No concept in Chomskian generative syntax is more fundamental than ‘syntactocentrism’ (terminology by Jackendoff Reference Jackendoff2003). This book re-examines this notion in the light of recent advances in Minimalism and compares the framework to alternatives that do not embrace the notion. The book starts with an exposé of the concept of syntactocentrism and its manifestation in Chomskyian generative grammar since the 1960s. Chomsky (Reference Chomsky1957: 21–25) noted that finite-state grammars are inadequate to model natural language, due to the fact that recursivity or self-embedding is central to natural language. The earliest attempt to encode this property in natural language was via so-called ‘rewrite rules’ (Chomsky Reference Chomsky1957). Later, Chomsky (Reference Chomsky, Jacobs and Rosenbaum1970) pointed out that lexical insertion introduces a certain redundancy into the system, as categorical and selection information is essentially doubled in the lexicon. Thus, the syntactic computation can be impoverished. This led to the introduction of the X-bar schema (Chomsky Reference Chomsky, Jacobs and Rosenbaum1970). In the 1990s, a further reduction of the system involved the introduction of bare phrase structure, where it was assumed that phrase structure consists of nothing other than features and objects available in the lexicon itself. Chomsky (Reference Chomsky1995) formulated this as the Inclusiveness Condition. The only two syntactic operations allowed are Merge and Move, operating over elements from the lexicon, listed in the numeration of a particular derivation.
In Chapter 2, Trotzke addresses cyclicity. In phonology, he explains, cyclicity is central. The original idea is attributed to Chomsky & Halle (Reference Chomsky and Halle1968), who proposed that stress assignment in English is accounted for by the Nuclear Stress Rule for syntactic concatenations and by the Compound Stress Rule inside words. The application of the rules takes place in accordance with the principle of the ’transformational cycle’, which means that stress assignment takes place from the inside out: to smaller concatenations first, and once the rules applied on a particular hierarchical level, they then apply to groupings on a higher level. The cyclic application of the rules can account for the difference in stress pattern for [[black board] eraser] vs. [black [board eraser]]. The cyclic system has more recently been generalised to apply to determining the strength of phonological junctures in prosodic structure by Wagner (Reference Wagner2005). He shows that in coordinate structures, prosodic boundaries can be determined accurately if one relies on a cyclic (i.e. bottom-up, cumulative) mapping of phonological groupings from the syntactic hierarchy. In a similar vein, Hornstein & Pietroski (Reference Hornstein and Pietroski2009) propose a minimalist view of the semantic computation, which is again cyclic in nature. Rooted in Katz & Fodor’s (Reference Katz and Fodor1963) early proposals, they provide a view of the semantic computation as a bottom-up process, which is compositional in the sense that more complex meaning is created from its parts.
In Chapter 3, the discussion starts with the description of the T-model, where the output of the syntactic derivation is simply read off by the semantic and phonological components, without the latter two having any direct communication. The discussion turns to the description of the cartographic framework within the syntax of information structure, relying on Rizzi’s (Reference Rizzi and Haegeman1997) well-known proposal. A sketch is provided of the translation of Frey’s (Reference Frey1994) analysis of the German left periphery in cartographic terms, using a KontrastP position as a target for elements that appear in the left-peripheral position with a contrastive interpretation, and FinP (Fin = finiteness) for elements that appear in the left periphery without any pragmatic import. There follows a critique of the cartographic approach to information structure. Trotzke cites Horvath (Reference Horvath2010) and Neeleman & Szendrői (Reference Neeleman and Szendrői2004) (but see also Szendrői Reference Szendrői2001, Fanselow Reference Fanselow, Brandt and Fuss2006, among others) for proposing the argument that information structure features, such as [+Focus], violate Chomsky’s (Reference Chomsky1995) Inclusiveness Condition, as they are not lexical features. For this reason they pose conceptual problems for the Minimalist architecture, and if possible, should be dispensed with.
Chapter 4 offers some empirical discussion, after charting the historical development of representations in the Chomskian tradition. This starts with the exposition of the motivation for the elimination of Deep and Surface structure, continues with the notion of the numeration (i.e. the set of lexical items that is the input of a derivation), and concludes with the motivations for multiple, cyclic spellout at phase edges. Trotzke notes that mutiple spellout allows for a certain convergence between the Chomskian model, which no longer limits phonological (i.e. PF) or semantic (i.e. LF) interpretation only at the end of the derivation, and the Jackendovian Parallel Architecture framework (Jackendoff Reference Jackendoff1997), where interfaces exist between the three independent computational modules, syntax, semantics and phonology.
The chapter continues with a summary of Fanselow & Lenertová’s (Reference Fanselow and Lenertová2011) account of subpart of focus movement in German. They argued that utterances like (1b) below do not involve focus movement, but rather – as part of the cyclic linearisation of the syntactic derivation (Fox & Pesetsky Reference Fox and Pesetsky2005) – that a syntactic operation allows fronting the closest accented element to the initial position for pragmatic effect. Pars-pro-toto focus movement (i.e. movement of a part of the focal constituent instead of the whole) would not make sense for parts of idioms.
