Spec T’ 3 AgrOP Restrictive Clause To 3 Spec Agr’ 3 vP Agro 3 Spec v’ 3 vo VP 3 Nuclear Spec V’ Scope 3 Comp Vo Diesing’s Mapping Hypothesis (1992) derives tripartite quantificational representations by splitting the syntactic tree into two parts. Under this procedure, a tripartite structure consisting of an operator (e.g., a quantifier), a restriction (the domain that quantifier quantifies over), and a nuclear scope (the remainder of the sentence). In her own words, Diesing explains the ramifications of the Mapping Hypothesis: …the Mapping Hypothesis gives the result that the VP forms the domain for default existential closure, and variables introduced in the material above VP are bound by the operator. This has the effect that different noun phrase interpretations are associated with distinct syntactic positions in the tree – existential interpretations are associated with VP-internal positions, which quantificational interpretations are associated with positions higher in the tree. (Diesing 1997:373-374).
The Mapping Hypothesis proposed in (2.102) above is intuitively straightforward: Existential indefinites require existential closure and must be mapped onto LF-interface for semantic interpretation.
Chapter 2. Properties of Scrambling
2.6
81
Interim Conclusion
The data and argumentation presented thus far in this chapter illustrate that middle field Scrambling in West Germanic (in particular German) cannot be classified ubiquitously as NP-movement, wh-movement or topicalization. Concerning Scrambling’s affiliation with NP-movement, accusative DPs that scramble into the middle field can enter into licit binding relations with anaphors; however, this appears to be impossible for dative objects. Scrambling in West Germanic also does not appear to exhibit the effects of WCOs; however, for the rest of the traditional A/A’-movement distinction, Scrambling shows a strong resemblance to A’-movement properties (i.e., licensing of parasitic gaps, reconstruction effects, etc.). Robust morphological Case was also shown not to be a triggering factor in licensing Scrambling. Parsing A’-movement types into operator (wh-movement) and non-operator (topicalization) afforded us the opportunity to observe more closely other characteristics of Scrambling. Middle field Scrambling in West Germanic exhibits a close relationship to Operator-Chains (wh-movement); however, as shown by Kidwai (2000), they inevitably do not target the same landing site cross-linguistically. From a discourse perspective, XPs that scramble into the middle field are similar to topics in that they bear reference to thematic discourse properties; i.e., previously established items in discourse. From the standpoint of syntax, however, they exhibit contrastive behavior: ‘True’ topics, i.e., XPs that move above the subject into the CP-layer are contrastive topics that indicate ‘aboutness’ and can bind cataphoric pronouns. Scrambled XPs into the middle field are ‘familiar’ and cannot bind cataphoric pronouns. As a result, topics and scrambled XPs do not target the same landing sites in the structure. At this juncture, I would like to draw two conclusions: First, middle field Scrambling in West Germanic is not licensed by any formal agreement features. Although Scrambling does display certain characteristics that are reserved for Agreement-Chains, proposing an analysis of Scrambling based on the necessity of licensing Case is extremely problematic and illogical. Unlike other elements of grammar, morphosyntactic agreement features must be licensed in the course of the derivation to insure grammatical outputs. Given the fact that definite DPs can remain in situ and do not need to scramble would force us to argue that agreement features in the same language occupy the status of both ‘strong’ and ‘weak’ (i.e., features that must be dealt with in the syntax ‘covertly’ and those that can be taken care post-syntactically ‘overtly’). Such optionality is a computational nightmare and should be avoided. Likewise, the logico-semantic feature [+wh] and discourse feature [Topic] are incompatible with Scrambling. What emerges as the one consistency at this point are the discourse requirements imposed by scrambled XPs. Introducing formal features in an ad hoc manner as ‘syntactic’ features without providing rationale for their existence as ‘syntactic’ elements is merely stipulative. At this point
82
Scrambling and the Survive Principle
we return to one of the core questions addressed in this work: Can Scrambling be interpreted as a syntactic operation? If this is possible, it must be done by other means than traditional agreement and logico-semantic features. The remainder of this chapter will investigate other grammatical and structural aspects of Scrambling. These data will complete the mosaic of what motivates middle field Scrambling in West Germanic and lead to the discussion to what extent Scrambling can be accounted for in syntactic theory. 2.7
Freezing and Anti-Freezing Effects
In general, the extraction of an XP α from another XP β is possible iff β is in situ, as schematically depicted below in (2.103a), and not if it has undergone movement itself, as shown in (2.103b). This effect is commonly referred to as the Freezing effect15, and I will refer to this movement type as such throughout this book. (2.103) Freezing effect (Müller 1998:124) a.
… α1 … [β … t1 …]2
b.
* … α1 … [β … t1 …]2 … t2
As pointed out by Diesing (1992) and Müller (1998), there appears to be a restriction placed on Scrambling from extraposed phrases, i.e., phrases that have already undergone prior movement. The sentences below from Diesing (1992:129) show that Subjacency (cf. Chomsky 1986), i.e., a violation of a locality condition on links of a movement chain, is violated when wh-items are extracted from a configuration that has already undergone movement from its base position. (2.104) Wasi hat [IP Otto immer [VP [NP ti für Romane] gelesen]]? what has Otto always for novels read ‘What kind of novels has Otto always read?’ (2.105) *Wasi hat [IP Otto [IP [NP ti für Romane] immer gelesen]]? what has Otto for novels always read ‘What kind of novels has Otto always read?’ Assuming that the temporal adverb immer ‘always’ is adjoined to the vP (VP in the examples above), we can see that the NP [Was für Romane] ‘what kind of novels’ has scrambled from its base position in the middle field. Müller 15
See also Ross’ (1967/1986:173) Frozen Structure Constraint, Wexler and Cullicover’s (1980:179) Freezing Principle and Diesing’s (1992:128) Revised Extraction Constraint.
Chapter 2. Properties of Scrambling
83
(1998:125) provides further evidence in support of the Freezing effect. The examples (2.106) and (2.107) below illustrate that the Scrambling of an object NP (that is transparent for extraction in situ) in front of an adverbial induces a Freezing effect, i.e., it prohibits extraction. (2.106) [PP Über wen]1 hat der Fritz letztes Jahr [NP ein Buch ti]2 geschrieben? about whom has the FritzNOM last year a bookACC written ‘Who did Fritz write a book about last year?’ (2.107) *[PP Über wen]1 hat der Fritz [VP [NP ein Buch ti]2 [VP letztes Jahr t2 geschrieben]]? a bookACC last year written about whom has the FritzNOM ‘Who did Fritz write a book about last year?’ An identical situation arises with Scrambling in front of the subject – even if the NP is transparent for extraction in its in situ position, it turns into an island if it is scrambled (again, data from Müller 1998:125). (2.108) [PP Worüber]1 hat keiner [NP ein Buch ti] gelesen? about what has no oneNOM a bookACC read ‘About what (topic) has no one read a book?’ (2.109) *[PP Worüber]1 hat [NP ein Buch ti]2 keiner t2 gelesen? about what has a bookACC no oneNOM read ‘About what (topic) has no one read a book?’ The definition provided for remnant movement (cf. (2.95)) appears to be a violation of the Freezing effect in a strict sense. In this regard, remnant movement involving Scrambling is often referred to as an Anti-Freezing effect. Consider the following example (data from Webelhuth and den Besten (1987:78)): (2.110) [VP t1 Gelesen] hat das Buch1 keiner t2. read has the book no one ‘No one read the book.’ Scrambling of the NP das Buch ‘the book’ into the middle field takes place from an already moved VP. Webelhuth and den Besten (1987) also point this fact out; namely, that an item may be scrambled from VP prior to the topicalization of that item. (2.111) [PRO t1 zu kaufen]2 habe ich das Brot1 der Helga versprochen t2. to buy have INOM the breadACC the HelgaDAT promised ‘I promised Helga to buy bread.’
84
Scrambling and the Survive Principle
The ‘problem’ with this structure and remnant movement structures in general is that the trace of das Brot ‘the bread’ (t1) is unbound since its antecedent does not c-command it. Such illegal antecedent-trace configurations are somehow ruled out, as to prevent non-cyclic argument lowering as a licit instance of XPmovement. Interestingly, if the remnant VP scrambled rather than topicalizes (as previously argued in this chapter), i.e., moves above the subject, the resultant structure is ungrammatical. (2.112) *Gestern habe ich [PRO t1 zu kaufen]2 das Brot1 der Helga t2 versprochen. yesterday have INOM to buy the breadACC the HelgaDAT promised ‘Yesterday I promised Helga to buy bread.’ Bear in mind that the Scrambling of the entire VP containing the direct object das Brot “the bread” is acceptable. (2.113) Gestern habe ich [PRO das Brot zu kaufen]1 der Helga t1 versprochen. yesterday have INOM the breadACC to buy the HelgaDAT promised ‘Yesterday I promised Helga to buy the bread.’ The difference between these two cases can be reduced to the difference in the character of the movement types involved: movement of the VP above the subject as shown in (2.110) and (2.111) constitutes topicalization, whereas movement of objects into the middle field, regardless if they are the result of remnant movement or the displacement of a finite clause as in (2.113), is the product of Scrambling. 2.8
Referentiality
Thus far we have seen that Scrambling patterns in West Germanic are not fully compatible with either movement for agreement (A-movement) or nonagreement purposes (A’-movement) purposes. In this respect, it is difficult, if not impossible, to isolate a formal morphosyntactic or logo-semantic feature that can license Scrambling. The only consistent conclusion to be drawn thus far is the discourse effects Scrambling patterns project. Structural constraints, such as Freezing effects, also face scrutiny. Fanselow’s (2001:413-414) data below are directed at thwarting any account of feature-driven Scrambling in German. (2.114)
Wen1 hat [t1 von den Musikern] fast jeder Student kennengelernt? whom.ACC has of the musicians nearly every student met ‘Which of the musicians has almost every student met?’
Chapter 2. Properties of Scrambling
85
(2.115)
* Wen1 hat fast jeder Student [t1 von den Musikern] kennengelernt?
(2.116)
Wen1 hat Karl [t1 von den Musikern] zweimal getroffen? whom.ACC has Karl of the musicians twice met ‘Which of the musicians has Karl met twice?’
(2.117)
*Wen1 hat Karl zweimal [t1 von den Musikern] getroffen?
These data suggest that quantifiers intervening between an operator and its trace block wh-movement. In such contexts wh-phrases can only be extracted from phrases occupying a scrambled position. Extraction from base positions is impossible. According to Fanselow, “phrases preceding their unmarked position tend to be referential topics, and the more referential a phrase is, the less transparent it is for movement. Island effects for scrambled XPs may thus reduce to referentiality rather than to positional barrierhood” (2001:413). As shown previously in this chapter, scrambled middle field XPs are not ‘true topics’ in that they can bind anaphoric pronouns, but exist as referential, ‘aboutness’ topics. Both Cinque (1990, 1999) and Stroik (2000) posit the existence of a referential feature, i.e., [+ Ref], in the syntax. Stroik (2000:77) provides the following data to test the existence of the [+ Ref] feature, although nonrestrictive relative clauses can modify referential DPs, they can never modify non-referential DPs. (2.118)
Sam, whom I like a great deal, hates me.
(2.119)
Everyone here, whom I like a great deal, hates me.
(2.120)
*No one, whom I like a great deal, hates me.
If scrambled XPs in West Germanic are referential in nature, we should expect them to be able to be modified by non-restrictive relative clauses. This assumption is confirmed by the following data (thanks to Kleanthes Grohmann and Winnie Lechner for their assistance here). Quantified DPs with non-restrictive relatives (such as ‘someone, whom I met yesterday’) can scramble in German. (2.121) Maria hat [jemanden, den ich neulich kennengelernt habe]i ihrem Freund ti vorgestellt. MariaNOM has someoneACC whomACC INOM recently got to know have her friendDAT introduced ‘Maria introduced someone, who I recently got to know, to her friend.’
86
Scrambling and the Survive Principle
(2.122) Wahrscheinlich würden [jemanden, den sie über das Internet kennengelernt haben]i nur Frauen mit gewisser Neigung ti heiraten wollen. probably would someoneACC whomACC theyNOM over the Internet got to know have only womenNOM with real affection marry would ‘Probably one woman would marry someone with real intent that they met over the Internet.’ In both (2.121) and (2.122) above, the quantified DP jemanden ‘someone’ scrambles from its base position in the VP and can be modified by a nonrestrictive relative clause. Following Stroik’s assertion that only referential DPs can be modified by non-restrictive relative clauses also holds for XPs that are scrambled into the middle field in West Germanic. Haider (1993) presents further evidence in support of the hypothesis that only referential XPs can scramble into the middle field. Wh-phrases wer ‘who’ and was ‘what’ can also be used as indefinites (truncation from irgendwer ‘someone’ and irgendwas “something”). Since these items are non-referential, i.e., do not have an exact real-world entity with which they are co-referenced, these weak indefinites should resist Scrambling (data from Winnie Lechner, p.c.). (2.123) Wer hat wem was gezeigt? whoNOM has whomDAT whatACC shown ‘Who has shown what to whom?’ (2.124) *Wer hat wasi [vP wem ti gezeigt]? whoNOM has whatACC whomDAT shown ‘Who has shown what to whom?’ In the absence of a specific real-world entity, scrambling of the truncated weak pronoun was (irgendwas) ‘what’ is not possible in German. In connection with the pragmatic/discourse interpretations available for XPs that scramble, here we encounter another consistent characteristic of Scrambling: Referential XPs can undergo Scrambling, whereas non-referential XPs cannot. 2.9
Scope Bleeding
Scrambling has the ability to change scope relations. If the scrambled element is a scope sensitive one (e.g., a weak NP), its scope domain gets wider by virtue of being moved to a position higher in the tree. If it moves across another scope-bearing element, scope ambiguities arise (cf. Frey 1993) (examples from Haider 2005a:227).
Chapter 2. Properties of Scrambling
(2.125) Q:
A:
(2.126)
Wie steht es mit der Beantwortung alle der Fragen? How stands it with the answering all of the questions ‘What about the answers to all the questions?’ Du weisst doch,… You know prt ‘You know,…’ dass mindestens einige Schüler fast alle Fragen beantworten konnte that at least some students almost all questions answer could ‘that some students could answer almost all the questions’ i.)
(2.127)
87
∃x ∀y
ii.)
*∀y ∃x
dass [mindestens einige Fragen]i fast alle Schüler ti beantworten konnten. That at least some questions all students answer could ‘that at least some questions almost all of the students could answer’ i.)
∃x ∀y
ii.)
∀y ∃x
In example (2.126) the existential quantifier takes wide scope, however in the latter example (2.127) scrambling the QP mindestens einige Fragen ‘at least some questions’ creates a scope ambiguity. This data set argues against the potential of a logico-semantic feature responsible for the scope properties of Scrambling. Although previous accounts (cf. Stowell and Beghelli 1997) have attempted to eliminate such ambiguities through the existence of a formal feature in the syntax, Scrambling in West Germanic creates ambiguities. In other words, Scrambling feeds scope ambiguities. As a result, no formal feature licensing scope in the narrow syntax can take care of middle field Scrambling in West Germanic. 2.10
Prosody
As demonstrated by the data in this chapter, the permutations of scrambled objects in the German middle field are relatively free. With that being said, multiple grammars, i.e., morphosyntactic, discourse/pragmatic and semantic requirements, exercise their preferences in the construction of licit grammatical sentences. This is covertly assumed by Lenerz’ constraints (cf. (2.86)) on Scrambling. Concerning morphosyntax, West Germanic on the whole prefers the preservation of objects as they appear in the predicate (vP) as an unmarked structure. From the standpoint of discourse configurationality, rhematic (i.e.,
88
Scrambling and the Survive Principle
discourse ‘old’ material) remains in the verb phrase, whereas thematic objects (i.e., discourse ‘new’ material) moves into the middle field and is recognized as a referential ‘familiar’” topic or to the CP-layer above the subject and receives a contrastive “aboutness” reading. Semantically, items that can be interpreted as generic/specific may scramble; however, items that only receive an existential reading cannot. Middle field XP-Scrambling in West Germanic – as observed by Lerenz (1977, 2002), Zubizarreta (1998), Abraham (2002, 2003), Molnárfi (2002, 2003), Abraham and Molnárfi (2003) and Dominguez (2004) among a host of others – results in the destressing of definite objects. In contrast, indefinite objects that scramble out of the vP receive prosodic stress (in the Dutch examples below, prosodic stress is indicated in bold typeface). (2.128)
dat Jan [vP gisteren het BOEK gelezen heeft] that JanNOM yesterday the bookACC read has ‘that Jan read the book yesterday.’
(2.129)
dat Jan het boeki [vP gisteren ti GELEZEN heeft] that JanNOM the bookACC yesterday read has ‘that Jan read the book yesterday.’
In example (2.128) the object DP het boek ‘the book’ remains in the vP and is thus marked for prosodic stress. We see the destressing effect on scrambled constituents in (2.129); in this sentence the object has scrambled into the middle field and does not receive prosodic stress. This is a perfect example of Lenerz’ constraint: Don’t scramble focus! (cf. (2.86c)). The notion that, cross-linguistically, Scrambling is perhaps prosodicallydriven has received much attention in recent studies. For example, Webelhuth (1992), Sekerina (1997, 2003), Kidwai (2000), and Abraham and Molnárfi (2003) have independently argued for a formal feature responsible for removing sentential stress from scrambled XPs in their respective analyses of this movement type. Such assumptions; however, come at a theoretical cost if we want to maintain a modular, ‘phonology-free’ syntactic system. Although the prosodic characteristics of middle field XP-Scrambling are indeed another consistent factor in determining scrambled word orders, the claim that some sort of prosodically-sensitive feature exists in the syntax must be carefully scrutinized. I will take up this task in full in Chapter 4. 2.11
Potential Triggers
In this section I test three candidates that could function as the triggering feature if Scrambling is indeed a feature-driven phenomenon:
Chapter 2. Properties of Scrambling
89
•
Scope features (similar in argument to Stowell and Beghelli 1997)
•
Prosodic features (cf. Abraham and Molnárfi 2002, Molnárfi 2003)
•
Pragmatic features
Although all of these aspects of Scrambling undoubtedly play a role at some point in the grammar of shaping grammatical outputs, in this section I argue that only the latter features (i.e., discourse/pragmatic features) have the ability to exist in the narrow syntax. 2.11.1 XP-Scrambling is not driven by a Scope Feature
Similar to Rizzi’s (1997) seminal work on the expanded nature of the CP-layer, Stowell and Beghelli (1997) introduce an extended structure of the middle field/TP-layer outside of Agr-phrases to account for scope relations. Stowell and Beghelli’s analysis of the middle field is empirically motivated by the lack of scope uniformity of quantifier phrases (QPs), with scope uniformity meaning that quantifier raising is applicable to all QPs and that neither quantifier raising nor any QP is landing-site selective. Most importantly for our purposes, Stowell and Beghelli (1997:16) believe that the position identified as Spec,ShareP in the middle field could function as a common landing site for scrambled XPs cross-linguistically. Under such an approach, scrambled XPs would leave the middle field in order to check a scopal feature in the middle field. Stowell and Beghelli’s account of Scrambling as a scope-driven phenomenon fails on two accounts. First, recall the data from early in this chapter (cf. (2.125) – (2.127) discussing scope bleeding (the data is repeated here for the sake of the reader; data from Haider 2005a:227). (2.130) Q:
Wie steht es mit der Beantwortung alle der Fragen? How stands it with the answering all of the questions ‘What about the answers to all the questions?’
A:
(2.131)
Du weisst doch,… You know prt ‘You know,…’ dass mindestens einige Schüler fast alle Fragen beantworten konnte that at least some students almost all questions answer could ‘that some students could answer almost all the questions’ i.)
∃x ∀y
ii.)
*∀y ∃x
90
(2.132)
Scrambling and the Survive Principle
dass [mindestens einige Fragen]i fast alle Schüler ti beantworten konnten. That at least some questions all students answer could ‘that at least some questions almost all of the students could answer’ i.)
∃x ∀y
ii.)
∀y ∃x
As shown in (2.131), scrambling an XP above another creates scope ambiguity, which is exactly what Stowell and Beghelli’s model/theory seeks to eliminate. Second, the scope projection for strong NPs would have to be above the posited scrambling position designated by Stowell and Beghelli in the middle field (e.g., Spec,ShareP) in order for scope ambiguity to exist in (2.132). If that is the case; however, it is unclear why (2.131) lacks the wide scope universal reading. It appears that scope may have to be computed by lowering, countercyclic movement in scope rigid languages. One piece of evidence that scope relations are fixed by lowering comes from quantifier scope interactions with more than two QPs.16 2.11.2 Is Scrambling Purely Prosodic? Next we can turn our attention to the argument that Scrambling as understood through prosodic properties. The theoretical shift in focus from GB-theory to minimalism and from proper government to interface interpretability forced theorists to ask pertinent questions regarding the true nature of the syntax-PFinterface. In earlier versions of minimalism (ca. 1993, 1995), the entire derivation proceeded until the hierarchy was shipped off to the interfaces via SpellOut. The LF and PF interfaces in early minimalism were regarded as the sole levels of representation, with all operations and principles being motivated by virtual conceptually necessity (VCN) or bare output conditions (BOC). With the onset of multiple Spell-Out models (cf. Uriagereka 1999) in the literature and an overall progressive development of interface phenomena (e.g., Distributed Morphology at the PF-interface (cf. Halle and Marantz 1993)) the PFbranch was attributed the task of licensing traditional morphological considerations. Furthermore as noted by Stroik (2000, 2007) and Embick and Noyer (2001) among others, operations applicable at the PF-branch are ordered with respect to one another. The model in (2.133) (taken from Grohmann 2006) represents the traditional a view of the PF-branch of the grammar ca. 2001 (cf. Embick and Noyer 2001). 16 The crucial point is that the DO can reconstruct below the subject but subject cannot participate in QR to the left of the DO (Winnie Lechner, p.c.). This provides further argumentation against an LF-lowering account of middle field Scrambling in West Germanic.
Chapter 2. Properties of Scrambling
91
(2.133) A View from the PF-Branch Lowering
arrangement of morphemes
Vocabulary Insertion
Linearization imposed by vocabulary insertion
Prosodic Domains
PF As will be discussed in full detail in Chapter 4, the main reason for rejecting the hypothesis that Scrambling is purely a prosodic operation is due to the ability for scrambled XPs to enter into licit binding relations with anaphora (cf. Section 2.3.2). Binding relations must remain intact at the end of the derivation to prevent any offensives to the C-I (i.e., LF) interface. The fact that middle field Scrambling in West Germanic exhibits both LF and PF reflexes eliminates the possibility this phenomenon being strictly prosodic. 2.11.3 Discourse/Pragmatics – Referential Features [+ Ref] Another consistent feature exhibited by XPs that scramble into the middle field in West Germanic is that they are referential in nature. By referential I mean that they refer to real-world entities, usually previously established rexpressions in the discourse.17 This is not to be confused with topics that appear above the subject in the CP-layer. As shown earlier in this chapter (cf. Section 2.5), ‘true’ topics are able to bind cataphoric pronouns and are more readily accessible to external discourse considerations (i.e., D-linking). If the scrambled XP is referential, it must move into the middle field. Failure to do such renders sentence highly marked. Consider the following Dutch examples. (2.134) Op tafel ligt een boterham met ham. Tom lust geen ham. Dus, zegt ie: on table lies a sandwich with ham Tom likes no ham thus says he ‘On the table lies a sandwich with ham. Tom doesn’t like ham, therefore he says:’ (2.135) Ik ga [de boterham]i niet ti opeten. I go the sandwich not up-eat ‘I am not going to eat the sandwich.’ 17
As stated in Chapter 1, I basically adopt Fodor and Sag’s (1982) definition of referentiality: A nominal expression is understood to be referential if it has a ‘fixed referent’ in the (model of) world, meaning that it can be identified by the speaker and/or by one of the people who prepositional attitudes are being reported.
92
Scrambling and the Survive Principle
(2.136) ??/*Ich ga niet [de boterham] opeten (maar het stukje pizza). I go not the sandwich up-eat rather the piece pizza ‘I am not going to eat the sandwich, but rather the piece of pizza.’ In the Dutch examples above, the object de boterham ‘the sandwich’ is a previously established entity (i.e., old information), therefore it must scramble into the middle field due to its referential nature (cf. (2.135)). On the contrary, failure to scramble into the middle field results in a highly marked/ungrammatical sentence (cf. (2.136)). The evidence above favors the potential existence of a [+ Ref] feature by demonstrating that non-specific, non-referential items never undergo Scrambling. It is also important to note that definiteness and referentiality are not (always) synonymous.18 (2.137) There are several men, all of whom I greatly admire, waiting for you outside. (2.138) Everyone has theories about the men who assassinated the President. In the first example (2.137), we have an indefinite subject that is referential in nature. Further evidence for the referential character of this subject can be inferred by the presence of a modifying non-restrictive relative clause. The object of the preposition in the men in (2.138) is definite and non-referential. If an interpretable [+ Ref] feature exists in the syntax, it can appear either on head of the nominal phase (Do) or can be assigned in the lexicon to the head of a bigger projection (Xo) (data from Fanselow 2001:413-414; this data set is repeated again for the sake of the reader). (2.139)
Wen1 hat [t1 von den Musikern] fast jeder Student kennengelernt? of the musicians nearly every student met whomACC has ‘Which of the musicians has almost every student met?’
(2.140)
* Wen1 hat fast jeder Student [t1 von den Musikern] kennengelernt?
(2.141)
hat Karl [t1 von den Musikern] zweimal getroffen? Wen1 whomACC has Karl of the musicians twice met ‘Which of the musicians has Karl met twice?’
(2.142)
*Wen1 hat Karl zweimal [t1 von den Musikern] getroffen?
18
Again, thanks to Tom Stroik (p.c.) for pointing this out to me.
Chapter 2. Properties of Scrambling
93
As noted by Fanselow (2001), these data suggest that quantifiers intervening between an operator and its trace block wh-movement. In such contexts whphrases can only be extracted from phrases occupying a scrambled position. Extraction from base positions is impossible. According to Fanselow, “Phrases preceding their unmarked position tend to be referential topics, and the more referential a phrase is, the less transparent it is for movement. Island effects for scrambled XPs may thus reduce to referentiality rather than to positional barrierhood” (2001:413). The fact that Scrambling is sensitive to any sort of island effects is significant evidence in favor of analyzing this displacement type as a syntactic operation. If referentiality is the potential grammatical feature that enduces what appears to be island effect violations for scrambled XPs, it falls out nicely that referentiality is a syntactic feature. Most importantly for this study, referentiality is the syntactic feature responsible for licensing middle field Scrambling in West Germanic languages. In conclusion, although scope ambiguity, prosody and discourse considerations factor into the grammaticality of representations, only the latter (i.e., discourse considerations) have the potential of being active in syntactic derivations. 2.12
Conclusion
The goal of this chapter was to supply a working profile of the complex nature of middle field Scrambling in West Germanic. Referring back to the interim conclusion, Scrambling does not fully align itself with NP-movement, whmovement, or topicalization. It is practically a lost cause to attempt to categorize Scrambling along the lines of traditional A- or A’-movement characteristics. Isomorphic reflexes for either movement type simply do not exist. In the latter section of this chapter, I brought forth more stable and consistent characteristics of Scrambling in West Germanic: First, DPs that can scramble in the West Germanic middle field are referential topics. Second, weak DPs that scramble into the middle field create scope ambiguities. Third, Scrambling follows strict prosodic considerations. Furthermore, middle field XP-Scrambling appears to follow structural constraints such as Freezing effects. Although these grammatical facets of Scrambling highlight consistent behavioral patterns of this phenomenon, a word of caution must be interjected here. As far as a syntactic analysis is concerned, the above-mentioned features (e.g., pragmatics, prosody and scope) traditionally fall outside the scope of syntax. Under most accounts they are viewed as post-syntactic phenomena. The task of this chapter was to see if it is possible (and if so, to what extent) to unite Scrambling under one formal feature or formal movement operation that can be analyzed as ‘syntactic.’ This feature was found to be the discourse/semantic notion of referentiality conceptualized by the [+ Ref] feature.
94
Scrambling and the Survive Principle
Before moving on to the theoretical portion of this work, the functional domains of the West Germanic clause – based on the data presented in this chapter – is presented and briefly discussed in (2.143) below. (2.143)
[CP
XP
[TP
XP
[vP
XP
]]]
Contrastive Topic
Referential Topic
Rhematic Information
Generic Indefinites
Generic Indefinites
Existential Indefinites
Stressed
Destressed
Stressed
The vP-layer, or predicate, is the domain responsible for thematic relations and from the perspective of discourse requirements, licenses rhematic information. Items that move to the middle field are specific (i.e., referential and/or generic) items that are not prosodically-marked. Constituents that move above the subject into the CP-layer are interpreted as contrastive ‘aboutness’ topics that are stressed. At this juncture it is difficult to make a definitive claim regarding the appropriate domain where agreement features are licensed. Recall that DPs in Scandinavian that participate in Object Shift move to the middle field for agreement purposes, hence their A-Chain characteristics. Scrambling in German shows a strong affiliation to O-chains. Perhaps this is an indication of a difference in phrase structure properties between Scandinavian and West Germanic languages. The following chapter will concern itself primarily with the theoretical implications of adopting middle field Scrambling as motivated by the [+ Ref] feature in a Survive-based model of minimalist inquiry.
CHAPTER 3 THEORETICAL CONSIDERATIONS The thorough analysis of the properties of middle field Scrambling in West Germanic languages and dialects provided in the previous chapter reinforces the insurmountable task of classifying Scrambling within the archaic A- vs. A’-movement (i.e., both operator and non-operator) typology. The mixed A/A’-movement properties are difficult to surmise under one (or a collection) of formal features. As shown with empirical data in Chapter 2 and reiterated in the tables (3.1) and (3.2) below, middle field Scrambling in German does not display an isomorphic A- or A’-movement reflex (adapted from Richards 2005:6). (3.1)
A’-movement Properties of Scrambling
A’-properties
Scrambling
Licenses Parasitic Gaps
3
Weak Crossover Effects (WCO)
2
Reconstructs (no binding)
2
Affects any category (XP)
2
Possible rightward movement
2
Long-distance movement
2
Discourse effect (topic/focus)
3
(3.2)
A-movement Properties of Scrambling
A-properties
Scrambling
Case-driven
2
Affects only arguments
2
Feeds binding
3
Fed by V-raising
2
Predicate formation (applies where
2
non-external θ)
96
Scrambling and the Survive Principle
The lack of an isomorphic trigger for middle field Scrambling in West Germanic is not a problem for a minimalist analysis within Surviveminimalism. The motivation for the iterative application of Merge and Remerge is to unite lexical items as one syntactic object recognized and interpreted by the external interfaces. This is achieved through Link!, an operation driven by feature interpretation. Thus the characteristics of scrambled constituents not being consistent with an isomorphic reflex that represents A- or A’-movement traits is acceptable, and perhaps even expected. Although Scrambling involves the interplay of multiple grammatical functions and features, I put forward and support the hypothesis that middle field Scrambling is driven by the discourse/semantic feature [+ Ref]. In this chapter I will illustrate and defend the strength of this assertion within the revised form of minimalist theory adopted in this study known as Survive-minimalism. Contra Putnam (2006a), who also supports the analysis of pragmatic referentiality as the most suitable syntactic trigger for middle field Scrambling in West Germanic, I show how easily Survive-minimalism can integrate the [+ Ref] feature in displacement typology without resorting to XP-adjunction.1 As a result of this study it can be demonstrated that calling upon XP-adjunction for pragmatic considerations constitutes a separate, isolated displacement type outside of the primitive Merge and Remerge set. One of the more attractive features of Survive-minimalism is that every single iterative application of Merge and Remerge generates fully interpretable structures at the external interfaces. An operation such as XP-adjunction, at least as it currently stands in the theory as a phrase structure-preserving mechanism that is invisible to the external interfaces yet somehow takes place in the course of syntactic derivation, shouldn’t exist in the MP. In this chapter, I support this claim with empirical and conceptual arguments. The challenge at hand, of course, is to address some of the more apparent issues that are the result of such bold assumptions. In this chapter, I seek to explain some of the more complicated issues associated with middle field Scrambling in West Germanic within the framework of Survive-minimalism. The first task I will take up is that of coming to terms with the definition of the exact nature of ‘syntactic’ features. Based on conceptual arguments as well as empirical data from English, Old Norse, German and Bantu languages (i.e., Xhosa), I present evidence in favor of classifying [+ Ref] as a bona fide syntactic feature. Essentially, the form of argumentation to be applied here begins with an empirical observation of the consistent appearance of a 1
Putnam (2006a) asserts that referentiality is at play in permitting middle field Scrambling in West Germanic; however, he does not label referentiality as a formal feature but only merely as a grammatical property. By labeling referentiality as a grammatical property, Putnam (2006a) fails to mention that [referentiality] can and does adhere to minimalist restrictions placed on constituent displacement such as Last Resort, Procrastinate, and Greed.
Chapter 3. Theoretical Considerations
97
particular lexical item or set of lexical items in an environment where there can also be distinguished a consistently correlated grammatical function or a consistently correlated set of grammatical properties. Here I will make the argument for the presence of referentiality in the middle field and for its legitimate status as a syntactic category. Secondly, I will demonstrate how my analysis of middle field Scrambling in West Germanic within Surviveminimalism has the ability to explain the anaphoric binding properties of scrambled objects in German. This also entails discussing the contextual properties of derivational units beyond the Merge/Remerge operation. Based on Putnam (2005, 2006a), I show how the anaphoric binding properties of scrambled objects in German can be accounted for by adopting Grohmann’s (2000a, 2003) notion of Prolific Domains and his Anti-Locality Hypothesis. The Anti-Locality Hypothesis will also provide a way and means to explain remnant movement and its relationship to Scrambling in West Germanic. I will also embark to discuss further implications that the notion of contextuallydefined larger derivational units such as Prolific Domains may have in Survive-minimalism. Third, I will revisit some of the more recent ideas surrounding adjunction in minimalist theory, with a particular focus on Fox’s (2003, 2004) analysis of Condition C asymmetries, and show that the unwanted ‘peek-a-boo’ effects of adjunction, i.e. its ability to be invisible and visible at different stages of the derivation, can be obviated through Survive (cf. Stroik and Putnam 2007). The shows further how the notion of adjunction obfuscates the notion of displacement in minimalist syntax. Equipped with this revision of adjunction in the MP, I show how an analysis of parasitic gaps resulting from middle field Scrambling in West Germanic – a construction that has traditionally been analyzed as involving some sort of adjunction – can be explained within Survive-minimalism. Lastly, as further evidence in favor of promoting [+ Ref] to the status of a syntactic feature, I introduce a novel syntactic solution explaining why English and Pennsylvania German lack middle field Scrambling based on the elimination (through syntactic change) of [+ Ref] in the middle field of these two aforementioned languages vs. German’s maintenance of marking [referentiality] as a syntactic feature licensed as a syntactic category (i.e., generated by Link!) in its middle field. 3.1
Syntactic Featurehood
What constitutes a grammatical property as a ‘syntactic’ feature vs. one that is deemed to be ‘non-syntactic’? Chomsky (2000:108) makes the distinction between features which he ultimately separates into two camps: core features (e.g., Case and φ-features) and peripheral features. Although the status of certain features such as [Focus] may vary as being interpreted as core or peripheral from one language to another, the key difference can be found in
98
Scrambling and the Survive Principle
the definition that the former (i.e., core features) must be valued during the derivation to ensure harmonious convergence with the external interfaces. Another way to approach this problem, as suggested by Kidwai (2000) and Putnam (2006a), is to view this distinction not between core and peripheral features, but between free and bound features. As argued by Putnam (2006a: 161), feature assignment of this nature occurs in the Numeration prior to the construction of any phrase structure in the syntax. Certain features such as agreement features (e.g., φ and Case) are inherently bound to the lexical items that they modify. Call them bound features. Kidwai (2000:121) notes that certain features exist that are not intrinsically related to any lexical items. She labels these features free features. (3.3)
Numeration: Merger of Free Features (Kidwai 2000:121) A free feature must be merged with a host for convergence, where α is a free feature if it is not intrinsically related to any lexical category.
This explains why different objects other than DPs can scramble: the feature [+ Ref] in West Germanic can be assigned to XPs other than DPs. The matching feature (in this case [+ Ref]) is then assigned to a maximal projection during the course of the derivation. The distinction between bound and free features is also easily integrated into Survive-minimalism which views the Numeration as an active component of the CHL and not just an idle lexical array containing items and features. However, based on this argument, do we have enough evidence yet to declare that [+ Ref] is a ‘syntactic’ feature? Another hurdle we must overcome is the lack of any sort of morphophonemic inflection on scrambled vs. nonscrambled constituents. Consider the following German examples: (3.4)
Ich habe gestern den Mann gesehen. I have yesterday the man seen
(3.5)
Ich habe den Manni gestern ti gesehen. I have the man yesterday seen ‘I saw the man yesterday.’
Although inflected for accusative Case, the direct object den Mann ‘the man’ that participates in Scrambling in (3.5) does not bear any morphological marker that distinguishes it from the object that remains in situ in (3.4). In order to determine whether or not a certain feature can be defined as a syntactic, the argument that morphological inflection is an indicator of featurehood rests on shaky ground. One needn’t look any further than English to prove this point. When one considers the lack of inflections in both the
Chapter 3. Theoretical Considerations
99
verbal and nominal morphology of English, it is difficult to argue that nouns are without morphological Case and that modal verbs do not possess φ-features due to a lack of inflection.2 The key definition for syntactic featurehood can be reduced to a feature’s ability to exhibit both LF and PF reflexes. (3.6)
Full Interpretation Revised A grammatical characteristic α is deemed a formal ‘syntactic’ feature iff it has reflexes at both the LF and PF interfaces.
According to the rubric established in (3.6), it is not possible to reduce Scrambling to an operation applying only at PF (i.e., post-syntactically) due to the fact that Scrambling displays both PF and LF (e.g., anaphoric binding relations) in the derivation. The arguments put forth in this chapter will supply evidence in defense of cataloguing [+ Ref] as a free syntactic feature assigned to concatenated lexical items in the Numeration. In close, I follow Åfarli (1995) in assuming that what counts as evidence for the presence of a specific functional projection in a given clause type is a visible effect – a functional morpheme, a morphological effect or a change in word order. Before moving on something more must be said regarding the Survive Principle in its current formulation. As it now stands the Survive Principle provides a straightforward account for how syntactic features are assigned to simple lexical items, but how do we account for complex XPs that also scramble? Stroik (2007) and Stroik and Putnam (2007) recognize this discrepancy in the most recent formulation of Survive. Stroik recognizes that the CHL involves more than just a Numeration N that feeds a Derivation D. Rather, if a Derivation D is restricted to the computation of an event (expressed in the V-v projections) and its temporal and speech event coordinates (expressed in the T-C projections), then D must be built from two different sources of syntactic objects – the Numeration N and a workspace (WS) not contained in D. This can be seen in a simple example such as (3.7). (3.7)
Diese Frau liebt mich. this woman loves me ‘This woman loves me.’
