This monograph consists of a revision of the author's doctoral dissertation, with the addition of extensions and related work by the author and collaborators. In this work Grohmann makes the interesting suggestion that along with STANDARD LOCALITY (the familiar upper bound on the length of syntactic dependencies), movement dependencies are subject to a lower bound, that is, ANTI-LOCALITY. This suggestion is implemented through a partition of the clause into three parts (roughly the theta-domain, the agreement domain, and the left periphery) and some simple conditions on movement between and within these domains. The result is a new candidate proposal in the realm of multiple/cyclic spell-out theories that provides a novel analysis of anaphora in several domains, as well as of left dislocation of various types. Although the proposal also has interesting theoretical implications (for example, a less stipulative ban on “improper movement”), it suffers from some empirical and theoretical shortcomings, which I will outline below.

The Anti-Locality thesis introduced in chapter 1 is implemented through the theory of PROLIFIC DOMAINS. According to this theory, the clause consists of three parts, or domains, defined over familiar syntactic categories: (1) the Θ-domain, in which θ-roles are discharged/assigned/checked (material at and below the vP level), (2) the Φ-domain, where agreement/φ-features are checked (material above the vP level including agreement projections (Agr), tense, and aspect), and (3) the Ω-domain, the location of syntactically represented discourse roles, such as topic and focus (a finely articulated left periphery along the lines of Rizzi 1997). Anti-Locality states that movement cannot take place within a single Prolific Domain.

To some extent, Anti-Locality could be viewed as a response to critics of the line of research that seeks to explain syntactic dependencies in terms of movement rather than construal, chains, or accidental binding (Hornstein 2001, Kayne 2002, among others). Since Hornstein (2001) makes crucial use of movement to θ-position, critics might ask why, if such movement is allowed, we do not observe sentences such as 1a with the meaning of 1b, given the possibility of the movement operation in 1c where John is merged as theme and moves to the external argument position.

The Anti-Locality framework provides a principled answer. Movement within a single Prolific Domain (in 1, movement from one theta position to another within the same VP) is too local. If independent evidence for Anti-Locality could be found, this critique of movement into theta position would then disappear. This book attempts to provide such evidence. However, it should be noted that were Hornstein to be wrong, much of Grohmann's proposal would still be tenable.

In chapter 2 the following is introduced as an explicit Anti-Locality condition:

• Condition on Domain Exclusivity (CDE) (p. 78)
• 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 ΠΔ.

Putting aside for the moment what “duplicity” and “drastic” mean, the CDE essentially states that a given object can only appear once in a Prolific Domain. Grohmann attempts to explain this restriction as a ban on ambiguity at the PF interface. A phonological form for a given Prolific Domain must contain unambiguous instructions. If an object appears twice in the domain, as it would if it moves domain-internally, ambiguity arises. However, there is one way to circumvent the ban on domain-internal movement imposed by the CDE: an object can appear more than once in a Prolific Domain if one copy of the object is pronounced differently from the other. Grohmann calls this escape hatch COPY SPELL-OUT. This is, presumably, what is meant by the vague “drastic effect” condition above.

While there is an intuitively appealing aspect to Prolific Domains and the CDE, Grohmann's explanation of the CDE in terms of a PF ban on duplicate spell-out suffers somewhat from his adoption of Nunes' (1995) approach to copy theory and deletion. Nunes argues that, due to the Linear Correspondence Axiom (Kayne 1994), only one copy of a given syntactic object can be spelled out phonologically. Other copies, formed through movement (envisioned as copy+merger), must be deleted. Lower copies are generally deleted since they contain more unchecked features. But if copies can be deleted, it should be simple to circumvent the CDE and allow all sorts of domain-internal movement, as long as all copies (but one) are deleted. Grohmann addresses this concern to some extent by saying that unpronounced copies are not deleted, but ignored. This is presumably due to a spell-out algorithm (unspecified in the monograph) that determines for a given domain where a given phrase with multiple occurrences is spelled out. If this is true, then the adoption of Nunes's approach to copy theory and deletion is only a hindrance. Grohmann's work could be done by a simple remerger account of movement, without the problems introduced by a copy operation and subsequent deletion.

Chapter 2 spells out the architecture of the system within which the Anti-Locality thesis is couched. The introductory material in the first two chapters might have benefited from additional editorial attention, as it may be vague and confusing at times. As such, it will be difficult to follow for readers not already familiar with recent work in minimalism. In fact, some of this material could have been omitted without affecting the book's main proposals, although some material in chapter 2 becomes crucial later in the book. First, Grohmann argues against Chomsky's (1995) proposal that little v is the locus of both external argument licensing and accusative case marking. Grohmann wishes to divorce the former, which is part of the Θ-domain, from the latter, which is a Φ-related property. Second, Grohmann argues against the existence of multiple specifiers. The clausal architecture assumed here allows a single specifier for a given XP (formed either through direct merger or movement) and any number of adjuncts, which appear above the specifier and can only be formed through (external) merger. This approach to specifiers and adjuncts allows an account of certain facts about left dislocation addressed later in the book. However, when this issue is discussed, the reader is left wondering why one would wish to rule out multiple specifiers. Unsupported statements, such as “multiple specifiers are undesirable, do not buy us much empirically, and can be banned from the grammar” (p. 53) are even more puzzling, especially given much recent work on multiple specifiers (see, for example, Richards 1999).

