Language

This volume is a revised version of Thomas Grano’s doctoral dissertation from the University of Chicago. The book comprises six chapters, along with a brief introduction and a conclusion. The introduction lays out the main proposal: ‘that restructuring is cross-linguistically pervasive and that, in virtue of its co-occurrence with some control predicates but not others, it evidences a basic division within the class of complement control structures’ (1). The division between restructuring and nonrestructuring predicates is related (originally by Wurmbrand 1998) to whether a predicate allows partial control or only exhaustive control (in the sense of Landau 2000). G associates this correlation with other empirical splits in different languages (‘the distribution of finite complementation in English, the availability of overt embedded subjects, temporal properties of controlled complements, as well as phenomena found in Mandarin Chinese and Modern Greek’ (1)) and then adopts Cinque’s (1999, 2004) cartographic approach to clausal structure to explain all of them. By taking all exhaustive control predicates to be functional verbs in a Cinquean hierarchy (and so all restructuring structures to be monoclausal), G accounts for all of the facts just mentioned: only exhaustive control predicates cause restructuring, they do not allow [End Page 729] finite complementation or embedded subjects, and so forth. The gist of the analysis is shown in the trees below.

(1)

(2)

In trying to show how restructuring could apply to languages as diverse as Chinese and Greek (besides Romance, where restructuring is more evident), the book is an excellent exercise in linguistic theorizing, solidly based on empirical data, and it works out important semantic details. The book is therefore an important reference in the studies of control and complementation.

Ch. 2 lays out the proposal of exhaustive control as functional restructuring. As evidence, G provides four ways in which exhaustive control predicates differ from partial control ones: they do not take finite complementation (in English and Romance, at least), their complements do not license an overt subject, they sometimes do not entail anything about their subjects, and their complements are transparent for licensing polarity items. To explain these properties, G sides with Cinque (2004), taking all exhaustive control predicates to be realized as functional projections, which would mean that restructuring is obligatory with those verbs. Rigid ordering effects (Cinque 2006) are attested with exhaustive control predicates in many languages, so Cinque’s hypothesis is plausible. The monoclausal hypothesis for exhaustive control structures, assumed by G, clashes with observations by Landau (2013:69–78), who maintains that control complements are always clausal and always include PRO. Although Landau shows convincingly that some nonfinite complements are clausal, he does not in my opinion make a good case for exhaustive control complements. Landau’s chief arguments for biclausality in the context of exhaustive control complements are the presence of nonfinite ‘complementizers’ in several languages (G mentions Romance di/de/a; Landau mentions Dutch om and Hebrew me) and the form of complements in Balkan languages. G argues that, since clitic climbing may obtain even in the presence of a putative complementizer in Italian (apud; Napoli 1981:863–64), the complementizer analysis of such particles is wrong, and he assumes, with Cinque (2006:45), that those apparent complementizers are functional heads in the extended projection of VP. G then looks at English verbal morphosyntax (in particular the behavior of to in ellipsis) and argues that the position of to is still elusive. One problem for G’s analysis is the supposition that all exhaustive control verbs (including try and manage) are functional restructuring predicates (the subject, therefore, raising over them on its way to T); the problem is that try and manage do not seem to be raising predicates by most tests.

Ch. 3 solves the problem just mentioned by assuming that ‘exhaustive control predicates contain as part of their meaning a variable that must be bound in the syntax: when the subject raises, it obligatorily binds this variable, giving the predicate semantic access to the subject’ (43). The rationale behind the proposal is that, when a verb is realized in a functional category (provided it has the right kind of meaning), if this verb has an individual argument, it is converted into a dependent variable. Since the subject raises to Spec T, it may bind the variable introduced by any functional verb that appears below T in Cinque’s hierarchy, but not those that are above T. This explains why some predicates restructure while others do not: restructuring takes place with predicates that match the meaning of a functional projection below T, where its dependent variable may be bound by the subject. By contrast, if the meaning of a verb corresponds to a functional projection above T, restructuring (i.e. realization of the verb at the functional category) would leave the dependent variable free, causing the derivation to crash. That is probably the best argument in G’s favor: it explains why restructuring takes place and why it takes place with that specific [End Page 730] group of predicates. However, G’s analysis further requires one to accept some key aspects of Cinque’s (1999, 2004) proposal, which can be controversial. Ch. 3 also proposes a semantics for the verb try in which it is a function from properties of events to properties of events.