In (3.7), although phrases like liebt mich ‘loves me’ can be formed by taking two lexical items LIs from the Numeration – liebt and mich – and merging them, it is not possible to merge the two LIs diese ‘that’ and Frau ‘woman’ 2
Although I readily admit that morphophonemic reflexes of concatenated features are in most cases a sign of syntactic featurehood, I merely wish to point out the crucial fact that just because it is not present does not bar a given feature from achieving the status of ‘syntactic’ in a given language.
100
Scrambling and the Survive Principle
directly in the Derivation from the Numeration. Instead, the DP diese Frau ‘this woman’ will have to be constructed elsewhere in a separate workspace (WS) prior to being merged in D. The Merge-domain for D, then, is the union of the Numeration with the WS. Stroik (2007) defines this union as the WorkBench (WB) for D, the ‘space’ that includes all the materials used in constructing D. Since the Merge-domain, and consequently the Survivedomain, for D is not restricted to the Numeration, we have to revise the Survive Principle accordingly, as in (3.8). (3.8)
The Revised Survive Principle If Y is a syntactic object (SO) in an XP headed by X, and Y has an unchecked feature <+F> which is incompatible with X, Y remains active in the WorkBench (WB).
This extended sense of the Survive-domain, as we shall see throughout this chapter, will play a prominent role in explaining some of the more puzzling properties of middle field Scrambling in West Germanic. 3.1.1
Multiple wh-movement and Relative Clauses in English (cf. Stroik 2000)
Up to this point, the definition for referentiality that I have made exclusive use of in this study is seriously constrained by pragmatic and semantic considerations. Following Cinque’s (1990) notion of referentiality: A referential element is one which receives a referential (argument) θ-role (as defined by Rizzi (1991)) and which can be linked to a pronoun that it does not c-command, Stroik (2000:84) introduces a way and means to understand the functionality of referentiality in a syntactic sense. Here’s an example: (3.9)
*Ich kaufte keinen Apfeli. Daher konnte ich deni nicht essen. I bought no apple. Therefore could I it not eat ‘I didn’t buy any apple. Therefore I could not eat it.’
(3.10) Ich kaufte jeden Apfel. Ich konnte sie essen. I bought every apple. I could them eat ‘I bought every apple. I could eat them.’ The quantifier keinen ‘no’ in (3.9) is non-referential, i.e. bears a [- Ref] feature, and cannot be co-indexed with a pronoun, in this case den ‘it’, that it does not c-command.3 On the other hand, the quantifier in (3.10) jeden ‘every’ can be 3
Based on Stroik and Putnam (2005b), perhaps a more appropriate alternative to a [- Ref] feature would be a feature denoting the attributive [attr] interpretation of non-referential XPs.
Chapter 3. Theoretical Considerations
101
interpreted due to its [+ Ref] feature and can be co-referenced with a pronoun that it does not c-command. The reality of the [+ Ref] feature is further demonstrated in multiple-wh construction in English. As an extension and improvement upon his earlier work (1992, 1995), Stroik (2000) argues that in English not all of the wh-elements in a multiple-wh construction are operators free to be assigned semantic referential values. Under such an analysis, the whitems in (3.11) are interpreted as an ordered pair <who, what> and in this particular pair <who> can be freely assigned any referential value (hence my declaration that this [+ Ref] is a free feature assigned in the Numeration), while the referential value of <what> can be determined/assigned only after a value for <who> has been selected (data in this section is taken from Stroik 2000:63). (3.11) Who bought what? Stroik points out that the relationship between the ordered pair of wh-items in (3.11), namely, <who, what>, reflects the same ordered-pair relationship that exists between the variable x and y in a mathematical function y = F(x). Explaining this in terms of range and domain of function, the range value of y is determined by the domain value affixed to x. What this means for the ordered pair of wh-items in (3.11) is that the semantic value assigned to the wh-item that remains in situ is set by the value assigned to the wh-operator. To support this claim about the reality of the [+ Ref] feature in multiple-wh construction in English, Stroik asserts that if a wh-operator is assigned a [- Ref] feature, the wh-item remain in situ must also bear the same feature due to the fact that the semantic range of this in situ wh-element (y) is fixed by the domain of the operator (x). Consider the following potential responses to the question in (3.11) above: (3.12) a.
No one bought anything.
b.
Everyone bought nothing.
c.
*No one bought everything.
Examples (3.12a) and (3.12b) are both acceptable responses to the question in (3.11): Response (3.12a) is acceptable because the wh-operator has been assigned a [+ Ref] feature, hence the in situ wh-item can be assigned any range value. Likewise, in (3.12b) is acceptable due to the fact that neither the operator nor the in situ wh-item have been assigned a [- Ref] feature. Example (3.12c), however, is an unacceptable response to question (3.11) based on the assumption that the wh-item in situ is assigned a [+ Ref] value while the wh-
102
Scrambling and the Survive Principle
operator does not possess such a feature. Consonant with Stroik’s analysis, in (3.12c) the range of the variable y exceeds its predetermined functional boundary and does not respect the boundaries established by the domain value present in x.4 As first mentioned in Section 2.8, Stroik provides further evidence for the theoretical reality of the [+ Ref] feature. To test the existence of the [+ Ref] feature, Stroik (2000:77) argues that although non-restrictive relative clauses can modify referential DPs, they can never modify non-referential DPs (these data originally appear as (2.118) – (2.120) but are repeated below). (3.13)
Sam, whom I like a great deal, hates me.
(3.14)
Everyone here, whom I like a great deal, hates me.
(3.15)
*No one, whom I like a great deal, hates me.
Since scrambled XPs in West Germanic are referential in nature, we should expect them to be able to be modified by non-restrictive relative clauses. This assumption is confirmed by the following data (thanks to Kleanthes Grohmann and Winnie Lechner for their assistance here). Quantified DPs with nonrestrictive relatives (such as ‘someone, whom I met yesterday’) can scramble in German (again, examples (2.121) and (2.122) are repeated below). (3.16) Maria hat [jemanden, den ich neulich kennengelernt habe]i ihrem Freund ti vorgestellt. MariaNOM has someoneACC whomACC INOM recently got to know have her friendDAT introduced ‘Maria introduced someone, who I recently got to know, to her friend.’ (3.17) Wahrscheinlich würden [jemanden, den sie über das Internet kennengelernt haben]i nur Frauen mit gewisser Neigung ti heiraten wollen. probably would someoneACC whomACC theyNOM over the Internet got to know have only womenNOM with real affection marry would ‘Probably one woman would marry someone with real intent, that they met over the Internet.’
4
Stroik (2000: Ch. 3) also develops a similar system to account for the scope properties of multiple quantifiers in English. The notion of an ordered pair in providing a theoretical account for these phenomena avoids the unwanted problem of having the final representational order resulting from successive applications of Merge and Remerge boiling down to how items are ordered in the Numeration prior to their participation in the derivation. (This possible issue was first brought to my attention by Peter Sells (p.c.)). By appealing to an ordered pair of multiple quantifiers as well as multiple wh-items, varying interpretations and word orders are still effectively determined by feature assignment and the Survive Principle.
Chapter 3. Theoretical Considerations
103
In both (3.16) and (3.17) above, the quantified DP jemanden ‘someone’ scrambles from its base position in the VP and can be modified by a nonrestrictive relative clause. Following Stroik’s assertion that only referential DPs can be modified by non-restrictive relative clauses also holds for XPs that are scrambled into the middle field in West Germanic. Coupled with Haider’s examples mentioned in Section 2.8, these facts provide a solid foundation in support of the existence of [+ Ref] feature and its relevance for this analysis of middle field Scrambling in West Germanic. 3.1.2
Xhosa
Xhosa, a Bantu language (Nguni subfamily) with an underlying SVO word order structure that also displays agglutinating morphology, makes use of the grammatical concept of referentiality in syntactic structures. Hendriske (1978) considers the morphosyntactic properties of DPs in Xhosa that are related to the presupposition of existence. In the environment of so-called ‘axiomatic negative predicates’, an object noun is non-referential. That is, a speaker does not presuppose the existence of the object to which a noun in the object position of an axiomatic negative predicate refers. In Xhosa such a noun drops its pre-prefix/initial vowel (data from Hendriske 1978:108-9). (3.18) Indoda ayisengi nkomo the man non-he-milks (any) cow ‘The man does not milk any cow at all.’ (3.19) *Indoda ayisengi inkomo the man non-he-milks (any) cow
The negative predicate ayisengi ‘not-he-milks’ entails a non-referential interpretation of the object nkomo ‘cow’. The inflectional prefix i on the object nkomo ‘cow’ in (3.19) indicates a referential entity, i.e. one that is presupposed to exist. In constructions with axiomatic negative predicates, the prefix denoting referentiality – brought forth through the derivational activation of [+ Ref] on the nkomo ‘cow’ through Link! – can not occur. Similar to the mathematical equation appealed to by Stroik in the previous section in explaining the range and domain of the nature of ordered pairs involving multiple wh-items and multiple quantifiers in English (i.e., y = F(x)), the negative morpheme in Xhosa determines the value of the pronoun, which obligatorily must be [- Ref] to preserve the derivation from stalling. This fact is further reinforced by the following data: Pronouns, which are obligatorily nonreferential as shown in the previous data set, also cannot co-occur with an
104
Scrambling and the Survive Principle
object noun following an axiomatic negative predicate (data from Hendriske 1978:109). (3.20) *Andiyiboni ndoda (I don’t see him no man)
(3.21) *Andiboni ndoda yona (I don’t see any man him)
Axiomatic negative predicates disallow the presence of referential items, i.e. those bearing the [+ Ref] feature and its subsequent morphological markings. In the surface structure subject position of the language, however, referential units are required in order to ensure grammatical outcomes (data from Hendriske 1978:110). (3.22) Indoda iyaphangela the man he-works ‘The man works.’
(3.23) *Ndoda iyaphangela [Any (X) man works] Lastly, consider the following comparison of active and passive voice structures in Xhosa (data from Hendriske 1978:110). Active (3.24) Ixhego aliyithengi imoto the old-man not-he-it-buys the car ‘The old man does not buy the car.’
Passive
lixhego (3.25) Imoto aylithengwa the car not-it-(is)-bought it-is-the-old-man ‘The car is not being bought by the old man.’ Active (3.26) Ixhego alithengi moto the old-man not-he-buys (any) car ‘The old man does not buy any car.’
Passive (3.27) *Moto ayithengwa lixhego not any car is being bought by the old man
Chapter 3. Theoretical Considerations
105
As indicated in (3.24) above, the object DP imoto ‘the car’ is referential as marked by the occurrence of the initial vowel i- and the object pronoun -yi-. The noun can be displaced to the surface subject position as shown in (3.25). Again, the object moto in (3.26) and (3.27) occur in the passive structure (3.25) because it follows an axiomatic negative predicate. Although a full-scale analysis of the morphosyntactic properties of Xhosa falls far outside the scope of our current investigation of middle field Scrambling, the data presented in this section suffices to supply further empirical evidence in support of the syntactic nature of the referentiality [+ Ref] feature in universal grammar. Based on our re-categorization of syntactic features either being bound or free features, a substantial argument can be made for promoting [+ Ref] to the status of a bound feature on DPs in Xhosa due to the morphophonemic reflex that accompanies the activation of this feature. 3.1.3
DP Structure of Old Norse
A final piece of evidence in support of the syntactic nature of [+ Ref] comes from Faarlund’s (2004) description of the noun phrase in Old Norse. Of particular worth for our current study is his discussion of the definite article and its relation to referentiality. The data cited in this section come directly from Section 4.1 of Faarlund’s (2004) textbook unless otherwise noted. In Old Norse, the definite article has two forms: It can appear either as an independent word, hinn or inn, or as a clitic, -inn. Both forms are inflected for number, gender, and case, agreeing with the head noun which they modify. Turning our attention to the independent form, the definite article is used in combination with an adjective, and precedes both the adjective and the noun. It may also be used when the noun is modified by a clause, as is the case in (3.28c). (3.28) a.
it fyrra sumar theNEU.ACC formerDEF summerACC ‘last summer/the summer before’ (Hkr II.281.11)
b.
en þo var hann hit mesta afarmenni but still was he theNEU.NOM greatestDEF outstanding-manNOM ‘But still he was a most outstanding man’ (Gunnl 1.8)
c.
en ek hefi hina skilning at engi sé þeira but I have theFEM.ACC opinionACC that nobodyMASC.NOM be their equalNOM ‘But it is my opinion that nobody is their equal’ (Nj 209.6)
106
Scrambling and the Survive Principle
Faarlund proposes that the definite article in Old Norse is the head of a functional projection which he labels the Reference Phrase (RP), since it is where the referential properties of the NP are determined. The phrase structure relation of (3.28a) between the definite article and the NP it modifies is presented in (3.29).5,6 (3.29)
RP 2 hit R’ 2 NP Ro 5 fyrra sumar
At first blush, the RP appears to strongly resemble the standard notion of DP (following Abney 1987); however, Faarlund provides evidence that reserves the notion of DP for still a higher projection above RP. In his analysis of nounphrase complements (cf. Section 5.1.1), Faarlund asserts that the demonstrative article is often followed by an NP (or RP), which is the complement of the determiner. As expected, the determiner agrees with the head noun of the following NP (data from Faarlund 2004:82). (3.30) a.
b.
sá orninn thatMASC.NOM eagleNOM-the ‘that eagle’ (Gunnl 4.16) þau in stóru skip thoseNEU.NOM the bigDEF shipsNOM ‘those big ships’ (Hkr I.437.13)
In the phrase sá orninn ‘that eagle’ in (3.30a), the demonstrative sá ‘that’ occurs before the R-position, where the noun orn ‘eagle’ has moved in order to acquire the definite article –inn ‘the’ by means of noun incorporation as illustrated in (3.31). 5
The phrase structure treelet in example in (197) is an adjusted version of Faarlund’s example (57) in Section 4.1.1 of his book that still makes use and reference of GB-theory terminology such as D- and S-structure. 6 For an alternative minimalist account of similar structures regarding the ontology of the DP in Germanic (in particular, Scandinavian languages and dialects), see Julien (2002) and Roehrs (2006).
Chapter 3. Theoretical Considerations
107
(3.31)
DP 2 sá RP 2 orni + inn NP 4 ti
In similar fashion, in example (3.30b) the neuter plural of the same demonstrative occurs before the independent definite article, which again means that it is in a position above R. (3.32)
DP 2 þau RP 2 in NP 2 stóru skip
Under this view, the specificer of the DP, that is the first position in the phrase, is a landing site for fronting, where elements from further down in the NP can be moved to and thus precede the demonstrative as in (3.33) (data from Faarlund 2004:84). (3.33) a.
b.
fé þat allt moneyACC that all ‘all that money’ (Eg 232.9) er nokkurr sá maðr á þingi, er þat kunni at segja is some that manNOM on assemblyDAT who thatACC can to tell ‘Is there any man at the assembly who can tell (us) that?’ (Hkr II.304.14)
Based on the RP-structure of the DP suggested by Faarlund (cf. (3.29)), in (3.33a) the demonstrative þat ‘that’ resides in the head Do position, allowing the NP fé ‘money’ to rise to Spec,DP, as illustrated by the derivational history in (3.34). (3.34) [DP féi D[þat] NP[allt ti]] As for the function of RP in his analysis, Faarlund (2004:57) suggests that the RP “is thus the nominal equivalent of the Inflectional Phrase (IP) in the verbal
108
Scrambling and the Survive Principle
projection.” Interestingly, however, Faarlund does not discuss how agreement epiphenomena (e.g., Case and φ-features) are accounted for in the syntax. No mention of functional projections with the specific intention of licensing agreement relations (i.e., AgrP) is included in Faarlund’s work. Although there may be a potential link to referentiality and robust Case morphology in diachronic Germanic languages (see Section 3.4), it is important to make a distinction between agreement licensing and referentiality based on our discussion of the sufficient lack of traditional A-movement characteristics involved in middle field Scrambling.7 Although the linguistic evidence provided in this section suggests the existence of a syntactic feature [+ Ref], i.e. a grammatical property with both LF and PF reflexes, is it possible to ascertain that such a feature also exists in the DP structure of other synchronic West Germanic languages? An anonymous reviewer points out that internal DP-Scrambling is not possible, and is most likely bound to the distinction between head-final and head-initial structures (data from Haider 2005b). (3.35) a. b. c.
[Mails an Kunden versenden] – [an Kunden Mails versenden] Mails to costumers send to costumers mails send das [Versenden von Mails an Kunden] - *das [Versenden an Kunden von Mails] das [Versenden der Mails an Kunden] – *das [Versenden an Kunden der Mails]
According to this reviewer, if a [+ Ref] feature exists in the German middle field and is the syntactic feature responsible for licensing Scrambling, we should expect this feature to also exist in the German DP. Such an assumption is faulty on both empirical and conceptual grounds. First addressing the empirical weaknesses of such an analysis, the data set in (3.35) – mentioned to me by the reviewer – is a poor choice to test the existence of [+ Ref] in the German DP. Each of these constructions exhibits genitive case, which requires strict adjacency in German in order for the NP to assign genitive case. Therefore, if the data in (3.35) were to involve any sort of Scrambling operation we would have to envision two movements (or two separate applications of Remerge): One which would be the NP-internal scrambling of the PP and the second involving the A-movement of the lower DP across the higher DP to check genitive case. The conceptual flaws in arguing that the analysis that middle field Scrambling is not driven by [+ Ref] based on the data set in (3.35) are as follows: First, it is not clear if the structure of DPs and 7
Of potential interest, in his defense of the maintenance of AgrP in syntactic theory, Adger (1994) suggests that the notion of ‘familiarity’ is closely associated with AgrP.
Chapter 3. Theoretical Considerations
109
natural clauses (CP) are completely symmetrical in structure and function (cf. Ticio 2003, Julien 2005 and Roehrs 2006 to mention a few). Second, if we are testing the potential existence of a [+ Ref] feature based on the data set above in (3.35), we would have to assume that such a feature is extremely low in the DP structure; as a matter of fact, lower than the adjective – which could also be taken as an adverb in the clause (cf. Roehrs 2006). In the end, the data and argumentation provided against a [+ Ref] feature can be easily dismissed. Assuming an internal structure of DPs based on Roehrs (2006:26) listed below in (3.36), the argument can be made that the internal structure of the German DP can include referentiality. (3.36) [DP Do [CardP QP Cardo [AgrP AP Agro [artP arto [NumP Numo [nP Poss no [NP No (PP)]]]]]]] The central claim of Roehrs’ analysis of the DP structure above is that articles (i.e., demonstratives and (in-)definite articles) are born in a functional projection which he labels artP (Julien’s (2005) αP) – a functional projection above the theta-domain of nouns and below adjectives – and move to DP in order to value features on D. Staying with the structure of DP put forward by Roehrs in (3.36) and remaining consonant with his analysis of the generation of articles in syntactic derivation, it is possible to provide a viable analysis of problematic data from an anonymous reviewer. (3.37) a.
b.
ein [den meisten an Kenntnissen überlegener] Kandidat a the most on proficienciesDAT regarded candidate ‘a candidate regarded to be the most proficient’ ein [an Kenntnissen den meisten überlegener] Kandidat
These data are relatively easy to handle: Again, let us assume with Cinque (1999) and most others these days that the adjective is an AP in the Spec of AgrP. Now, the elements in (3.37) are arguments of the adjective (and not the noun). Therefore, this instance of Scrambling has nothing to do with the DP per se but the AP inside Spec,AgrP. In other words, APs – just like CPs – have a referential feature in the middle field of the extended projection of the DP (for some discussion of the extended AP, see Zamparelli (2000)).8 In the end, 8
Dorian Roehrs (p.c.) points out that the fact that referentiality (or something close to it) also appears in natural clauses (i.e., the middle field) is not problematic for my analysis. Since it is one and the same DP that is referential (but two instances of Ref-checking), perhaps one could argue that it (Ref) is an +interpretable feature that gets checked inside the DP (deriving its referentiality proper) and then triggering movement/displacement in the clause. Since Ref is on D, there is no internal DP-Scrambling, but the feature is accessible to clausal Scrambling. Such
110
Scrambling and the Survive Principle
the argument that the lack of internal DP-Scrambling is solely based on linearization principles (i.e., VO vs. OV) has been shown to be untenable under closer scrutiny. As will be discussed in the remainder of this chapter, I show that the [+ Ref] feature plays an active role outside of the vP in the West Germanic clause. In the analysis that follows, I will represent the concatenation and linkage of the functional head bearing the [+ Ref] feature and the Remerger into the middle field of the constituent also bearing this matching feature by the mnemonic RefP9 to remain consistent with minimalist principles dealing with phrase structure rules. 3.2
Prolific Domains and their Potential Role in Survive-minimalism
In the similitude of Epstein and Seely’s (1998, 2002) dismissal of phases and other rich ontological commitments in the course of derivational syntax by means of their ‘level-free’ approach, Survive-minimalism also does away with such units a priori based on the fact that they are, once again in the words of Brody, weakly representational. With that being said, some sort of structural spine is necessary to determine the timing of when the operations Structural/Functional Merge and Lexical (Re)Merge are to take place in the narrow syntax and what particular features are responsible for triggering each successive application of concatenation. The thoroughly researched structure of natural clauses throughout the history of transformational grammar has been shown to stand the proverbial test of time. In Survive-minimalism, I, too, make use of the three functional heads of v, T and C as representing benchmarks in the derivational history of computation. The fact that the logical argument structure of the verb must be accounted for first is easily illustrated in the following examples. (3.38) **Bob saw the dog the piano. (3.39) *Him looks nice. Clearly both sentences are ungrammatical in English, however, native speakers will remark that (3.38) is markedly worse than (3.39). Why is that the case? If we assume that the items housed under the functional v-head are responsible for thematic relations (and potentially other things, too), this contrast is easily explained. The utterance in (3.38) is considerably worse due to the fact that the an analysis fits perfectly within Survive-minimalism: All features – including + Ref – remain +interpretable throughout the entire derivation and will Merge and Remerge (Link!) when required. 9 Also see Kiss (1996) and Mohr (2005) for arguments and data that support the appearance of RefP in the Germanic middle field.
Chapter 3. Theoretical Considerations
111
argument structure of saw is super-saturated: It can only license one external and one internal argument. The fact that two internal arguments exist means that one of them does not Merge into the derivation to link its θ-role to the interfaces. The sentence will stall before the vP is fully completed. In (3.39), all of the θ-roles are completed, however, there is a Case error: The subject of the utterance him has erroneously received accusative rather than nominative Case. This error is less costly due to the fact that at least one successful application of Link! has taken place in the course of the derivation. Making use of the C >> T >> v functional hierarchy, Survive-minimalism makes use of a functional, tripartite structure similar to Grohmann’s (2000a, 2003) postulation of Prolific Domains. A key, fundamental difference however remains preventing full acceptance of Prolific Domains into Surviveminimalism: Viewing Prolific Domains as larger derivational chunks that are ‘independent’ at the interface due to the iterative application of multiple SpellOut (cf. Uriagereka 1999); Survive-minimalism interprets the immediate domains surrounding these benchmark functional heads as merely contextually-specified for particular predetermined relations and functions. Nonetheless, there exists a number of significant similarities between Prolific Domains and the idea of dynamic computation as understood within Surviveminimalism that warrant an in depth discussion of the properties and functionality of Prolific Domains in this framework. The composition of syntactic structure through the iterative application of Merge and Remerge will generate a perfect representation within Survivesyntax and be shipped off to Spell-Out for interface interpretation. Recall that a given language possesses the set pair (π, λ) and must enter into local relationship to converge at the interfaces. But are there smaller derivational units specified for particular relations? To properly answer this question we have to consider the development of clause structure over the past forty years. Below are representations of clause structure according to EST (Chomsky 1965) and its replacement within the Barriers-framework (Chomsky 1986). (3.40) [S’ Comp [S NP (Aux/Infl) VP]] (3.41) [CP XP Co [IP XP Io [VP (XP) Vo … ]]] Towards the end of the GB-era and in early minimalism (cf. Larson 1988, Pollock 1989, Chomsky 1993 among others) the rise of functional projections within the clause structure are easily noticed. For example, the VP was enlarged to include the predicate-internal subject hypothesis (Larson 1988) and a light verb and positions specifically intended for arguments once they had moved out of their base-generated thematic positions. The structure in (3.42)
112
Scrambling and the Survive Principle
sketches the overall consensus of the composition of clause structure at this time (from Grohmann 2006:3). (3.42) [CP XP Co [AgrSP XP AgrSo [TP XP To [AgrOP XP AgrOo [vP XP vo VP ]]]]] From this point on much research was directed at mapping out the left periphery of the clause, for example, by splitting the CP-layer into independent functional projects (cf. Rizzi 1997 and subsequent work). At this point in the theory the structure presented above (3.42) was extended by an array of functional projections ((3.43) below is taken from Grohmann 2006:4).
(3.43) a. b. c.
CP = ForceP > TopP > FocP > (TopP > ) FinP IP = AgrSP > TP > NegP > AgrIOP > AgrDOP > AspP VP = vP > VP (or any other implementation of VP-shells)
Accordingly, what the proliferation of the clause along the lines of multiple functional projections hides is an intuitive tripartition of the clause. As a matter of fact, the structure in (3.43) strongly resembled derivational grammar within the GB-framework: D-structure was the stage of lexical insertion and subsequently dealt with thematic relations. S-structure accounted for agreement relations (among other things) and at LF semantic and discourse relations were established covertly. The key difference between the GB-model and the minimalist version of phrase structure above in (3.43) is the outright elimination of internal interface levels in the Minimalist Program. Most recent work in minimalism resembles a return to structure similar to the Barriers-framework (Chomsky 1986). With the introduction of multiple specifiers to the theory (Richards 1997), phrase structure has once again become minimalized: (CP > TP > vP > VP). Within the phase-based framework of minimalism (Chomsky 2000, 2001, 2004, 2005), the heads v and C are assumed to represent strong phases. This is an implementation of multiple Spell-Out (cf. Uriagareka 1999) where derivational units interact directly at certain cut-off points during the dynamic assembly of structure. (3.44) [CP XP* Co [TP XP* To [vP XP* Vo … ]]]] PHASE PHASE The structure provided above in (3.44) (from Grohmann 2006:4) is in concordance with current minimalism. As stated above, these relative cyclic units within the framework of orthodox minimalism are called phases, which are assumed to be the verb (vP) and complementizer (CP) in their strongest
Chapter 3. Theoretical Considerations
113
form and tense phases (TP) in the weakest. At these points where propositional units are completed, material at both strong phases (vP, CP) is regarded to be independent at the interfaces. Recent challenges to the concept of strong phases; however, have proven these constructions not to be optimal tools. For example, Epstein and Seely (2002) discuss at length why the specification of vP and CP as phases is potentially problematic. Chomsky (2000, 2001, 2004) attempts to motivate vP and CP as phases on the grounds that they are relatively independent at the interface levels. Although it is unclear exactly what Chomsky means by ‘relatively independent’ in this context, Epstein and Seely (2002:78) raise the following key question: “How can we know that they are relatively independent at the interface if Spell-Out applies before the interface is reached, and without access to interface properties?” Epstein and Seely (2002:78) comment further that “…it is a potential architectural paradox to hypothesize that vP and CP are spelled out cyclically, internal to the narrow syntax by virtue of them having the property being later, relatively independent at the interface.”10 Grohmann (2000a, 2003 and subsequent work) offers an alternative to a phase-base system while maintaining a multiple Spell-Out view of dynamic syntax. According to this line of thought, the ideal derivational system based on contextual information rather than mere ontological, structural commitments should include not only CP and vP as chief benchmarks of systematic evaluation, but also TP. Based on Chomsky’s (1986) earlier conceptions of clause structure, Platzack (1991) and Grohmann (2000a, 2003) elucidate the natural tripartite division of labor exerted by natural clauses, with the vP layer responsible for thematic relations, the TP layer licensing agreement and the CP-layer (cf. Rizzi 1997) available for discourse properties. Grohmann’s Prolific Domains (2000a, 2003) function as the ideal units of evaluation for the construction of syntactic structure. Derivation qua Prolific Domains not only achieves a natural functional tripartite structure of clauses, but also establishes a tripartite structure of layers relevant to contextual information: thematic context (in terms of vP/VP or separate v/V-projections), agreement context (vis-à-vis split Infl: AspP, AgrP, TP etc.), and discourse context (viz. Split-CP: TopP, FocP, CP, etc.). The individual layers responsible for thematic, agreement and discourse contextual relations are referred to as Prolific Domains (ΠΔ): (3.45) Prolific Domain (Grohmann 2003:78) Let a Prolific Domain ΠΔ be a contextually defined part of CHL: (i) Each ΠΔ spells out its context information and (ii) Spell Out feeds the PF and LF interface levels 10
For further criticisms of phases see Grohmann and Boeckx 2004, Legate 2002, Epstein and Seely 2002, Bošković 2002, Matushansky 2004, Hiraiwa 2003, Bouchard 2002.
114
Scrambling and the Survive Principle
The generation of Prolific Domains and their existence in the CHL is chiefly motivated by their intimate connection with contextual information and its direct relation with phrasal hierarchy. We can understand the notion of context information as follows: (3.46) Context Information (Grohmann 2003:78) Context Information is determined by context values: (i) |Θ| ranges over thematic relations; (ii) |Φ| ranges over agreement properties; (iii) |Ω| ranges over discourse information. Grohmann employs the mnemonics Θ, Φ and Ω to be context values, that are part of the information contained within each head. In order to stay consistent with previous work dealing within this theoretical framework, I, too, will henceforth adopt these symbols in this work. Accordingly, V and v, for example, come equipped with the context value |Θ|, and Top or C with the context |Ω|. This information is relevant in two aspects: First, it groups various projections into a single ΠΔ; second, each ΠΔ contextually identified in this way ships the information applicable for the specific context to the interfaces. This convention identifies a number of heads, and by extended projections, as part of the same context (Grohmann 2003:78). In connection with introducing the context variables in (3.46), we can rename the phrasal units to conform to their interpretable identity; accordingly, the vP-layer is properly known as the Θ-domain, the IP-layer as the Φ-domain and lastly the CP-layer as the Ωdomain. Prolific Domains depart from orthodox minimalist desiderata in the stipulation that these contextually-valued metric units serve as the hierarchical units of evaluation that are cyclically assessed and shipped off to Spell-Out (although as stated previously, Survive-minimalism refutes the notion of multiple Spell-Out). The clausal hierarchy established through Prolific Domains is schematized in (3.47) (adapted from Grohmann 2003).
Chapter 3. Theoretical Considerations
115
(3.47)
CP Contextual information: 2 inherent index |Θ|, |Φ|, |Ω| on lexical Spec C' and functional heads 2 TopP C0 2 ΩΔ Spec Top' 2 Top0 FocP 2 Spec Foc' 2 TP Foc0 2 Spec T' 2 T0 AgrOP 2 Spec AgrO' 2 AgrO0 vP ΦΔ 2 Spec v' 2 VP v0
ΘΔ From a purely theoretical standpoint, Prolific Domains provides us with a metric by which we can approach former GB-principles and integrate them into minimalist inquiry. Rather than appealing to some sort of structural filters, derivation via Prolific Domains from their outline evoke an image of derivational syntax that focuses solely on contextual information and its representation at the interfaces, which is ultimately the same goal sought by Survive-minimalism. Prolific Domains, however, do not come without some serious baggage. As discussed by Putnam (2006a) in the latter half of his dissertation, Prolific Domains are difficult to reconcile to the functional clausal structure of West Germanic without serious adjustments to the rigid nature of Prolific Domains. Regardless of these shortcomings, some version of Prolific Domains has the potential to find a home in Survive-minimalism. 3.2.1
Anti-Locality
A chief motive in implementing Prolific Domains in the CHL is their inherent ability to evaluate the displacement of items in the syntax with regards to their locality status. Locality is conceived as a restriction on the maximum distance
116
Scrambling and the Survive Principle
that an element may move in the derivation in one cyclic step. Locality conditions are imposed on all relevant dependencies (i.e., those created by Move/Attract) and delimit the upper-bound of a well-formed dependency between two positions created by movement. Grohmann (2000a, 2003) argues in favor of a lower-bound restriction on movement requiring displacement operations to travel a minimal distance that he labels Anti-Locality. Under this guise, Anti-Locality is violated when a particular movement chain is deemed to not move ‘far enough.’ Prolific Domains serve as the ideal contextually defined units to evaluate Locality and Anti-Locality.11 (3.48) Anti-Locality (Grohmann 2000a, 2003) … [αΔ … XP … ]] (anti-local movement) [αΔ XP Yo {α = any Prolific Domain}
Movement within any given Prolific Domain would violate Anti-Locality. In the event that anti-local movement occurs in the derivation, Grohmann (2000a, 2003) proposes a repair mechanism that is implemented by the CHL to prevent the derivation from crashing. Since the goal of the Minimalist Program is to refine a system to include the least amount of exceptions and principles to achieve successful derivations, Grohmann reformulates the concept of ‘XP uniqueness’ in regards to anti-locality effects through the Condition on Domain Exclusivity (CDE). (3.49)
Condition on Domain Exclusivity (CDE) (Grohmann 2003:80) For a given Prolific Domain ΠΔ, an Object O in the phrase-marker must receive an exclusive interpretation at the interfaces, unless duplicity of O yields a drastic effect on the output of that ΠΔ.
The object O refers to any element of the phrase-marker within a ΠΔ, i.e. all substantive and functional heads and projections.12 Pending a more detailed discussion of interface interpretation, the notion understood here regarding O is that O signals its presence to the interfaces in terms of a set of instructions that LF and PF pass on to the respective interfaces (Grohmann 2003:81). The ‘drastic effect on the output’ correlates to the PF-matrix of O at the PFinterface. In most cases the higher copy will spell out at PF. This makes sense, 11
Abels (2003) also construes a version of anti-locality banning the movement of the complement of a head from merging in the specifier position of the same local head. The primitive mechanisms Merge and Remerge in Survive-minimalism naturally prevent Abels’ phrasal anti-locality. 12 This lower-bound restriction on a minimal required distance of traversal is unique only to phrasal constituents and does not apply to minimal projections (Xo). Moving a head Xo to adjoin to the next highest head Yo, results in the creation of a complex head [Xo-Yo]-Yo rather than substitution and the building of new syntactic structures.
Chapter 3. Theoretical Considerations
117
given that higher copies of an object O will possess one less uninterpretable feature than its lower copies. If, however, a multiple occurrence of O involves two phonetically distinct copies of O, the CDE is not violated.13 Therefore, the CDE can be upheld if movement within a ΠΔ takes place. The structural representation in (3.50) below illustrates the effects of Anti-Locality on movement chains qua the CDE. (3.50)
YP 2 xP (X3) 2 (X2) x’ 2 xo XP Anti-Locality 2 Violation Xo (X1)
The movement chain (X1, X2) above occurs within the maximal projection xP. Assuming that xP represents a contextually defined Prolific Domain, this intradomain movement chain (X1, X2) violates Anti-Locality. The higher link in the movement chain (X2, X3) targets a superior domain and therefore averts any potential violation of Anti-Locality. A view of syntactic composition along the lines of Prolific Domains eliminates the ‘escape hatch’ necessitated in a phase-based system reminiscent of earlier stages in generative syntax; i.e., the Barrier-framework (cf. (3.41)). The infinite proliferation of phase edges can also be eliminated by Prolific Domains, thus regaining the intimate locality effects of pure sisterhood relations advocated by Fitzpatrick (2002) and Stroik (2007). Consequently, if the scrambled XP were to move to a multiple specifier of vP, according to the Anti-Locality Hypothesis, it would incur a lower-bound movement violation. Assuming that predicate adverbials are adjoined to vP in German, the scrambled DP must leave, i.e., survive the contextually-defined domain for thematic properties (vP). This, of course, would take place to link the scrambled XP bearing a [+ Ref] feature with a functional head in the middle field with a matching feature. If the Anti-Locality Hypothesis is indeed correct, then, we should expect all scrambled XPs to leave the vP.
13 As pointed out by Putnam (2006c), Grohmann’s definition of the exact nature of O is sorely needed to account for unbounded topicalization and the multiple Case assignment that takes place in anti-local movement chains.
118
Scrambling and the Survive Principle
Focusing our attention on the lower link of the movement chain in (3.50), at this juncture of the derivation, two outcomes are possible. Due to the CDE, which will force both the higher (X2) and lower copy (X1) of the first link in this multiple movement chain to be spelled out as two separate phonological entities at the PF-interface, the system can either accept the appearance of each individual object or the CDE will subsequently cause the system the crash. It may seem intuitive to posit that violations of Anti-Locality will unambiguously cause the system to crash, there are, however, instances in which the spelling out of both copies of the movement chain (X1, X2) is required to generate fully grammatical structure interpretable at the interfaces. Hornstein (2001) puts forward the argument that anaphora are generated in the run of the derivation and are not borne in the Numeration.14 This generalization is neatly captured and explained within the framework of Anti-Locality and the CDE. Example (3.51) represents a sample application of the CDE generating anaphora. (3.51) a. b.
* Sara likes. #[vP Sara vo [VP likes (Sara)]] [AG] ([TH])
Based on the assumption that vP is the layer in which thematic contextual information is both processed and accounted for (hence, ΘΔ), we know that any XP-movement within this Prolific Domain is illegal due to the fact that it violates the Anti-Locality Hypothesis, unless it is followed by Copy Spell Out of the lower XP qua the CDE. The only way to salvage (3.51b) is through the application of the CDE. (3.52) [ΘΔ Sara v [VP likes Sara
herself]
In the derivation above (3.52), the diacritic represents the spelling-out of the lower copy of <Sara> at PF. Sara doesn’t move far enough (i.e., outside of the ΠΔ it is generated in, in this case ΘΔ) which is a clear violation of the AntiLocality Hypothesis. Here we see that it is possible, and in this particular case, expected that we have two PF-matrices within the same ΠΔ. This example suggests that this approach is indeed conceivable, in that it correctly predicts the possible ways of expressing reflexivization qua the Anti-Locality Hypothesis (cf. Hornstein 2001). Repulsion to another position within the Θdomain would imply the existence of two distinct θ-roles. This is reminiscent of Lidz’s (2001) Condition R constructed to account for the binding differences between anaphors and logophors. 14
Neither Hornstein nor Grohmann maintains a distinction between reciprocals and reflexives.