Following Chomsky (2000:116), Grohmann attempts to derive fundamental specifier-head, head-head, and head-complement relations through the use of what he calls the “Natural Relations” provided by merger. However, this attempt does not quite go through. While Grohmann is correct that the combination of SISTER and IMMEDIATELY-CONTAIN, two relations that arguably fall out from the operation Merge, provides a natural account of the specifier-head relation, this same relation should hold between a head and the specifier of that head's complement. Therefore, while Grohmann's approach includes the Spec-Head relation, it might be too permissive. Grohmann also argues for a greed-based over an attraction-based treatment of movement based on these relations. The Spec→Head relation, viewed as Immediately-Contain(Sister-Of (XP)), fits into a greed-based framework since it is the XP in the specifier position that checks features on the head, not the other way around. However, this deduction seems spurious to me. It is based on the assumption that the Mother→Daughter relation (Immediately-Contain()) is fundamental, but the Daughter→Mother relation (Mother()) is not. Furthermore, the discussion is open to Chametsky's (2003:200–201) critique that more “natural relations” are definable from these assumptions than are needed.

Chapters 3 through 6 form the empirical core of the book. Chapter 3 presents a treatment of local anaphora as the result of too-local movement within the Θ-domain, with Copy Spell-out of the lower trace as an anaphor, as in 2, where Copy Spell-out is shown as (→X).

This analysis is only available if movement to a theta position is possible, as it is in a system where theta roles are treated as features to be checked through merger or movement. This treatment of anaphora serves as a proof-of-concept of the CDE and Copy Spell-out in the Θ-domain. It differs from other movement-based analyses of binding (see, for instance, Kayne 2002) in that the anaphor is not present prior to Copy Spell-out. In this chapter, anaphora in exceptional case marking contexts are also treated through Copy Spell-out, but in the Φ-domain.

The PF explanation of the CDE is challenged further if an analogous analysis applies to sentences that contain multiple anaphora, as in John is protecting himself from himself. Here one would expect the CDE to rule out multiple Copy Spell-out of John as himself. Grohmann acknowledges this problem, but provides no clear solution.

Chapter 4 contains perhaps the most interesting empirical case study of the CDE and Copy Spell-out, namely that of left dislocation, including topicalization as in 3a, contrastive left dislocation (CLD) as in 3b, clitic left dislocation (CLLD) as in 3c, and two types of hanging topic left dislocation (HTLD), shown in 3d,e.

According to Grohmann, topicalization, CLD, and CLLD suggest that the left-dislocated XP is derived through movement from a clause internal position, while this is not true of HTLD. Evidence for this, some of it well-known, includes Case connectivity and reconstruction effects (binding of anaphora, bound variable readings, obviation of weak crossover) in the first three, but not in HTLD. Turning to CLD, we can ask how the left-dislocated XP could be moved from a clause-internal position when the resumptive pronoun (RP) is in the prototypical preverbal topic position, presumably moved from a clause-internal position. Grohmann suggests that the RP is the result of Copy Spell-out of a lower occurrence of XP within the Ω-domain, as shown in 4.

Here the dislocated XP moves to topic position (within the left-peripheral Ω-domain), and then moves on to a “quasi-extra-sentential” CP projection. While it is not clear what this CP projection is, this movement is too local and only Copy Spell-out can make the derivation legitimate by the CDE. The assumptions regarding specifiers and adjuncts from chapter 2 allow an account of the ordering among HTLDed and CLDed XPs, where the former always appear to the left of the latter. One appealing aspect of Grohmann's analysis of the demonstrative pronoun in CLD as the spell-out of a copy due to Anti-Locality is that resumptive pronouns now appear when standard (maximum-distance) locality is violated, as with islands for movement, and also when Anti-Locality is violated. This treatment of CLD might also provide an argument against chain formation as a plausible alternative to movement. In this case, one would have to posit a difference between chains in CLD and coreference (presumably also encoded with coindexation) in HTLD.

Chapter 5 returns to CLLD and analyzes this phenomenon as Copy Spell-out in the Φ-domain. Grohmann suggests that the CLLDed XP moves through two positions in an Agreement projection: one adjoined to the Agr head and one in the specifier of this head. The occurrence adjoined to the Agr head is spelled out as a clitic. The XP in [Spec,AgrP] then moves on to a left-peripheral topic position.