Ch. 4 deals with the mixed properties of the verb want and its crosslinguistic kin, which usually support partial control but tend to also be able to restructure—unexpectedly, according to G’s logic. The chapter argues that want, besides its normal use as a restructuring predicate, may combine with a silent predicate HAVE, which (when taking a clausal complement) may allow partial control readings. The chapter also discusses the semantics of HAVE.

In Ch. 5, G investigates the temporal properties of controlled complements in English, building predominantly on Wurmbrand 2014, and argues that only the complements of partial control predicates may contain a T node. Then, Chs. 6 and 7 extend the analysis to two languages that, at first sight, seem to create problems for the proposal: Chinese and Greek. G shows that the debate about whether Mandarin has a finite/nonfinite distinction is easily reinterpretable as reflecting a biclausal/monoclausal split, thus supporting his overall claim that exhaustive control predicates actually constitute monoclausal (raising) structures. Greek offers the reverse problem, since all control structures seem to be biclausal. Following Giannakidou (2009), G argues that, since the particle na in Greek is also used in veridical complements to predicates meaning ‘see’ and ‘start’ (besides nonveridical/subjunctive complements), na structures may correspond to structures smaller than clauses, such as bare infinitives and gerunds, in which na occupies a position equivalent to de/di and other ‘prepositional complementizers’. According to G, a morphological syncretism between nonpast tense and tenselessness in Greek causes the appearance of a lack of nonfinite complementation. G then shows that Greek exhaustive complements also pass tests for monoclausality (polarity licensing and inverse scope readings), which corroborates his analysis. Finally, Ch. 8 offers a conclusion.

Although G explains the presence of nonpast tense in Greek nonfinite complements, his analysis leaves the presence of number and person agreement quite open. In order to go through, the analysis must dissociate agreement features from T, which is supposedly absent in exhaustive control complements. One would have to assume that some lower projection (probably AspP) may contain agreement features, which are then valued by the subject before it moves to the controller position. The analysis of Greek is probably the most unconvincing part of the book, especially when compared to Landau’s (2015) theory of control by predication under exhaustive control predicates.

Relevant to this discussion is the use of inflected infinitives in Brazilian Portuguese (BP), described in Modesto 2010, 2016. Until recently, BP was considered to be one of the Romance languages in which restructuring did not take place at all, since there is no clitic climbing in BP. However, applying tests involving scope and licensing of polarity items, Modesto 2016 shows that all exhaustive control predicates in BP behave differently from partial control ones, in a way that seems to indicate the possibility of restructuring. In addition, nonfinite inflection is usually excluded from exhaustive control complements, but allowed in partial control complements (with either an exhaustive or a partial control reading). If, unlike in Greek, agreement features are barred from being generated in Asp in BP, the lack of T in exhaustive control complements would explain why nonfinite inflection is not used, giving empirical support to G’s analysis.

G’s analysis has a lot of explanatory power. It explains why restructuring effects obtain only with exhaustive control predicates: because only those give rise to monoclausal structures; temporal dependence and the lack of partial control readings are trivial consequences of the fact that the controller and controlled positions are members of the same A-chain in a monoclausal structure. It also predicts the behavior of the exhaustive control class of predicates in any language (a testable hypothesis). Additionally, the availability of partial control with the other class of control predicates becomes a fully general property of PRO, due to its status as a bound pronoun. However, G’s proposal requires some extra syntactic and theoretical machinery (the universality of Cinque’s hierarchy, and the assumption that it is more economical to realize those predicates in functional projections than as lexical verbs, for example). Also problematic is the fact that G does not address the behavior of exhaustive object control predicates, which do not seem to restructure. Such problems, instead of undermining the analysis, may serve to stimulate further research, [End Page 731] and I strongly recommend G’s book to anyone interested in the syntax and semantics of control.