Chapter 3. Theoretical Considerations
119
(3.53) Condition R (Lidz 2001:14) λx [P(x,x)] (θ1 = θ2) semantics theta-grid The condition states that if a predicate is semantically reflexive, then it must be lexically reflexive. Similarly, if a predicate is lexically reflexive, then it must be semantically reflexive as well. A clear difference between the Anti-Locality Hypothesis and Lidz’ Condition R is that in the former the syntactic object that undergoes movement bears two different theta roles, whereas in Condition R they are identical. The desired result; namely, that the application of MergeSurvive-Remerge involving the same lexical item within a given Prolific Domain will force the lower copy of the moved item to receive prominence at PF (qua the CDE), is achieved. 3.2.2
Anaphoric Binding
Derivation by Prolific Domains and the lower-bound restriction on XPmovement enforced by their presence (Anti-Locality) offers a more unified explanation to the anaphoric binding characteristics of middle field Scrambling. Recall from Section 2.3.2 that although accusative objects that scramble outside of the DP can create new binding relations with anaphors, while dative objects cannot. Putnam (2005) delivers a syntactic analysis of this problem utilizing Prolific Domains and Anti-Locality. Focusing on the internal structure of the vP-layer/Θ-domain, similar to Larson (1988), I adopt a version of the light verb approach and take the structure of the thematic layer to consist of ‘shells’ – in particular VP (hosting the lexical verb), usually dominated by vP (or the light verb). An issue of much debate, however, is the position of the direct (THEME) and indirect objects (GOAL) within the vP, or ΘΔ. Agreeing in principle with Grohmann (2003:127), the merging of object DPs from the Numeration into the vP-layer theoretically need not occur in a set order, although I do advocate the general order of indirect (GOAL) objects being merged into a structural higher position than direct (THEME) objects. If we assume that Θ-features are the features that motivate Merge, it is possible to view the internal structure of vP/ ΘΔ as ‘unfixed’ regarding the THEME >> GOAL >> THEME ordering. In other words, the verb – being the head of the VP – possesses a Θ-feature to be assigned to the object that merges with it, thus projecting further as VP. Under such assumptions, it would not matter if the direct object or the indirect object were to merge as Comp,VP. It is conceivable that the initial merger of all arguments (Pure Merge) checks a Θfeature. The internal structure of vP/ ΘΔ is illustrated below in (3.54):
120
Scrambling and the Survive Principle
(3.54) a.
[vP Ext-Arg V-vo [VP GOAL tv THEME]]
b.
[vP Ext-Arg V-vo [VP tv GOAL]]
c.
vP 3 EA/SU v’ 3 VP vo 3 GO/IO V’ 3 6 Vo … TH/DO …
According to the structural outline of the vP-shell/Θ-domain in (3.54c) above, arguments Merge into the thematic layer of the narrow syntax in the following positions: Comp,VP, Spec,VP or Spec,vP. Based on Horstein’s assumption within the framework of Anti-Locality as conceived by Grohmann, in order for reflexives to be generated in the derivation, Survive and Remerge within the vP-layer must take place. In this section the true nature of the problem regarding why scrambled indirect objects cannot bind anaphors is revealed: indirect objects (GOAL) are inaccessible to any potentially empty nodes within the vP-shell to generate reflexives. The sentences in (3.55) and (3.58) contrast the ability of displaced direct (THEME) and indirect (GOAL) objects and their ability (or lack thereof) of entering into licit binding relations with reflexives. (3.55) …dass der Fritz die Gästei einander ti vorgestellt hat.
Chapter 3. Theoretical Considerations
(3.56)
121
CP 2 C’ 2 dass TP 2 der Fritz T’ 2 AgrOP T 2 die Gäste vP 2 v’ tFritz 2 VP hat 2 tdie Gäste V’ 2 tdie Gäste vorgestellt
The movement chain (merely used as a mechanism to chart the derivational history of the object) involving the direct object (THEME) die Gäste ‘the guests’ violates Anti-Locality by moving from Comp,VP to Spec,VP. Crucially, there is only one DP that scrambles possessing two distinct θ-roles, not two separate DPs that are extracted from the Numeration. Remerger of a maximal projection within a contextually-defined domain responsible for thematic properties violates Anti-Locality and will be interpreted by the CHL as remerger to a Θ-position, resulting in both the lower and upper copy of the displaced XP being exclusively interpreted as independent objects at the interfaces. What makes this step in the derivation legal is the generated empty position Spec,VP: The fact that no indirect object (GOAL) or its copy occupies this position makes it a desired landing site of the subject die Gäste ‘the guests’. This operation then invokes CDE effects that in turn generate the reflexive anaphor einander ‘each other’. (3.57) [ΘΔ die Gäste v [VP die Gäste
einander]
Indirect objects (GOAL); however, are not afforded the same structural opportunities to generate reflexives as are their direct object (THEME) counterparts. Consider the ungrammatical sentence in (3.58) and its subsequent derivational history in (3.59).
122
Scrambling and the Survive Principle
(3.58) *…dass der Fritz den Gästeni einander ti vorgestellt hat. ...that the FritzNOM the guestsDAT each other introduced has ‘…that Fritz introduced each other to the guests.’ (3.59)
*CP 2 C’ 2 dass TP 2 der Fritz T’ 2 AgrOP T 2 den Gästen vP 2 tFritz v’ 2 VP hat 2 tden Gästen V’ 2 vorgestellt
If indirect objects (GOAL) are base-generated (or adjoined) to Spec,VP, there does not exist any possibility for the indirect object (GOAL) to land in an open specifier in the vP-shell. The lower copy of the subject der Fritz ‘the Fritz’ remains in Spec,vP, thus preventing any attempt of the indirect object (GOAL) from generating a reflexive anaphora by means of the CDE. As established in Chapter 2 of this work, agreement features are not responsible for licensing Scrambling, therefore it is highly unlikely that Amovement of any kind is involved in middle field Scrambling.15 Lastly, by allowing reflexives and reciprocals to be generated in the derivation via AntiLocality and CDE effects the theory needn’t develop a fixed canonical basegenerated position for anaphora. Rather than arguing for separate landing sites in the middle field for accusative and dative objects that participate in Scrambling, this approach explains the different binding characteristics in a more unified approach: Scrambled dative objects can’t bind anaphors, because 15 Constituents scrambled to the middle field in German can be iterative; both crossing and nesting paths are possible constructions. This is a crushing blow to any theory interpreting middle-field scrambling as A-movement based on Chomsky’s (1995) assertion that movement from a position responsible for Case to a higher position responsible for the same grammatical purpose is illegal.
Chapter 3. Theoretical Considerations
123
they can’t create them. Critically, this appears only to be an issue for scrambled dative objects and their ability to bind reflexives. (3.60) …dass der Jörg jedem [seinen Vater] gestern [t gezeigt hat]. …that the JörgNOM everyoneDAT his fatherACC yesterday shown has ‘…that Jörg has shown his(x) father to everyone(x) yesterday.’ Example (3.60) demonstrates that a scrambled dative object jedem ‘everyone’ can bind an accusative object seinen Vater ‘his father’. The following sentence (3.60) creates a problem for Kidwai’s (2000:108) re-analysis of binding theory, which claims that indirect objects can never serve as licit binders due to their adjoined position at the end of the derivation. Kidwai’s revision of the binding theory rests upon archaic constructions such as chains, L-marking, binding domains, etc. Admittedly, although a comprehensive treatment of binding theory and its properties in a Survive-minimalist framework is currently lacking, the sketch of a formal approach outlined here in this section does not need to call upon movement chains to determine anaphoric binding. The inability for indirect (dative) objects to bind anphora is determined by the initial Merge position of arguments in the vP (i.e., Θ-domain) (cf. Baker 1988). It is thus possible to view the failure of licit binding relations between indirect objects and anaphora as a syntactic problem. Furthermore, the data and analysis presented here make a case for contextually-defined derivational units similar in design and function to Grohmann’s (2000a, 2003) Prolific Domains. 3.2.3
Remnant Movement (Anti-Freezing Effects)
Another aspect of middle field XP-Scrambling in West Germanic that appears to adhere to the hypothesis of Anti-Locality and subsequently displays CDE effects is instances of remnant movement. Recall that remnant movement is movement of an XP β from which extraction of α has taken place, (cf. (3.61)). (3.61) [β … ti … ]2 … [ …α1 … [ … t2 …]] As discussed in the previous chapter, remnant movement, i.e., extraction of an XP from a unit that has already itself undergone movement from its base position, must not have both items that resided in the original structure occupying the same field after substitution (movement) has taken place. kaufen]2 habe ich das Brot1 der Abby versprochen t2. (3.62) [PRO t1 zu to buy have INOM the breadACC the AbbyDAT promised ‘I promised Abby to buy bread.’
124
Scrambling and the Survive Principle
(3.63) *Gestern habe ich [PRO t1 zu kaufen]2 das Brot1 der Abby t2 versprochen. yesterday have INOM to buy the breadACC the AbbyDAT promised ‘Yesterday I promised Abby to buy bread.’ First, a derivational account that obeys the strict cycle is necessary. Müller (1998) explains that the scrambled XP, i.e., the one that appears in the middle field upon the completion of the clause, is extracted from the clause that will eventually be topicalized. The derivational history of (3.61) is sketched out in (3.64) below. (3.64) Step 1: [CP … [TP … XP1 … [ vP [YP … t1 …]]]] Step 2: [CP … YP2 … [TP …XP1 … [vP …t2 …]]] As indicated by example (3.63), if both the scrambled and topicalized XPs move into the middle field, this operation results in a crash in the system as illustrated in (3.65). (3.65) *[CP … [TP [YP … t1 …]2 XP1 [vP …t2 …]]] The derivational history in (3.65) looks suspiciously similar to previous examples of Anti-Locality. As laid out in the previous chapter, topicalization and Scrambling are not identical operations, chiefly because they target separate landing sites in the clause, as evidenced once again in (3.65). AntiLocality effects are the product of the same XP illegally surviving and remerging with a functional head within the same Prolific Domain. The first XP (XP1) – created in the WorkBench (WB) and then merged into the Derivation D – scrambles from YP bearing the [+ Ref] feature. This, of course, stands in complementary distribution with a [Topic] feature that is checked in the CP-layer, i.e., Ω-domain. Since the remaining infinitival phrase no longer contains the constituents that needed to scramble, its movement into the middle field is unmotivated, especially if we maintain that the middle field is contextually defined as the Prolific Domain specifically responsible for agreement checking. This, as pointed out by Putnam (2006a), is a major hurdle that Prolific Domains has yet to overcome, namely, figuring out a way and means to allow the licensing of referentiality and other discourse properties in the middle field.16 On the other hand, the Anti-Locality Hypothesis provides us 16 Jayaseelan (2001) and Mohr (2005) advocate the presence of discourse-oriented functional projects (e.g., Topic and Focus) in the Germanic middle field. Accordingly, scrambled XPs move into the middle field to check discourse features. Mohr (following Kiss (1996)) also supports the presence of a referential phrase (RefP) in the middle field for subjects. The analysis developed in this study partially agrees with both Jayaseelan (2001) and Mohr (2005);
Chapter 3. Theoretical Considerations
125
with a unified mechanism to explain why scrambled XPs in German can enter into anaphoric binding relations while simultaneously displaying a strong preference for A’-movement characteristics as well as account for remnant movement effects brought forth via Scrambling. With this distinction between [Topic] and [+ Ref] features and the different landing sites in the clause that they target in mind, this can help account for the following data set (provided by an anonymous reviewer based on (3.62) and (3.63) above). (3.66) a.
*[Der Abby ti zu schenken] habe ich das Broti versprochen. the AbbyDAT to send have I the breadACC promised Intended: ‘I promised to send bread to Abby.’
b.
[Zu kaufen ti tj versprochen] habe ich der Abbyi das Brotj versprochen.
c.
*[Den Kindern ti zu erzählen] hat er die Märcheni begonnen. the childrenDAT to tell has he the fairy talesACC begun Intended: ‘He began telling fairy tales to the children.’
d.
[Zu erzählen ti tj begonnen] hat der die Märchenj den Kinderni.
According to this reviewer, these structures are an instance of long-distance Scrambling. I maintain, on the contrary, that these structures receive a straightforward explanation in the analysis adopted and developed in this study. Why are the structures in (3.66a) and (3.66c) ungrammatical? The answer is quite simple: The dative objects der Abby ‘the Abby’ and den Kindern ‘the children’ respectively possess a [+ Ref] feature that was not presented to the external interfaces prior to the end of the derivation (i.e., this feature is ‘unchecked’). When these dative objects are permitted to scramble from the larger VP – as is the case in (3.66b) and (3.66d) – the sentence is fully grammatical. Therefore, in (3.66a) and (3.66c) the concatenative integrity of the [+ Ref] feature of these dative objects is not respected; allowing them to remain inside the topicalized VP in (3.66a) and (3.66c) without Remerging into the derivation in the middle field creates an ungrammatical structure. Second, we can explain the topicalization of VP without having to refer to longdistance Scrambling, something that has been shown not to exist in West objects that scramble into the Germanic middle field are referential topics. This distinction is critically necessary or else objects would never scramble above the subject due to the natural economy considertations (i.e., why would an object move to a higher, more distant functional projection when it could check its discourse feature in a more local domain? – Shortest Move.) Second, both subjects and objects are referential in the Germanic middle field, thus suggesting the possible presence of RefPs for both subjects and objects.
126
Scrambling and the Survive Principle
Germanic. A simpler solution would be to argue that the completed VP is marked with a [Topic] feature (as a free feature) and must raise above the infinitive – most likely in FinPo in the extended CP-layer (cf. Abraham 1997) – in order to properly license this feature. 3.3
Eliminating XP-Adjunction
Since the characteristics of middle field XP-Scrambling in West Germanic do not fully align themselves with NP-movement, wh-movement and/or topicalization, it is worth probing the claim that Scrambling is a form of adjunction in a version of minimalist syntactic theory that makes use of Internal Merge/Move to explain displacement phenomena. Positing that Scrambling is a form of XP-adjunction finds its origin in theoretical literature during the GB-era: •
To create extraction barriers
•
To insure structure preservation; and
•
To disrupt c-command relations.
In Survive-minimalism, all of these three characteristics of XP-adjunction are removed from the theory. The notions of extraction and c-command chains formed in the derivation refer back to constituent displacement via Internal Merge/Move and are no longer thought to be necessary to the theory. The principle of structure preservation also falls to the wayside: The actual constituent participating in the derivation remains in the lexicon. The concatenate syntactic objects that are formed through Merge, Remerge and Link! then consist of functional heads and the relevant matching feature at this local point of phrasal fusion. In other words, the long-held theoretical properties thought to distinguish substitution (i.e., feature-driven displacement) and adjunction are severely blurred in the derivational revision of the Minimalist Program supported in this book (cf. Section 1.4.4). Within a view of the Minimalist Program that makes use of the Internal Merge/Move approach, there are decided pros and cons to the line of thinking that middle field Scrambling is an instance of XP-adjunction. As pointed out by Haider (2005a), however a desirable advantage to explaining middle field Scrambling in West Germanic as a displacement type driven by a formal feature would align this movement type with other forms of movement in syntactic theory. The following list (taken from Putnam 2006a:117) outlines the ‘pros’ (indicated by ☺) and the ‘cons’ (marked by ) to a feature-driven approach.
Chapter 3. Theoretical Considerations
☺
127
Adopting the view that all movement in the narrow syntax is featuredriven is economically desirable; this would help classify all movement types (including Scrambling) under the same licensing mechanism. In the same light it is difficult to provide evidence for a morph syntactic trigger of a functional head that would account for the multitude of diverse characteristics of Scrambling in West Germanic. The problem of optional movement. The mixed A/A’-properties of Scrambling.
The purported disadvantages of a feature-driven approach to middle field Scrambling can easily be explained away and shown to no longer be problematic in Survive-minimalism. First addressing the lack of morphosyntactic inflection on scrambled XPs in West Germanic, our brief, yet concise discussion of the syntactic featurehood of the [+ Ref] feature in typologically distinct (i.e., Bantu/Xhosa) and those closely related (i.e., English and Old Norse) to West Germanic provided ample evidence in support of understanding [+ Ref] as a syntactic feature. Besides morphosyntactic and logico-semantic features a strong case can be made for pragmatic features (at least in some languages (cf. Parafita 2005)) playing an active role in syntactic derivation. The presence of both LF and PF effects on scrambled XPs encourage this analysis. Furthermore not all relevant ‘syntactic’ features provide morphophonemic reflexes in English but are traditionally considered to be ‘syntactic’ by nature. Take, for example, the logico-semantic features [+ wh] and [Q]. While it is uncontested that the interplay of both features provides the proper word order and semantic (i.e., scope) interpretation of whitems, neither feature upon being valued results in inflectional morphology in English. The [+ Ref] feature residing in the middle field of German is of similar design. Criticism of these sorts of features rests on shaky ground. Second, the initiative of labeling Scrambling as ‘optional’ movement returns us to the standard minimalist assumption that pragmatic readings are merely the result of successful syntactic derivations. Again, consider the interplay of logico-semantic features such as [+ wh] and [Q]. The absence of [Q] will result in an echo-question interpretation in English. Although it is technically a possibility, would anyone dare purport that this, too, is an instance of optionality? Certainly not. Lastly, the mixed A/A’-movement characteristics exhibited by scrambled XPs can simply be understood as the movement attributes associated with pragmatic/semantic features such as [+ Ref]. In his thorough discussion of the cartography of the CP-layer, Rizzi (2004) notes that topics that move into this discourse-domain are a ‘special’ instance of A’movement bearing the exact same mixed A/A’-movement qualities as middle
128
Scrambling and the Survive Principle
field Scrambling. If, then, [Topic] and [+ Ref] – both pragmatic/semantic features – display similar, if not identical, movement characteristic, the hypothesis put forward here of reducing these effects to an instance of feature identity is supported and upheld. When one fails to see the A/A’-movement dichotomy in Scrambling as being the result of the feature identity and displacement characteristics generated through pragmatic/semantic features, many adjustments must take place in the theory. As pointed out by Webelhuth (1992), attempting to interpret Scrambling as feature-driven movement to a functional projection would force us to split this operation into two separate movement types licensed by two unique features: one responsible for A-movement properties (e.g., anaphoric binding) and one to account for A’-movement characteristics (e.g., parasitic gaps and the fact that other grammatical categories other than DP can undergo Scrambling).17 To quote Webelhuth (1992:210): “Since many Scrambling languages allow both DPs and categories other than DP to move, it is desirable for the theory of grammar to make one unified process to individual languages from which all properties follow. The phrase structure theory fails in this respect and should be rejected.” This, of course, comes at the cost of making adjoined positions able to A-bind anaphora, which due to their non-c-commanding position should not be possible in an adjoined (via Pair Merge) union. The argument that Scrambling can and should be reduced to XPadjunction is an attempt to sweep a messy topic under the rug. The Surviveminimalism framework can tackle this problem head on. As a result, however, rather than blurring the distinction between adjunction and substitution, the bold claim can be made that XP-adjunction for displacement purposes can be eliminated from minimalism outright. In this section, I will provide a decisive argument in favor of removing adjunction from the Minimalist Program based on Stroik and Putnam’s (2007) critique of Fox’s (2003, 2004) treatment of Condition C asymmetries. In their analysis, Stroik and Putnam strongly criticize Chomsky’s (2001, 2004) recent amendments to adjunction and offer a crushing blow to this construction’s further existence in the theory. Equipped with this knowledge, I will turn my attention to Parker’s (1999) analysis of parasitic gaps in Dutch (based largely on an adjunction analysis) and show how Survive-minimalism can account for her findings without calling upon adjunction.
17
An alternative to proposing an A’-feature would be perhaps able to establish an Amovement feature and adjunction.
Chapter 3. Theoretical Considerations
3.3.1
129
Condition C Asymmetries and ‘Peek-a-boo’ Effects18
Following Stroik and Putnam (2007), I illustrate in this section the conceptual and empirical disadvantages of calling upon XP-adjunction to account for constructions such as Condition C asymmetries. Here I will establish a superior way and means of dealing with these data within the Survive-minimalist framework. This is important for our analysis of middle field Scrambling, for it will equip us with tools to deal with parasitic gaps in relationship to middle field Scrambling in West Germanic. First, let’s consider Condition C asymmetries (see (3.67) and (3.68)). These constructions pose the problem of explaining why some, but not all, the DPs contained in the bracketed whphrases can be co-referential with the pronoun he. (3.67) [which picture of Bill that John likes] did he buy? *Bill…he Ok John…he (3.68) a. he likes [everything that John writes] b. [everything that John writes] he likes
*John…he Ok John…he
To account for asymmetries such as those in (3.67) and (3.68), Lebeaux (1991, 1995), Chomsky (1993), Epstein and Seely (1998), Rubin (2003), and Fox (2003, 2004) make use of the Merge operation, however they propose two separate applications of this same operation. Accordingly, one application of Merge is argued to take place cyclically, whereas the other applies in noncyclic fashion (i.e., after the application of the Move/Internal Merge or Remerge operations). Furthermore, although these authors maintain that arguments must be merged cyclically, adjuncts enjoy the unique status of being able to be Merged into the narrow syntax either cycically or non-cyclically. Consider the following derivational history of example (3.67) above. (3.69) a. he did buy [which picture of Bill] – wh-movement b. [which picture of Bill] did he buy [which picture of Bill] – adjunctMerge c. [which picture of Bill [that John likes]] did he buy [which picture of Bill] The derivational history above in (3.69) allows us to explain the co-referential relations in (3.69). According to Stroik and Putnam (2007:1-2), “the DP Bill in 18
Unless otherwise noted, data in this section is taken from Stroik and Putnam (2007).
130
Scrambling and the Survive Principle
(3.69c) cannot be co-referential with the pronoun he because the pronoun ccommands the most embedded DP copy Bill, whereas the DP John can be coreferential with the pronoun because the late merger of the relative clause prevents the pronoun from ever c-commanding the DP.” An analysis of this sort is capable of accounting for the aforementioned data containing Condition C asymmetries (cf. (3.67) and (3.68)), and in this sense it is descriptively and explanatorily adequate. With that being said, such an analysis is extremely costly in other ways. As stated throughout this book, a principle tenet of Survive-minimalism is the outright elimination of non-cyclical operations in the narrow syntax. The presence of such non-cyclical operations dramatically complicates the processing responsibilities of CHL, since such operations force derivations to reconstruct and recompute structures that have already been built and interpreted (cf. Frampton and Gutmann 2002). In a label-free, crash-proof version of the MP that Survive-minimalism strives to be, such look-back and look-ahead operations are strictly prohibited. With his elimination of non-cyclical Merge operations, Chomsky (2001) simplfies processing however ‘orthodox’ minimalism must still confront the task of accounting for the Condition C asymmetries shown in (3.67) and (3.68). The proposed solution to this problem once again calls for a bifurcation of the function of cyclical Merge. Accordingly, these two separate Merge operations involve the Merger of arguments (Set Merge) and the Merger of adjuncts (Pair Merge). Although this readjustment of the properties of Merge is superior to previous analyses in that only cyclical Merge applies, they differ quite distinctly in their structural properties. Whereas Set Merge is visible at all stages of syntactic computation, the output of Pair Merge is critically not visible at all times in the narrow syntax. Structures created through the application of Pair Merge are not inherently visible and must await conversion to a Set Merge output via an operation called Simpl. Structures generated via Pair Merge are deemed to be complex, hence they are ‘not simple’ and, as a result, not structurally visible in the narrow syntax. These units must be simplified by the operation Simpl to resemble syntactic objects SOs created by Set Merge. The result of this argumentation is the somewhat paradoxical outcome that somehow the CHL generates certain concatenate units that are invisible (i.e., uninterpretable) to the narrow syntax, an outcome which clearly defies the MP’s reliance on bare output conditions. The derivational history below (3.70) is taken to represent, once again, the example (3.67). At this stage, the relative clause is merged into the derivation (3.70a), but it remains invisible until the derivation reaches the next application of cyclic operations in (3.70b) (in (3.70) the italicized clause in syntactically invisible).
Chapter 3. Theoretical Considerations
131
(3.70) a. he did buy [which picture of Bill that John likes] – wh-movement b . [which picture of Bill that John likes] did he buy [which picture of Bill that John likes] – Simple-operation . c . [which picture of Bill that John likes] did he buy [which picture of Bill that John likes]. It is the structural invisibility of the most embedded relative clause in (3.70a) and (3.70b) that keeps the DP John from being c-commanded by the pronoun. Stroik and Putnam (2007:2) point out that “Though Chomsky explains the coreferential relations in (3.67), we should note the fact that to account for these coreferential data, without appealing to non-cyclical Merge operations, Chomsky must complicate his system of operations, in a rather bizarre way, by adding a non-simple adjunct operation (Pair Merge) and a Simpl operation that later undoes (makes simple) the output of the adjunct operation.” The operation Simpl is thus an oxymoron, for it complicates the processing load of the CHL far more than reducing it. The operation Simpl allows adjunction to display what Stroik and Putnam (2007) call ‘peek-a-boo’ effects, i.e. a construction is allowed to be invisible to notions such a c-command and binding and later in the same derivation magically ‘reappear’ and become visible. The challenge for Survive-minimalism is simple (pun intended): If it can be proven that in instances where Simpl applies, making the ‘adjoined’ structure both visible and invisible at different stages of the same derivation, that adjoined structure is visible throughout the entire derivation, this would be the source of strong evidence for the elimination of adjunction and Simpl from the theory. Stroik and Putnam (2007) demonstrate that in Survive-minimalism these Condition C asymmetries can receive a straightforward explanation without calling upon non-cyclic Merge or ‘peek-a-boo’ effects induced through Simpl. To illustrate this point, consider the Condition C asymmetry in (3.71). (3.71) a. [which picture of Bill that John likes] did he buy? *Bill…he/OKJohn…he b. He bought [a picture of Bill that John likes]. *he…Bill/*he…John The crux of the matter in understanding constructions exhibiting Condition C asymmetries from the standpoint of the narrow syntax lies in determining how and when adjuncts enter the syntax. Fox’s (2004) operation Late Merge also
132
Scrambling and the Survive Principle
falls short in providing an adequate, minimalist explanation to these data. In its current formulation, the operation Late Merge requires the ‘tucking in’ of the adjunct [that John likes] into the wh-items. This in itself is also an undesireable result, because it is similar to the Simpl operation that somehow hides adjuncts at stages in the derivation from visible interpretation. Stroik and Putnam (2007) advance the argument that Fox’s (2004) version of Late Merge is faulty in design due to its reliance on the ‘movement’ of lexical items regulated by Move/Internal Merge. In Survive-minimalism, the displacement of syntactic objects does not take place within the Derivation D, but rather involves the iterative mapping of information and objects from the Numeration N into the Derivation D (i.e, N → D). Adhering to Survive-minimalist desiderata, we can avoid ‘tucking in’ the adjunct (qua Merge) to the complex wh-item by arguing that the adjunct resides in the Numeration and adjoins to the wh-item [what picture of Bill] prior to its remerging into the syntax since it survives and returns back to the lexicon due to its [Q]-features which must be checked in CP. For the sake of argument, call this operation Late Num Merge.19 This analysis couched within the Survive-minimalism framework is conceptually superior and much simplier in design than its predecesors. First, the adjunct [that John likes] is created in a WorkBench (WB) in the Numeration. This is critical, because its concatenation with [what picture of Bill] cannot be labeled non-cyclic and does not fall victim to the same ‘tucking in’ effects that Fox’s (2004) version of Late Merge does. Second, the cyclic application of our reformulation of Fox’s Late Merge – our Late Num Merge – requires that the adjuncts be visible to the narrow syntax at all stages upon its introduction into the Derivation D. Let’s now return once more to example (3.67). “Since the DP John was not a part of the original complex wh-item [what picture of Bill] that merged into the VP prior to its repulsion there is no point in the derivation during which the pronoun he could potential c-command John, thus explaining how John and he can be co-referential in (3.71a). The derivational history in (3.72) below highlights the pivotal steps in the composition of (3.71a)” (paraphrased from Stroik and Putnam 2007:14-15). (3.72) a.
b.
Merge {buy, [which picture of Bill] } → buy which picture of Bill Survive [which picture of Bill] ([Q]-feature) →
19 We take an agnostic view at this time as to whether adjunction is motivated by some sort of feature or feature-like entity active in the Numeration or Derivation (also see Putnam 2006a:Ch. 4 and Rubin 2003).
Chapter 3. Theoretical Considerations
133
c.
Merge {he, buy [which pictures of Bill] } → he buy which picture of Bill
d.
Merge {did, {he, {T, {buy [which picture of Bill] }}} → did he buy which picture of Bill
e.
Merge {C , {did, {he, {T, {buy [which picture of Bill] }}}} → C did he T buy which picture of Bill
f.
Late Num Merge { [which pictures of Bill] , [that John likes] }
g.
Remerge { [which pictures of Bill that John likes] , {C , {did, {he, {T, { buy [which picture of Bill] }}}}} → which pictures of Bill that John likes C did he T buy which picture of Bill
‘Which pictures of Bill that John likes did he buy?’ The non-cyclic application of Late Num Merge (3.72f) in the Numeration rather than in the course of the Derivation provides a straightforward explanation of Condition C asymmetries within the framework of our revised minimalism. Furthermore, it provides influential evidence in support of eliminating adjunction from syntactic theory. 3.3.2
Parasitic Gaps
Recall that parasitic gaps, i.e. the licensing of an empty category by the presence of an A’-trace, is designated as an A’-movement phenomenon. This property of objects in Dutch and German is illustrated in Chapter 2 with parasitic gaps represented by the empty position e.20 (3.73) Jan heeft die boekeni zonder ei te bekijken weggelegd. John has the books without to inspect away-put ‘John put away the books without having looked at them.’ ohne ei zu lesen weggeworfen hat. (3.74) …weil er die Zeitungi …because he the newspaper without to read away-thrown has ‘…because he threw away the newspaper without having read it.’
20
These examples are taken from Bennis and Hoekstra (1985:68) and Haeberli (2002:110) respectively.
134
Scrambling and the Survive Principle
Although parasitic gaps are canonically regarded as A’-movement phenomena, Webelhuth (1989) and Neeleman (1994:403) argue that the licenser of a parasitic gap in Dutch can also bind an anaphor, which is a property of Amovement, as illustrated in (3.75) and (3.76) below (Data from Bennis and Hoekstra (1985) and Parker (1999)).21 (3.75)
...dat Jan de rivaleni namens elkaar zonder ei aan te kijken feliciteert that Jan the rivals on behalf of each other without at to look congratulates ‘…that John congratulates the rivals on behalf of each other without looking.’
(3.76)
Ik ben mijn oomi [na ti jaren niet gezien te hebben] gisteren weer tegen gekomen. I am my uncle after year not seen have yesterday again against come ‘Yesterday I met my uncle again after year of not having seen (him).’
The main argument put forward by Neeleman and Webelhuth above exploits the false prediction that licensing conditions for parasitic gaps are identical for English, German and Dutch. Dutch appears to manifest a peculiar type of construction that contains parasitic gaps where no licensing gaps are present. If the parasitic gap in (3.76) refers to mijn oom ‘my uncle’, it would have to be bound by an element in an A-position. Disagreeing with Neeleman and Webelhuth, I posit that classifying parasitic gaps as an A-movement phenomenon is problematic. First, it is an attempt to create undesirable language specific rules. Why does it appear cross-linguistically that parasitic gaps exhibit A’-movement characteristics for every language expect German and Dutch? Second and most importantly, are the Dutch examples above authentic parasitic gaps? Parker (1999) elucidates that these constructions in Dutch are indeed both parasitic gaps and accordingly license A’-movement properties. To follow Parker’s argument, consider the data with regards to the underlying structure of the Dutch VP (from Parker (1999:163)).
21
Neeleman (1994) observes two additional differences between Dutch and English with respect to the licensing of parasitic gaps. First, the subject of a passive sentence can license a parasitic gap in Dutch. Second, parasitic gaps can even be licensed in nominalized constituents.
Chapter 3. Theoretical Considerations
135
(3.77) Claudia denkt dat Melanie op zondag het boek zal lezen. Claudia thinks that Melanie on Sunday the book will read ‘Claudia thinks that Melanie will read the book on Sunday.’ (3.78) Claudia denkt dat Melanie het boek op zondag zal lesen. Claudia thinks that Melanie the book on Sunday will read ‘Claudia thinks that Melanie will read the book on Sunday.’ A Dutch VP [het boek op zondag zal lesen] ‘the book on Sunday will read’ containing both an NP complement to the verb het boek ‘the book’ and an adverbial clause op zondag ‘on Sunday’ can appear in two different orders, as seen in the two previous examples. The adverbial clause can appear either immediately before or after the NP complement. Through a series of tests involving ECM verbs constructions and scope marking, Bennis and Hoekstra (1985) reach the conclusion that Dutch possess an optional movement property whereby a DP complement to a verb may be moved leftward and adjoined. Parker refers to this ‘optional’ movement in terms of Move α, which is reminiscent of GB-Theory. Based on what we now know about the true motivation behind middle field Scrambling in West Germanic, we can safely assume that the DP complement het boek ‘the book’ in examples (3.77) and (3.78) is repelled and remerges into the middle field due to the presence [+ Ref] feature on Do. Based on early instantiations of transformational syntax, this leftward adjunction in Dutch will leave behind a trace in Comp,VP. This step provides us with a perfectly legitimate parasitic gap construction: the antecedent is located in an A’-position, from where it binds a trace in the preverbal position. This trace then licenses a parasitic gap, because neither gap c-commands the other (Parker 1999:164). Therefore, the fact that Dutch allows Scrambling from tensed adjunct clauses that license parasitic gaps may be typologically eccentric in comparison to other contemporary West Germanic languages, however the way in which Dutch licenses parasitic gaps in accord with A’-movement properties remains rather universal. According to the established principles of Survive-minimalism, the explanation for these parasitic gap mysteries in Dutch can be significantly simplified. Similar to the analysis imparted for the Condition C asymmetries in the previous section, adjunction in this case can also be understood best as another reflex of the noncyclic application of Late Num Merge in the Numeration. As such, an explanation of how constituents that scramble into the middle field while simultaneously having the ability to license parasitic gaps without calling upon A/A’-movement distinctions or adjunction does exist. An abbreviated derivational history of the Dutch example (3.78) highlighting the most important steps in the derivation is sketched out below in (3.79).
136
Scrambling and the Survive Principle
(3.79) a.
Merge {[het boek op zondag], lezen} → het boek op zondag lezen
b.
Survive [het boek] ([+ Ref] – feature) →
c.
Merge {[op zondag], zal, lezen} → op zondag zal lesen
d.
Remerge/Late Num Merge {[het boek], [op zondag], zal, lesen} → het boek op zondag zal lezen
Steps (3.79a) and (3.79d) in the derivational history segment illustrated above look very familiar to one another, although in traditional theory these are viewed as separate displacement types. The complex XP het boek op zondag ‘the book on Sunday’ is assembled in the WorkBench (WB), enters the Derivation D, and survives (is repelled) into the middle field in order to license [+ Ref]. This returns us once again to the original dichotomy presented between structure generated by means of adjunction and those created by means of substitution, i.e. XP-movement. Building on the proposals of Kidwai (2000), Rubin (2003), Putnam (2006a) and Stroik and Putnam (2007) to name but only a few, the distinction between adjunction and formal movement of a syntactic object to the specifier of a functional head can be eliminated (at least in the case of pragmatic information). All of the terminology and problems associated with previous stages of development of transformational syntax, i.e. A/A’-movement, the licensing of parasitic gaps, etc., can be reduced to [+ Ref] and the effects that the repulsion and activation this feature has on the West Germanic middle field. The end effect is that a label-free syntax such as Survive-minimalism is no longer reliant upon adjunction to attempt to explain issues such as Condition C asymmetries and parastic gaps. 3.4
[+ Ref] in English, Pennsylvania German, and German and the Diachronic Loss of Scrambling
It is well noted that synchronic English does not allow the separation of any objects from the predicate verb (3.80), whereas object scrambling is perfectly fine in German (3.81). (3.80) a. b.
*I did the mani yesterday not ti see. *He will the storei tomorrow ti open.
Chapter 3. Theoretical Considerations
(3.81) a. b.
137
Ich habe den Manni gestern nicht ti gesehen. Er wird den Ladeni morgen ti öffnen.
In earlier stages of its development, both English and Pennsylvania German resembled their Germanic sister languages and dialects by exhibiting middle field Scrambling. An obvious question that desires an attempt of an explanation is how to account for the synchronic discrepancy of Scrambling in modern English in comparison to its Germanic relatives. Here I will sketch out a potential novel solution to this problem by making use of Surviveminimalism desiderata and the hypothesis supported here that middle field Scrambling in West Germanic is motivated by the pragmatic/semantic feature [+ Ref]. The diachronic loss of middle field Scrambling is often attributed to the word order shift in English from OV to VO. In my view, this word order shift was also accompanied by, perhaps even motivated by, a shift in the representation of pragmatic information in the formal clausal structure. In modern English, middle field referential topics cannot exist. In the SVO order in English, pragmatic principles prefer that ‘given information’ be arranged before ‘new information’ – this is often referred to as the Theme-Rheme articulation. Czepluch (2004:169) illustrates this fact with the following English examples: (3.82) a.
The man kissed a girl. Topic < Comment Theme < Rheme
b.
??A girl was kissed by the man. ?Topic < Comment Rheme < Theme
c.
??A man kissed the girl. ?Topic < Comment Rheme < Theme
d.
The girl was kissed by a man. Topic < Comment Theme < Rheme
The fact that ‘given information’ takes clausal precedence over ‘new information’ explains why English speakers prefer the active in (3.82a), but the passive in (3.82b). These forms better comply with discourse-functional preferences. Based on Stroik’s (2000) analysis of multiple wh-items in English outlined in Section 3.1.1, [+ Ref] is a feature on an operator which must be checked in
138
Scrambling and the Survive Principle
Spec,CP. Stroik also assumes that wh-elements identified by Rizzi (1991) and Cinque (1990) as referential can optional be assigned a [+ Ref] in the lexicon, while those identified by these authors as non-referential lack this feature. Operatoring under these assumptions, Stroik explains properties of multiple wh-elements in English. Take, for example, the data below (taken from Stroik 2000:78). (3.83) a. b.
Who bought what? *What did who buy?
(3.84) a. b.
*Why did you buy what? *What did you buy why?
The well-formedness of (3.83a) is simple: If both wh-elements bear the [+ Ref] feature, they will both appear in Spec,CP at the conclusion of the derivation to be interpreted at LF. The ungrammatical reading associated with (3.85b) rests on inherent economy principles: The functional Co must recall (via Remerge) the wh-item that will maintain the base order of wh-items (similar to Shortest Move). In the examples presented in (3.85), the wh-adjuncts cannot bear the [+ Ref] feature. Differing with Stroik, we can view this as a lexical deficiency of the wh-item why rather than a structural constraint based on adjunction. These data illustrate two key concepts to us: First, wh-items in English bear [+ Ref] features that must be checked in the CP-layer. Second, and most important for our immediate discussion, it appears that what is barring middle field Scrambling in Pennsylvania German and English is the lack of presence of the syntactic category RefP in its middle field. Following Borer (1984) and Stroik and Putnam (2005a), parametric variation is a facet of the lexicon. The lexicon provides the information determining the mapping of lexical items onto syntactic structure for interface interpretability. Roberts and Roussou (2003) and Van Gelderen (2004) denote that syntactic symbols that can be mapped to PF as +p, and those that have an LF mapping as +l. If the +p and +l properties of features are lexically determined those that are interpreted as –p will not instantiate displacement of a syntactic object.22 Roberts and Roussou notate a functional feature F that requires a PF realization as F*. The PF realization of F* can be achieved two ways: either by means of Merge or Move (Internal Merge). The option selected depends on what the lexicon makes available, but the most economical is always preferred. If the lexicon provides a morphophonological matrix for F*, then the matrix with be F*’s realization, and Move is unavailable. Conversely, if the lexicon has no such matrix for F*, 22 Pace Chomsky (1995, 2000, 2001), Roberts and Roussou (2003), and Van Gelderen (2004), there are no uninterpretable functional categories or principles. As a consequence, it is possible to maintain that all such features are LF-interpretable. If so, then there are no [-l] features.
