Nine original contributions make up this rich and well-edited volume. Six chapters deal with constraints on wh-movement of subjects or from subject position, two chapters study the positions targeted by wh-movement, and one chapter considers the categorial properties of complementizer words.
E. Phoevos Panagiotidis's well-written and succinct ‘Introduction: Complementizers and their phase’ is both a useful overview of the Complementizer Phrase (CP), from its introduction in Chomsky's (Reference Chomsky1986)Barriers to its incarnation as a phase in contemporary Minimalism, and a presentation of the theoretical issues taken up in the various contributions.
Luigi Rizzi's ‘On some properties of criterial freezing’ is the opening chapter. It extends the author's (2006) perspective on chain delimitation to criterial freezing effects on movement at Logical Form (LF) – as, for example, the non-topicalizability of the second wh-phrase in an English multiple wh-construction – and to the non-extractability of subjects.
Rizzi elaborates the basic idea behind the subject criterion, starting from the claim that the functional backbone of the clause contains an obligatory SubjectP, which encodes a criterial relation corresponding, in the default case, to aboutness topicality. Phrases attracted to the specifier of SubjP are blocked from further movement by criterial freezing. In English, objects but not subjects can be extracted from a that-clause. According to Rizzi, subject extraction but not object extraction is blocked by criterial freezing, since Universal Grammar has a subject criterion but no object criterion (see also Rizzi & Shlonsky Reference Rizzi, Lai-Shen Cheng and Corver2007). The criterial approach to subject extraction restates the EPP requirement (‘clauses must have subjects’) and the ECP perspective on subject extraction (‘traces must be properly governed’) in terms of an interface condition, subject to economy constraints.
The final section of this chapter briefly discusses subextraction from subjects. The idea here is that if criterial freezing is limited to the feature probed by Subj0, it is predicted that material pied-piped along with the subject XP be amenable to further movement. The data supporting this prediction is drawn from the realm of combien ‘how many/much’ extraction in French. The prediction is only partially met, however. First, the relevant example (from Obenauer Reference Obenauer1976), is (perhaps too generously) marked with ‘?’.
(1) ?Combienveux-tuque[[ __ de personnes]viennentàtonanniversaire]?how.manywant-youthatof peoplecometoyourbirthday‘How many do you want that of people come to your birthday?’
If subextraction were unhindered in (1), its grammaticality status should be the same as that of a sentence with extraction from object position. It is not. Second, Gallego's contribution to the volume under review, discussed below, cites Spanish data from Uriagereka (Reference Uriagereka1988), which illustrate a clear contrast between subextraction from preverbal subjects (ungrammatical) and postverbal subjects (grammatical). The Spanish facts point to the opposite conclusion from Rizzi's. Matters are compounded when one takes into account the observation, discussed in Kotzoglou's chapter, that in Greek, subextraction is licit both from preverbal and from postverbal subject position. One might argue that these cases are not comparable, that is, that in French, a quantifier is extracted from a Determiner Phrase, in Spanish, a post-nominal possessor phrase and in Greek, a pre-nominal possessor. But if Stepanov (Reference Stepanov2001) is right, then subextraction from subjects varies cross-linguistically in ways that remain mysterious, and Rizzi's point remains moot.
George Kotzoglou's ‘(Non)-extraction from subjects as an edge phenomenon’ argues that licit violations of the Subject Condition in Greek (illustrated, for example, by the grammaticality of ‘Whose did the patience persuade you?’ in Greek) are not due to a relaxation of the conditions on extraction but rather on the conditions on copy spellout. Greek arguably has two positions for subjects, SpecvP and SpecTopicP. Kotzoglou proposes a principle according to which only one silent copy of a moved constituent is licit per phase. Since subjects never move to or through SpecT(ense)P, the single unpronounced copy of the subextracted element in the CP phase appears in the base position of the subject in SpecvP. If movement proceeded via SpecTP, as it presumably does in English and in other non-pro drop languages, there would be two unpronounced copies of the subextracted element in the CP phase, in violation of his proposed Condition on Copy Reduction. On the face of it, however, Kotzoglou's principle is too strong, as it rules out stepwise movement of subjects from SpecvP to SpecTP, as demonstrated by agreement on multiple verbs in, for example, Bantu compound tense constructions (Carstens Reference Carstens2001), the distribution of floated quantifiers in English, or the myriad positions that subjects can occupy with respect to adverbs (Cinque Reference Cinque1999).
There are two parts to Ángel J. Gallego's ‘Subextraction from phase edges’. In the first part, he argues that illicit subextraction from subjects follows from a ‘morphological effect related to maximal phi-feature checking in SPEC-T’ (52), that is, from the Activity Condition. In the second and more original part (Section 4.3), Gallego discusses subextraction from SpecCP. The data on which this discussion is based is, admittedly, murky, with significant unexplained cross-linguistic variation. Nevertheless, the author develops an interesting perspective on the familiar Spanish example from Torrego (Reference Torrego1985) (in English: (?)‘Of which author don't you know which translations have won international prizes?’), suggesting that ‘of which author’ is not extracted from inside the constituent ‘which translations of which author’ in the embedded CP, but rather from a proleptic Prepositional Phrase merged above the embedded CP – which he labels an ‘aboutness’ phrase.