This analysis of CLLD is perhaps the book's most problematic part. First, it predicts that all languages with CLLD should have clitic doubling. There is nothing inherent in 5 that would force XP to move on to a topic position, so the derivation should be possible in any clause. However, Italian is a classic example of a language with CLLD but no clitic doubling. Furthermore, since the left-dislocated XP does undergo movement to the Φ-domain (presumably the domain of A-bar movement), the analysis falls to Cinque's (1990) argument that if CLLD involves A-bar movement, it should license parasitic gaps, contrary to fact. The advantage of the CLD and CLLD analyses is that they would explain several observed differences between the two, as well as moving toward an account of why and in which languages these types of left dislocation should occur.

Why does Copy Spell-out emerge as an anaphor in some cases, a pronoun in others, and a clitic in still others? Grohmann comes close to explaining this distinction by appeal to different domains: anaphors arise in the Θ-domain, clitics in the Φ-domain (when the language has Agrrelated clitics), and pronouns in the Ω-domain. However, the analysis of ECM anaphors as arising in the Φ-domain makes this line of reasoning less consistent. Furthermore, the author provides no explanation for why only lower copies undergo Copy Spell-out as a pronoun/anaphor/clitic. Grohmann alludes to a possible feature-based explanation along the lines of Nunes 1995, but, as noted above, Nunes’ assumptions may be problematic for the CDE theory.

Chapter 6 summarizes Grohmann and Haegeman 2003 and extends the Copy Spell-out paradigm to the DP level, where Grohmann argues that the DP contains multiple Prolific Domains (analogous to clausal structure), and that domain-internal movement within the DP also triggers Copy Spell-out. The latter is used to account for pronominal possessor doubling in Germanic, as in German dem Vater sein Auto ‘the father's car’ (lit. the father his car). If movement (and Copy Spell-out) is at play here, Hornstein's (2003) suggestion that movement does not underlie control in DPs (as in John's attempt PRO to leave) may be at risk.

Chapter 7 presents some of the most interesting ideas of the book. Grohmann reviews the long history of research into syntactic cyclicity and develops a view of cyclicity in the Prolific Domains framework in the form of two generalizations.

The combination of 6a and 6b, together with the unnoted assumption that a syntactic object must have an occurrence in each clause between the highest and lower occurrence of that object, ensure a strongly falsifiable type of cyclicity. Moving phrases always target either the next highest clause-internal domain or the same domain in next highest clause. The result is a satisfying account of so-called “improper movement” (the ban on A-bar movement followed by A-movement) without simply stipulating that this type of movement is “improper”.

Grohmann adopts a MULTIPLE SPELL-OUT approach that differs in crucial ways from both Uriagereka's (1999) and the phase-based approach put forth by Chomsky (2000). In the Prolific Domains framework, each domain—Θ (vP), Φ (IP), and Ω (CP)—is submitted to PF and LF computation after it is built. Cyclicity in movement is viewed as a fundamental property of this system, not as an accidental property of intervening heads, as in theories that incorporate the EPP as a trigger for movement.

The principles in 6 predict that raising verbs embedded under control verbs should behave as if they were control verbs, a prediction that Grohmann claims is born out in English. For example, in 7a John is raised clause-internally from its first theta position to the Φ-domain, from where it raises clause-internally to the Φ-domain of the raising verb seem. However, from here John would have to move to the Θ-domain of the control verb hope, a move ruled out by 6b. Therefore, the derivation must be as in 7b, where John moves from Θ-domain to Θ-domain until it reaches hope. This entails that John must be an argument of seem, and so seem must behave like a control verb, a prediction Grohmann claims is born out.

But this prediction does not hold up cross-linguistically. Rizzi (1986) shows that a coreferential anaphor can appear between an NP and a controlled PRO, as in 8a, but not between an NP and its trace position in raising, as in 8b.

The discussion of 7 suggests that when embedded under a control verb (such as sperare), raising verbs (such as sembrare) should pattern with 8a. However, as shown in 9, this is not the case.¹

Further problems arise with this strict view of cyclicity if wh-adjuncts in the Θ-domain must pass through the Φ-domain on the way to their surface Ω-domain position. The question is what type of agreement forces adjunct movement to the Φ-domain since most (all?) languages lack agreement with adjuncts.

As a whole, the proposals of the book are thought provoking and appealing. Grohmann does a good job of motivating his proposals independently, so Anti-Locality does not become an ad hoc patch required by Hornstein's (2001) treatment of control as movement. However, since it adopts many of Hornstein's assumptions, the book is open to criticism along the lines of Landau 2003 and Culicover and Jackendoff 2001. The tripartite division of the clause has intuitive appeal, and Grohmann's treatment of cyclicity within the three-domain system provides an interesting alternative in the realm of multiple spell-out theories. It also provides a satisfying ban on improper movement. While empirical and theoretical problems remain, they are outweighed by the book's virtues and should serve as challenges to those pursuing a theory of Anti-Locality.