Chapter 3. Theoretical Considerations
139
material from elsewhere must be moved to F (subject to the usual constraints on movement). The underlying principle in this analysis of syntactic change is that LF remains a universal constant for CHL and relegates the parametric variation found in human language to PF. With the reorganization of the placement of temporal adverbials in both English and Pennsylvania German from being adjoined to vP to moving to the rightmost periphery of the clause also served the function of ‘confusing’ PF when completed grammatical strings were to be evaluated. (3.85) L1:
[TP [RefP XP [vP Adv [vP [VP XP]]]]]
(3.86) L2:
[TP [RefP XP [vP [VP XP Adv]]]]
Language L1 in (3.85) represents a language that licenses middle field Scrambling with the [+ Ref] feature whose temporal adverbials adjoin to Spec,vP whereas in Language L2 (3.86) temporal adverbials appear as sisters of V. Although the movement operation proposed in (3.86) is indeed possible it comes at the cost of appearing as a vacuous movement at PF at the evaluation stage of the entire grammatical string. With this hypothesis I am not proposing that the loss of middle field Scrambling in English and Pennsylvania German is a correlate of adverbial displacement; rather, what I hope to convey is that the reassignment of temporal adverbials to become sisters of V is but one part of a larger shift in the pragmatic structure of both English and Pennsylvania German. As a result, scrambling constituents into the middle field has the potential of being vacuous and hence undetectable to future generations of L1 acquirers of these respective languages.23 Without reliable input to suggest that referentiality requires a PF realization F* (i.e., the displacement must take place) in the middle field, neonates acquiring English and Pennsylvania German will have no positive evidence to scramble constituents into the middle field. As a result, the grammatical category [referentiality] is at risk of losing its status as a syntactic feature – one that is interpreted at both external interfaces (LF, PF) – in languages such as L1 in (3.85) according to our revised definition of Full Interpretation (cf. (3.6)). In such a language [referentiality] will cease to be a true syntactic feature and will be licensed and marked by other means by the CHL.24 Before leaving the topic of the relationship of diachronic Germanic and the placement of the [+ Ref] feature, I would like to address two final 23
Also see Hinterhölzl (2003) for a similar argument of vacuous Scrambling in English. Another important point to bear in mind is that the ability (or lack thereof) of a given language to license middle field Scrambling is not dependent on an underlying OV-word order. As discussed in Section 1.4.4., Diesing (1997) provides solid evidence against labeling Yiddish as an SOV-language. This strengthens our position that middle field Scrambling is indeed licensed by semantic/discourse requirements in West Germanic. 24
140
Scrambling and the Survive Principle
tangentially relevant points in a speculative manner for future research. The first has to do with the relationship (or lack thereof) between robust morphological Case and middle field Scrambling. In Section 2.3.5, we effectively dismissed the direct connection between the licensing of Case and Scrambling, however from a functionalist and typological standpoint, the connection between a lack of Case marking (i.e., most notably a ‘true’ accusative Case) results in the absence of word order freedom associated with Scrambling. If XP-Scrambling survives to a higher XP, which structure is the most likely host bearing the matching [+ Ref] structures? Is it possible that the [+ Ref] feature could also be closely associated with agreement phrases in the middle field? Putting aside the lack of A-movement characteristics exhibited by middle field Scrambling, the alternative that [+ Ref] resides in an AgrOP may impart a unique way of accounting for the tangential connection between Case morphology and Scrambling in diachronic West Germanic. In Germanic languages (as well as many other language families of IE ancestry) definite articles developed from demonstratives. This linguistic shift is believed to be the result of an interaction between lexical specification and syntactic structure: The loss of the +demonstrative feature is responsible for the development of the definite article (Roberts and Roussou 2003:133). Following Philippi (1997), the feature +referential can be expressed either directly on N, in which case we see a different Case marking, or as a distinct lexical item, namely a definite article. According to Giusti (2001), the underlying assumption is that Case features reside in D and need to be checked by N-to-D movement. Bearing this distinction in mind, the following feature matrices for N are possible. (3.87) a.
[+ Ref; + Case]
b.
[- Ref; + Case]
The absence of [+ Ref] could have conceivably had a profound effect on the Nto-D movement for Case licensing in the development of West Germanic: the absence of the +referential feature forced a reanalysis of how Case is licensed in the narrow syntax. The shift from aborting Case features on N-to-D by means of overt movement in favor of directly Merging them onto D is a direct result of the absence of [+ Ref]. Consequently, the Case features now present on D in synchronic West Germanic are not as strong as their predecessors and can be licensed by means of non-local operations and potentially at PF under McFadden’s (2004) approach. Admittedly, this brief analysis is speculative in nature; however, based on the hypotheses put forward thus far, it is based on logical assumptions that would benefit from future research. Similar to Faarlund’s (2004) posited Referential Phrase in the DP structure of Old Norse
Chapter 3. Theoretical Considerations
141
(cf. Section 3.1.3), perhaps this diachronic change has more to do with the development of the internal structure of the DP in English and German respectively. Secondly, I would like to at least mention how language change, in particular syntactic change, can be described and addressed in the Surviveminimalist model. The fact that Afrikaans still exhibits [+ Ref] at the phrasal level similar to other West Germanic languages is the reason that Scrambling, even to a limited extent, is still possible, whereas it is no longer legal in Pennsylvania German. This argument provides further validity to the idea that morphological Case is a post-syntactic phenomenon handled at the PF-node, or at the very least must be separated from referentiality: With the removal of the feature [+ Ref], perhaps acting in tandem with prosodic requirements, morphological Case over time would eventually give way to more rigid word order. Koch (1974), for instance, argues that the stabilization of word order is a necessary precursor to the loss of morphological Case; that is, that erosion of Case endings can only occur after their functions have been transferred to some other form of grammatical marking such as rigid word order. If we continue to follow Koch’s assertion that syntactic change precedes morphological inflectional syncretism, we arrive at a hypothesis that has received little attention in the literature to date; namely, the possibility that alterations to the syntactic structure of a language function as a prerequisite for morphological changes. Burridge’s (1993) examination of syntactic change in Middle Dutch (1300-1650 C.E.) notices the gradual grammaticalization of word order. Departing from most traditional theories accounting for this shift, Burridge contends that the shift in dominant word order patterns (e.g., verbsecond vs. verb-final word orders for matrix and subordinate clauses) represents a stage in the language of pragmatic (i.e., discourse) word order prominence, during which morphosyntactic principles were much less adhered to. For example, according to Burridge’s research many constructions in Middle Dutch (e.g., gapping between unlike constituents and surface coding of the topic) that are no longer grammatical in Modern Dutch can be attributed to this shift from pragmatic-prominence to morphosyntactic-prominence.25 This theory is not too far-fetched to apply to other historical West Germanic dialects. For example, Old English had both rich Case morphology 25 Also of historical relevance is that fact that the pragmatic use of word order variation seems to be associated with the dissolution of the strong grammaticalized SOV word order and an intermediate phase during which at least some ex-SOV typologies seem to develop a considerable freedom in the pragmatic use of word-order variation. Burridge (1993:223) states that clear links between verb-final order and pragmatic factors can be found in historic West Germanic dialects (e.g., Old High German and Middle Dutch) in clauses headed by the conjunction want ‘for/because’. In Middle Dutch a pattern emerged in which in was found that verb-final order occurred consistently in those clauses carrying presupposed, topical material. Contemporary discourse analyses of modern German also support this dichotomy (Galato 2002).
142
Scrambling and the Survive Principle
and Scrambling. Although Scrambling was lost by the Middle English period, it did not disappear overnight. Allen (1995:178, 415-431) shows that even after paradigmatic leveling had long set in since the 12th century, that Scrambling constructions are attested for a century later. This argumentation finds support from Faarlund’s (2001:9) hypothesis that syntactic change, unlike morphological leveling, is much more gradual, lasting a minimum of a century. Allen’s thorough analysis of thematic properties of nominals and their correlation to lexical and abstract Case investigated many possible catalysts to paradigmatic leveling. After dispensing with phonological factors as being the primary trigger, she focused on another change which took place around the middle of the fourteenth century which supports the current approach argued for in this manuscript: The loss of freedom of constituent order (e.g., the topicalization of an indirect object not introduced by a preposition) (Allen 1995:417). Rather than viewing Case syncretism towards the end of the Old English period as being a prerequisite for Scrambling restrictions in the Middle English period, the opposite position is supported. In addition, it is instructive to note that a phonological reduction of final syllables also took place in the historical development of German (between Old High German and Middle High German) without a simultaneous loss of Scrambling or paradigmatic reductions of morphological Case. In sum, the proposal of diachronic syntactic change outlined in the works of Roberts and Roussou (2003) and Van Gelderen (2004) can be easily integrating into Survive-minimalism: The [+ Ref] feature in earlier instatiations of English was once a syntactic feature licensed in the middle field, hence, we can label it with the diacritic F*. With the weakening of this feature (along with Case features), Move was no longer necessary to license [+ Ref]; based on the principle of economy Merge became the only sufficient mechanism necessary to license [+ Ref] in the narrow syntax. In this sense the proposal of syntactic change found in the works of Roberts and Roussou (2003) and Van Gelderen (2004) correlate nicely with the key principles of Surviveminimalism. 3.5
Conclusion
The purpose of this chapter was to illustrate the theoretical consequences of labeling [referentiality] as a bona fide syntactic feature (in certain languages) and its relationship to middle field Scrambling in West Germanic. The first task at hand was to illustrate that [referentiality] can be labeled in West Germanic – as well as other languages and language families – as a syntactic feature. By adhering to our revised notion of Full Interpretation (cf. (3.6)) that states that a grammatical category can be considered a ‘syntactic’ feature iff it is necessary for both LF and PF interpretation, this chapter provides evidence
Chapter 3. Theoretical Considerations
143
supporting the claim that [referentiality] is a syntactic category in the West Germanic middle field in languages that license Scrambling. The second task was to test the proposal that larger derivational chunks – at least larger than pure syntactic objects SOs created through Merge/Remerge and Link! – bore some sort of contextual identity relevant for interface interpretation. As a point of departure, I elected to test whether some form of Grohmann’s (2000a, 2000b, 2003) Prolific Domains could function as optimal tools in describing the interpretive nature of natural clauses. An added attraction that accompanies a minimalist analysis of clause structure qua Prolific Domains is the lowerbound restriction on constituent distal known as Anti-Locality that is naturally enforced on displacement procedures. In other words, XPs must move – or survive – to a higher syntactic position that isn’t too local. Rather than regulating this sense of ‘syntactic distance/closeness’ by means of structure, this marking of distance is maintained through the contextual identity of Prolific Domains. Movement within a given Prolific Domain would force the CHL to recognize two features of identical nature on the same XP, thus forcing PF to generate two separate instatiations of a given object O in the grammatical string. Adopting Putnam’s (2005, 2006a) analysis of anaphoric binding discrepancies of scrambled accusative and dative objects in German, we demonstrated how a derivational system making use of Prolific Domains can explain not only these inconsistencies in anaphoric binding in purely syntactic terms, but it also provides a straightforward, unified account of remnant movement and Anti-Freezing Effects and their relationship with middle field Scrambling in West Germanic. The remainder of this chapter explained how parasitc gaps and their relationship with Scrambling can be accounted for without calling upon XP-adjunction, a structure that is troublesome and unwanted in any label-free syntactic system. Making use of Stroik and Putnam’s (2007) analysis and explanation of Condition C asymmetries within Survive-minimalism, I show how parasitic gap structures in West Germanic can be explained in the framework supported and developed here. Lastly, I suggest a novel explanation regarding the diachronic loss of middle field Scrambling in English and Pennsylvania German along the lines of Roberts and Roussou’s (2003) and Van Gelderen’s (2004) work in syntactic change. This chapter honed in on some of the more problematic issues that riddle minimalist analyzes of middle field Scrambling and demonstrates how Surviveminimalism can take these issues head-on, providing a clear, economic explanation of these inter-related epiphenomena. This, of course, is done without referring to extraneous conditions placed on the syntax such as anaphoric binding as a possibility from adjoined positions, making use of A/A’-movement chains, and similar arguments. In Survive-minimalism the analysis of middle field Scrambling can be minimally derived as follows: Middle field Scrambling in West Germanic is a syntactic operation driven by
144
Scrambling and the Survive Principle
the syntactic feature [referentiality] that is licensed as a syntactic category in the middle field. As demonstrated by our retooled definition of Full Interpretation (cf. (3.6)) and our brief discussion of syntactic change within Survive-minimalism in the previous section, the analysis of grammatical strings according to the core desiderata developed here provides us with the necessary analytic tools to determine which grammatical properties are ‘syntactic’ from those that are not. In this sense, our theory is also observationally adequate. Two immediate challenges faced in the early stages of generative grammar were to (i) find a way to account for the phenomena of particular languages (i.e., descriptive adequacy), and (ii) to explain how knowledge of these facts arises in the mind of the speaker-hearer (e.g., explanatory adequacy). In this quest to create and improve upon generative models to mold them to conform to criteria of descriptive and explanatory adequacy, we often pay little if any attention to a third component of linguistic theory referred to by Davies and Dubinsky (2004) as observational adequacy. In their own words, Davies and Dubinsky (2004:154) define observational adequacy as: […O]bservational adequacy involves the not always trivial task of determining which are the well-formed expressions in a language, and which are not (and presumably being able to state whether the ill-formedness, where it occurs, is syntactic, semantic, or pragmatic).
This third and equally valuable facet of linguistic adequacy does not receive adequate attention in the literature, primarily due to a lack of understanding of where and when these various modules of grammatical competence intersect and where they remain autonomous unto themselves. Survive-minimalism is a derivational theory of syntax that is descriptively, explanatorily and finally observationally adequate in design. In the following chapter I will make particular use of the observational adequacy of this framework in discussing the rather delicate/intricate relationship between the narrow syntax and prosody in the licensing of middle field Scrambling. In the discussion that follows, I will examine the data and provide a clear delimitation between the narrow syntax and prosody while maintaining the analysis supported throughout this manuscript.
CHAPTER 4 THE PROSODIC SIDE OF SCRAMBLING1 One of the larger issues addressed in this manuscript is the nature of syntactic featurehood in order to determine what is and what isn’t ‘syntactic movement’. Another way of phrasing this question in this theoretical discussion is what constitutes a bona fide functional projection in the narrow syntax? From the standpoint of explanatory adequacy, it is desirable, if not essential, that a minimalist theory of language computation and evaluation make readily available a set of functional units and features sufficient for the successful acquisition of L. Contra Abraham et al. (1996), I have made the case for the existence of a referential topic feature (e.g., [+ Ref]) as the principle motivating feature licensing middle field Scrambling in West Germanic even though it is not overtly associated with inflectional morphology. My strongest evidence for this claim comes in the form of reconstruction effects and anaphoric binding relations displayed by scrambled XPs. The fact that this displacement type shows LF and PF reflexes allows us to classify Scrambling as a ‘syntactic’ operation according to my revised definition of Full Interpretation (cf. (3.6)). Discourse phenomena, as pointed out by Parafita (2005), can take the form of morphological, syntactic or prosodic features, active in relevant modular units of the CHL. Through the process of acquisition, the experience gained by a native L speaker will have his grammatical parameters established by the feature identities set by the innate rules of grammar enforced and maintained by native speakers of L. Another question that stems from any discussion of the acquisition of features and functional projections is the depth of the pool of options a neonate L speaker has to choose from. In the theory of functional projections, it has been the widespread assumption that the inventory of functional categories and projections could, and perhaps, should be, universal across all languages. Cinque (1999) extends this hypothesis even further by speculating that the inventory of functional categories and projections might also be universal across not only all languages but also all clause types. Regardless if one buys into the concept 1 Portions of this chapter are the result of collaborative work (Grohmann and Putnam 2004; 2005; forthcoming) and lengthy discussions with Kleanthes Grohmann. Materials involving the modifications on the model of the grammar were presented at the Michigan Linguistics Society’s annual symposium (2004) and GLAC 11 (2005). I am grateful for the comments from the participants – especially Acrisio Pires, Richard Wiese and John te Velde.
146
Scrambling and the Survive Principle
of a theory of universal inventory or not, some sort of constraining mechanism is necessary for effective, successful L1 acquisition. Thráinsson (1996:254) notes that: Clausal architecture is determined by UG in the sense that UG defines a set of functional categories, {F1, F2, … Fn} that languages “select” from. For any functional categories Fi and Fj, the sequence will be uniform whenever they occur, i.e. if L1 and L2 each instantiate both Fi and Fj and Fi c-commands Fj in L1, then F1 c-commands Fj in L2.
Of course, not every functional category or feature is subject to selection in a given L, however according to Thráinsson’s quote, the possibility is potentially available to each and every category and feature at the beginning stages of acquisition.2 Our revised derivational approach to minimalist syntax once again has the ability to substantially simplify our outlook on these processes. Stroik and Putnam (2005a) eliminate the existence of uninterpretable and unvalued features on the conceptual grounds that uninterpreted features by nature cannot be acquired by L1 learners due to their unspecified, and in some cases, imaginary nature. The removal of Internal Merge/Move from the narrow syntax as well as other non-local feature-checking configurations such as Attract and Agree (Probe-Goal) has also paved the way for the elimination of these uninterpretable variants of syntactic features. The acquisition of lexical items also entails the acquisition of the interpretable features that accompany them, regardless if they are free or bound to a given XP.3 Herein lies the key doctrine of our revision of the Minimalist Program according to the Survive Principle: The lexicon – viewed in ‘orthodox’ minimalism to be separated from the computation of syntactic operations – is now included in the process. The features present in the lexicon are borne +interpretable, i.e. these features have access to the external interfaces. What is then necessary from an acquisitional standpoint is the learning of features and lexical items and their associations with one another. The phrase structure building operations of Merge and Remerge are the introduction of a functional minimal projection Xmin bearing the feature α that activates a syntactic object with the same matching feature.4 Their union 2
The strongest prediction borne out by Thráinsson’s (1996) notion of features and their relationship with syntactic structure is that non-catographic in design. Under such assumptions, all languages have the same functional features (such as Tense and Agreement), but the syntactic expression of these features is parameterized – they may project separate heads in some languages, and ‘fused’ heads in others. Thráinsson’s (1996) proposal along these lines regarding the structure of IP is an excellent example of the application of such a line of thinking. 3 Putnam and Stroik (in preparation) note that another facet of Survive-minimalism is that it enables head movement to once again be viewed as syntactic movement. Concerning acquisition, the learning of heads and their attached features and properties is no different that full XPs (also see Ferguson 1996). 4 Concerning Merge, Stroik and Putnam (2005b) and Stroik (2007) make the further distinction between Lexical Merge and Structural Merge. For a more detailed account of the individual properties of these operations, the reader is referred to Stroik (2007:Ch. 3).
Chapter 4. The Prosodic Side of Scrambling
147
is automatically shipped to the external interfaces. The iterative application of the primitive operations Merge-Survive-Remerge/Link! computes the syntactic structure and creates one perfect representation at the end of the derivation. Of course, some sort of order or hierarchy of these constructions is necessary to arrive at a grammatical representation. This leads us back to a system similar to Grohmann’s Prolific Domains (2000a, 2003) which provide the skeletal structure of the functional heads C > T > v.5 Although we have provided a clear and direct definition of syntactic featurehood and have illustrated how these notions adhere to and improve upon the current theoretical charge of explanatory adequacy, something that remains to be discussed in this forum is the role of prosody in the narrow syntax. In particular, the challenge is to determine when and if prosodic-movement (hereafter referred to as p-movement) is an operation licensed by a formal feature in the narrow syntax or a post-syntactic phenomena that takes place after syntactic objects (structure) has been composed. In order for p-movement to be ‘syntactic’ it must possess two crucial qualities to make its case: First, not only must this movement type be visible to PF but also LF (cf. Full Interpretation Revised (3.6)). Second, some sort of formal feature γ must exist on a functional head in the lexicon to properly license this feature through the application of Link! If p-movement is purely prosodic, it should fall outside of the framework of the narrow syntax, which is the working hypothesis I will adopt at the beginning of this chapter. In the discussions that follow, I will attempt to address the prosodic nature of middle field Scrambling and its relationship to the narrow syntax according to Survive-minimalism. Following the arguments put forward by Büring (2005) and Grohmann and Putnam (in press), I maintain that information structure (hereafter IS) requirements are an aspect of syntactic representation that interfaces with PF by rules of IS-realization, and receives its meaning via rules of IS-interpretation. What this means for our current analysis of middle field Scrambling within Survive-minimalism is that certain linear orders and properties of this displacement type will be rendered ungrammatical by the CHL not by means of some sort of syntactic error, but rather at the level of prosodic structure. In this chapter, I will provide both empirical and conceptual arguments to support this claim. 4.1
Introduction to the Prosody-Syntax Interface
Over the past four decades, many controversial theories have been proposed in the search of an empirically and theoretically adequate account for the focus5
Although it is theoretically possible to attempt to start out with another functional head other than v, the system would ultimately crash/stall badly enough times that the affixed order C > T > v is most likely the result of a learned behavior (thanks to Hamid Ouali for pointing this out to me).
148
Scrambling and the Survive Principle
prosody relation. While the fundamental questions – what is the relation between focus and intonation, and which principles determine its interpretation – have yet to receive a fully satisfactory answer, the picture that has emerged over time shows that focus is a “multidimensional concept characterized by the interrelatedness of syntax, argument structure, formal semantics, and pragmatics” (Winkler and Göbbel 2002:1185). The relationship between focus and metric phonology is believed to be governed by the Relative Prominent Rule, hence connected to notions of rhythm, alternation, and organization of features into syllables, and syllable into larger units, etc. In other words, by the Relative Prominent Principle, stress cannot be manifested without a structure. While the prosodic structure, at the lexical level, is the organization of moras, syllables and feet, at the phrasal level the prosodic organization is essentially syntactic, and this is captured by the Nuclear Stress Rule (hereafter NSR) formulated in Liberman and Prince 1977, represented as follows (4.1): (4.1)
[A B]c NSR: if C is a phrase, B is strong.
The goal of Zubizarreta’s (1998) seminal work pursues the formulation of “an adequate theory of nuclear stress” (18). Building upon the NSR formulated by Liberman and Prince (4.1), Zubizarreta arrives at a ‘modularized NSR’ (MNSR). This MNSR comprises both a positional NSR, which is sensitive to constituent ordering (C-NSR), and a rule that is sensitive to selectional ordering (S-NSR) and thus incorporates the insights of the argument-structural account. The C-NSR is an extension of Cinque’s (1993) Theory of Sentence Stress and is in essence a correspondence between the asymmetric c-command ordering of the syntax and prosodic ordering. Below (4.2) the primary tenets of Cinque’s proposal and (4.3) Kayne’s (1994) Asymmetric c-command Theory are summarized: (4.2)
Theory of Sentence Stress (Cinque 1993): No parameterization for NSR
•
Eliminates the projection parameter: headedness of stress projection is predictable from word order variety (i.e., head-parameter).
•
The Nuclear Stress Rule (NSR) assigns main stress to the most deeply embedded phrase in a sentence.
•
When two phrases are in a sister relation, i.e., when they are at the same level of depth of embedding, the embeddedness is determined by
Chapter 4. The Prosodic Side of Scrambling
149
the direction of recursion of syntactic projection, i.e., the right node bears stress in right-branching structures; the left node bears stress in left-branching structures. Ck 2 Ci Cj (Where Ci and Cj are sisters) (4.3)
Asymmetric c-command Theory (Kayne 1994) Given two terminals α and β, α precedes β iff there exists a constituent A dominating α and a constituent B dominating β such that A asymmetrically c-commands B. A 2 α B 2 β …
As far as focus in verb-final structures is concerned, Zubizarreta provides an account for the interaction of nuclear stress and focus in German according to which nuclear stress falls on the complement immediately to the left of the verb in verb-final structures and on the last constituent in verb-second structures. The NSR is considered as a device that determines phrasal prominence by providing the most embedded argument with nuclear stress. The relation between prosody and focus then is determined by the Focus Prosody Correspondence Principle that states that the focused constituent (or F-marked constituent) of a phrase must contain the intonation nucleus of that particular phrase.6 The F-structure of the sentence is constrained by the Focus Prominence Rule (FPR) outlined below in (4.4): (4.4)
Focus Prominence Rule (Zubizarreta 1998:21) Given two sister categories Ci (marked [+F]) and Cj (marked [-F]), Ci is more prominent than Cj.
Naturally, the coexistence of the FPR and the NSR in the grammar gives rise to cases in which the output of the NSR contradicts the output of the FPR. Conflict arises in such structures where the focalized items are not in the position 6
This feature [F] above is not to be confused with the discourse-oriented functional feature ‘focus’ which acts as a probe in the syntactic structure and attracts an F-marked constituent motivating movement in the narrow syntax (cf. Zubizarreta 1998:178).
150
Scrambling and the Survive Principle
where they receive prominence via the NSR. In an SOV-language like German, this situation occurs in sentences with non-contrastive focus when relevant defocalized material intervenes between the focused element and the verb (F-marked items appear in bold typeface throughout): (4.5)
a.
Wem hast du das Geld gegeben? whomDAT have youNOM the moneyACC given ‘To whom did you give the money?’
b.
Ich habe dem KASSIERER das Geld gegeben. the moneyDAT given INOM have the cashierDAT ‘I have given the money to the cashier.’
Further data from West Germanic V2 languages that license Scrambling also display this discrepancy. For instance, the Dutch examples in (4.6a) and (4.6b) below illustrate another scenario in which the NSR and FPR conflict: (4.6)
a.
…dat Jan gisteren het BOEK gelezen heeft that JanNOM yesterday the bookACC read has ‘…that Jan read the book yesterday.’
b.
… dat Jan het boek gisteren GELEZEN heeft has that JanNOM the bookACC yesterday read ‘…that Jan read the book yesterday.’
Given that the complement het boek ‘the book’ in the first example (4.6a) is the most embedded element in the phrase structure in tandem with its sisterhood relationship with the participle gelezen ‘read’ allows the S-NSR to select the VP complement as the item that receives exhaustive focus. In (258b) the DP het boek ‘the book’ has scrambled into the middle field. Traces are believed to be ‘metrically invisible’ to any stress assignment, so the participle gelezen ‘read’ must assume the neutral stress of the utterance provided that it is the most embedded element in the derivation. In Germanic (and French) Zubizarreta deals with this existing contest between the NSR and FPR by treating defocalized constituents (i.e., constituents marked [-F]) as ‘metrically invisible’ for the NSR. In an updated treatise on the subject matter, Zubizarreta and Vergnaud (2000) argue that the perceived parametric optionality of whether or not a language family makes its defocalized arguments available for the application of the NSR lies in the fact that these aforementioned language’s prosodic structure is based on purely metric trees – exhibiting only ‘visible’ units to the NSR – whereas languages utilizing the NSR map their syntax-prosody interface from standard syntactic hierarchy.
Chapter 4. The Prosodic Side of Scrambling
151
This invisibility condition present in metric trees is best captured by the formulation of the Structural Removing Condition (Feng 2003): (4.7a)
L 3 H Z 3 X Y(invisible)
(4.7b)
L 3 H Z X
Structural Removing Condition Remove all the prosodically invisible elements (with their syntactic branches) from the tree structure, before the application of the NSR. As can be derived by the diagram in (4.7), (4.7a) represents a standard morphosyntactic hierarchical structure. On the other hand, (4.7b) has removed all ‘invisible’ material from the tree so that only ‘visible’ items, i.e., constituents that are θ-marked, remain to play a role in determining the prosodic alignment of clauses. We will shortly return to the Structural Removing Condition, however for the time being, the main premise of the principle is comprehensible. In our reanalysis of the Minimalist Program based on the Survive Principle, constructs such as the invisibility condition and the Structural Removing Condition (4.7) are naturally accounted for: Recall that there truly are no ‘traces’ or ‘copies’ that remain in the CHL – these are nothing more than derivational labels to assist linguists in analyzing the syntax. What is left behind is information relevant for LF and PF based on the feature nexus shared between Xmin and Xmax matching and linking the feature α. The question still remains how other languages deal with the conflict of the NSR and FPR if their defocalized elements are not metrically invisible. One strategy to resolve this divergence is to move the defocalized element far enough making the focus position available for the most embedded constituent to receive prosodic marking. This would result in a movement chain as illustrated below in (4.8), where the item B is marked [-F]. (4.8)
[A
B]c
In Romance languages where all phonologically overt constituents are metrically visible, the defocalized constituent undergoes movement, i.e., these languages have recourse to a particular strategy to avoid the clash between the output of the FPR and that of the NSR. As explained by Zubizarreta (1998:89),
152
Scrambling and the Survive Principle
in (4.9a) the NSR applies to the sister nodes (DP, VP), assigning main prominence to the VP (which will become visible on una manzana ‘an apple’, as a result of the reapplication of the NSR inside the VP). The FPR is functional between the sister nodes (DP, VP) and allocates main prominence to the DP subject (which will become visible on Juan, as a result of the reapplication of the NSR inside the DP subject). Thus in (4.9b) the defocalized constituent comio una manzana undergoes movement in order to leave the Focused phrase Juan in a position to receive NS via the NSR. In this way, the Focused phrase ends up being the lowest constituent in the asymmetric c-command ordering. Zubizarreta (1998) assumes this kind of movement to be prosodically motivated and thus she calls it p-movement, as opposed to traditional movement types that are triggered by feature-checking matters. The purpose of pmovement is for this reason: To repair a prosodically conflicting situation. In Zubizarreta’s approach, non-focused constituents are allowed to move to ensure that the focused phrase always appears in final position.7 (4.9)
a.
*JUAN Juan
comió una ate an
b.
Comió una manzana Ate an apple ‘Juan ate an apple.’
manzana8. apple
Spanish
JUAN Juan (Zubizarreta 1998:20)
P-movement is not triggered by feature checking per se under standard minimalist assumptions, but rather, is considered to be a local syntactic operation that affects two metrical sister nodes in order to alter a prosodically ill-formed structure. In this sense, the phrasal hierarchy present in syntactic computation feeds the NSR. Therefore, since the NSR is sensitive to hierarchical structures, it is assumed that p-movement is active in the narrow syntax rather than existing as only a post-syntactic phenomenon (Zubizarreta 1998:141). As one would expect, such suppositions have an altering effect on the traditional model of grammar. To accommodate the active presence of p-movement in the syntax, Zubizarreta proposes the following amendments to the standard model of grammar.
7
It remains unclear how Zubizarreta’s notion of p-movement can explain scrambling-type movement in Hungarian, which is movement to focalize an XP at the left periphery of vP in a phase-based system. 8 Zubizarreta (1998:20) points out that (45a) with main prominence on a phrase-internal constituent is only possible if it is interpreted as contrastive or emphatic. (e.g. JUAN comió una manzana (no Pedro); (Juan ate an apple, not Pedro)).
Chapter 4. The Prosodic Side of Scrambling
(4.10)
153
(sets of phrase markers, feature checking) Σ-Structure
(unique phrase marker) (F-marking, NSR, FPR, p-movement)
LF PF
Assertion Structure (Zubizarreta 1998:32)
According to the architecture presented in (4.10), p-movement interacts in a feeding relationship with LF. If p-movement does indeed supply processed units to the LF-interface, are we to assume that Scrambling could actually be movement to eliminate prosodically uninterpretable (or at the very least illformed) structures? If this is the case, we wouldn’t need to argue for a [+ Ref] feature as we have thus far in this study to force Survive, so from an economy stand point, this at first glance might be an attractive stance to take. For example, Kidwai (2000:232) emphasizes the fact that scrambled XPs tend to be interpreted as topics, which might explain the selfish reason for moving these items for discourse configuration requirements that are amenable to syntactic and prosodic analysis. What Kidwai’s statement fails to take into account, however, is the distinction between ‘aboutness’ and ‘familiar’ topics discussed in Chapter 2. There are clear pragmatic and syntactic (the ability or lack thereof of being bound to a cataphoric pronoun) traits that elucidate this distinction. Therefore the necessity of the [+ Ref] feature is upheld. If p-movement does exist in the syntax as argued by Zubizarreta, it is yet to be determined to what functional projections these XPs merge with higher in the structure. To extend this argument, if formal features are not involved in pmovement, i.e., if this form of displacement is not driven by a feature, is it possible that we are dealing with some form of adjunction rather than movement? Grewendorf and Sabel (1999) adopt this line of thought, namely, that Scrambling is a form of adjunction, based primarily on anaphoric binding data (based largely on the binding properties of scrambled accusatives). This argument notwithstanding, indirect objects that are scrambled in the middle field in German are focused, not destressed; therefore, this is an instance of movement to focalize rather than to move an element ‘out of the way’. The fact that definite accusative objects can enter into new binding relations indicates – which relations must hold at LF – Scrambling is movement to a structural position that allows the scrambled items to c-command anaphors and cannot be purely determined by means of prosodic structure. To argue that Scrambling in West Germanic is purely an instance of pmovement, however would call for us to remove the caveat on this trigger only existing in Romance languages. This position has been advocated by linguists
154
Scrambling and the Survive Principle
such as Domínguez (2004)9, Grewendorf (2005), Abraham (2003) and Molnárfi (2002, 2003).10 As a matter of fact, the AntiFocus hypothesis put forward by Abraham and Molnárfi (2002) is in actuality an attempt to integrate Zubizarreta’s p-movement into Germanic. Not only that, the AntiFocus hypothesis assumes that the formal feature [AF] calls the shots not only for the prosodic status of a given XP, but also for semantic and/or discourse interpretation of an object. If an XP does not undergo Scrambling even though the [AF] feature is frustrated and cannot Merge to a superior position in the hierarchy, the system does not crash, but rather initiates prosodic marking – an in some cases a morphophonemic reflex – of the frustrated XP. Abraham and Molnárfi classify [AF] as a third kind of trigger that can communicate with both the LF and PF interfaces, which according to our revised version of Full Interpretation (cf. (3.6)) is the defining characteristic that we use to classify all syntactic features. In Section 4.2.1 I will provide ample argumentation against the adoption of the [AF] trigger as the sole syntactic feature responsible for middle field Scrambling. 4.2
Minimalist View of the Prosody-Syntax Interface
Next we can turn our attention to the argument that Scrambling can potentially be licensed by prosodic properties. The theoretical shift in focus from GBtheory to minimalism – from proper government to interface interpretability – forced theorists to ask pertinent questions regarding the true nature of the syntax-PF-interface. In earlier versions of minimalism (cf. Chomsky 1993, 1995), the entire derivation proceeded until the hierarchy was shipped off to the interfaces via Spell-Out. The LF and PF interfaces in early minimalism were regarded as the sole levels of representation, with all operations and principles being motivated by virtual conceptually necessity (VCN) or bare output conditions (BOC). With the onset of multiple Spell-Out models (cf. Uriagereka 1999) in the literatures and an overall progressive development of interface phenomena (e.g., Distributed Morphology at the PF-interface (cf. Halle and Marantz 1993)) the PF-branch was attributed the task of licensing traditional morphological considerations. Furthermore as noted by Stroik (1999) and Embick and Noyer (2001) among others, operations applicable at the PF-branch are ordered with respect to one another. The model in (4.11) (taken from 9
Domínguez (2004) also attempts to establish a connection between p-movement and scrambling in German. She views both operations utilized to remove elements from the main stress position so that the stress falls on a different constituent. 10 Although Grewendorf (2005) advocates the presence of p-movement in the narrow syntax for Germanic languages, he explicitly states that he defends “the idea that scrambling should be reanalyzed as movement driven by discourse configuration requirements.” Similarly to Abraham and Molnárfi’s (2002) AntiFocus hypothesis, Grewendorf views p-movement in this light as a ‘defocusing device’ that he analyzes as topic movement.
Chapter 4. The Prosodic Side of Scrambling
155
Grohmann 2006) represents the traditional a view of the PF-branch of the grammar ca. 2001 (cf. Embick and Noyer 2001). (4.11) A View from the PF-Branch
Lowering
arrangement of morphemes
Vocabulary Insertion
Linearization imposed by vocabulary insertion
Prosodic Domains
PF Grohmann (2003) interprets the syntactic derivation as part of the CHL that feeds the interfaces cyclically through Prolific Domains. Subscribing to this view of syntactic derivation, it is possible to conclude some operations licensed at PF are potentially detectable prior to their presentation at the interface. In a way, we have oversimplified things a bit up to this point; assuming a direct mapping of structure and features from dynamic syntax directly to PF fails to answer key theoretical inquiries such as the nature of permissible structure at PF, i.e., how much structure is there if any at all? In their thorough study, Ackema and Neeleman (2004) argue that the interface level known as PF is a complex level at which syntactic representations are mapped onto an initial prosodic structure, which is itself mapped onto a phonological representation. In (4.12) below, Ackema and Neeleman (2004:185) outline the intricate steps of this mapping process. (4.12) a. b. c. d.