Anna Roussou's chapter, ‘Subjects on the edge’, is a very rich study of the relationship between complementizers and clausal subjects. Her point of departure is that elements such as to, that, and for are nominal (see Manzini's contribution for a development of this idea). To is a locative element in the lower left periphery of the clause, which then associates with a variable corresponding to the EPP. That requires an overtly realized EPP-slot within its complement clause. The that-trace effect is a consequence of the fact that once the subject is extracted, that fails to take the right complement and the result is ungrammatical, unless that is also absent. Languages other than English utilize a variety of strategies for the realization of the EPP to circumvent the ban on subject extraction.
Chapter 6 is Clemens Mayr's ‘On the necessity of phi-features: The case of Bavarian subject extraction’. Mayr argues for a link between complementizer agreement and subject extraction. He shows that in Bavarian, subjects can be extracted from a finite clause headed by an overt complementizer; hence there is no subject–object asymmetry in wh-movement in Bavarian and no that-trace effect. Mayr's account of this observation exploits the manifestation of phi-agreement on C, arguing that agreement connects the subject wh to the projection line on which the probing head is located. Since agreement on the TP level is rendered invisible by spellout of C's complement, it must be reinstated at the CP level so as to connect the wh-expression to the projection line.
Brazilian Portuguese hyper-raising – raising of the subject of a finite clause over an overt complementizer (as in ‘The boys seem.3pl that traveled.3pl’) – and its relation to apparent hyper-raising (‘The boys seem.3pl that they traveled.3pl’) is the subject of Ana Maria Martins & Jairo Nunes' ‘Apparent hyper-raising in Brazilian Portuguese: Agreement with topics across a finite CP’. The gist of their proposal is well summarized in the conclusion:
[H]yper-raising structures arise when an embedded finite T fails to assign Case to its subject position, allowing the latter to enter into an agreement relation with the matrix T. Given that in B[razilian]Portuguese] topics can independently agree with T, a possibility arises that matrix T agrees with an embedded topic. (163)
Rita Manzini's ‘The structure and interpretation of (Romance) complementizers’ argues that the Romance che-type complementizer is not a verb-related functional category, but an independent nominal head, which satisfies the argument slot of the matrix verb and which takes the embedded sentence as its complement. Both as a wh-phrase and as a complementizer, Italian che introduces a variable. If it introduces an individual variable, the wh-phrase reading arises; if it introduces a propositional variable, it is read as a so-called complementizer. In a somewhat similar vein, Roussou argues in her contribution that the English complementizer that is the demonstrative that. Jointly, these two chapters would seem to suggest that there are no dedicated lexical items for the C head. This is a tantalizing idea, which has empirical support in Romance and Germanic. However, there are many languages in which complementizers resemble neither nouns, demonstratives or prepositions.
Omer Preminger's ‘Nested interrogatives and the locus of wh’ re-assesses Reinhart's (Reference Reinhart, Belletti, Brandi and Rizzi1982) claim that Hebrew has two positions for wh-elements in CP, thus explaining the absence of wh-island violations in the language. Assuming that the final landing site of Hebrew wh-words is SpecFocusP, Preminger argues that Hebrew can simultaneously exploit a higher ‘escape hatch’ position through which wh-words can move. Both positions can host only a single phrase. Thus, wh-movement out of an indirect question is licit because the escape-hatch position can serve a second wh-phrase. However, if both wh-phrases need to move out of an embedded declarative, an island effect ensues.
The final paper in this volume is Jeroen van Craenenbroeck's ‘Complex wh-phrases don't move: On the interaction between the split CP hypothesis and the syntax of wh-movement’. The author summarizes some familiar and some less familiar asymmetries between simple (e.g. what) and complex (e.g. which boy) wh-expressions, arguing that the latter are merged in a high position in CP (as specifiers of a head bearing the [Q] feature) and are associated with a null operator which moves to a lower specifier position (perhaps Rizzi's SpecFocP, although the author does not commit himself to this label). Among the asymmetries that follow from this analysis are the following:
(i) the absence of superiority violations with complex wh;
(ii) the positional differences (with respect to various C heads) between simple and complex wh-expressions in a number of Dutch dialects;
(iii) the possibility of swiping in sluicing with simple but not with complex wh-expressions;
(iv) the possibility of merging a demonstrative to the right of a sluiced simple wh-word, but not a complex one, in some Dutch dialects (a phenomenon dubbed ‘spading’ by the author).
The analysis is elegant but, as the author himself notes, it does not square with the evidence that complex wh-expressions manifest reconstruction effects.
Collections of articles suffer from an inherent defect in that the theoretical and methodological assumptions of the individual contributions are not, and cannot be fully compatible with one another. Often, the incompatibility reflects a theoretical debate, the nature of which emerges upon a comparative reading of the papers. This is well exemplified in the present volume by the tension between the notions of phase and phase edge, on the one hand, and the detailed maps drawn by the cartography approach, on the other. Chomsky's Minimalism works with a relatively simple structure, with a single v and a single C. Under cartography, there is no C as such, but an expanded left periphery with a number of hierarchically ordered categories. Wh-phrases, according to Chomsky, move to the edge of CP (its outer specifier), but which one of ForceP, IntP or FocP constitute the edge? In the absence of an integrated theory, incorporating the insights of the research agendas of Minimalism and cartography, the theoretical contributions of the papers constituting this volume are difficult to evaluate.