Linearization of syntactic terminals Initial prosodic phrasing, on the basis of syntactic information Application of context-sensitive allomorphy rules Spell-out of terminals
According to the derivational order in (4.12) above, derived prosodic domains determine the application of the rules of allomorphy. These rules apply at PF, and can change the feature content of a terminal in the presence of another terminal in the same prosodic domain. Ackema and Neeleman (2004) refer to this mechanism responsible for altering the feature content of a terminal at PF as Prosodic Checking. Although this concept bears close affinity to Distributed Morphology (cf. Halle and Marantz 1993), Ackema and Neeleman (2004:186) point out that although both models adopt the notion that there are postsyntactic allomorphy rules that adjust the feature content of terminals in par-
156
Scrambling and the Survive Principle
ticular environments, the notion of Prosodic Checking implies that these postsyntactic allomorphy rules are sensitive to initial prosodic phrasing. In other words, in a Distributed Morphology-model, (4.12c) would always proceed (4.12b). Ackema and Neeleman (2004:190-193) distinguish two general types of prosodically-conditioned allomorphy rules. The first rule states that morphosyntactic features of a terminal contained in the same phonological phrase as a certain other terminal are deleted. Such deletion necessarily requires agreement between the features of the terminals to which the rule applies on the basis of recoverability. (4.13) {[A F1 F2] … [B F1 F3] …} Feature deletion: {[A F2] … [B F1 F3] …} Spell-out: [A F1 F2] → /a/ [A F2] → /a’/ The deletion of features obviously changes the phonological realization of the terminal whose content has changed. The second type of allomorphy rule states that if a particular terminal appears in the same prosodic domain as some other terminal, its phonological realization is altered. According to Ackema and Neeleman, this type of allomorphy rule is responsible for cliticization: the presence of a syntactic head endowed with the required features forces a pronoun to be realized a simple clitic, i.e., something smaller than a phonological words (such as a σ, a syllable). (4.14) { …A… [B F1 F3] …} → …} Spell-out: [B F1 F3] → /bσ/ A clear difference between (4.13) and (4.14) is that the model of Prosodic Checking in the former does not delete any morphosyntactic features in its target. Chocano (2004) unites both of these versions of Prosodic Checking to account for phonological properties of Scrambling in West Germanic. For the time being; however, I would like to turn our attention to Abraham and Molnárfi’s (2002, 2003) claim that a ‘third kind of feature’ interpretable at both the LF and PF interface exists. In the following section I will outline their deriva-
Chapter 4. The Prosodic Side of Scrambling
157
tional approach to Scrambling in West Germanic (with a particular focus on shibboleth morphology in Afrikaans) utilizing the feature [AF]. I take the stance that Ackema and Neeleman’s version of Prosodic Checking advocating the deletion of features within the same prosodic domain accounts for this phenomenon in way more compatible to both orthodox minimalist thought and Prolific Domains a priori. If Scrambling is driven by a pragmatic/semantic, referential feature, the prediction is made that the feature [AF] is +interpretable, yet not syntactic. What constitutes a syntactic feature is its ability to generate a plausible syntactic licensing relation that can both legitimate that functional element and spread to a lexical feature that requires such licensing (cf. Chametzky 2000:104). Contra Abraham and Molnárfi (2002, 2003), I dispose of a functional projection responsible solely for the licensing of the feature AntiFocus [AF]. Agreement features can be ‘rescued’ by non-local configurations such as Agree and Attract; however, the same luxury is not afforded to formal features responsible for discourse properties. The presence of [AF] in its current form does not improve our understanding of the ‘optionality problem’ involving West Germanic Scrambling; on the contrary, it is unfalsifiable and ad hoc in its application. 4.2.1
AntiFocus
Abraham and Molnárfi employ the concept of AntiFocus to explain shibboleth pragmatic morphology in West Germanic, in particular in Afrikaans. In modern spoken Afrikaans a morphological marker – the preposition vir derived from the German preposition für and Dutch voor – serves as a rheme marker for definite DPs.11 The origin for such morphology is believed to have entered the language by means of language contact with pidgin languages (e.g., forms of Low Portugese and Malay)12 during the early stages of Afrikaans’ development.13 Ponelis (1993:272) perceives that the link between Afrikaans’ objective ‘vir’ and Germanic proper transitive structure is extremely tenuous. Objective ‘vir’ is most probably, according to Polenis, a complete novelty, the origin of which has to be sought beyond Dutch and Germanic in general.
11
Den Besten (2000:949) attests that objective vir in Afrikaans can also marginally be attributed to nominal indirect objects. A more thorough investigation of objective vir in Afrikaans is necessary for this study in order to properly ascribe syntactic theory to account for its accurate application in the spoken language. 12 Ponelis (1993:272) notes the striking distributional similarities between objective per in varieties of Low Portuguese and objective vir in Afrikaans. Furthermore, Ponelis (1993:273) also argues that objective vir does not appear in documents from the Colonial Dutch period. 13 Swiss and Southern Austrian dialects exhibit a construction that builds dative nominals with prepositions without definite articles. It is unclear, however, if this is unique for definite DPs that mark rhematic XPs. Therefore, it is uncertain that these constructions are related with objective ‘vir’ in Afrikaans (Schirmunski 1962:439-40).
158
Scrambling and the Survive Principle
Data set (4.15) demonstrates definite D/NPs bearing the cluster [- TH, + Base] (TH = thematic) that undergo XP-Scrambling are destressed, whereas (4.16) these XPs with the [- TH, + Base] matrix that remain in situ are not only marked by prosodic stress but also the shibboleth rhematic marker vir. In other words, the connection between rhematic definites and prosodic stress assignment – albeit not universal for all languages families – is present in West Germanic. (4.15) a. b.
(4.16) a. b.
dat ek die meisie [VP gister gesoen het] that I the girl yesterday kissed have dat ek haar [VP gister gesoen het] that I her yesterday kissed have ‘that I have kissed her yesterday’ (vir) die MEISIE gesoen] Ek het [VP gister I have yesterday (for) the girl kissed Ek het [VP gister *(vir) HAAR gesoen] I have yesterday (for) her kissed ‘I have kissed her yesterday.’
The stressability conditions of scrambled and non-scrambled definites seem somewhat relaxed if compared to German and Dutch. The particle vir surfaces whenever definites do not scramble, and it is omitted whenever definites are realized to the left of adverbials, i.e. in situ. According to Molnárfi (2002), the use of vir is always obligatory with pronouns to the right of adverbs and is clearly preferred with definite full nouns in such positions, at least in the spoken variants of Afrikaans where written language standardization has little influence. Polnelis differs slightly with his description of the distribution of objective ‘vir ‘ in modern spoken Afrikaans. Polnelis (1993:266) argues that the “variables governing the use of objective vir include: category, animacy and position of the noun-phrase/object; emotivity and formality.” He claims that the NP category hierarchy for vir is: proper nouns >> pronouns >> substantives. Crucially, Polnelis (1993:267-68) shows that “the occurrence of objective ‘vir’ is more restricted within the clausal core. Proper nouns, which almost invariably combine with vir, and pronouns may occur without vir in the core…Outside the sentence core, vir is obligatory.” Essentially, both Polnelis and Molnárfi accurately attribute objective ‘vir’ to D/NPs that rank ‘high’ on a definiteness scale (which will correlate more strongly with any animacy scale) and prosodic prominence assignment. Assuming that the basic model of the grammar is inherently correct, we face a dilemma in accurately describing how and when objective ‘vir’ enters
Chapter 4. The Prosodic Side of Scrambling
159
the derivation. We cannot simply postulate that this rheme marker is inserted after movement has taken place. Such an assumption would violate the concepts of Strict Cycle and the Inclusiveness Condition (Chomsky 1995). This is where AntiFocus comes into play. Abraham and Molnárfi argue that West Germanic possesses a destressing phrase – AntiFocus [AFP] – which subsequently exhibits the feature [AF] that motivates movement of definite DPs to [Spec,AFP] for checking purposes. AntiFocus [AF] is assumed by Abraham and Molnárfi to be the prosodic/discourse trigger motivating Scrambling in West Germanic languages:
(4.17)
[SpecAF α domain of destressing
[ADV [VP α domain of focus
V]]]
Attempting to unify their theory with current minimalism, Abraham and Molnárfi argue that the interpretable/uninterpretable distinction is too narrow to accommodate syntactically relevant peripheral features in a minimalist movement. (In other words, they seek to label this phenomenon as a 3rd kind of trigger.) By this they hope to achieve that AntiFocus is understood in “a strictly derivational sense” (Molnárfi 2003:152). This approach advocates that [AF] may encode instructions for both the core component of grammar and the interface levels. If the higher copy is spelled out (triggering overt Scrambling), checking of the [AF] feature in the relevant Spec-position and subsequent PF-deletion of the base copy will eliminate all prosodic features: (4.18)
[SpecAF α +AF
checking
[AF
+AF
[ADV
[VP
α +AF
V]]]
PF-deletion
In a nutshell, the principal raison d'être for the existence the [AF] trigger is to assign a feature to account for the optionality of movement chains with regards to Scrambling in languages such as West Germanic. Abraham and Molnárfi’s AntiFocus approach is not entirely unique to modern approaches to this phenomenon. For example, Sekerina (1997, 2003) argues for the trigger sentential stress [±SS] to account for Scrambling in Russian. Both Sekerina’s and Abraham and Molnárfi’s methodology strongly resemble Zubizarreta’s (1998) concept of p(rosodic)-movement, which states that when generated outcomes are prosodically ill-formed (i.e., when the Nuclear Stress Rule and the Focus
160
Scrambling and the Survive Principle
Prominence Rule come into conflict) destressed items will move to the right in order for another constituent to then be available to receive nuclear stress.14 In support of the AntiFocus approach, Abraham extends the feature matrix of EPP on T’ to possess [+N, + presupposed] (2003:19). This assertion is not too far from the advancement of a pragmatic/semantic feature responsible for licensing Scrambling. Fundamentally, a pragmatic feature such as [+ Ref] and [AF] are quite similar, with the only crucial difference being the licensing of discourse morphology of the latter. With this feature Abraham and Molnárfi successfully achieve the integration of discourse features in the lexicon without violating the Inclusiveness Condition. With this in mind, Molnárfi (2003:149) assumes the guiding intuition of early minimalism (ca. 1993) that is, relating the occupation of functional specifiers to definiteness within the D-feature of NPs was on the right track. Although I find Abraham and Molnárfi’s argument regarding the narrowness of the feature dichotomy interpretable/uninterpretable features intriguing and worthy of further research, problems exist under such an analysis. First, it is unclear how to account for the Scrambling of generic indefinites – which can and do scramble in West Germanic languages – within this framework. If the feature matrix [+N, + presupposed] on Agr-heads interacting with the feature [AF], we would have to find an alternative way to motivate the movement of generic indefinites. This also highlights the danger of putting too much stock into one non-lexical trigger. Second, if the [AF] trigger is universal in West Germanic that licenses Scrambling, how is shibboleth morphology triggered in Afrikaans yet not in other languages like Dutch, Frisian or German? Lastly, the derivational histories in (4.17) and (4.18) fail to explain how examples of pronominal fronting in Afrikaans – (4.19) and (4.20) – are licit: (4.19)
Vir hom vra ons niks. for him ask we nothing ‘We ask him nothing.’
(4.20) Vir haar terg hulle gedurig. for her tease they always ‘They are always teasing her.’
(Polnelis 1993:268)
If objective ‘vir’ is truly associated with rhematicy, then we can observe disharmony with Abraham and Molnárfi’s (2002) discourse properties based on natural clause partition due to the fact that definite DPs occupying a position in the CP-layer are [+TH]. The only possible solution to this situation would be to 14
Zubizaretta (1998) does not apply p-movement to German; rather, she asserts that defocalized XPs are metrically invisible to the Nuclear Stress Rule in German. Grewendorf (2005) argues that something similar to p-movement should exist in German as well. Essentially, this is the core of Abraham and Molnárfi’s AntiFocus-argumentation, too.
Chapter 4. The Prosodic Side of Scrambling
161
assume that objective ‘vir’ was generated in the derivation prior to moving further. It is unclear whether the fronted pronouns in (4.19) and (4.20) even reach a specifier position in the CP-layer (i.e., they could be in (Spec,TP), although it is unlikely.) If that is the case, it would make sense why the shibboleth morphology would still be generated. The occurrence of an AntiFocus FP at the left periphery of the vP also respects Anti-Locality; therefore, it is theoretically sound in this respect. Unfortunately, the optionality of movement in Scrambling that this claim seeks to obviate, or in a weakened state reduce, still seems to persist. In addition, the correspondence between information structure and prosodic stress assignment is consistent throughout for the movement chains of definites; however, it encounters difficulties in explaining the movement and prosodic behavior of indefinites. Recall from our discussion on the properties of Scrambling in West Germanic in Section 2.5.1 that definites will be destressed and thematic if staying in their own discourse chain outside of VP, while contrastively stressed and rhematic if staying inside of VP. Indefinites, on the other hand, are always rhematic (if not generic), being structurally identified by grammatical clausal accent (GA) if within VP and carrying heavy pitch accent if outside VP (Abraham and Molnárfi 2002:17-21). Therefore the [AF] feature in its current state does not solve all the problems that it originally was designed for when viewed in isolation. Mostly likely any pragmatic feature involved would remain agnostic to the prosodic properties of scrambled and non-scrambled XPs. Therefore, something in the theory is necessary to ascribe the correct distribution of prosodic stress in sentences exhibiting Scrambling, whether that is in the form of a formal feature in the derivation or not is the question we must attempt to at least come to terms with. By assuming that both a pragmatic feature such as [+ Ref] and an AntiFocus [AF] feature exist on definite objects that undergo Scrambling in West Germanic eliminates the problem of having a prosodic feature [AF] interpreted at both the LF and PF interfaces. Definites that remain in the VP lack the scrambling feature, and will thus be interpreted as rhematic, i.e., non-referential. If we argue that [AF] is also interpretable similar to [+ Ref], we can put forward the argument that this feature technically no longer needs to be checked in the narrow syntax. Consequently, we no longer need the functional project AFP at the left periphery of vP, either. If the functional projection that houses [+ Ref] in the TP-layer is also the home for [AF], the disassociation of both variants of [AF] can be viewed as a frustration of sorts: The feature [AF] cannot on its own motivate movement; only the referential Scrambling feature can elicit constituent displacement. This frustration then leads to the assignment of prosodic stress due to the fact that [AF] could concatenate with an identical [AF] feature higher in the syntax. The main reason I do not want to consolidate the features [AF] and [+ Ref] is based on the following simple fact: [AF] cannot accurately explain the Scrambling characteristics of indefinites from VP. An alternative,
162
Scrambling and the Survive Principle
yet somewhat radical, argument would be that it is the prosodic feature [AF] that calls the shots as far as movement is concerned, and the referential feature is simply along for the ride. The principle behind this assumption is that prosodic requirements hold top priority in the derivation (a la Büring 2001a, 2001b, Feng 1995, 2003 among others). The exact role of prosody in the model of the grammar will be discussed later in this chapter; however, it suffices to reiterate at this point that the feature [AF] is non-syntactic in that it does not have the potential of contributing to the construction of new projections. Therefore, a theory advocating a destressing feature [AF] and specified functional projection for it at the left periphery of vP should be not advanced. Returning to the central claim of this work – that middle field Scrambling in West Germanic is licensed by a +referential feature – definites and indefinites will be repelled from their base position in the VP into a higher domain at the level of prosodic structure by the presence of this feature. If the XP undergoing Scrambling licensed by the presence of [+ Ref] is definite, the nonsyntactic feature [AF] essentially piggybacks itself onto the displaced constituent in the middle field. Appealing to Ackema and Neeleman’s version of Prosodic Checking that advocates the deletion of agreement allomorphy in the same prosodic domain, I propose that this mechanism provides a more parsimonious account to the licensing of shibboleth morphology in Afrikaans than stipulating [AF] as a grammatical operation handled in dynamic syntax. (4.21) {ΦΔ [A AF F2]j … [B AF F3]k …} {ΘΔ tj tk} Feature deletion of [AF]: {[A AF F2] … [B AF F3] …}
Spell-out of ΦΔ: [A F2] → /a’/ In the derivation schematized above in (4.21), the phonological phrases A and B are both present in the middle field, Grohmann’s Φ-domain. Assume that A represents the scrambled definite XP whereas B could hypothetically represent another XP, but it is most probable that the feature [AF] in B resides on the head on a functional project in the middle field. The syntactic Scrambling of A brings both A and B within the same Prolific Domain (i.e., the Φ-domain) prior to PF. In (4.21) let us further assume that scrambling from the vP to the middle field also represents movement to a higher prosodic domain. The absence of [+ Ref] on a definite XP will render it –referential and thus prevent Scrambling from occurring. The failure of this XP to scramble into the middle field will also prevent the deletion of [AF] at PF.
Chapter 4. The Prosodic Side of Scrambling
163
(4.22) {ΦΔ … [B AF F3] …} {ΘΔ … [A AF F2] …} Feature preservation [AF] at PF: {ΦΔ … [B AF F3] …} {ΘΔ … [A AF F2] …} Spell-out of ΘΔ: [A AF F2] → /a/ Again, assume that the phonological phrase A represents a definite XP; however in contrast to (4.21), A in (4.22) lacks the pragmatic feature [+ Ref]. As a result of a lack of the pragmatic feature, A remains in its base position in the vP, or Θ-domain. Feature deletion qua Prosodic Checking is barred in (4.22) based on the fact that the phonological phrases A and B exhibiting the [AF] feature exist in different prosodic domains at the completion of each respective Prolific Domain. The existence of [AF] post-prosodic restructuring triggers the existence of shibboleth morphology, in this case objective ‘vir’. Under this view, Scrambling in Afrikaans is identical to the middle field Scrambling present in other West Germanic languages. The status of [AF] is a morphological feature unique to Afrikaans, acquired most likely through language contact during its linguistic development. My explanation of the true nature of [AF] and its active role at the PF-interface rather than in the narrow syntax renders the existence of an additional functional phrase responsible for licensing [AF] at the left periphery of vP as irrelevant on the grounds of virtual conceptual necessity (VCN) considerations. My analysis of the true role of [AF] and other features/elements that initiate p-movement effects bring relevant questions to the surface about the role of larger derivational chunks, multiple Spell-Out, and the timing of the application of prosodic properties to name but only a few. First addressing the question of multiple Spell-Out in combination with the idea of larger derivational units (e.g., phases and Prolific Domains) in tandem, recall that the operation Link! concatenates two syntactic objects to form a unit that is both automatically LF and PF interpretable. As a result, this information is immediately shipped off to the external interfaces for computation. The revised Minimalist Program according to the Survive Principle allows us to eliminate the notion of these larger derivational units with regard to the ‘interface independence’ status. In other words, multiple Spell-Out is an unnecessary component of the grammar that we can rightly discard. Second, the properties of p-movement must be discerned. Although p-movement invokes a reaction at PF, is it visible at LF? This, of course, depends on our definition of LF: Traditionally LF has been viewed as an interface component concerned solely with truth-conditional semantics. If we wish to maintain this idea, we are forced to the conclusion that pragmatic interpretations – and the features that motivate such readings –
164
Scrambling and the Survive Principle
are not strictly properties of LF. According to my analysis of discourse/semantic features, namely [+ Ref], this is absolutely correct: [+ Ref] being a syntactic feature has both a LF and PF reflex; hence, it’s not purely an LF epiphenomenon. Here we see a fundamental distinction once again between prosodic features such as [AF] and pragmatic ones like [+ Ref]: The former lack all-important LF-reflexes. According to our revised principle of Full Interpretation (cf. (3.6)), p-movement and its motivating features fall outside the scope of standard syntactic operations. Lastly, the challenge of properly determining when p-movement takes place in the course of the derivation is perplexing: Since p-movement has been shown to occur within strictly local domains, should we assume that p-movement occurs in a cyclic, derivational manner quite similar to syntactic operations, or can the assignment of prosodic stress be postponed until the completion of the entire representational unit? At this point, I would just like to plant this seed of a question and let it fester until I provide a more detailed answer in Section 4.5. 4.3
Permuted Word Orders in the Middle Field
One of the most pervasive issues confronting any analysis of middle field Scrambling that advocates a fixed position/established functional head is the possibility of permuted word orders in German. (4.23) [CP dass [TP der Bewährungshilferi dem Mannj das Geldk [vP gestern ti tj tk zurückgegeben hat]]] that the probation officerNOM the manDAT the moneyACC yesterday backgiven has (4.24) [CP dass [TP der Bewährungshilferi das Geldk dem Mannj [vP gestern ti tj tk zurückgegeben hat]]] that the probation officerNOM the moneyACC the manDAT yesterday backgiven has ‘that the probation officer gave the money back to the man yesterday.’ As discussed in Chapter 2, German allows the permutation of objects in the middle field even if they differ from the base order of objects in the vP. Although a symmetric representation of the objects in the middle field as they appear in the vP is the unmarked order (cf. (4.23)), alternate orders are possible (cf. (4.24)). In a phase-based derivational system, XPs could move to a multiple specifier of vP, i.e., to the ‘edge’ of a phase. This makes XPs eligible for further operations in the next immediately dominating strong phase.
Chapter 4. The Prosodic Side of Scrambling
(4.25) a. Step One (lexical insertion): b. Step Two (multiple specs):
165
[vP n o p] vP 2 n vP 2 o vP 2 p vP 2 vo …
c. Step Three (permuted orders): [CP n o p], [CP n p o] … Since multiple specifiers are not interpreted as separate functional projections building onto the structure, constituents that reside in multiple specifiers are defined as being equidistant from an attracting head and/or feature in the next strong phase. Prior to the construction of an immediately dominating higher phrase, arguments positioned in multiple specifiers are equidistant from both their governing head (in this case vo) and any operations that will take place in the next phase (cf. (4.25b)). With the status of multiple specifiers in the current theory under investigation in questionable status (due to Stroik’s (2007) strict understanding of locality consisting of only one true specifier and a functional head), it is unclear how to attain these permutations in the middle field. In a more local sense, it is difficult to support an analysis of movement to a unique specifier for featurechecking, or in the revision of minimalism support here, for linking purposes. To circumvent this shortcoming, Putnam (2006a) suggests a potential alternative solution outside of the syntax involving prosodic restructuring. According to this assumption, the base order lexical insertion of objects is respected in movement from the vP (i.e., Θ-domain) to the middle field/TP-layer (i.e., Φdomain) and that permuted word orders are the result of linearization principles at PF. This view of permuted word orders resulting from middle field Scrambling – which may in the end be correct – removes the responsibility from the syntax and makes these permuted word orders a result of PF linearization principles rather than structural concerns. Under this view, perhaps both items moved to fixed positions in the middle field and participate in a re-linearization process to conform to sentential stress patterns. Considering the apparent order-preservation effects holding between XPs, as illustrated in (4.23) and (4.24) above, one way of approaching this problem would be to follow Bobaljik (2002, 2004) and Anagnostopoulou (2004) in assuming that such effects boil down to syntactic intervention rather than PF-
166
Scrambling and the Survive Principle
linearization. The very fact that Holmberg's Generalization (i.e., verb-object shape) is inviolable (like linearization itself!) whereas both nesting and crossing patterns are possible (albeit one of them is slightly more degraded) suggests that they have a different source/explanation. Logic would predict that crossing paths should be preferred over nesting paths; therefore an intervention account for the preference of crossing over nesting along the lines of Anagnostopoulou would be in order. Thus, assuming a base structure where the indirect object (IO) is structurally higher than the direct object (DO), i.e. [IO V DO] (cf. 3.54c), IO will intervene for movement (i.e., Scrambling) of the DO; however, if they both scramble, as in examples above in (4.23) and (4.24), then the IO intervener is displaced ‘out of the way’ first, to Spec,vP, allowing the DO to then also scramble to Spec,vP. The preference for the crossing order over the nesting order in may then simply reduce to something like a preference for Norvin Richards-style ‘tucking-in’ over not tucking-in. Like Anagnostopoulou (2004), Richards (2005:8) assumes that tucking-in is what is responsible for order-preservation effects amongst arguments/XPs targeting the same head. This answer, of course, presupposes the existence of multiple specifiers in the narrow syntax. The question of whether structures such as multiple specifiers can or should exist in Survive-minimalism is yet to be addressed. On the one hand, with the elimination of the distinction between XPadjunction (at least from discourse/semantic purposes) and traditional movement of an XP to the specifier of a functional projection, the potential to arrive at structures with multiple specifiers that also blur the specifier vs. adjunct contrast (cf. Kayne 1994) does exist. However remaining consonant with Stroik’s (2007) idea of strict locality in the Minimalist Program, the creation of syntactic objects occurs only by means of an immediately local (i.e., between an Xmax and Xmin sharing the same feature)15 operation forming a larger syntactic object that is interface interpretable qua Link! At this point I leave the existence of multiple specifiers open for debate; however, I wish to parse the distinction between multiple specifiers and the notion of a phonological edge that purportedly exists at the edge of strong phases. The notion of a phonological edge, which according to Chomsky (2006) is shaped and force constituent distal by means of ‘edge features’16, is not evidenced in other biological processes. In the section that proceeds, I will demonstrate on conceptual grounds how the concept of a phonological edge at the right periphery of vP fails to account for natural data. Perhaps the solution to understanding the permuted word order possibilities of XPs in the middle field lies outside the realm of formal syntactic opera15
However see Putnam and Stroik (in progress) for a unification of head movement and XPdistal. 16 It is unclear at the present time how these ‘edge features’ compare and contrast with the previous notion of EPP.
Chapter 4. The Prosodic Side of Scrambling
167
tions and in the phonological linearization principles as argued by Büring (2005) and Grohmann and Putnam (in press) most recently. Richards (2005:7) proposes that the defining property of Scrambling is the preservation of linear head-complement ‘shape’ (ordering relations)17, which emerges as a direct consequence of a symmetrical system in which base order (Verb-Object [VO] and Object-Verb [OV]) is a linearization effect imposed only at PF. As pointed out to me by Marc Richards (p.c.), such a linearization/shape algorithm as employed in his dissertation strictly applies only to sisters (i.e., the headcomplement relation) and therefore has nothing to say about apparent orderpreservation effects holding between XPs. The Scrambling of both objects in ditransitive verb phrases is highly marked, regardless of whether or not a crossing or nesting path is involved. Although a head-complement shape preservation constraint may not address this problem, an explanation of this phenomenon may still fall within the boundaries of PF restructuring. Finally, it is unclear how Prolific Domains can address this situation. One potential solution would be to argue that each Prolific Domain is uniquely equidistant from one another. This solution is, however ad hoc and increases the computational workload of the CHL, which is hardly desirable in a minimalist view of computation. In a phase-based account of syntax, the only elements available for further participation in any operations are those that exist at the ‘edge’ of the phase, i.e. the head of the strong phase and any constituents positioned in multiple specifiers of that particular strong phase. By arguing that Prolific Domains are innately equidistant from one another would require keeping an entire contextual unit alive in the derivation while the system actively composes another. This would be necessary to allow all of the potential permuted word order variation in the middle field of a language such as German that licenses such structures. This idea runs contrary to Grohmann’s (2000a, 2000b, 2003, 2006) vision of how Prolific Domains interact with a multiple Spell-Out version of the grammar; Grohmann assumes that upon their completion Prolific Domains locally and immediately undergo Spell-Out. Herein lies one of the principle problems with the notion of multiple Spell-Out and the ‘weakly representational’ status of current ‘orthodox’ minimalism as exhaustively pointed out by Brody (1995, 1998, 2002). The derivation operates in a dynamic fashion creating interface interpretable concatenative units by means of extremely local primitive operations. In this regard, the derivation is always ‘open’, however according to Survive-minimalist desiderata; the system’s reliance on any look-ahead or look-back operations can be done away with once and for all. The bigger issue that confronts Prolific Domains is in regard to the nature of the contextual identity of the middle field: It is not possible to envision a rigorous definition of the middle field contextualized strictly 17
See, e.g., Collins and Thráinsson 1996, Müller 2000, 2001, Fox and Pesetsky 2003, Williams 2003, Bobaljik 2004.
168
Scrambling and the Survive Principle
for agreement relations, especially in a language like German that as successfully argued by Abraham (1997) and later subsequent works, exhibits a clausal type driven by discourse configurationality. In the final section of this chapter (Section 4.5) and in the concluding chapter of this book (Chapter 5), I will turn my attention directly to the task of properly redefining the middle field and attempt to show how an ontological system similar to Prolific Domains can survive (pun intended) and remain in Survive-minimalism. 4.4
The Coherent Infinitive Puzzle
Before discussing the issues that surround coherent infinitives and their relationship to middle field Scrambling, a brief introduction into the word order properties of pronominals must be introduced. Unlike full DPs, pronouns must adhere to a strict linear order as illustrated in examples (4.26) and (4.27) below (from Boeckx and Grohmann 2004: 149). (4.26) [Seine Mutter]i, [der Wagen]j, [der Tobi]k his mother the car the Tobi (4.27) a.
ihnj ihmk geschenkt hat. ...weil siei because sheNOM him(Tobi)ACC itDAT given has ‘because she gave it to him.’
a.
*...weil sie ihm ihn geschenkt hat.
b.
*...weil ihn sie ihm geschenkt hat.
c.
*...weil ihn ihm sie geschenkt hat.
d.
*...weil ihm sie ihn geschenkt hat.
e.
*...weil ihm ihn sie geschenkt hat.
In (4.26), the nominative forms of all of the full DPs are provided. As shown in (4.27a), the order of the pronouns in unmarked structures must strictly adhere to the SUBJ > DO > IO linear pattern. The following examples show that any derivation in this pattern results in an ungrammatical result. In double object constructions that consist of both a full DP and a pronominal element, the pronoun also precedes the full DP, as shown in (4.28) below. (4.28) a.
...weil seine Mutter ihm den Wagen geschenkt hat. has because his motherNOM himDAT the wagonACC given ‘because his mother gave him the wagon.’
Chapter 4. The Prosodic Side of Scrambling
b.
169
*....weil seine Mutter den Wagen ihm geschenkt hat. has because his motherNOM the wagonACC himDAT given
The reason for introducing pronouns into the discussion at this point is due to their phonological and prosodic properties. Unless these function as demonstratives, pronouns are prosodically weak items. This leads into our discussion of coherent infinitives and their interaction with middle field Scrambling. The puzzling behavior of German coherent infinitives with respect to islandhood properties (cf. Grewendorf and Sabel 1994, Müller 1998) provides another example at the complex interaction of the prosody-syntax interface. Extraction out of coherent infinitives in their base position is permitted independently of any kind of movement. Thus, as originally shown by Grewendorf and Sabel and Müller, from whom the following data are taken, scrambling of a nominal (4.29a) or pronominal (4.29b) argument as well as topicalization (4.29c) and wh-movement (4.29d) result in grammatical structures, provided that the coherent infinitive appears adjacent to the selecting main verb (cited from Chocano 2004:44). (4.29) a.
…dass das Buchi keiner [ ti zu lesen ] versucht hat. that the bookACC no-oneNOM to readINF tried has
b.
…dass esi keiner [ ti zu lesen ] versucht hat. that itACC no-oneNOM to readINF tried has ‘… that no one tried to read the book / it.’
c.
Das Buchi hat keiner [ ti zu lesen ] versucht. The bookACC has no oneNOM to readINF tried ‘The book, no one tried to read.’
d.
Wasi hat keiner [ ti zu lesen ] versucht? whatACC has no-oneNOM to readINF tried ‘What did no one try to read?’
When we consider scrambled coherent infinitives, a different situation arises. Although fully nominal elements cannot scramble from coherent infinitives, it is still possible for pronominal arguments to participate in this type of displacement, as the contrast between (4.30a) and (4.30b), originally from Müller (1998), as well as (4.30c) shows, where both subject and object are pronominal (an adverb may appear after es ‘it’, for example, but not before) (data taken from Chocano 2004:44-45). (4.30) a.
* …dass [ ti zu lesen]j das Buchi keiner tj versucht hat. that to readINF dasBuchACC no-oneNOM tried has
170
Scrambling and the Survive Principle
b.
…dass [ ti zu lesen]j esi keiner tj versucht hat. that to readINF itACC no-oneNOM tried has
c.
esi tj versucht hat. …dass [ ti zu lesen]j er that to readINF heNOM itACC tried has ‘… that no one / he tried to read the book / it.’
According to Müller, the difference in grammaticality between (4.29a) and (4.30a) boils down to a single constraint: (4.31) Principle of Unambiguous Domain (Müller 1998:241) An α-trace must not be α-dominated. Under this constraint, an ‘α-trace’ is understood to be a “trace with a (not necessarily c-commanding) antecedent in a position of type α” and ‘α-dominated’ means “dominated by a category in a position of type α” (Müller 1998:241), where ‘type α’ is determined by the different landing sites (e.g., Spec,CP, XPleft adjoined positions, and so on). (4.32) a. b.
…dass das Buchi keiner [ ti zu lesen ] versucht hat. that the bookACC no-oneNOM to readINF tried has * [ ti zu lesen ] hat das Buchi keiner tj versucht. to readINF has the bookACC no-oneNOM tried ‘… that no one tried to read the book.’
According to the Principle of Unambiguous Domain, (4.32b) is ill-formed because the infinitive occupies a position of a type different from the one in which the scrambled object appears (according to Müller, Spec,CP vs. adjunction to IP). As argued by Chocano (2004:49), however, the Principle of Unambiguous Domain should also rule out (4.33): (4.33) … dass [ ti zu lesen] esi keiner tj versucht hat. that to readINF itACC no-oneNOM tried has ‘… that no one tried to read it.’ To solve this problem, Müller claims that the difference between moving a full DP object such as das Buch ‘the book’ and a pronominal object such as es ‘it’ is that the latter does not undergo Scrambling but targets a landing site other than adjunction to IP. What Chocano claims is that the mechanism involved in determining the grammaticality of these linear strings relies purely upon PF-
Chapter 4. The Prosodic Side of Scrambling
171
processes. A licit ‘double scrambling’ construction — i.e., scrambling of the infinitival clause along with scrambling of one of its internal arguments — is possible, if and only if it affects pronominal objects. Chocano postulates that the crucial difference between ungrammatical and grammatical double Scrambling structures (cf. (4.32b) vs. (4.33)) is that phonological processes repair the damage caused by the application of chain reduction to the infinitival clause, turning it into a phonologically complete constituent again. This is similar to Selkirk’s (1986) universal theory of phrasing, according to which the edge (be it right or left) of a phonological phrase φ is always aligned with the edge of a syntactic XP depending on the head parameter of a given language. In her analysis of the phonological aspects of Scrambling in German, Chocano (2004) employs a phase-based version of minimalism. Following Chomsky (2001), Chocano envisages DISL(ocation) as a phonological operation that strips the (dislocated) constituent of its phonological features, transferring them to PF. This differs from regular Spell-Out procedures in that it is not linked to phase completion, and caused the dislocated string to be phonologically realized in a position preceding the hypothetical syntactic one, i.e. at the phonological edge of the strong phase vP. This proposal – as postulated by both Chocano and Chomsky – is flawed and incompatable with Surviveminimalism based on the following arguments: First, according to Chomsky (1993, 1995 and later work), all copies of displaced constituents must be identical. If they are forced to share the same bundle of features, the same lexical item and its features must participate in the iterative application of Merge. As argued thoroughly throughout this manuscript, this assumption that the syntactic operations insert the exact same lexical item iteratively throughout the Derivation (by means of Internal Merge/Move) leads to a weakly representational computational system. Although items remain alive (i.e., they survive successive Merge-Remerge-Link!-Survive operations) in the Numeration, the actual lexical item and all of its corresponding features remain in the Numeration throughout the course of the derivation. The lexicon is thus contained within the interfaces. What does appear in the course of narrow syntactic operations is the relevant feature identity necessary for the immediately local application of Remerge and Link! Second, in order for the DISL-operation to work for Chocano she is forced to propose that DISL raises the phonological features of the scrambled object at the edge of the lower strong phase to a higher one (CP), which explains why the shifted object appears phonologically realized past the vP phase edge. To illustrate DISL in action, consider the grammaticality contrast between (4.34a) and (4.34b) (taken from Chocano 2004:57). (4.34) a.
…dass das Buchi keiner [ti zu lesen] versucht hat. that the bookACC no-oneNOM to read tried has
172
Scrambling and the Survive Principle
b.
*…dass [ti zu lesen] das Buchi keiner ti versucht hat. that to read the bookACC no-oneNOM tried has ‘that no one tried to read the book’
Chocano’s claim is that such a contrast as observed in (4.34) is due primarily to one of the core properties of DISL, namely the requirement that elements undergoing any of the operations involved in DISL possess phonological features. The constraint is respected in (4.34a), but violated in (4.34b), as the derivations below demonstrate (adapted from Chocano 2004:57-60). (4.35) …dass das Buchi keiner [ti zu lesen] versucht hat. that the bookACC no-oneNOM to read tried has
Narrow Syntax: a.
vP-cycle: V raises to v, das Buch is pied-piped to the vP edge to satisfy [EPP]
b.
vP[das
c.
CP-cycle: merging of T and C (Aux projection and Aux-to-T movement omitted)
d.
CP[ C[dass TP[ T[ vP[das
Buch sucht]]]
vP[keiner v[
VP[[das
Buch zu lesen] versucht] ver-
Buch vP[keiner v[ VP[ [ das Buch zu lesen] versucht] versucht]]] hat]]]]
PF: e.
Spell-Out: phase-bound transfer to PF (vP) das Buch zu lesen] versucht] veruscht]]
vP[keiner v[ VP[[
f.
Dislocation 1: elimination of phonological features from the narrow syntax (hereafter NS): vP[ das Buch vP[ keiner v[ VP[[das Buch zu lesen] versucht]]]
g.
Chain reduction and PF output for PF at CP: das Buch keiner das Buch zu lesen versucht versucht
h.
Spell-Out phase bound transfer to PF (CP) hat]]]]
CP[ C[dass TP[ T[
Chapter 4. The Prosodic Side of Scrambling
173
i.
PF-input (Spell-Out of CP plus PF of vP) dass das Buch keiner das Buch zu lesen versucht versucht hat
j.
Dislocation 2: (raising of phonological features): dass das Buch 0 keiner zu lesen versucht hat
k.
Chain reduction (vacuous): dass das Buch keiner zu lesen versucht hat
(4.36) *…dass [ti zu lesen] das Buchi keiner ti versucht hat. that to read the bookACC no-oneNOM tried has ‘that no one tried to read the book’
Narrow Syntax: a.
vP-cycle: V raises to v, das Buch and das Buch zu lesen are pied-piped to the vP edge to satisfy [EPP]
b.
vP[das
c.
CP-cycle: merging of T and C (Aux projection and Aux-to-T movement omitted)
d.
CP[ C[dass TP[ T[vP[das
Buch zu lesen vP[das Buch vP[keiner v[ VP[[das Buch zu lesen] versucht] versucht]]]]
Buch zu lesen vP[das Buch VP[[das Buch zu lesen] versucht] versucht]]]] hat]]]]
vP[keiner v[
PF: e.
Spell-Out: phase-bound transfer to PF (vP): Buch zu lesen] versucht] versucht]]
vP[keiner v[ VP[[das
f.
Dislocation 1: elimination of phonological features from NS: Buch zu lesen vP[das Buch vP[keiner v[ VP[[das Buch zu lesen] versucht] versucht]]]]
vP[das
g.
h.
Chain reduction and PF output for PF at CP: das Buch zu lesen das Buch keiner das Buch zu lesen versucht versucht
Spell-Out phase-bound transfer to PF (CP): hat]]]]
CP[ C[dass TP[ T[
174
Scrambling and the Survive Principle
i.
PF-input (Spell-Out of CP plus PF of vP): dass das Buch zu lesen das Buch keiner das Buch zu lesen versucht versucht hat
j.
Dislocation 2: (raising of phonological features): dass das Buch zu lesen das Buch 0 0 keiner versucht hat
k.
Chain reduction (vacuous): dass zu lesen das Buch keiner versucht hat
The critical difference between (4.35) and (4.36) is as follows: The notable distinction relates to chain reduction at the stage of the vP-cycle: Scrambling the object das Buch ‘the book’ makes it the head of its respective movement chain, which entails that the phonological features of the rest of the copies of this movement chain must be eliminated. Here we see first hand the complexity of the phase-based derivational approach to syntactic derivation. Not only is there a shift of phonological features/elements that follows the completion of vP, but there is also a necessity of chain reduction that must occur (4.35g, k) and (4.36g, k) to license this construction. There exists a much simpler way of accounting for the grammaticality of (4.35) in comparison and the ungrammaticality of (4.36). As could be expected, this difference between these two structures does not violate syntactic requirements, but rather phonological ones. If pronouns are classified as prosodically deficient elements, this provides the proper constraint necessary to explain how pronouns are licit in double Scrambling configurations. (4.37) a.
…dass zu machen’s jeder versucht hat. that to doINF-CL.ACC everyoneNOM tried has
b.
…dass zu machen das jeder versucht hat. that to doINF thatACC everyoneNOM tried has
c.
* …dass zu machen der Peter das versucht hat. that to doINF the PeterNOM thatACC tried has
d.
…dass zu machen er es versucht hat that to doINF heNOM itACC tried has ‘… that everyone / Peter / he tried to do that / it.’
The assumption is that the phonological phrase der Peter in (4.37c) prevents phonological alignment with the infinitival phrase at PF. Leaving aside a detailed account of phonological reduction at PF, we see that Scrambling into the middle field in double Scrambling structures is sensitive to both phonological and syntactic considerations. Independent phonological phrases cannot land in
Chapter 4. The Prosodic Side of Scrambling
175
the TP-layer in derivations involving remnant movement. The subject der Peter in (4.37c) merges into the vP to expel its agentive θ-role. The DP then survives into the middle field to check its Case feature on T. In this sense, the derivation from a syntactic perspective is well-formed. The completed grammatical string is evaluated as ill-formed at PF. Rather than relying on operations such as chain reductions and rich ontological commitments such as strong phases, Survive-miniamlism is observationally adequate in design, meaning this system allows us to see that although (4.37c) is syntactically acceptable, it is phonologically misconfigured at PF. Perhaps of prime importance, we see that certain phenomena such as permuted word orders in the middle field and Scrambling from coherent infinitives are marked as ungrammatical at the level of prosodic structure rather than by the evaluation of syntactic structure alone. 4.5
Implications for the Model of the Grammar18
The dynamic computation of syntactic structure according to Surviveminimalism generates syntactic objects that are at once interpretable at both the PF and LF interfaces. Again, in the spirit of Epstein and Seely (2006:13), each application of Merge and Remerge is itself an independent manifestation of the traditional Y-model of the grammar. The question remains as to where we should properly house prosody in such a model of the grammar. On the one hand, as shown by a host of scholars over the last four decades, the assignment of prosodic stress based on syntactic structure occurs in extremely local environments. The reason for not introducing purely prosodic features into the narrow syntax is clear given our revised definition of Full Interpretation (cf. (3.6)) that requires ‘syntactic’ features and operations to create structures that are interpreted at both interfaces. Prosodic features are by default purely PF when they stand alone, however as demonstrated in my analysis of the AntiFocus feature hypothesis in Section 4.2.1, these features may act in tandem with other ‘syntactic’ features. In this regard, prosodic features and procedures should be separated from other PF operations (cf. (4.11)), and be interpreted as grammatical processes that are computed outside of the narrow syntax.19 With this being said, I do not wish to separate prosody from PF, but recognize that the ‘road to PF’ is a complex one, with many levels of which prosody is but only one of them (see Grohmann and Putnam (in press)). The next pertinent question regarding the prosody-syntax interface is the issue of when these prosodic features and properties interact (or are recognized by the human language faculty) with the syntactic structure. Due to the adher18
Many of the ideas for the reformulation of the contextual identity of Prolific Domains were developed through conversation and collaborative work with Tom Stroik. 19 However, see Feng (1995, 2003) who puts forward the radical claim of a bi-directional model of the grammar involving interaction between syntactic and prosodic structures.
176
Scrambling and the Survive Principle
ence to locality rules at the level of the syntax-prosody interface, one potential argument is to view prosodic assignment as a derivational procedure. According to a theoretical system of minimalism that advocates the existence of interphase-independent propositional units known as phases, movement to the ‘edge’ of a strong phase can be induced for solely prosodic purposes. In his most recent treatment of phases, Chomsky (2006) boldly suggests that movement to the edge of these larger derivational units is for purely phonological reasons motivated by the existence of ‘edge features’, whose purpose and identity are less than clearly defined. As stated throughout, calling upon the existence of abstract larger propositional units is taxing to the faculty of human language in many ways. Where this system fails to account for the prosodysyntax relation involving scrambled middle field constituents is its inability to explain reconstruction effects and anaphoric binding relations. If movement to the edge of the phase is purely phonological, one would not expect any remnant A-movement characteristics such as those just mentioned. Linearization processes enacted by prosodic considerations should not generate syntactic structure, which in our model of minimalism is only possible with the existence of a syntactic head (Xmin) bearing a syntactic feature. Once again, the phase-based version of minimalism fails on both conceptual and empirical grounds. The challenge of whether we should interpret the prosody-syntax interface as a derivational or representational mechanism is a rather difficult one. For the time being, I would like to support the claim made by Grohmann and Putnam (in press) supporting a representational view of dynamic stress assignment and its interaction with a derivational model of syntax. The basis of their claim rests on two fundamental assumptions: First, prosodic domains are read off of syntactic structure, mediated, of course, through IS-interpretation requirements. The formal grammatical features that are active at the level of prosodic structure are latent syntactically. Take, for example, the AntiFocus [AF] feature. Second, Accent Domains (cf. Büring 2001a, 2001b) are interpreted as the optimal mapping units between syntactic and prosodic structure and are not synonymous with Prolific Domains. Speaking first on the role of prosodic features in determining dynamic stress assignment, Putnam (2006a) and Grohmann and Putnam (in press) make use of Gussenhoven’s (1983, 1992) work that proposes as a solution the Sentence Accent Assignment Rule (SAAR). Based on previous studies (cf. Pierrehumbert 1980, Beckman and Pierrehumbert 1986, Hayes and Lahiri 1991, Hayes 1995, among others), the SAAR rests on the assumption that phrasal stress is assigned first, and then serves as the anchor for pitch accents at the intonation level. The SAAR is thus composed of two parts: First the assignment of prosodic stress relative to the syntax takes place and then the assignment of pitch accent at the intonation phrase level to each element carrying
Chapter 4. The Prosodic Side of Scrambling
177
prosodic stress is carried out. Gussenhoven presents a view of accent assignment that refers to intonational groups and the argument/adjunct distinction. The basic properties of his original formulation of the SAAR can be summarized as in (4.38): (4.38) Basic Properties of the SAAR (Krifka 2001/2:2) (i) Constituents in focus are marked by underlining. (ii) A, P, C stand for “argument”, “predicate”, “condition” (roughly, adjunct), X and Y for constituents in general. (iii) Accent domains (intonational groups) are marked by brackets [ ]. (iv) Accent is marked by *. As just mentioned, the SAAR proceeds in two steps: First a complex constituent is grouped in domains, then the pitch accent is assigned within domains. The definition of the SAAR below illustrates the basic derivational steps in the composition of domains and subsequent prosodic assignment that occurs. (4.39) Sentence Accent Assignment Rule (Gussenhoven 1983:391) a. Domain assignment: i. P(X)A → [ P(X)A ] A(X)P → [ A(X)P ] ii. Y → [Y] b. Accent assignment i. [] → [*] ii. In [ A(X)P ] or [ P(X)A ], accent A.
Krifka (2001/2:2), who minimally modifies what is now (4.39b.ii) of Gussenhoven’s original definition, explains the rules in (4.39) as follows: Rule (a.i) says that if a predicate and an argument is in focus (even if there is an intervening constituent that is not in focus), then predicate and argument (with the intervening material) form a domain. Rule (a.ii) says that every focused constituent forms a domain. (Probably a.ii applies only if rule a.i does not apply.) Rule (b.i) says that stress has to be assigned in every domain. Rule (b.ii) says that stress is realized on the argument within a predicateargument constituent that forms a domain.
Adopting the SAAR as the evaluative metric as far as prosodic stress assignment is concerned regarding syntactic hierarchy draws further support for the hypothesis put forward by Grohmann and Putnam (2004, 2005, in press) that Accent Domains function potentially as ideal intermediary units that map lower
178
Scrambling and the Survive Principle
levels of prosodic stress (PWd) to the intonation level. Consider the German examples in (4.40) that illustrate our adoption of the SAAR (taken from Grohmann and Putnam (in press)): (4.40) a.
Der Mann(a) hat(b) die Katzei(c) gestern(d) ti gesehen(e). seen the manNOM has the catACC yesterday ‘The man saw the cat yesterday.’ ΠΔ assignment: AD assignment: iP representation:
b.
[ΦΔ a b c ] … [ΘΔ d ti e ] [ΦΔ a* b c ] … [ΘΔ d ti e* ] [ *]
Der Mann(a) hat(b) gestern(c) die Katzei(d) gesehen(e). seen the manNOM has yesterday the catACC ‘The man saw the cat yesterday.’ ΠΔ assignment: AD assignment: iP representation:
[ΦΔ a b ] … [ΘΔ c d e ] [ΦΔ a* b ] … [ΘΔ c d* e ] [ * ]
As observed by Büring (2001a, 2001b), the argument will always receive prominent stress in any accent domain shared by a verb and its argument. The idea of adopting SAAR to explain the checking of prosodic features – such as AntiFocus [AF] (cf. Section 4.2.1 – clarifies the connection between syntactic and prosodic structure. So exactly what purpose do Prolific Domains serve in dynamic stress assignment? As a working hypothesis Grohmann and Putnam (in press) suggest that Prolific Domains serve as potential edge boundaries at the syntaxphonology (prosody) mapping interface. To illustrate this point, let’s revisit the claim that in German when both the direct and indirect objects of a ditransitive predicate scramble into the middle field focused direct objects cannot precede unfocused indirect objects, though unfocused direct objects can precede focused indirect objects (data adapted from Büring 2005:27) (Note: iP = intonation phrase, AD = accent domain and PWd = prosodic word level). (4.41) a.
( * )iP ( * )AD (Er hat) (dem PilotenF) (die Passagiere) (vorgestellt)PWd the passengersACC introduced heNOM has the pilotDAT
Chapter 4. The Prosodic Side of Scrambling
b.
179
( * )iP ( * )( * )AD (Er hat) (die Passagiere) (dem PilotenF) (vorgestellt)PWd heNOM has the passengersACC the pilotDAT introduced ‘He introduced the passengers to the pilot.’ (IOF – DO/ DO-IOF)
In the examples in (4.41), a focused IO in German can become right-peripheral within the iP by either forming an over-size AD with everything to its right, or be inverting canonical IO-DO order to DO-IO and forming an AD with the final verb only as in (4.41b). On the other hand, a focused DO forms a rightperipheral AD in its canonical, verb-adjacent position, and non-canonical order would actually create a more marked prosodic structure (i.e., one where DO, IO and V would have to form on AD). The supersaturated AD in (4.42b) is the result of the instatiation of this overly marked prosodic structure and is a blatant violation of the Accent Domain Formation constraint put forward by Büring (2001a: 82). (4.42) a.
( * )iP ( * )( * )AD (Er hat) (dem Piloten) (die PassagiereF) (vorgestellt )PWd
b.
( * )iP ( * )AD *(Er hat) (die PassagiereF) (dem Piloten) (vorgestellt)PWd ‘He introduced the passengers to the pilot.’ (IO – DOF/ *DOF – IO)
In sum, “focus on the IO yields either canonical order with marked prosody, or marked order with canonical prosody, while focus on DO ‘freezes’ the canonical word order, which is unmarked in all respects (cf. Lenerz 1977; Büring 2001a)” (Büring 2005:27). Again this also demonstrates that the permuted word orders in the middle field that result from Scrambling are prosodically rather than syntactically licensed. One last piece of evidence that convincingly shows that permuted word orders of scrambled objects into the middle field is purely a PF-regulated phenomenon is observed in the binding properties of these aforementioned objects. Recall Stroik’s argument that while referential objects can bind a pronoun, those that receive an attributive reading cannot. As illustrated in the (4.43) and (4.44), the IO der Frau ‘the woman’ receives a referential interpretation – and thus can be argued to exhibit the [+ Ref] feature necessary for Scrambling – and can bind a pronoun from either position.
180
Scrambling and the Survive Principle
(4.43) a.
b.
(4.44) a.
b.
Ich weiss, dass er der Fraui das Buch gegeben hat. I know that heNOM the womanDAT the bookACC given has Ich habe siei gestern angesprochen. I have herACC yesterday to-spoken ‘I spoke with her yesterday.’ Ich weiss, dass er das Buch der Fraui gegeben hat. I know that heNOM the bookACC the womanDAT given has Ich habe siei gestern angesprochen.
Since binding relations must hold at LF and the optional order of the IO in (4.43) and (4.44) does not have an adverse effect on either linear order, permutation of object order in the German middle field appears to be an operation determined solely by prosodic structure requirements. As mentioned in the introduction of this chapter and periodically throughout this entire book, Grohmann’s (2000a, 2000b, 2003) tripartition of natural clauses into three contextually defined derivational units is indeed meritous of further inspection. Although our revision of minimalism here no longer makes use of any sort of multiple Spell-Out of larger derivational chunks, some sort of spinal cord, as noted by Putnam and Stroik (in progress), must exist to ensure the proper order of Merge and Remerge operations will take place. Let us follow this assumption and suggest that the functional heads C > T > v similar to Grohmann’s mnemonics function as place holders for the application of both Lexical and Functional Merge as well as all other syntactic operations (i.e., Remerge and Link!). Furthermore let us assume that each of these main functional heads define the contextual identity for the immediate domains which they govern over. According to Grohmann, this would have the vP-layer responsible for thematic relations, the TP-layer (i.e., middle field) in charge of agreement relations and the CP-layer (i.e., left periphery) responsible for discourse properties. Putnam (2006a:Ch. 4) criticizes Grohmann’s characterization of the middle field as a domain strictly responsible for agreement phenomena, especially in a language such as German that relies so heavily on discourse configurationality in its clausal structure (see also Abraham 1997 and Meinunger 2000). The assumption that the middle field alone is responsible for the activation of agreement features seems to be in error. This hypothesis leads to the oft-touted mistake in the literature that Scrambling is an ‘optional’ operation disconnected from the syntax. (4.45) a.
Ich habe das Bieri gestern ti getrunken. I have the beer yesterday drunk
Chapter 4. The Prosodic Side of Scrambling
b.
181
Ich habe gestern das Bier getrunken. I have yesterday the beer drunk ‘I drank the beer yesterday.’
Assuming that the adverbial gestern ‘yesterday’ is adjoined to vP, the object das Bier ‘the beer’ never enters the middle field in (4.45b). If agreement relations are only to be found in the middle field (Grohmann’s Φ-domain), the structure in (4.45b) should be ungrammatical due to its failure to account for the φ-features of the object. This, however, is not the case since (4.45b) is a perfectly grammatical sentence. Even in a relatively closely related language family such as Germanic, it is evident that φ-features have the potential of appearing on each of the core functional heads, e.g. C, T and v. For example, in the examples of Scandinavian Object Shift discussed in Section 2.4, it appears from a quick glance of the data that Scandinavian clause structure permits the checking of agreement features in the middle field. Other Germanic (e.g., Bavarian German, Dutch dialects, Frisian and West Flemish to name a few) languages and dialects license complementizer agreement, i.e. morphological inflection to appear on complementizers. If each of these primary functional heads has the ability to possess φ-features, Grohmann’s assumption that the middle field is strictly contextually defined as an ‘agreement domain’ is incorrect. Of course, based on the hypothesis supported in this manuscript that middle field Scrambling is driven by the [+ Ref] feature, this was an inevitable conclusion. So what do we make of the middle field in Prolific Domains? Following the basic assumptions on syntactic structure established in our revision of minimalism, namely that every iterative operation produces structures that are interface interpretable; it makes sense that each domain must be linked to the performance systems. The middle field, as a syntactic domain, has the specific C-I role to establish the temporal/referential frame from which the thematic relations (accounted for in the vP-layer) are evaluated. Events (as defined by thematic relations) never stand alone; if this were the case, a sentence such as (4.46) would be a well-formed sentence, since it is a well-formed thematicallydefined event. (4.46) *Sam to leave. Rather, events are placed into two evaluative contexts – one is the temporal/referential domain defining the ‘real world’ and the other evaluates the social/speech-acts, or ‘discourse world’. Since the C-I system interprets events, it comes as no surprise that one of the Prolific Domains in the syntax would be devoted to deriving and interpreting such events. The domain responsible for this function would be the vP-layer, or Θ-domain. However, what the C-I sys-
182
Scrambling and the Survive Principle
tem requires for an event to be interpreted is context: both world/reference context and social/discourse context. The former context is sensitive to temporal, locative, and other world-reference indices. This context appears to be derived in the middle field – defined around TP.20 The latter context defines, among other things, speech act, topicalization, etc. These relations are determined in the left periphery (i.e., in the CP-layer/Ω-domain). Although Grohmann is correct in his observations about the tripartite structure of syntactic derivations, it appears that his interpretation of Prolific Domains has not connected this division rigorously enough to the interpretive designs of the performance systems (also see Putnam and Stroik (in progress) for a more extended version of this argument). To improve upon Grohmann’s notion of Prolific Domains as larger derivational units that are ‘interface independent’, I suggest the removal of the label Φ-domain in place of Ε-domain; the introduction of this new name is in harmony with the reinterpretation of the true contextual identity of the middle field. In (4.47) below I sketch out the revised role of the context values assigned to the various core functional heads in natural clauses (cf. (3.46)). (4.47) Revised Context Information of Clausal Tripartition Context Information is determined by context values: (i) |Θ| ranges over thematic relations; (ii) |Ε| ranges over temporal/event/ “real world” relations; (iii) |Ω| ranges over speech-act/discourse information. We now have a basic architecture of the order of the bare-bones necessary syntactic operations in order to facilitate the easiest and efficient mapping of the completed representation of grammatical sentences to the performance systems upon the completion of each successful iterative derivational step: Syntactic operations proceed to first account for the thematic properties/argument structure of the predicate. The lexical items that have entered the narrow syntax at this initial stage of the derivation (i.e., in the vP-layer) will then participate in further structure-generating procedures in the middle field (i.e., TP-layer) for temporal, i.e. ‘real world’ events and at the left periphery (i.e., CP-layer) of the clause to be interpreted in the ‘discourse world’. The licensing of agreement features can take place in any of these aforementioned domains and no longer needs to be housed exclusively in the middle field.21 20
For a similar view of the syntactic middle field as a contextually-defined event structure, see also Borer 1994, 1996; Ramchand 1997 and Ritter and Rosen 2001 among others. 21 Further support for my reanalysis of the tripartite structure of natural clauses put forward in this chapter can be found in the field of psycholinguistics. In his discussion of the minimalism program (with a particular focus on the psychological reality of uninterpretable features and displacement), van Valin (2001 and earlier work) maintains that displacement phenomena in
Chapter 4. The Prosodic Side of Scrambling
183
To provide further evidence in support of my reanalysis of the contextual identity of natural clauses according to (4.47), consider the following data from von Stechow (1995, 1996). In this data set, von Stechow illustrates that the adverbial again offers a probe into the semantic and syntactic composition of predicates. (4.48) a.
Sally opened the door again.
b.
Sally opened the door, and she had done that before. (repetitive)
c.
Sally opened the door, and the door had been opened before. (restitutive)
For both interpretations of the predicate in relation to again in (4.48b) and (4.48c), what makes the sentence in (4.48a) appropriate is some previous eventuality. On the repetitive reading, the event has to be a previous opening of the door by Sally. On the restitutive reading, by contrast, that event is the door’s being opened. Rather than regulating these different scopal readings solely to the LF interface, von Stechow supports the hypothesis that these different predicate interpretations of again are primarily syntactic in nature. Most important for von Stechow’s solution in support of the syntactic argument is the fact that the availability of the restitutive reading hinges on syntactic context. Following Beck and Johnson (2004) who also use again (in German: wieder) as an indicator of syntactic structure, I briefly introduce and review this argument. Based on the following German word order facts, a restitutive reading of wieder is only possible when wieder follows the direct object. If wieder precedes the object, only the repetitive reading is available (data from Beck and Johnson 2004:105). (4.49) a.
b.
(weil) Satoshi die Tür wieder öffnete. (because) Satoshi the doorACC again opened (repetitive and restitutive readings) (weil) Satoshi wieder die Tür öffnete. (because) Satoshi again the doorACC opened (repetitive reading only)
the narrow syntax are enacted by semantic requirements. Under this view, the truly universal component of UG is semantically-driven. The only deviations to this program are discoursepragmatically oriented. The only way that optimal syntax-semantic correspondence can be undone is by the introduction of discourse-pragmatic factors into the grammar. Although van Valin does not discuss the ontology or structure of natural clauses in his work (2001), these assumptions fall naturally from my reanalysis of Prolific Domains introduced in this chapter.
184
Scrambling and the Survive Principle
Von Stechow’s explanation for these facts is, in a nutshell, that the direct object in German obligatorily moves to a fairly high structure22 (Spec,AgrOP for von Stechow; a multiple Spec,vP in a phase-based system, and a concatenate structure with a functional head bearing the [+ Ref] feature in Surviveminimalism). Accordingly, that position must be higher that the element in the syntactic structure that expresses the CAUSE BECOME component, or v. Thus when wieder precedes, it is adjoined in a higher position than the direct object, and must take scope over CAUSE BECOME. These data show that a proper analysis of restitutive-wieder must rely on syntactic structure; as demonstrated in (4.49a), scrambling the direct object die Tür ‘the door’ has a profound effect on the event interpretation of the predicate in relation to the adverbial wieder. In this limited corpus of data, it is clear that the Scrambling of the direct object over the adverbial wieder in (4.49a), contributes to the event interpretation of the predicate. Most importantly for our analysis and the revision of the content values of Prolific Domains championed in (4.47), the example (4.49b) somehow manages to check its φ-features in the vP without overtly moving to the middle field. These data provide strong evidence for the analysis of licensing middle field Scrambling in West Germanic according to the [+ Ref] feature and for understanding the middle field as the syntactic domain responsible for event interpretability.23 4.5.1
Language Acquisition of Dutch Object Scrambling and Italian Clitic Placement
The reanalysis of the contextual identity assigned to Grohmann’s Prolific Domains in (4.47) finds further support in Schaeffer’s (2000) study of the acquisition of direct object Scrambling in Dutch and clitic placement in Italian. Building on Szabolsci (1983) and Abney (1987), Schaeffer proposes that the feature [referentiality] is semantic feature (which can be syntactically licensed in certain languages) present on Do in DPs. The syntactic notion of referentiality marking is defined by Schaeffer as in (4.50) below. (4.50)
22
Referentiality marking The D-head of a nominal expression with a referential reading bears the syntactic feature [referential].
Also see Heim (1982) and Diesing (1990, 1992) for similar arguments. Another possibility for typological variation in languages and the contextual identity they project onto different units of syntactic structure is that languages can potentially differ in what kinds of features motivate the displacement of syntactic objects. This raises the question on what is realized as a core or peripheral (in the sense of Chomsky 1995, 2000) feature. Contra Chomsky (1995), Abraham (1997) proposes that certain languages (such as German) rely on discourse-oriented features to motivate syntactic displacement rather than traditional agreeemnt (i.e., morphosyntactic) features. 23
Chapter 4. The Prosodic Side of Scrambling
185
Making use of several of the ideas proposed by Sportiche (1992), Schaeffer asserts that both middle field Scrambling in Dutch and clitic placement in Italian undergo movement into the middle field to account for their referentiality. Critically, Schaeffer (2000) and Sportiche (1992) make note of the fact that a further distinction needs to be made to differentiate referential DPs that scramble into the CP-layer above the subject and those that only land in the middle field. Both aforementioned authors formally recognize this distinction as items that are discouse-related referential items (i.e., those in the CP-layer) and nondiscourse-related referential units.24 Following this line of arugment, Schaeffer develops the following structural schemata to adopt the general Spec-Head licensing mechanism that Sportiche suggests for the licensing of [referentiality]. (4.51)
DiscP 3 Disc’ 3 Disc TP 3 RefP 3 Ref’ 3 Ref VP
According to the treelet in (4.51), DiscP represents a Spec-Head configuration that is responsible for licensing discourse-related referential XPs (i.e., topics) whereas RefP houses constituents that are referential though not linked to the external discourse. One can easily see how the reconfiguration of Prolific Domains (cf. (4.47)) advanced here supports such an analysis. Schaeffer (2000) and Sportiche (1992) predict that XPs – depending on whether they are Dlinked – will make use of the specifier positions (i.e., Spec,DiscP and Spec,RefP) while referential clitics will attach themselves to either of the head projections. Accordingly, a clitic or DP syntactically marked for [referentiality] must move to the middle field (to RefP) if it is not D-linked (or else it would be forced to move to DiscP).25 24 I make essentially the same argument in Section 2.5, however I also provide evidence in the syntactic behavior of topics in the CP-layer (i.e., these topics can bind cataphoric pronouns whereas non-discourse-related items in the middle field cannot). 25 As a minor point of criticism, Schaeffer (2000) purports that the cyclic movement of Dlinked referential object DP through AgrOP > RefP > DiscP represents an A-chain, however as suggested in Chapter 2 of this work, such an assertion cannot be successful upheld given the mixed A/A’-movement characteristics of middle field Scrambling.
186
Scrambling and the Survive Principle
(4.52) Object clitic placement/Object Scrambling26 TP 3 ClitVoice 3 Spec Clit’ 3 Clit VP [+ Ref] 3 V’ 3 V pro/DP [+ Ref] The functional projection labeled ClitVoice was created by Sportiche (1992) to be a designated projection for what he labels ‘specific’ (i.e., referential) object DPs and clitics. Sportiche’s ClitVoice corresponds with the functional project RefP proposed in this study. In the case of object clitics, Sportiche maintains that an object clitic’s [referential] feature is licensed under the strictly local Spec-Head agreement with a [referential] XP in the specifier of the same projection. This is formally stated in his Clitic Criterion in (4.53). (4.53)
Clitic Criterion (adopted from Sportiche 1992) At LF: (i) A [referential] clitic must be in a Spec-Head relationship with a [referential] XP; (ii) A [referential] XP must be in a Spec-Head relationship with a [referential] clitic.
Consider the following data displaying the distribution of object clitics in Dutch and Italian (taken from Schaeffer 2000:25-6). (4.54) a.
b.
Anna mangia le mele/*le. Anna eats the apples/it ‘Anna is eating the apples/them’
Italian
Anna le/*le mele mangia. Anna them/the apples eats
26 For ease of representation, I adopt a universal SVO (e.g., Spec-Head-Comp) underlying structure when discussing the effects of referentiality in the middle field of both Dutch and Italian.
Chapter 4. The Prosodic Side of Scrambling
c.
Anna ha mangiato le mele/*le. Anna has eaten the apples/them
d.
Anna le/*le mele ha mangiate. Anna them/the apples has eaten
(4.55) a.
dat Meike ‘t niet gezien heeft. that Meike it not seen has ‘that Meike didn’t see it’
b.
*dat Meike niet ‘t gezien heeft. that Meike not it seen has
187
Dutch
As we have seen throughout this book, referential items occupy a higher structural position than VP-internal objects. In (4.54b), we see that the full DP object le mele ‘the apples’ cannot occupy the same position as the direct object clitic le ‘them’. In the same tack, the object clitic‘t ‘it’ in the Dutch examples (4.55) must proceed negation (e.g., RefP > NegP > VP) to be properly licensed, as observed in the contrast between examples (4.55a) and (4.55b). Under the analysis developed and supported in this book, these predictions fall easily and gain natural support in our system. Sportiche’s Clitic Criterion (4.53) can be subsumed under Survive and Link! Any and all features must enter into a unique sisterhood relation for LF and PF interpretation. Seconly, our reformulation of the middle field as a structural unit responsible for licensing temporal event structure finds reciprocity in Sportiche’s (1992) and Schaeffer’s (2000) earlier proposals involving the syntactic licensing of [referentiality] in the Dutch and Italian middle field. The experimental study presented in Schaeffer’s (2000) work shows that scrambling of direct objects in obligatory contexts is optional in the language of Dutch and Italian 2-year olds. Dutch 2-year olds optionally scramble overt direct objects, and Italian 2-year olds optionally scramble pro direct objects, resulting in the optional realization of direct object clitics. The argument is that the semantic notion of [referentiality] has not yet been fully acquired by Dutch and Italian child, thus leading to the optionality in these aforementioned constructions. Between the ages of 2 and 3, the ability of Dutch children to correctly scramble referential direct object DPs rose from 30% to 72%. This nontrival percentage provides evidence supporting the stance that principles of UG are available to children from the beginning. These results make a strong argument for the existence of RefP in early grammars. In the end, the jump in success from a success rate of 30% at age 2 to 72% at age 3 for Dutch children in their ability to scramble object DPs into the middle field not only provide evidence that referentiality is optionally marked in early child grammars, but
188
Scrambling and the Survive Principle
also provide a unified theory of object Scrambling and object clitic placement along the lines of Sportiche (1992). 4.6
Conclusion
Maintaining a modular view of syntactic computation in its role as subcomponent of the CHL, this chapter provided a rough overview of the prosodysyntax interface in the framework of Survive-minimalism adopted here in regard to middle field Scrambling in West Germanic. Here I argued against a purely prosodic analysis of Scrambling as well as the introduction of prosodic features (e.g., AntiFocus [AF]) based on both conceptual and empirical grounds. From a theoretical standpoint, prosodic features are not syntactic features in that they do not contribute to phrase structure building principles and violate Full Interpretation (i.e., they are not interpreted at LF). Empirically, adopting the AntiFocus feature as a functional element active in the narrow syntax – or any prosodic features for that matter – cannot explain the scrambling or prosodic behavior of indefinites or larger XPs in Afrikaans (or West Germanic in general). Following Grohmann and Putnam (2004, 2005, in press) and my analysis of the prosodic properties of the middle field with regard to coherent infinitives and their relation to middle field Scrambling as well as object order permutations, I insinuate both of these aforementioned constructions are evaluated at the level of prosodic structure. This allows the narrow syntax to maintain its prosodically myopic status. Finally, my revision of Grohmann’s Prolific Domains – in particular, my revision of the contextual identity of the middle field – sheds further light on the true structure of the middle field as well as the true nature of Scrambling. In the concluding chapter of this work, I will revisit the central claims of this text, namely the hypothesis that middle field Scrambling in West Germanic is driven by the syntactic feature [+ Ref]. In line with my reanalysis of the contextual structure of the middle field within the Prolific Domains framework, I will unite previous aspects of my investigation of Scrambling with this essential reanalysis of Prolific Domains as potential visible derivational units in the narrow syntax. This will also allow us to take an in depth look at the structure of the middle field, its function in the narrow syntax and, most importantly for this immediate study, what this means for our analysis of middle field Scrambling in West Germanic. In this final chapter I will engage in a brief discussion of how the MP guided by the Survive Principle can account for Scrambling phenomena in cross-linguistically diverse languages such as Japanese and Russian.
CHAPTER 5 CONCLUSION The grammatical operation known as Scrambling is a multi-faceted phenomenon that has received a myriad of different generative analyses. As summarized by Kiss (1994:223), there are four basic approaches to this phenomenon broadly categorized as Scrambling. These approaches are shown in her tree diagram: (5.1)
scrambling qp stylistic syntactic phenomenon (PF) qp adjunction base-generated qp hierarchical structure flat structure
Similar in design to most investigations in the recent literature on this matter, the central purpose of this work has been to find out which of these proposals represents the correct analysis of Scrambling. On the basis of Hungarian data, Kiss argues for the existence of Scrambling as a base-generated (i.e., flat or arbitrary argument order) phenomenon. However, she differentiates Scrambling from movement involving operators in the narrow syntax (e.g., focusing and Q-raising). Kiss (cf. 1994:228) posits the rule that the operation labeled as Scrambling does not affect the semantic meaning of a sentence. Such a statement is unfounded for German, in that good evidence exists against the basegeneration approach. First, as pointed out by Clahsen (1982), German children acquire ‘scrambled’ orders at a later stage than ‘standard’ ones, a fact which squares with their later acquisition of wh-movement, since they initially form questions solely with the use of intonation.1 Second, the suggestion that scrambled objects are merged into the middle field and then participate in some sort of lowering operation (cf. Bošković and Takahashi 1998) is countercyclic in design and hence should be avoided. The hypothesis developed and supported in the previous chapter (Chapter 4) that prosodic considerations are not the ultimate syntactic trigger for Scrambling in West Germanic, eliminates 1
As pointed out by Hoyer (2004:3), Webelhuth (1992) argues strongly against a basegeneration approach because it contradicts the phrase structure requirements of his Saturation Condition.
190
Scrambling and the Survive Principle
the possibility that Scrambling is solely a post-syntactic, phonological linearization process. This claim is based on the fact that objects that scramble into the middle field continue to display A-movement characteristics. If prosody were the principle feature involved in this operation, anaphoric binding, some instances of WCOs and the sensitivity to generic vs. existential indefinites would not be possible, because LF under such a theory is predicted to be blind to purely phonological phenomena. In stating that prosodic considerations are not the primary cause of Scrambling, I certainly do not want to misinform the reader that sentential stress plays no role at all in determining the grammaticality of scrambled objects. On the contrary, the purpose of this work was to determine the role of the narrow syntax in achieving its proper function of creating and evaluating scrambled orders. According to Kiss’ model in (5.1), the only option left for a syntactic definition of Scrambling rests on XP-adjunction. Most recently, Putnam (2006a) has championed this analysis, suggesting that middle field Scrambling involves adjunction to vP where a discourse component of the CHL will interpret these displaced items. In his dissertation, Putnam also finds Prolific Domains as deficient minimalist tools in their current state based on his adjunction analysis of Scrambling in West Germanic languages. With the revision of the Minimalist Program along the lines of Stroik’s (1999, 2000, 2007) Survive Principle, it is possible to dispense with the distinction between substitution (i.e., traditional movement to a specifier of a functional projection) and adjunction: The creation of concatenate syntactic objects involves a functional head possessing a matching feature which calls a lexical item active in the Numeration to merge/remerge in the narrow syntax.2 Based on the battery of tests probed in this study, the most likely candidate for a syntactic feature responsible for licensing middle field Scrambling in West Germanic is the +interpretable feature [+ Ref] that represents referentiality. So what becomes of the structural functions of adjunction in early instantiations of minimalism? Rather than rely upon the representational notions of movement chains and c-command, under the approach adopted here, these ‘structure idiosyncrasies’ can be reduced to the characteristics of particular features, in this case [+ Ref]. Although Putnam’s analysis of middle field Scrambling in West Germanic as an instance of XP-adjunction is in error (largely due to his reliance upon Move/Internal Merge as the chief displacement operation available to the CHL), it is a logical conclusion to be reached in ‘orthodox’ minimalism. My approach, adopting the principles of Survive-minimalism, eliminates the barriers that once separated Scrambling from being considered a ‘movement type’ similar in design to other related epiphenomena.
2
In the creation of the VP, the V-head, although not functional in design, is assumed to also possess a [Vform] feature to maintain the general structure of the vP-shell.
Chapter 5. Conclusion
191
Putnam was also correct with his abandonment of Prolific Domains in an analysis of middle field Scrambling. While Putnam and Stroik (in progress) see great potential in derivational approaches that utilize the contextual properties of the tripartite structure of natural clauses, they critically differ with Grohmann’s notion of the functional responsibilities of the middle field. Putnam and Stroik discard Grohmann’s idea that the middle field is solely responsible for agreement phenomena and suggest that the middle field exclusively licenses temporal, locative, and other world-reference indices. As a point of adjustment, in the previous chapter (Chapter 4) I renamed the middle field in Grohmann’s framework of Prolific Domains the Ε-domain to more accurately reflect the true contextual identity of the middle field. This is, of course, a strong revision of Grohmann’s notion of the middle field (i.e., Φ-domain) being uniquely identified as the derivational unit of syntactic structure solely responsible for the licensing of agreement phenomena. For example, in Scandinavian languages such as Swedish that license Object Shift (see Chapter 2, Section 4), there is a ubiquity of A-movement characteristics associated with this displacement type. Therefore, in such languages and language families perhaps the middle field can be identified as a unit responsible for the licensing of both temporal event structure as well as agreement features. Such a line of thinking is reminiscent of Abraham’s (1997) proposal that rather than inflectional features (cf. Chomsky 1995) perhaps discourse-oriented features function as the impetus of syntactic ‘movement’ in particular languages and language families (such as German). Research incorporating data from typologically diverse languages will improve upon our understanding of the contextual identity of the middle field, however based on this study and its focus on West Germanic languages it is untenable to maintain that the middle field in this language family is responsible for the licensing of agreement phenomena in the narrow syntax.3 In this concluding chapter, I will focus in a little more detail on the structure of the West Germanic middle field with regards to its relationship to Scrambling. Closely associated with this investigation of the structure of the middle field, I will also take another look at the exact nature and definition of the [+ Ref] feature. For example, throughout this text I have continually referred to this feature as a pragmatic feature, however, based on the revision of the middle field to accommodate and license temporal/events and ‘real world’ relations, is it perhaps more accurate to label this feature as a semantic feature? Lastly, I provide a brief sketch of an analysis of how a minimalist version of a derivational system based on the main desiderata of the Survive Principle could account for Scrambling in Japanese and Russian.
3
See also Bobaljik (2005) and Sigurðsson (2007) for arguments that morphosyntactic agreement could possibly be regulated as a purely PF-phenomenon external to syntactic operations.
192
5.1
Scrambling and the Survive Principle
[+ Ref] and the Design of the Middle Field
Is it possible that semantic features exist in the narrow syntax? After all, purely semantic features would elevate the status of the narrow syntax beyond being ‘semantically myopic’ in the sense of Hornstein (1995). Butler (2003) builds upon the proposal of Adger and Ramchand (2003) by providing a semantic characteristic of EPP features (cf. Chomsky 2000), where they are treated as syntactic features [Λ] on argument-related heads that map directly to the semantics as a predicate abstraction operator λ. According to this assumption, [Λ] features enter into binding dependencies via [ID] features on V, which are interpreted as variables; they may, according to Butler, be interpreted as a featural reinterpretation of θ-roles. In this section I will show that Butler, following Adger and Ramchand (2003), is basically on the right track with his analysis of the semantic identity of EPP, however, an account of constituent displacement motivated by EPP cannot be upheld in Survive-minimalism.4 As thoroughly discussed by Stroik (2007: Ch.3), the EPP exists as a mechanism to ensure that Internal Merge/Move will not overgenerate itself in the course of the derivation. If you remove Internal Merge/Move there is no need to keep the EPP to police the syntax from performing long-distance actions. In Surviveminimalism, economy constraints as they existed in earlier stages of minimalism (e.g., Last Resort, Greed, Procrastination and the like) are naturally observed and respected in every iterative application of Survive-Remerge-Link! Secondly, the EPP doesn’t contribute anything to either the A-P or C-I interfaces for interpretation, but simply exists as a safeguard mechanism. Let’s further assume that the decomposition of the verbal complex into V and v layers is on the right track. If we take V to be the lexical root (cf. Pesetsky 1995; Marantz 1997), which is largely unspecified, and the v head to introduce a subsituation into the representation: stative, causative, inchoative, etc., we can also extend this theory to suggest that v makes further suggestions to derivational operations on how the syntax should properly saturate itself with sufficient structural considerations to achieve interface harmony. In this sense, I am not claiming that v – or any other functional head for that matter – has look-ahead planning properties, however, based on the assertion that natural clauses demonstrate the tripartite skeletal structure C > T > v, this is only a naturally following assertion not requiring additional hardware or mechanisms in the CHL. The structure of the Ε-domain (i.e., middle field) would require an intricate, probing study that would potentially transcend and surpass the scope and mission of this text. This is clearly the task of future research; however for the time being I would like to end with some speculative comments and thoughts on the structure of the middle field with regards to its interpretive relationship 4
See also Epstein and Seely (2006) for conceptual and empirical reason why EPP should be removed from derivational syntax.
Chapter 5. Conclusion
193
to Scrambling. Based on Travis (1994), I support the hypothesis of inserting a functional head denoting event interpretation in the middle field above vP. For the sake of ease of reference, let’s call this EventP.5 Similar to the fact that v constitutes a cause-denoting head on top of the predicate; Evento determines the temporal/event properties of the proposition (vP) that is interpreted at the external interfaces. Another safe assumption is to suggest that either EventP represents a domain of existential closure (for the interpretation of indefinites and QPs) or that EventP is dominated by a functional projection ∃P. For the time being I will take an agnostic approach to this matter and relegate it to future research considerations. So is it possible, and perhaps preferred to refer to [+ Ref] as a semantic feature rather than a pragmatic one? As pointed out by Büring (2005), ISinterpretation (including pragmatic readings) may not primarily affect the truth condition of utterances; however they do affect more elusive aspects of their meaning such as their implicatures and felicity conditions. Therefore for the time being it is best to classify [+ Ref] as a +interpretable feature interpretable at both LF and PF to both pragmatic and semantic operations. As for the contextual structure of the middle field, future research should explore the connection between other aspectual elements in German syntax such as middle voice constructions and middle field Scrambling. Furthermore it will be interesting to discover how ‘semantically myopic’ (see Hornstein 1995) the middle field in universal clausal architecture truly is. 5.2
De re/de dicto distinction
The distinction between de re and de dicto semantic readings are directly influenced by middle field Scrambling in German. In this section I demonstrate the degree to which middle field Scrambling plays a direct role in the de re/de dicto readings attributed to DPs in the middle field and those above the subject in the CP-layer. When a DP appears inside the clausal or VP complement of a modal predicate, there is often a so-called de re-de dicto ambiguity. To illustrate this point, consider (5.2) below in which the DP a garbage collector is contained inside the infinitive complement want. (5.2)
Littlejohn wants to marry a garbage collector.
According to the de dicto reading, every possible world in which Littlejohn gets what he wants is a world in which there is a garbage collector whom he 5
In their discussion of the expanded middle field in determining the scope properties/readings of the quantifiers each and every, Stowell and Beghelli (1997) suggest that their ShareP may be a potential home for scrambled XPs in the middle field.
194
Scrambling and the Survive Principle
marries. According to the de re reading, there is a garbage collector in the actual world whom Littlejohn marries in every world in which he gets what he wants. We can imagine and create situations in which one of the readings is true and one of them false. For example, suppose Littlejohn thinks that garbage collectors make ideal spouses, because they make a lot of money. Although he hasn’t met a garbage collector so far in his life, he is definitely sure he would like to marry one. In this scenario, the de dicto reading is true, whereas the de re reading is false. What all of Littlejohn’s desire-worlds have in common is that they have a garbage collector getting married to Littlejohn in them. The key difference is that it is not the same exact garbage collector in each of these worlds. As a matter of fact, there is no particular individual (actual garbage collector or other), which he actually marries in any of those desire-worlds. In another scenario, let’s assume that Littlejohn has fallen in love with Victoria and wants to marry her. Victoria just so happens to be a garbage collector, but Littlejohn is unaware of this fact. Who knows, perhaps Littlejohn would call off the marriage if he found out about Victoria’s true profession, however regardless of this fact, the de re reading in this situation holds true and the de dicto is rendered false: Victoria is a bona fide garbage collector who gets married to Littlejohn in every world in which he gets what he wants. The de dicto reading is false, because the worlds that conform to Littlejohn’s wishes actually do not have him marrying a garbage collector in them. In his favorite, preferred world he marries Victoria, who is not a garbage collector in any of those particular worlds. An interesting question to raise at this point is whether or not middle field Scrambling induces or somehow inhibits de re and/or de dicto readings. Lechner (1997:44-47) points out the following differences in semantic interpretations between DPs headed by a quantifier that scramble into the middle field (5.3) and those that move about the subject into the CP-layer (5.4).6 (5.3)
…dass Maria viele Lösungen gesucht hat. that Maria many solutions looked for has ‘…that Maria look for many solutions.’
According to Lechner, (5.3) is ambiguous in three ways: It can be read (weak) cardinal de dicto, (weak) cardinal de re or proportional de re, but the strong, proportional de dicto interpretation of the object is not available. The following chart (5.4) illustrates the de re reading of (5.3).
6
All data and discussion cited in this section comes from Lechner (1997) unless otherwise noted.
Chapter 5. Conclusion
195
Cardinal de re reading of (5.4): Semantics [[solutions-seek] [[ti solutions seek]
= λxλy[solution(x) & look for(↑↑(x)(y)] = = λxλy[solution(x) & look for(λz[z=x])(y)] = λy[solution(ti) & look for(λz[z=ti])(y)]
[[manyi [λi [tk ti solutions-seek] = Ǝx[solution(x) & look for(λz[z=x])(Mary) & many (x)] In contrast to the quantified DP occupying the middle field in (5.3), scrambling the object above the subject disrupts the cardinal de re reading of the object. (5.5)
...dass viele Lösungeni jeder in seiner Studierstube ti gesucht hat. that many solutions everybody in his study looked for has ‘…that everybody has looked for many solutions in his study.’
The Scrambling of the quantified object viele Lösungen ‘many solutions’ over the subject in (5.5) interferes with the availability of the cardinal de re, but not the cardinal de dicto reading of the scrambled object. Lechner argues that this makes sense due to the fact that the cardinal de re reading, but importantly not the cardinal de dicto reading, is dependent, since Scrambling has been shown to inhibit syntactic reconstruction (SynR). According to Lechner (p.c.), what makes it extremely difficult to examine whether the de re/de dicto reading contrast plays a decisive role in determining/licensing Scrambling is the lack of an intentional operator in most predicates. It is not easy to test the meaning of the common nouns in the absence of an intentional operator (i.e., in verbs such as seek, need and want), because we simply do not know which evaluation situation we should use to cash out the context of the noun (cf. Fodor 1970). There may indeed be unembedded de re/de dicto readings in relation to the Scrambling operation. A relevant question, for example, might be whether or not Scrambling has an impact on opacity, as opposed to specificity/ presupposionality / givenness and other concepts. This illustrates, at least in German, that the syntax is not as ‘semantically myopic’ as once conceived. In the final section of this chapter in my brief discussion of Scrambling in Japanese and Russian, I provide a proposal of how different scope readings can be derived within the framework of Survive-minimalism.
196
5.3
Scrambling and the Survive Principle
A Sketch of Japanese and Russian Scrambling in Survive-Minimalism
In this final section, I introduce a novel, and all too brief, sketch of Scrambling in Russian and Japanese, based primarily by the works of Bailyn (2001, 2006) and Saito (2003) respectively. Both Bailyn and Saito support the idea that Scrambling is the result of some sort of movement operation in the narrow syntax. More specifically, they argue that such an interpretation of Scrambling is not only conceptually but also empirically superior given certain assumptions about the derivational nature of movement and its interaction with interpretation, based on Epstein et al. (1998), Zubizaretta (1998), and Saito (2003). Bailyn’s (2006) proposal that the apparent optionality associated with Scrambling is a side-effect of a derivational account of Information Structure (IS), is similar in design and scope to our vision of the narrow syntax-PF interface put forward in the previous chapter. To derive some of the fundamental characteristic distinctions between Japanese and Russian Scrambling patterns and behaviors, Bailyn suggests that formal features are bundled in various ways in various languages. Thus in Japanese, wh-features are considered to be separable from the nominal they attach to. Crucially under such an account, Operator-Variable relations and relative Quantifier Scope are determined derivationally before surface order readings, which only alter pure IS-relations. Conversely, in Russian, reorderings that occur in the narrow syntax feed LF relations; the mechanism is the same, but the semantic effect is predicted to differ in just the observed fashion. This fact is illustrated in the following data set (taken from Müller and Sternefeld (1993) and Bailyn (2006)). (5.6)
(5.7)
a.
*Kto ty videl kogda [IP ti podežžal] who you saw when came Intended: ‘Who did you see when came?’
b.
Ty doktor videl kogda [IP ti podežžal] you doctorNOM saw when came ‘The doctor did see you when (he) came?’
Ty každuju devuškui videl kogda [IP kakoj-to malĕik celoval ti]? you [every girl]ACC saw when some boyNOM kissed “Did you see when some boy kissed every girl?” i.) *∃x ∀y ii.) ∀y ∃x (high scope for the scrambled NPACC)
The data above in (5.6) and (5.7) cast serious doubt on any sort of LF-lowering account (cf. Bošković and Takahashi (1998) and Bošković (2004)) of Scrambling. Example (5.6b) illustrates that any sort of lowering mechanism to derive
Chapter 5. Conclusion
197
Scrambling patterns in Russian does not account for the fact that example (5.6a) contains a wh-island and yet it does not prevent reordering. Therefore, the supposed interpretive vacuity of LF-lowering into a θ-position – which is predicted to hold for both Russian and Japanese – is shown to be false, provided the scope readings attributed to (5.7). Further evidence against an operation involving LF-lowering to explain Scrambling is put forward in (5.8) and (5.9) below (data from Bailyn 2001). (5.8)
a.
xočet, čtoby Boris uvidel ti. [Každogo malčika]i kto-to [every boy]ACC someoneNOM wants that Boris saw ‘Every body someone wants Boris to see.’ i.) *∃x ∀y ii.) ∀y ∃x (high scope for the scrambled NPACC)
b.
[Daremo-ni]i dareka-ga [Mary-ga ti atta to] ometteiru. EveryoneDAT someoneNOM MaryNOM met that thinks ‘Someone thinks that Mary met everyone.’ i.) ∃x ∀y ii.) *∀y ∃x (low scope for the scrambled NPDAT)
(5.9)
*Borisai ty pozvonil [agentu [kotoryj ljubit ti] ] BorisACC youNOM phone spyDAT who loves Intended: ‘It’s Boris who you phoned a spy who loves!’
The contrast in scope readings between the Russian example (5.8a) and the Japanese example (5.8b) supports the argument against LF-lowering. At the core of the mechanism LF-lowering is the prediction that there is neither surface interpretive effects nor locality constraints. Example (5.9) shows that Russian Scrambling over the subject is still sensitive to subjacency violations. Therefore, these counter-examples to LF-lowering clearly illustrate that Scrambling is a syntactic operation that is sensitive to island effects. Lastly, Japanese Scrambling – just like in Russian and German – exhibits distinct discourse effects. The fact that (5.10b) below is not a suitable answer to the question (5.10a) proves this point (data from Bailyn 2006). (5.10) a.
John-wa [Mary-ga dakode ringo-o katta to] omotteiru no? Q JohnTOP MaryNOM where appleACC bought C think ‘Where does John think that Mary bought apples?’
198
Scrambling and the Survive Principle
b.
#Ringo-o [John-wa [Mary-ga SAC-de katta to] omotteiru applesACC JohnTOP MaryNOM SAC-at bought C think ‘Apples, John thinks that Mary bought at SAC.’
As stated above, to account for these variations Bailyn (2006) proposes the following feature-bundling typology in (5.11) below. (5.11) Parameters of Scrambling: a. b. c.
Open A-position Morphological discourse marker WH/Scope Effect
(Jap: +, Russian: +, SC: -, German: -) (Japan: +, Russian: -) (Japan: +, Russian: -)
Although testing and deriving the exact discourse effects of both Japanese and Russian Scrambling is a task that falls outside the scope of this work, I would like to make some suggestions on how a minimalist approach to XPdisplacement based on the Survive Principle could further simplify the differences between Japanese and Russian Scrambling. First and foremost, some sort of syntactic feature, i.e. a feature that has a potential reflex at both PF and LF (cf. Full Interpretation (3.6)) is the motivating factor behind Scrambling in Japanese and Russian. Let us refer to this feature as Σ for the time being. Disregarding the different morphological Case markings of the scrambled objects in (5.8) and (5.9), we see the contrastive scope readings in the Russian and Japanese examples of Scrambling: Whereas Scrambling in Russian (5.8) allows the scrambled object to receive high scope (*∃x ∀y > ∀y ∃x), the exact opposite behavior is observed in the Japanese example (5.9) (∃x ∀y > *∀y
∃x). The fact that items that scramble in Japanese must receive their scope interpretation from the lowest operator is often referred to as the ‘undoing’ property of Scrambling. Constituent displacement enacted by the Survive Principle offers a unified account of Scrambling in Japanese and Russian in spite of the different scope readings evidenced in (5.8) and (5.9). If both Russian (R) and Japanese (J) possess the Scrambling feature Σ, which is most likely some kind of discourse feature similar to West Germanic [+ Ref], then the different scope readings in R and J respectively are the result of different realizations of the scrambled objects at both LF and PF. The proposed derivational histories in (5.12) and (5.13) will elucidate this point:
Chapter 5. Conclusion
(5.12) Russian a. PF:
199
[CP XPΣ Subj … [vP … XPΣ …] [CP XPΣ Subj … [vP … XPΣ …]
b.
LF:
c.
Scope interpretation: *∃x ∀y > ∀y ∃x
(5.13) Japanese a. PF:
[CP XPΣ Subj … [vP … XPΣ …] [CP XPΣ Subj … [vP … XPΣ …]
b.
LF:
c.
Scope interpretation: ∃x ∀y > *∀y ∃x
Note the key difference between the LF derivational history of R and J in steps (5.12b) and (5.13b) respectively. Although both scrambled constituents are recognized at PF in their respective scrambled positions, the LF interpretations of the scrambled items differ considerably. Based on the key tenets of Surviveminimalism, a version of ‘reconstruction’ is no longer necessary in syntactic theory: All ‘reconstructions’ involve non-covert concatenations. In other words, nothing is actually ‘reconstructed’. The scope reconstruction that emerges in (5.13b) exists because the dislocated, i.e. remerged, DPs had previous instances of Merge that are interpretable at LF.7 Admittedly, there are many other aspects of Scrambling in both Russian and Japanese (as well as other languages) that must be investigated (e.g., the A-position properties of Japanese and Russian in comparison with German), however this brief sketch of how Scrambling can be accounted for in Surviveminimalism shows promise for improving our understanding of this displacement type cross-linguistically as well as gaining a better understanding of its parametric variation. 5.4
Conclusion
The radical generative approach to Scrambling advanced and supported in this book makes a convincing case for the reality and validity of two key concepts. First, the grammatical notion of referentiality – encapsulated in the syntactic feature [+ Ref] – represents the ideal candidate to license the syntactic aspects of middle field Scrambling in West Germanic languages. Secondly, the revi7
As mentioned to me by John Hale (p.c.), such an interpretation of scope reading differences imposes a great deal of computational complexity on the PF-interface. Such an assumption falls out nicely if we assume that LF represents a constant in CHL (i.e., all language require meaning interpretation) and that parametric variation in languages (e.g., linearization and inflectional morphology) reside and are determined at PF.
200
Scrambling and the Survive Principle
sion of the Minimalist Program according to Stroik’s Survive Principle and its accompanying principles and constraints can be upheld. Much needed crosslinguistic research on Scrambling in other languages and language families utilizing this framework will not only provide a clearer picture of this linguistic phenomenon, but will also further test and improve upon the revision of the Minimalist Program (i.e., Survive-minimalism) supported here.
REFERENCES Abels, K. 2003. Successive cyclicity, anti-locality and adposition stranding. PhD dissertation, University of Connecticut. Abney, S. 1987. The English noun phrase in its sentential aspect. PhD dissertation, MIT. Abraham, W. 1986. Word order in the middle field of the German sentence. In Topic, Focus, and Configurationality: Papers from the 6th Groningen Grammar Talks, W. Abraham & S. de Meij (eds.), 1-38. Amsterdam: John Benjamins. Abraham, W., Epstein, S.D., Thraínsson, H. & Zwart, C.J.-W. (eds.) 1996. Minimal Ideas-Syntactic Studies in the Minimalist Framework. Amsterdam: John Benjamins. Abraham, W. 1997. The base structure of the German clause under discourse functional weight: Contentful functional categories vs. derivative ones. In German: Syntactic Problems – Problematic Syntax, W. Abraham & E. van Gelderen (eds.), 11-43. Tübingen: Max Niemeyer. Abraham, W. & van Gelderen, E. (eds.). 1997. German: Syntactic Problems – Problematic Syntax. Tübingen: Max Niemeyer. Abraham, W. 2003. The syntactic link between thema and rhema: The syntaxdiscourse interface. In Optionality in Syntax and Discourse Structures – Aspects of Word Order Variation in (West) Germanic and Other Languages, W. Abraham & L. Molnárfi (eds.). Folia Linguistica XXXVII(1-2): 1-34. Abraham, W. & Molnárfi, L. 2002. German clause structure under discourse functional weight: Focus and antifocus. In Issues in Formal German(ic) Typology, W. Abraham & C.J-W. Zwart (eds.), 1-38. Amsterdam: John Benjamins. Abraham, W. & Molnárfi, L. 2003. Introduction. In Optionality in Syntax and Discourse Structure, Folia Linguistica Special Issues 37/1-2, W. Abraham & L. Molnarfi (eds.), 1-11. Berlin: Mouton de Gruyter. Ackema, P. & Neeleman, A. 2004. Beyond Morphology: Interface Conditions on Word Formation. Oxford: OUP. Adger, D. 1994. Functional heads and interpretation. PhD dissertation, University of Edinburgh. Adger, D. & Ramchand, G. 2003. Predication and equation. Linguistic Inquiry 34: 325-359. Allen, C. L. 1995. Case Marking and Reanalysis: Grammatical Relations from Old to Early Modern English. Oxford: Clarendon Press. Anagnostopoulou, E. 2004. Holmberg’s generalization and locality. Handout. GLOW 2004. Åfarli, T. A. 1995. Seeds of functional projections. In De Grammatica Generativa, P. Goenaga (ed.), 139-150. Gasteiz: Universidad del País Vasco. Bailyn, J. 2001. On scrambling: A response to Bošković & Takahashi. Linguistic Inquiry 32(4): 635-658. Bailyn, J. 2006. Against the anti-movement scrambling movement. Formal Approaches to Slavic Linguistics 14(1): 1-22. Banerji, S. 2003. Focus, word order and scope effects in Bangla. In Optionality in Syntax and Discourse Structure, Folia Linguistica Special Issues 37/1-2, W. Abraham & L. Molnárfi (eds.), 35-75. Berlin: Mouton de Gruyter.
202
Scrambling and the Survive Principle
Beck, S. & Johnson, K. 2004. Double objects again. Linguistic Inquiry 35(1): 97123. Beckman, M.E. & Pierrehumbert, J.B. 1986. Intonation structure in Japanese and English. Phonology Yearbook 3: 255-309. Bennis, H. & Hoekstra, T. 1985. Gaps and parasitic gaps. The Linguistic Review 4(1): 29-87. Bierwisch, M. 1963. Grammatik des deutschen Verbs. Berlin: Akademie Verlag. Bobaljik, J. 2002. Realizing Germanic inflection: Why morphology does not drive syntax. Journal of Comparative Germanic Linguistics 6(2): 129-167. Bobaljik, J. 2004. Re: CycLin and the role of PF in object shift. Theoretical Linguistics 31: 1111-1125. Bobaljik, J. 2005. Where’s phi? Ms., University of Connecticut. Boeckx, C. 2003. Islands and Chains. Amsterdam: John Benjamins. Boeckx, C. & Grohmann, K. 2004. Left dislocation in Germanic. In Studia Typologica 6: Focus on Germanic Typology, W. Abraham (ed.), 139-152. Berlin: Akademie Verlag. Borer, H. 1984. Parametric Syntax. Dordrecht: Foris. Borer, H. 1994. The projection of arguments. In Functional Projections [University of Massachusetts Occasional Papers 17], E. Benedicto & J. Runner (eds.), 19-47. Amherst MA: Department of Linguistics, University of Massachusetts. Borer, H. 1996. Structuring Sense. Oxford: OUP. Bošković, Ž. 2002. A-movement and the EPP. Syntax 5: 167-215. Bošković, Ž. 2004. Topicalization, focalization, lexical insertion, and scrambling. Linguistic Inquiry 35: 613-638. Bošković, Ž. & Takahashi, D. 1998. Scrambling and last resort. Linguistic Inquiry 29: 347-366. Bresnan, J. 2001. Lexical-Functional Syntax. Oxford: Blackwell. Brody, M. 1995. Lexico-Logical Form: A Radically Minimalist Theory. Cambridge MA: The MIT Press. Brody, M. 1998. Projection and phrase structure. Linguistic Inquiry 29: 367- 398. Brody, M. 2002. On the status of representations and derivations. In Derivation and Explanation in the Minimalist Program, S. Epstein & T.D. Seely (eds.), 19-41. Oxford: Blackwell. Bouchard, D. 2002. Adjectives, Number, and Interfaces. Why Languages Vary. London: Elsevier. Buffington, A.F. & Barba, P.A. 1954. A Pennsylvania German Grammar. Allentown PA: Schlechter’s Publishing Company. Burridge, K. 1993. Syntactic Change in Germanic. Amsterdam: John Benjamins. th
Büring, D. 1997. The Meaning of Topic and Focus: The 59 Bridge Street Accent. London: Routledge. Büring, D. 2001a. Let’s phrase it! In Competition in Syntax, G. Müller & W. Sternefeld (eds.), 69-105. Berlin: Mouton de Gruyter. Büring, D. 2001b. What do definites do that indefinites definitely don’t? In Audiatur Vox Sapentiae, C. Féry & W. Sternefeld (eds.), 70-100. Berlin: AkademieVerlag. Büring, D. 2005. Semantics, intonation and information structure. Ms., UCLA. Butler, J. 2003. On having arguments and agreeing. In York Papers in Linguistics Series 2, J. Butler, D. Morgan, L. Plug & G. Walker (eds.), 1-27.
References
203
Chametzky, R.A. 2000. Phrase Structure: From GB to Minimalism. Oxford: Blackwell. Chocano, G. 2004. Narrow syntax and phonological form: Scrambling in the Germanic languages. PhD dissertation, Universidad Autónoma de Madrid. Choi, H.W. 1999. Optimizing Structure in Context – Scrambling and Information Structure. Stanford CA: CSLI. Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge MA: The MIT Press. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986. Barriers. Cambridge MA: The MIT Press. Chomsky, N. 1993. A minimalist program for linguistic theory. In The View from Building 20: Essays in Honor of Sylvain Bromberger, K. Hale & S.J. Keyser (eds.), 1-52. Cambridge MA: The MIT Press. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step: Essays on minimalist syntax in Honor of Howard Lasnik, R. Martin, D. Michaels & J. Urigereka (eds.), 89-115. Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale: A Life in Language, M. Kenstowicz (ed.), 1-52. Cambridge MA: The MIT Press. Chomsky, N. 2004. Beyond explanatory adequacy. In Structures and Beyond: The Cartography of Syntactic Structures 3, A. Beletti (ed.), 104-31. Oxford: OUP. Chomsky, N. 2005. Three factors in language design. Linguistic Inquiry 36(1): 1-22. Chomsky, N. 2006. On Phases. Ms., MIT. Chomsky, N. & Halle, M. 1968. The Sound Pattern of English. New York NY: Harper and Row. Cinque, G. 1990. Types of A-bar Dependencies. Cambridge MA: The MIT Press. Cinque, G. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24: 239-298. Cinque, G. 1999. Adverbs and Functional Heads. Cambridge: CUP. Clahsen, H. 1982. Spracherwerb in der Kindheit: Eine Untersuchung zur Entwicklung der Syntax bei Kleinkindern. Tübingen: Narr. Collins, C. 1995. Towards a theory of optimal derivations. MIT Working Papers in Linguistics 27: 65-103. Collins, C. 1997. Local Economy. Cambridge MA: The MIT Press. Collins, C. 2002. Eliminating labels. Ms., MIT Working Papers in Linguistics 22: 26-52. Collins, C. & Thráinsson, H. 1996. VP-internal structure and Object Shift in Icelandic. Linguistic Inquiry 27(3): 391-444. Culicover, P. & Rochemont, M. 1991. English Focus Constructions and the Theory of Grammar. Cambridge: CUP. Czepluch, H. 2004. Reflections on the form and function of passives in English and German. In Studia Typologica 6: Focus on Germanic typology, W. Abraham (ed.), 169-194. Berlin: Akademie Verlag. Davies, W. & Dubinsky, S. 2004. The Grammar of Raising and Control. Oxford: Blackwell. den Besten, H. 2000. The slaves’ languages in the Dutch Cape colony and Afrikaans vir*. Linguistics 38(5): 949-971. Diesing, M. 1990. The syntactic roots of semantic partition. PhD dissertation, University of Massachusetts-Amherst. Diesing, M. 1992. Indefinites. Cambridge MA: The MIT Press.
204
Scrambling and the Survive Principle
Diesing, M. 1997. Yiddish VP order and the typology of movement in Germanic. Natural Language and Linguistic Theory 15: 369-429. Domínguez, L. 2004. Mapping focus: The syntax and prosody of focus in Spanish. PhD dissertation, Boston University. Embick, D. & R. Noyer. 2001. Movement operations after syntax. Linguistic Inquiry 32: 555-595. Engel, U. 1972. Regeln zur Satzgliedfolge: Zur Stellung der Elemente im einfachen Verbalsatz. Linguistische Studien der Gegenwart 19: 17-75. Epstein, S. & Seely, T.D. 1998. A Derivational Approach to Syntactic Relations. Oxford: OUP. Epstein, S. & Seely, T.D. 2002. Rule applications as cycles in a level-free syntax. In Derivation and Explanation in the Minimalist Program, S. Epstein & T. D. Seely (eds.), 65-89. Oxford: Blackwell. Epstein, S. & Seely, T.D. 2006. Derivations in Minimalism. Cambridge: CUP. Etzenberger, J. 1979. On the sentential nature of prenominal adjectives in German. Folia Linguistica XX: 341-380. Faarlund, J. T. 2001. Introduction. In Grammatical Relations in Change, J. T. Faarlund (ed.), 1-15. Amsterdam: John Benjamins. Faarlund, J.T. 2004. The Syntax of Old Norse. Oxford: OUP. Fanselow, G. 2001. Features, theta roles, and free constituent order. Linguistic Inquiry 32: 404-437. Feng, S. 1995. The syntax-prosody interface. PhD dissertation, University of Pennsylvania. Feng, Shengli. 2003. Facts of prosodic syntax in Chinese. Ms., University of Kansas and Harvard University. Ferguson, K.S. 1996. A Feature-Relativized shortest move requirement. PhD dissertation, Harvard University. Fitzpatrick, J. 2002. On minimalist approaches to the locality of movement. Linguistic Inquiry 33(3): 443-63. Fodor, J.D. 1970. The linguistic description of opaque contexts. PhD dissertation, MIT. Fodor, J. D. & Sag, I.A. 1982. Referential and quantificational identities. Linguistics and Philosophy 5: 335-598. Fox, D. & Pesetsky, D. 2003. Cyclic linearization and the typology of movement. Handout. GLOW 2003. Fox, D. 2000. Economy and Semantic Interpretation. Cambridge MA: The MIT Press. Fox, D. 2003. On logical form. In Minimalist Syntax, R. Hendrick (ed), 82-123. Oxford: Blackwell. Fox, D. 2004. Condition A and scope reconstruction. Linguistic Inquiry 35: 475-485. Frampton, J. & Gutmann, S. 2002. Crash-proof syntax. In Derivation and Explanation in the Minimalist Program, S. Epstein & T. D. Seely (eds.), 90-105. Oxford: Blackwell. Frank, R., Y.Lee & Rambow, O. 1996. Scrambling, reconstruction and subject binding. Rivista di Grammatica Generativa 21: 67-106. Frege, G. 1884. Die Grundlagen der Arithmetik. Eine logisch mathematische Untersuchung über den Begriff der Zahl. Breslau: Koebner. Frey, J. W. A Simple Grammar of Pennsylvania Dutch. 2nd ed. 1945 [1981]. (Forward by C. R. Beam). Lancaster: John Baers and Son.
References
205
Frey, W. 1993. Syntaktische Bedingungen für die semantische Interpretation. Berlin: Akademie Verlag. Frey, W. 2004. A medial topic position for German. Linguistische Berichte 198: 153-190. Frey, W. & Pittner, K. 1998. Zur Positionierung der Adverbiale im deutschen Mittelfeld. Linguistische Berichte 176: 489-534. Fuller, J. 1997. Pennsylvania Dutch with a Southern touch. PhD dissertation, University of South Carolina. Galato, A. 2002. Grammar and interaction: Reported discourse and subjunctive in German. Zeitschrift für Sprachwissenschaft 22(1): 24-55. Grewendorf, G. 1984. Reflexivierungsregeln im Deutschen. Deutsche Sprache 1: 1430. Grewendorf, G. 1988. Aspekte der deutschen Syntax. Tübingen: Narr. Grewendorf, G. 2005. The discourse configurationality of scrambling. In The Free Word Order Phenomenon. Its Syntactic Sources and Diversity, J.Sabel & M. Saito (eds.), 75-135. Berlin: de Gruyter. Grewendorf, G. & Sternfeld, W. 1990. Scrambling theories. In Scrambling and Barriers, G. Grewendorf & W. Sternfeld (eds.), 3-37. Amsterdam: John Benjamins. Grewendorf, G. & Sabel, J. 1994. Long scrambling and incorporation. Linguistic Inquiry 25: 263-308. Grewendorf, G. & Sabel, J. 1999. Scrambling in German and Japanese: Adjunction versus multiple specifiers. Natural Language and Linguistic Theory 17: 1-65. Grohmann, K.. 2000a. Prolific peripheries: A radical view from the left. PhD dissertation, University of Maryland. Grohmann, K. 2000b. Prolific domains. University of Maryland Working Papers in Linguistics 9: 42-70. Grohmann, K. 2003. Prolific Domains: On the Anti-Locality Movement Dependencies. Amsterdam: John Benjamins. Grohmann, K. 2006. Towards PF. Paper presented at the 17th International Symposium on Theoretical and Applied Linguistics, Aristotle University of Thessaloniki, April 15-17. Grohmann, K. & Boeckx, C. 2004. Putting phases in perspective. Ms., University of Cyprus and Harvard University. Grohmann, K. & Putnam, M.T. 2004. Prosodic stress assignment in dynamic computations. Handout. Michigan Linguistic Society Annual Conference, University of Michigan, Flint. Grohmann, K. & Putnam, M.T. 2005. Prosodic stress assignment in dynamic computations. Handout. GLAC 11, University of California-Davis. Grohmann, K. & Putnam, M.T. In press. Dynamic prosodic stress assignment. In Dynamic Interphases [Linguistic Analysis], K. Grohmann (ed.), 32: 3-4. Giusti, G. 2001. The birth of a functional category: From Latin ILLE to the Romance article and personal pronoun. In Current Studies in Italian Syntax: Essays Offered to Lorenzo Renzi, G. Cinque & G. Salvi (eds), 157-171. Amsterdam: North Holland. Gussenhoven, C. 1983. Focus, mode and the nucleus. Journal of Linguistics 19: 377417. Gussenhoven, C. 1992. Sentence accents and argument structure. In Thematic Structure: Its Role in the Grammar, I.M. Roca (ed.), 79-106. Dordrecht: Foris.
206
Scrambling and the Survive Principle
Haag, E. C.1982 A Pennsylvania German Anthology. Cranbury PA: Associated University Presses. Haeberli, E. 2002. Features, Categories and the Syntax of A-Positions: CrossLinguistic Variation in the Germanic Languages. Dordrecht: Kluwer. Haftka, B. 1995. Syntactic positions for topic and contrastive focus in the German middlefield. In Göttingen Focus Workshop. Arbeitspapiere des SFB 340 Bericht Nr. 69, I. Kohlhof, S. Winkler & H.B. Drubig (eds.), 137-157. Haider, H. 1993. Deutsche Syntax – Generativ. Tübingen: Narr. Haider, H. 2005a. Middlefield phenomena (Scrambling in Germanic). In The Blackwell Companion to Syntax (Volumes I-IV), M. Everaert, R. Goedenmans, B. Hollebrandse, & H. van Riemsdijk (eds.), 204-274. Oxford: Blackwell. Haider, H. 2005b. Ho w to turn German into Icelandic – and derive the VO/OV contrasts. The Journal of Comparative Germanic Linguistics 8: 1-53. Haider, H. & Rosengren, I. 2003. Scrambling: Nontriggered chain formation in OV languages. Journal of Germanic Linguistics 15(3): 203-267. Halle, M. & Marantz, A. 1993. Distributed morphology and the pieces of inflection. In The View from Building 20: Essays in Honor of Sylvain Bromberger, K. Hale & S. K. Keyser (eds.), 111-1176. Cambridge MA: The MIT Press. Hayes, B. 1995. Metrical Stress Theory: Principle and Case studies. Chicago IL: The University of Chicago Press. Hayes, B. & Lahiri, A. 1991. Bengali intonational phonology. Natural Language and Linguistic Theory 9: 47- 96. Heim, I. 1982. The semantics of definite and indefinite noun phrases. PhD dissertation, University of Massachusetts-Amherst. Hendrikse, A.P. 1978. Topics in Xhosa Relativization: Some Traditional Analyses Re-examined. Grahamstown: Rhodes University. Hinterhölzl, R. 2003. Parametric variation and scrambling in English. In Studies in Comparative Germanic Syntax (Proceedings of CGSW 15 in Groningen, 2000), W. Abraham & C.J.-W. Zwart (eds.), 131-150. Amsterdam: John Benjamins. Hiraiwa, K. 2003. Cyclic locality. Ms., MIT. Hoekstra, E. 2001. Licensing conditions on phrase structure. PhD dissertation, Rijksuniversiteit Groningen, Groningen. Holmberg, A. 1986. Word order and syntactic features in the Scandinavian languages and English. PhD dissertation, University of Stockholm. Hornstein, N. 1995. Logical Form: From GB to Minimalism. Oxford: Blackwell. Hornstein, N. 2001. Move! A Minimalist Theory of Construal. Oxford: Blackwell. Hornstein, N. & Weinberg, A. 1981. Case theory and preposition stranding. Linguistic Inquiry 12: 55-91. Hoyer, C. 2004. Information structure and the left periphery in German. MA thesis, Oxford University. Höhle, T. 1982. Explikation für ‘normale Betonung’ und ‘normale Wortestellung.’ In Satzglieder im Deutschen, W. Abraham (ed.), 223-249. Amsterdam: John Benjamins. Huffines, M.L. 1987. The dative case in Pennsylvania German: Diverging norms in language maintenance and loss. Yearbook of German-American Studies 22: 173181. Jayaseelan, K.A. 2001. IP-internal topic and focus phrases. Studia Linguistica 55: 39-75. Jäger, G. 1996. Topics in dynamic semantics. PhD dissertation, Humboldt University, Berlin.
References
207
Julien, M. 2002. Determiners and word order in Scandinavian DPs. Studia Linguistica 56: 264-315. Julien, M. 2005. Nominal Phrases from a Scandinavian Perspective. Amsterdam: John Benjamins. Kayne, R. 1994. The Anti-Symmetry of Syntax. Cambridge MA: The MIT Press. Keel, W.D. 1994. Reduction and loss of case marking in the noun phrase in GermanAmerican speech islands: Internal development or external interference? In Sprachinselforschung: Eine Gedenkschrift für Hugo Jedig, N. Berend & K. J. Mattheier (eds.), 93-103. Berlin: Peter Lang. Keiser, S. H. 1999. An historical phonology of Holmes County Pennsylvania German. Lecture, University of Dayton. Kidwai, A. 2000. XP-Adjunction in Universal Grammar. Scrambling and Binding in Hindi-Urdu. Oxford: OUP. Kiparsky, P. 1997. The rise of positional licensing. In Parameters of Morphosyntactic Change, A. van Kemenade & N. Vincent (eds.), 460-494. Cambridge: CUP. Kiss, K.É. 1994. The Syntactic Structure of Hungarian. New York NY: Academic Press. Kiss, K.É. 1996. Two subject positions in English. The Linguistic Review 13: 119142. Kobele, G. 2006. Generating copies: An investigation into structural identity in language and grammar. PhD dissertation, University of California Los Angeles. Koch, M. 1974. A demystification of syntactic drift. Montreal Working Papers in Linguistics 3: 63-114. Krifka, M. 1992. A compositional semantics for multiple focus constructions. In Informationsstruktur and Grammatik, Linguistische Berichte Sonderheft 4: 1753, J. Jacobs (ed.). Krifka, M. 2001/02. Topik und Fokus. Class Notes. Kuno, S. 1972. Functional sentence perspective: A case study for Japanese and English. Linguistic Inquiry 3: 269-336. Larson, R. 1988. On the double-object construction. Linguistic Inquiry 19: 355-391. Lasnik, H. 1999. Minimalist Analysis. Malden MA: Blackwell. Lebeaux, D. 1991. Relative clauses, licensing, and the nature of the derivation. In Perspectives on Phrase Structure: Heads and Licensing, S. Rothstein (ed), 209239. San Diego CA: Academic Press. Lebeaux, D. 1995. Where does the binding theory apply? University of Maryland Working Papers in Linguistics 3: 63-88. Department of Linguistics:University of Maryland. Lechner, W. 1997. On semantic and syntactic reconstruction. Ms., University of Massachusetts-Amherst. Legate, J. 2002. Some interface properties of the phase. Linguistic Inquiry 34: 506516. Lenerz, J. 1977. Zur Abfolge nominaler Satzglieder im Deutschen. Tübingen: Narr. Lenerz, J. 2002. Scrambling and reference in German. In Issues in Formal German(ic) Typology, W. Abraham & C.J.-W. Zwart (eds.), 179-192. Amsterdam: John Benjamins. Liberman, M. & Prince. A. 1977. On stress and linguistic rhythm. Linguistic Inquiry 8: 249-336. Lidz, J. 2001. Condition R. Linguistic Inquiry 32: 123-140.
208
Scrambling and the Survive Principle
Lightfoot, D.W. (ed.). 2002. Syntactic Effects of Morphological Change. Oxford: OUP. Lopez, L. 2002. Towards a grammar without TopP or FocP. Paper presented at Workshop on the Syntax-Semantics Interface in the CP-Domain. Louden, M.L. 1989. Syntactic variation and change in Pennsylvania German. In Studies on the Languages and Verbal Behavior of the Pennsylvania Germans, W.Enninger, J. Raith & K.Wandt (eds.), 29-40. Stuttgart: Franz Steiner Verlag. Lötscher, A. 1981. Abfolgeregeln für Ergänzungen im Mittelfeld. Deutsche Sprache 29: 44-60. Mahajan, A. 1990. The A/A-bar distinction and movement theory. PhD dissertation, MIT. Mahajan, A. 1994. Against the relevance of subjacency at LF. Linguistic Inquiry 25(1): 171-179. Marantz, A. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. University of Pennsylvania Working Papers in Linguistics 4(2): 201-225. Matushansky, O. 2004. Going through a phase. In Proceedings of the MIT Workshop on EPP and Phases, M. McGinnis & N. Richards (eds.), 67-80. Cambridge MA: MITWPL. McFadden, T. 2004. The position of morphological case in the derivation: A study on the syntax-morphology interface. PhD dissertation, University of Pennsylvania. Meinunger, A. 1995. Discourse dependent DP (de-)placement. Groninger Arbeiten zur gemanistischen Linguistik (GAGL) 39. Meinunger, A. 2000. Syntactic Aspects of Topic and Comment. Amsterdam: John Benjamins. Miyagawa, S. 1997. Against optional scrambling. Linguistic Inquiry 28(1): 1-25. Miyagawa, S. 2000. EPP, scrambling, and wh-in-situ. In Ken Hale: A life in language, M. Kenstowicz (ed.), 293-338. Cambridge MA: The MIT Press. Mohr, S. 2005. Clausal Architecture and Subject Positions: Impersonal Constructions in the Germanic Languages. Amsterdam: John Benjamins. Molnár, V. 1991. Das TOPIK im Deutschen und Ungarischen. Stockholm: Almqvist & Wiksell International. Molnár, V. 1998. Topic in focus – On the syntax, phonology, semantics and pragmatics of the so-called ‘contrastive topic’ in Hungarian and German. Acta Linguistica Hungarica 45(1-2): 89-166. Molnárfi, L. 2002. Focus and antifocus in modern Afrikaans and West Germanic. Linguistics 40: 1107-1160. Molnárfi, L. 2003. On optional movement and feature checking in West Germanic. In Optionality in Syntax and Discourse Structure, W. Abraham & L. Molnarfi (eds.), 129-162. Berlin: Mouton de Gruyter. Moltmann, F. 1991. Coordination and comparatives. PhD dissertation, MIT. Müller, G. 1998. Incomplete Category Fronting. Dordrecht: Kluwer. Müller, G. 2000. Shape conservation and remnant movement. Ms., University of Tübingen. Müller, G. 2001. Order preservation, parallel movement, and the emergence of the unmarked. In Minimality Effects in Syntax, G. Legendre, J. Grimshaw & S. Vikner (eds.), 279-313. Cambridge MA: The MIT Press. Müller, G. & Sternefeld, W. 1993. Improper movement and unambiguous binding. Linguistic Inquiry 24(3): 461-507.
References
209
Neeleman, A. 1994. Scrambling as a D-structure phenomenon. In Studies on Scrambling: Movement and non-movement approaches to free-word-order phenomena, N. Corver & H. van Riemsdijk (eds.), 387-429. Berlin: Mouton de Gruyter. Newmeyer, F. 2003. On split-CPs, uninterpretable features, and the ‘perfectness’ of language. Ms., University of Washington, Seattle. O’Grady, W. 2005. Syntactic Carpentry. Mahwah NJ: Lawrence Erlbaum Associates. Parafita Couto, M.C. 2005. Focus at the interface. PhD dissertation, University of Kansas. Parker, A.R. 1999. Parasitic gaps in the Germanic languages. MA thesis, National University of Ireland, Dublin. Pesetsky, D. 1982. Paths and categories. PhD dissertation, MIT. Pesetsky, D. 1995. Zero Syntax: Experiencers and Cascades. Cambridge MA: The MIT Press. Philippi, J. 1997. The rise of the article in the Germanic Languages. In Parameters of Morphsyntactic Change, A. van Kemenade & N. Vincent (eds.), 62-93. Cambridge: CUP. Platzack, C. 1991. Comp, Infl and Germanic word order. In Topics in Scandinavian Syntax, L. Helland & K. Koch Christensen (eds.), 185-234. Dordrecht: Foris. Pierrehumbert, J.B. 1980. Synthesizing intonation. Journal of the Acoustical Society of America 70: 985-995. Pittner, K. 1999. Adverbiale im Deutschen. Untersuchungen zu ihrer Stellung und Interpretation. Tübingen: Narr. Pollock, J.Y. 1989. Verb-movement, universal grammar and the structure of IP. Linguistic Inquiry 20: 365-424. Pollard, C. & Sag, I. 1994. Head Driven Phrase Structure Grammar. Standford CA: CSLI. Ponelis, F. 1993. The Development of Afrikaans. Frankfurt: Peter Lang. Putnam, M.T. 2002. Case morphology syncretism in Ohio Pennsylvania German: Internal change or external influence? MA thesis, University of Kansas. Putnam, M.T. 2003. The prepositional case in German-American dialects. Focus on German Studies 10: 207-224. Putnam, M.T. 2005. An anti-local account of why scrambled datives in German can’t bind anaphors. SKY Journal of Linguistics 18: 287-309. Putnam, M.T. 2006a. Scrambling in West Germanic as XP-adjunction: A critical analysis of prolific domains. PhD dissertation, University of Kansas-Lawrence. Putnam, M.T. 2006b. Eliminating agree. Ms., Michigan State University. Putnam, M. T. 2006c. Re-thinking anti-locality. Ms., Michigan State University. Putnam, M.T. & Stroik, T. In progress. The Survive manifesto: A derivational guide to minimalist syntax. Ms., Carson-Newman College and the University of Missouri-Kansas City. Ramchand, G. 1997. Aspect and Predication. Oxford: Clarendon Press. Raposo, E. 1987. Case theory and Infl-to-Comp: The inflected infinitive in European Portuguese. Linguistic Inquiry 18: 85-109. Reinhart, T. 1981. Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27: 53-94. Reinhart, T. 1995. Interface strategies. Ms., Technical Report OTS Working Papers, Utrecht University.
210
Scrambling and the Survive Principle
Reis, M. 1986. Die Stellung der Verbargumente im Deutschen. Stilübungen zum Grammatik-Pragmatik-Verhältnis. Paper presented at the 5th Lund Symposium “Sprache und Pragmatik,” May 12-16. Richards, M.D. 2005. Object shift and scrambling in North and West Germanic: A case study in symmetrical syntax. PhD dissertation, Robinson College. Richards, N. 1997. What moves where when in which language? PhD dissertation, MIT. Rettig, L. 1976. Amana Today. Ames IA: Iowa State Press. Ritter, E. & Rosen, S.T. 2001. The interpretative value of object splits. Language Sciences 23: 425-452. Rizzi, L. 1991. Relativized Minimality. Cambridge MA: The MIT Press. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar, L. Haegeman (ed.), 281-337. Dordrecht: Kluwer. Rizzi, L. 2004. The Structure of CP and IP: The Cartography of Syntactic Structures, Vol. 2. Oxford: OUP. Roberts, I. & Roussou, A. 2003. Syntactic Change. Cambridge: CUP. Rochemont, M. 1986. Focus in Generative Grammar. Amsterdam: John Benjamins. Roehrs, D. 2006. The Germanic noun phrase: Determiners MOVE into the determiner phrase. PhD dissertation, Indiana University. Ross, J. 1967. Constraints on variables in syntax. PhD dissertation, MIT. Rubin, E. 2003. Determining pair-merge. Linguistic Inquiry 34: 660-668. Russ, C. et al. 1989. The Dialects of Modern German. Oxford: OUP. Sabel, J. 2005. String vacuous scrambling and the effect on output condition. In The Free Word Order Phenomenon: Its Syntactic Sources and Diversity, J. Sabel & M. Saito (eds.), 1-23. Berlin: Mouton de Gruyter. Saito, M. 1989. Scrambling as semantically vacuous A’-movement. In Alternative Concepts of Phrase Structure, M. Baltin & A. Kroch (eds.), 182-200. Chicago IL: The University of Chicago Press. Saito, M. 1992. Long distance scrambling in Japanese. Journal of East Asian Linguistics 1: 69-118. Saito, M. 2003. A derivational approach to the interpretation of scrambling chains. Lingua 113: 481-518. Sauerland, U. 1999. Erasability and interpretation. Syntax 2(3): 161-188. Schaeffer, J. 2000. The Acquisition of Direct Object Scrambling and Clitic Placement: Syntax and Pragmatics. Amsterdam: John Benjamins. Schirmunski, V. 1962. Deutsche Mundartkunde. Berlin: Akademie-Verlag. Sekerina, I. 1997. The syntax and processing of split scrambling constructions in Russian. PhD dissertation, CUNY Graduate School and University Center. Sekerina, I. 2003. Scrambling and processing: Dependencies, complexities and constraints. In Word Order and Scrambling, S. Karimi (ed.), 301-324. Malden MA: Blackwell. Selkirk, E. 1986. On derived domains in sentence phonology. Phonology 3: 371-405. Sigurðsson, H. 2004. Agree and agreement: Evidence from Germanic. In Focus on Germanic Typology, W. Abraham (ed.), 61-103. Berlin: Akademie Verlag. Sigurðsson, H. 2007. Remarks on features. In Phases at the Interfaces [Interface Explorations], K. Grohmann (ed.), pp.-pp.. Berlin: Mouton de Gruyter. Sportiche, D. 1992. Clitic constructions. Ms., University of California Los Angeles. Starke, M. 2001. Move dissolves into Merge: A theory of locality. PhD dissertation, University of Geneva.
References
211
Stowell, T. & Beghelli, F. 1997. Distributivity and negation. In Ways of Scope Taking, A. Szabolcsi, (ed.), 71-107. Dordrecht: Kluwer. Stroik, T. 1992. Middles and movement. Linguistic Inquiry 23: 127-137. Stroik, T. 1995. On middle formation: A reply to Zribi-Hertz. Linguistic Inquiry 26: 165-171. Stroik, T. 1999. The Survive Principle. Linguistic Analysis 29: 278-303. Stroik, T. 2000. Syntactic Controversies. München: Lincom. Stroik, T. 2007. Locality in Minimalist Syntax. Cambridge MA: The MIT Press. Stroik, T. & Putnam, M.T. 2005a. The lexicon at the interfaces. Paper presented at LASSO 35. Lubbock, Texas. Stroik, T. & Putnam, M.T. 2005b. One step closer to a crash-proof syntax. Paper presented at DEAL. Berlin, Germany. Stroik, T. & Putnam, M.T. 2007. Surviving reconstruction. In Phases at the Interfaces [Interface Explorations], K. Grohmann (ed.), pp.-pp.. Berlin: Mouton de Gruyter. Szabolsci, A. 1983. The possessor that ran away from home. The Linguistic Review 3: 89-102. Thráinsson, H. 1996. On the (non-)universality of functional categories. In Minimal Ideas-Syntactic Studies in the minimalist framework, W. Abraham, S.D. Epstein, H. Thráinsson & C. J-W. Zwart (eds.), 253-281. Amsterdam: John Benjamins. Ticio, E. 2003. On the structure of DPs. PhD dissertation, University of Connecticut. Travis, L. 1994. Event phrase and a theory of functional categories. Proceedings of the 1994 Annual Conference of the Canadian Linguistic Society, 559-570. Urigereka, J. 1999. Multiple spell-out. In Working Minimalism, S.D. Epstein & N. Hornstein (eds.), 251-282. Cambridge MA: The MIT Press. van Gelderen, E. 2004. Grammaticalization as Economy. Amsterdam: John Benjamins. van Valin, R.D. 2001. Minimalism and explanation. Ms., The State University of New York at Buffalo. von Stechow, A. 1995. Lexical decomposition in syntax. In The Lexicon in the Organization of Language, U. Egli, P.E. Pause, C. Schwarze, A. von Stechow, & G. Wienold (eds.), 81-118. Amsterdam: John Benjamins. von Stechow, A. 1996. The different readings of wieder “again” – a structural account. Journal of Semantics 13: 87-138. Webelhuth, G. 1989. Syntactic saturation phenomena and the modern Germanic languages. PhD dissertation, University of Massachusetts-Amherst. Webelhuth, G. 1992. Principles and Parameters of Syntactic Saturation. Oxford: OUP. Webelhuth, W. & den Besten, H. 1987. Adjunction and remnant topicalization in the Germanic SOV-languages. Paper Presented at 10th GLOW Colloquium, Venice. Wexler, K. & Culicover, P.W. 1980. Formal Principles of Language Acquisition. Cambridge MA: The MIT Press. Williams, E. 2003. Representation Theory. Cambridge MA: The MIT Press. Winkler, S. & Göbbel, E. 2002. Focus, p-movement, and the nuclear-stress rule: A View from Germanic and Romance. Linguistics 40(6): 1185-1242. Yang, C. 1999. Unordered merge and its linearization. Syntax 2(1): 38-64. Zamparelli, R. 2000. Layers in the Determiner Phrase. New York NY: Garland. Zhang, N. 2003. Move is remerge. Ms., Berlin: ZAS.
212
Scrambling and the Survive Principle
Zubizarreta, M.L. 1998. Prosody, Focus and Word Order. Cambridge MA: The MIT Press. Zubizarreta, M.L. & Vergnaud, J. 2000. Phrasal stress, focus and syntax. Ms.,USC. Zwart, C. J.-W. 1993. Dutch syntax: A Minimalist approach [Groningen Dissertations in Linguistics 10]. Groningen: University of Groningen. Zwart, C. J.-W. 1997. Morphosyntax of Verb Movement: A minimalist approach to the syntax of Dutch. Dordrecht: Kluwer.
SUBJECT INDEX -AAdjunction, 34-38, 67, 76, 126-33, 153, 166, 190 Affect α, 28-29 Afrikaans, 48, 66-68, 140, 157-160 Agree (Probe-Goal), 21, 22, 26, 146, 157 Agreement features, 4-5, 14-15, 27, 31, 67, 81, 98, 108, 122, 140, 146, 157, 180-182, 191 Dynamic Agreement, 19, 36 Amana German, 63-65 A-movement, 5, 24, 33, 37, 46, 5368, 95, 108-109, 128, 140, 176 A'-movement, 5, 29, 31, 37, 46, 5368, 95, 127-129, 136, 186 Anti-focus, 44, 74-76, 154, 157164 Anti-locality, 115-119, 121, 124, 161 -BBinding, Anaphoric Binding, 37, 54-55, 119-123, 153, 176, 192 Condition C asymmetries, 128133 -CCase, 4, 24, 30, 58-68, 97-99, 111, 122-124, 139-140, 175, 198 Chains, 13, 21-23, 116-118, 126, 151, 159, 161, 171-175, 185, 190 Agreement, 68-71 Object, 68-71, 73 Clitic, 157, 184-188 Complementizer Phrase (CP), 17, 26-27, 33, 47, 52-53, 64, 72, 7778, 81, 88-89, 91, 94, 112-115, 124, 160, 171, 182, 193 Computation of Human Language (CHL), 14, 24, 26, 29-30, 98-99,
114, 130, 138, 145, 151, 155, 167, 188, 190, 199 Condition on Domain Exclusivity (CDE), 116, 118, 121-123 Coherent infinitives, 168-175 Copy Theory, 19-20, 22, 24, 151 -DDerivation, 6, 8, 11, 14-20, 48, 81, 91, 99-100, 124, 154-155, 159, 161-164, 171, 176 Descriptive Adequacy, 144 Determiner Phrase (DP), 47-48, 5354, 68-72, 78, 98-100, 105-110, 168, 193-195 Discourse (Pragmatics), 8-10, 3233, 37, 41, 43, 47, 73-78, 84-86, 91-94, 112-114, 137, 153-159, 161, 181-182, 191 Dutch, 9, 31, 47-48, 57, 65-66, 88, 91, 133-136, 150, 187 -EE-domain, 182, 192-193 Economy, 138, 142, 153, Constraints, 34 Greed, 8, 13, 26, 34, 96, 192 Last Resort, 8, 12-13, 29, 3435, 96, 192 Procrastinate, 96, 192 English, 100-102, 111, 136 EPP-feature, 22, 27, 160, 166, 172173, 192 Explanatory Adequacy, 44, 144147 -FFeature Checking, 13, 35-37, 146, 153-154, 167 Feature(s), 4, 6-13, 15-20, 35-37, 47, 51, 81-82, 145-147, 153-164, 192, 196 Featurehood, 97-100, 127 Focus, 31-33, 60-61, 74-76, 90-91, 99, 127, 147-154, 177-179
214
Focus Prominence Rule (FPR), 167, 178 Freezing Effects, 82-84, 123 Full Interpretation, 13, 23, 25, 99, 139, 142, 145, 147, 154, 175, 188, 198 Functional Projection (FP), 7-8, 31, 33, 99, 106-109, 112, 145, 153, 157, 161, 165-166, 186, 193 -GGerman, 2-3, 7, 32, 49-50, 51-58, 69-79, 82-88, 92, 98, 102-103, 108-109, 120-125, 136, 150, 164, 168-174, 178-181, 183, 194-195 Government and Binding Theory (GB), 5, 11, 29, 34, 112, 115, 154 -HHead, 1, 6, 8, 10-11, 13, 15, 17-18, 20, 25, 27, 34, 79, 92, 106-108, 110-115, 146-147, 156, 164 Head-Parameter, 38-41, 108-110, 148-149 Hungarian, 34 -IIndefinites, 74-80, 86, 94, 160-161, 193 Information Structure (IS), 147, 161, 178-180, 196 Islands, 83, 85, 93, 169, 197 Italian, 184-188 -JJapanese, 29-30, 72, 196-199 -LLexicon, 4, 13, 15, 22-23, 26, 75, 92, 126, 132, 138, 146-147, 160, 171 LF-lowering, 28-31, 196-197
Scrambling and the Survive Principle
Light Verb Phrase (vP), 7, 15, 17, 22, 26-28, 30, 94, 110-115, 117120, 162-165, 182, 193 Linearization, 110, 155, 165-167, 176, 199 Link!, 20-24, 26, 30-31, 33, 96, 110, 147, 163, 171, 180, 187, Logical Form (LF), 8, 11-13, 1924, 28-31, 33, 35, 37, 41, 99, 108, 112, 138, 153, 161, 163, 175, 193 -MMapping Hypothesis, 78-80 Merge, 5-6, 10-15, 19-23, 96, 100, 102, 110-111, 116, 124, 129, 146, 171 Late Num Merge, 132-133, 135 Pair Merge, 128, 130-131 Set Merge, 130 Minimalist Program (MP), 3, 1113, 19, 24, 34-35, 112, 116, 126, 128, 151, 163-165, 180-182 Model of the Grammar, 153, 175184, 190-191 Move, 3, 6, 8, 14, 19, 26, 109, 116117, 126, 129, 132, 138, 142, 146, 171 Internal Merge, 5-6, 10, 15, 19, 22-23, 35-36, 126, 129, 132, 138, 146 , 171 Move α, 3, 8, 29, 31-34, 135 -NNarrow Syntax, 6-8, 10, 15, 19-20, 22-23, 25-26, 28, 110, 113, 127, 129-130, 132, 140, 147, 152154, 166, 175, 182, 191, 196 NP-movement, 46-47, 53-68 Numeration, 6, 15, 19-20, 22, 75, 98-100, 118, 132, 135, 171, 190 -OObject Shift, 50, 68-71, 191 Observational Adequacy, 143-144, 175
Subject Index
Old Norse, 105-108 Omega domain (Ω-domain), 114115, 124, 182, 191 -PParasitic Gaps, 57-58, 70, 133-136 Pennsylvania German, 60-63, 65, 67, 136-142 Phases, 17-25, 26-28, 32-33, 110113, 117, 152, 163-167, 171176, 184 Phi-domain (Φ-domain), 27, 114115, 180-182, 191 Phi-features (φ-features), 10, 46, 97-98, 181, 184 Phonological Form (PF), 8, 11-13, 20, 22-24, 73, 90-91, 153-157, 159, 161-167, 174-179, 198-199 Preposition Stranding, 71 Prolific Domains, 25-27, 110-124, 147, 155, 157, 164-167, 177184, 190-191 Prosodic Checking, 155-157, 163 Prosodic Movement (p-movement), 152-155, 160, 163-164, Prosody, 48, 61, 66, 87-88, 148157, 162, 169, 175-176 -RReconstruction, 55-56, 176, 195, 199 Reference Phrase (RP, RefP), 106107, 124, 140, 184-188 Referentiality, 9-10, 33, 36, 41-42, 84-86, 97, 100-110, 139, 184188 Relative Clauses, 85, 100-102, Non-restrictive, 85-86, 102-103 Remerge, 6, 15-17, 19-23, 26-27, 29, 96-97, 102, 109-111, 126, 142, 171, 175, 190 Remnant Movement, 78, 83-84, 123-126 Rhematic Marker, 66, 157-164 Rheme, 76, 137 Russian, 196-199
215
-SScope, 79-80, 86, 102, 135, 184, 192, 195-199 ambiguity, 87-90 Scrambling, 1-4, 8-12, 27-28, 4145, 168-170, 195-199 Above the subject, 3, 33, 124125, 195, 196-199 Loss of, 136-142 Middle field, 6-11, 27, 120-125, 158, 164, 168-170 Properties of, Chap. 2 Strong, 47-50, Weak, 47-50, Selectional Properties of Verbs, 5153 Semantics interpretations, de dicto reading, 193-195 de re reading, 193-195 repetitive readings, 183-184 restitutive readings, 183-184 Sentence Accent Assignment Rule (SAAR), 176-178 Simpl, 130-132 Spanish, 152 Spell-Out, 12, 14, 26, 39, 51, 90, 111, 113-114, 118, 154-155, 163, 167, 171 Subjacency, 82, 197 Subject-Object-Verb (SOV), 28, 39, 70, 139, 141, 150, 167, 186 Subject-Verb-Object (SVO), 28, 39-41, 70, 103, 137, 167, 186 Survive, 6-8, 14-20, 99-100, 132, 171, 199 Survive-minimalism, 24-27, 31, 33, 41-44, 96-98, 110-114, 127-130, 115, 130, 171, 180-182, 190-191 Swedish, 69-71 -TTense Phrase (TP), 27, 50, 77, 80, 89, 94, 112-115, 161, 165, 180182
216
Theta-domain (Θ-domain), 114115, 118, 120-122, 163, 165, 181-182, 191 Topic, 73-78, 124-127, 137-139, 153, 185 aboutness, 76-78,153 contrastive, 94, familiarity, 76-78,153 Topicalization, 73-78, 84, 117, 124-126, 142, 169, 182, Trace, 25, 53, 57, 84, 85, 93, 133, 135, 150-151, 170, 195 -UUniversal Grammar (UG), 11-12, 35, 76, 105, 183
Scrambling and the Survive Principle
-VVerb Phrase (VP), 8, 25, 32, 120122, 150-152, 161-162 -WWeak Crossover Effects (WCOs), 54 West Germanic, 5, 7-11, 27-28, 31, 41-43,50-51, 58, 66, 71-73, 76, 84-86, 94-97, 100, 115, 150, 156-159, 180-182 -XXhosa, 103-105 -YYiddish, 39-40
Linguistik Aktuell/Linguistics Today A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 118 Jäger, Agnes: History of German Negation. x, 351 pp. Expected January 2008 117 Haugen, Jason D.: Morphology at the Interfaces. Reduplication and Noun Incorporation in Uto-Aztecan. xv, 250 pp. Expected January 2008 116 Endo, Yoshio: Locality and Information Structure. A cartographic approach to Japanese. x, 239 pp. Expected October 2007 115 Putnam, Michael T.: Scrambling and the Survive Principle. 2007. ix, 217 pp. 114 Lee-Schoenfeld, Vera: Beyond Coherence. The syntax of opacity in German. 2007. viii, 206 pp. 113 Eythórsson, Thórhallur (ed.): Grammatical Change and Linguistic Theory. The Rosendal papers. vi, 458 pp. + index. Expected November 2007 112 Axel, Katrin: Studies on Old High German Syntax. Left sentence periphery, verb placement and verbsecond. 2007. xii, 364 pp. 111 Eguren, Luis and Olga Fernández Soriano (eds.): Coreference, Modality, and Focus. Studies on the syntax–semantics interface. xii, 236 pp. + index. Expected November 2007 110 Rothstein, Susan (ed.): Theoretical and Crosslinguistic Approaches to the Semantics of Aspect. vi, 439 pp. + index. Expected November 2007 109 Chocano, Gema: Narrow Syntax and Phonological Form. Scrambling in the Germanic languages. 2007. x, 333 pp. 108 Reuland, Eric, Tanmoy Bhattacharya and Giorgos Spathas (eds.): Argument Structure. v, 241 pp. + index. Expected November 2007 107 Corver, Norbert and Jairo Nunes (eds.): The Copy Theory of Movement. 2007. vi, 388 pp. 106 Dehé, Nicole and Yordanka Kavalova (eds.): Parentheticals. 2007. xii, 314 pp. 105 Haumann, Dagmar: Adverb Licensing and Clause Structure in English. 2007. ix, 438 pp. 104 Jeong, Youngmi: Applicatives. Structure and interpretation from a minimalist perspective. 2007. vii, 144 pp. 103 Wurff, Wim van der (ed.): Imperative Clauses in Generative Grammar. Studies in honour of Frits Beukema. 2007. viii, 352 pp. 102 Bayer, Josef, Tanmoy Bhattacharya and M.T. Hany Babu (eds.): Linguistic Theory and South Asian Languages. Essays in honour of K. A. Jayaseelan. 2007. x, 282 pp. 101 Karimi, Simin, Vida Samiian and Wendy K. Wilkins (eds.): Phrasal and Clausal Architecture. Syntactic derivation and interpretation. In honor of Joseph E. Emonds. 2007. vi, 424 pp. 100 Schwabe, Kerstin and Susanne Winkler (eds.): On Information Structure, Meaning and Form. Generalizations across languages. 2007. vii, 570 pp. 99 Martínez-Gil, Fernando and Sonia Colina (eds.): Optimality-Theoretic Studies in Spanish Phonology. 2007. viii, 564 pp. 98 Pires, Acrisio: The Minimalist Syntax of Defective Domains. Gerunds and infinitives. 2006. xiv, 188 pp. 97 Hartmann, Jutta M. and László Molnárfi (eds.): Comparative Studies in Germanic Syntax. From Afrikaans to Zurich German. 2006. vi, 332 pp. 96 Lyngfelt, Benjamin and Torgrim Solstad (eds.): Demoting the Agent. Passive, middle and other voice phenomena. 2006. x, 333 pp. 95 Vogeleer, Svetlana and Liliane Tasmowski (eds.): Non-definiteness and Plurality. 2006. vi, 358 pp. 94 Arche, María J.: Individuals in Time. Tense, aspect and the individual/stage distinction. 2006. xiv, 281 pp. 93 Progovac, Ljiljana, Kate Paesani, Eugenia Casielles and Ellen Barton (eds.): The Syntax of Nonsententials. Multidisciplinary perspectives. 2006. x, 372 pp. 92 Boeckx, Cedric (ed.): Agreement Systems. 2006. ix, 346 pp. 91 Boeckx, Cedric (ed.): Minimalist Essays. 2006. xvi, 399 pp. 90 Dalmi, Gréte: The Role of Agreement in Non-Finite Predication. 2005. xvi, 222 pp. 89 Velde, John R. te: Deriving Coordinate Symmetries. A phase-based approach integrating Select, Merge, Copy and Match. 2006. x, 385 pp. 88 Mohr, Sabine: Clausal Architecture and Subject Positions. Impersonal constructions in the Germanic languages. 2005. viii, 207 pp.
87 86 85 84 83 82 81 80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65 64 63 62 61 60 59 58 57 56 55 54
Julien, Marit: Nominal Phrases from a Scandinavian Perspective. 2005. xvi, 348 pp. Costa, João and Maria Cristina Figueiredo Silva (eds.): Studies on Agreement. 2006. vi, 285 pp. Mikkelsen, Line: Copular Clauses. Specification, predication and equation. 2005. viii, 210 pp. Pafel, Jürgen: Quantifier Scope in German. 2006. xvi, 312 pp. Schweikert, Walter: The Order of Prepositional Phrases in the Structure of the Clause. 2005. xii, 338 pp. Quinn, Heidi: The Distribution of Pronoun Case Forms in English. 2005. xii, 409 pp. Fuss, Eric: The Rise of Agreement. A formal approach to the syntax and grammaticalization of verbal inflection. 2005. xii, 336 pp. Burkhardt, Petra: The Syntax–Discourse Interface. Representing and interpreting dependency. 2005. xii, 259 pp. Schmid, Tanja: Infinitival Syntax. Infinitivus Pro Participio as a repair strategy. 2005. xiv, 251 pp. Dikken, Marcel den and Christina M. Tortora (eds.): The Function of Function Words and Functional Categories. 2005. vii, 292 pp. Öztürk, Balkız: Case, Referentiality and Phrase Structure. 2005. x, 268 pp. Stavrou, Melita and Arhonto Terzi (eds.): Advances in Greek Generative Syntax. In honor of Dimitra Theophanopoulou-Kontou. 2005. viii, 366 pp. Di Sciullo, Anna Maria (ed.): UG and External Systems. Language, brain and computation. 2005. xviii, 398 pp. Heggie, Lorie and Francisco Ordóñez (eds.): Clitic and Affix Combinations. Theoretical perspectives. 2005. viii, 390 pp. Carnie, Andrew, Heidi Harley and Sheila Ann Dooley (eds.): Verb First. On the syntax of verbinitial languages. 2005. xiv, 434 pp. Fuss, Eric and Carola Trips (eds.): Diachronic Clues to Synchronic Grammar. 2004. viii, 228 pp. Gelderen, Elly van: Grammaticalization as Economy. 2004. xvi, 320 pp. Austin, Jennifer R., Stefan Engelberg and Gisa Rauh (eds.): Adverbials. The interplay between meaning, context, and syntactic structure. 2004. x, 346 pp. Kiss, Katalin É. and Henk van Riemsdijk (eds.): Verb Clusters. A study of Hungarian, German and Dutch. 2004. vi, 514 pp. Breul, Carsten: Focus Structure in Generative Grammar. An integrated syntactic, semantic and intonational approach. 2004. x, 432 pp. Mišeska Tomić, Olga (ed.): Balkan Syntax and Semantics. 2004. xvi, 499 pp. Grohmann, Kleanthes K.: Prolific Domains. On the Anti-Locality of movement dependencies. 2003. xvi, 372 pp. Manninen, Satu Helena: Small Phrase Layers. A study of Finnish Manner Adverbials. 2003. xii, 275 pp. Boeckx, Cedric and Kleanthes K. Grohmann (eds.): Multiple Wh-Fronting. 2003. x, 292 pp. Boeckx, Cedric: Islands and Chains. Resumption as stranding. 2003. xii, 224 pp. Carnie, Andrew, Heidi Harley and MaryAnn Willie (eds.): Formal Approaches to Function in Grammar. In honor of Eloise Jelinek. 2003. xii, 378 pp. Schwabe, Kerstin and Susanne Winkler (eds.): The Interfaces. Deriving and interpreting omitted structures. 2003. vi, 403 pp. Trips, Carola: From OV to VO in Early Middle English. 2002. xiv, 359 pp. Dehé, Nicole: Particle Verbs in English. Syntax, information structure and intonation. 2002. xii, 305 pp. Di Sciullo, Anna Maria (ed.): Asymmetry in Grammar. Volume 2: Morphology, phonology, acquisition. 2003. vi, 309 pp. Di Sciullo, Anna Maria (ed.): Asymmetry in Grammar. Volume 1: Syntax and semantics. 2003. vi, 405 pp. Coene, Martine and Yves D’hulst (eds.): From NP to DP. Volume 2: The expression of possession in noun phrases. 2003. x, 295 pp. Coene, Martine and Yves D’hulst (eds.): From NP to DP. Volume 1: The syntax and semantics of noun phrases. 2003. vi, 362 pp. Baptista, Marlyse: The Syntax of Cape Verdean Creole. The Sotavento varieties. 2003. xxii, 294 pp. (incl. CD-rom).
53 Zwart, C. Jan-Wouter and Werner Abraham (eds.): Studies in Comparative Germanic Syntax. Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27, 2000). 2002. xiv, 407 pp. 52 Simon, Horst J. and Heike Wiese (eds.): Pronouns – Grammar and Representation. 2002. xii, 294 pp. 51 Gerlach, Birgit: Clitics between Syntax and Lexicon. 2002. xii, 282 pp. 50 Steinbach, Markus: Middle Voice. A comparative study in the syntax-semantics interface of German. 2002. xii, 340 pp. 49 Alexiadou, Artemis (ed.): Theoretical Approaches to Universals. 2002. viii, 319 pp. 48 Alexiadou, Artemis, Elena Anagnostopoulou, Sjef Barbiers and Hans-Martin Gärtner (eds.): Dimensions of Movement. From features to remnants. 2002. vi, 345 pp. 47 Barbiers, Sjef, Frits Beukema and Wim van der Wurff (eds.): Modality and its Interaction with the Verbal System. 2002. x, 290 pp. 46 Panagiotidis, Phoevos: Pronouns, Clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. 2002. x, 214 pp. 45 Abraham, Werner and C. Jan-Wouter Zwart (eds.): Issues in Formal German(ic) Typology. 2002. xviii, 336 pp. 44 Taylan, Eser Erguvanlı (ed.): The Verb in Turkish. 2002. xviii, 267 pp. 43 Featherston, Sam: Empty Categories in Sentence Processing. 2001. xvi, 279 pp. 42 Alexiadou, Artemis: Functional Structure in Nominals. Nominalization and ergativity. 2001. x, 233 pp. 41 Zeller, Jochen: Particle Verbs and Local Domains. 2001. xii, 325 pp. 40 Hoeksema, Jack, Hotze Rullmann, Víctor Sánchez-Valencia and Ton van der Wouden (eds.): Perspectives on Negation and Polarity Items. 2001. xii, 368 pp. 39 Gelderen, Elly van: A History of English Reflexive Pronouns. Person, Self, and Interpretability. 2000. xiv, 279 pp. 38 Meinunger, André: Syntactic Aspects of Topic and Comment. 2000. xii, 247 pp. 37 Lutz, Uli, Gereon Müller and Arnim von Stechow (eds.): Wh-Scope Marking. 2000. vi, 483 pp. 36 Gerlach, Birgit and Janet Grijzenhout (eds.): Clitics in Phonology, Morphology and Syntax. 2001. xii, 441 pp. 35 Hróarsdóttir, Thorbjörg: Word Order Change in Icelandic. From OV to VO. 2001. xiv, 385 pp. 34 Reuland, Eric (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. xii, 255 pp. 33 Puskás, Genoveva: Word Order in Hungarian. The syntax of Ā-positions. 2000. xvi, 398 pp. 32 Alexiadou, Artemis, Paul Law, André Meinunger and Chris Wilder (eds.): The Syntax of Relative Clauses. 2000. vi, 397 pp. 31 Svenonius, Peter (ed.): The Derivation of VO and OV. 2000. vi, 372 pp. 30 Beukema, Frits and Marcel den Dikken (eds.): Clitic Phenomena in European Languages. 2000. x, 324 pp. 29 Miyamoto, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 2000. xiv, 232 pp. 28 Hermans, Ben and Marc van Oostendorp (eds.): The Derivational Residue in Phonological Optimality Theory. 2000. viii, 322 pp. 27 Růžička, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. x, 206 pp. 26 Ackema, Peter: Issues in Morphosyntax. 1999. viii, 310 pp. 25 Felser, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions. 1999. xiv, 278 pp. 24 Rebuschi, Georges and Laurice Tuller (eds.): The Grammar of Focus. 1999. vi, 366 pp. 23 Giannakidou, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. xvi, 282 pp. 22 Alexiadou, Artemis and Chris Wilder (eds.): Possessors, Predicates and Movement in the Determiner Phrase. 1998. vi, 388 pp. 21 Klein, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. x, 232 pp. 20 Laenzlinger, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and clause structure in Romance and Germanic. 1998. x, 371 pp. 19 Josefsson, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998. ix, 199 pp. 18 Alexiadou, Artemis: Adverb Placement. A case study in antisymmetric syntax. 1997. x, 256 pp.
17 Beermann, Dorothee, David LeBlanc and Henk van Riemsdijk (eds.): Rightward Movement. 1997. vi, 410 pp. 16 Liu, Feng-hsi: Scope and Specificity. 1997. viii, 187 pp. 15 Rohrbacher, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and prodrop. 1999. viii, 296 pp. 14 Anagnostopoulou, Elena, Henk van Riemsdijk and Frans Zwarts (eds.): Materials on Left Dislocation. 1997. viii, 349 pp. 13 Alexiadou, Artemis and T. Alan Hall (eds.): Studies on Universal Grammar and Typological Variation. 1997. viii, 252 pp. 12 Abraham, Werner, Samuel David Epstein, Höskuldur Thráinsson and C. Jan-Wouter Zwart (eds.): Minimal Ideas. Syntactic studies in the minimalist framework. 1996. xii, 364 pp. 11 Lutz, Uli and Jürgen Pafel (eds.): On Extraction and Extraposition in German. 1996. xii, 315 pp. 10 Cinque, Guglielmo and Giuliana Giusti (eds.): Advances in Roumanian Linguistics. 1995. xi, 172 pp. 9 Gelderen, Elly van: The Rise of Functional Categories. 1993. x, 224 pp. 8 Fanselow, Gisbert (ed.): The Parametrization of Universal Grammar. 1993. xvii, 232 pp. 7 Åfarlí, Tor A.: The Syntax of Norwegian Passive Constructions. 1992. xii, 177 pp. 6 Bhatt, Christa, Elisabeth Löbel and Claudia Maria Schmidt (eds.): Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences. 1989. ix, 187 pp. 5 Grewendorf, Günther and Wolfgang Sternefeld (eds.): Scrambling and Barriers. 1990. vi, 442 pp. 4 Abraham, Werner and Sjaak De Meij (eds.): Topic, Focus and Configurationality. Papers from the 6th Groningen Grammar Talks, Groningen, 1984. 1986. v, 349 pp. 3 Abraham, Werner (ed.): On the Formal Syntax of the Westgermania. Papers from the 3rd Groningen Grammar Talks (3e Groninger Grammatikgespräche), Groningen, January 1981. 1983. vi, 242 pp. 2 Ehlich, Konrad and Jürgen Rehbein: Augenkommunikation. Methodenreflexion und Beispielanalyse. 1982. viii, 150 pp. With many photographic ills. 1 Klappenbach, Ruth (1911–1977): Studien zur Modernen Deutschen Lexikographie. Auswahl aus den Lexikographischen Arbeiten von Ruth Klappenbach, erweitert um drei Beiträge von Helene MaligeKlappenbach. (Written in German). 1980. xxiii, 313 pp.