In addition, they argued, as the ungrammatical (1c) shows, fronting can only target the closest accent-bearing element (i.e FLInte, and not KORN in (1a)).
Trostke calls into question the generalisation that accentuation would be the crucial factor determining the mobility of idiom parts. The example in (2a) below, he argues, has the same accentuation pattern as (1a), with two accented elements.
But unlike (1c), either element can be fronted in the case of the idiom in (2a), as shown in (2b) and (2c), although these important examples are not actually provided by Trotzke. Thus, (2c) would potentially constitute fronting of an accented element over another, contrary to Fanselow & Lenertová’s (Reference Fanselow and Lenertová2011) proposal.
Instead, Trotzke proposes that in idioms ‘where a non-idiomatic reading is not (easily) accessible to the hearer” there is more syntactic flexibility and “the ordering constraint can ... be violated ..., without destroying the idiomatic reading”’ (90). Trotzke offers a test to measure the accessibility/plausibility of the literal meaning: turning the idiom into an imperative (targeting the literal reading) should sound natural for idioms with an accessible literal meaning. These, in turn, should be syntactically less flexible than less accessible idioms, by assumption. Concretely, Trotzke argues that the idiom in (2a) is inaccessible, as apparently the imperative test illustrated in (3a) yields only the idiomatic meaning, according to him. He then cites (3b) to illustrate, that, as expected under his proposal, argument reordering is possible with this idiom.
According to Trotzke, (3a) contrasts with (4) below, where the literal meaning is supposed to be accessible, which in turn is proposed to explain the ungrammaticality of (1c), because an accessible idiom is expected to show more syntactic rigidity.
But there are several problems with this account. First, the possibility of the ordering in (3b) actually provides an alternative base order for the fronted order that is unexpected on Fanselow & Lenertová’s (Reference Fanselow and Lenertová2011) account, i.e. (2c). If (3c) is possible, then it can be the underlying order for (2c), so in fact no accented element needs to be moved over another accented element in (2c), and thus (2c) is not a counterexample to Fanselow & Lenertová’s account. Note that the corresponding alternative base order is not grammatical for the idiom in (1), so the unavailability of (2c) is still explained.
Second, it is not obvious that the literal meaning is actually more accessible or plausible in (3b) compared to (4). According to what measure are we supposed to determine accessibility and plausibility? Individual judgement? Why should these coincide across speakers on such subtle matters? Should we not expect considerable speaker variation then with respect to the syntactic flexibility of such idioms? Given these issues, I conclude that Fanselow & Lenertová’s account is superior to Trotzke’s proposal regarding subpart of focus fronting of idioms.
The remainder of Chapter 4 addresses certain island violations in syntax and argues that they constitute a case of pragmatic influence on syntax. The idea is that such opacity effects arise ‘when a derivation layer contains two syntactic objects whose deictic interpretation clashes at the syntax–discourse interface’ (99). Trotzke goes on to state that this explanation rests on the assumptions that ‘(i) unbounded dependencies are placed in discourse-prominent slots; (ii) constituents conveying new information allow extraction most easily (see Erteschik-Shir 1973 et seq.); (iii) elements occurring later in the string (usually the object) are canonically more likely to be interpreted as focus, whereas earlier constituents (usually the subject) are canonically interpreted as topics’ (99). But Trotzke provides no motivation for these assumptions, or any evidence why they might hold, and no explanation how the island effects actually follow from these assumptions, so it is impossible to evaluate this part of the proposal in more detail.
Chapter 5 compares the Minimalist proposal that assumes cyclic spellout to Jackendoff’s Parallel Architecture and to Cognitive Linguistics, as two alternatives to the syntactocentric view of Minimalism. I will only concentrate on the first of these, in the interests of space. (The second comparison argues that Minimalism and Cognitive Linguistics share substantial common ground with respect to language evolution, and the interested reader can peruse the details of this in the book.) After introducing the main tenets of Jackendoff’s (Reference Jackendoff1997) system, Trotzke proposes that once cyclic spellout is accepted into the Minimalist grammar, there is in fact no syntactocentrism. It is not clear what he means by this. One way to understand this would be in the sense that syntactocentrism disappears because later syntactic steps of a particular derivation can be dependent on the PF or LF representations of the previous cycle. This, however, does not seem to be Trotztke’s interpretation. As the discussion on page 117 reveals, rather, it seems that he states that Minimalism (despite the presence of cyclic spellout and cyclic linearisation) must nevertheless give up syntactocentrism in that it must allow direct communication between PF and LF to be able to account for the contrastive interpretation of contrastively accented phrases. But then direct PF–LF communication is not an inherent property of the Minimalist system or of cyclic linearisation: it is an additional assumption that is forced on us by the need to provide an account for empirical data. In other words, it seems that syntactocentrism in this area simply cannot be maintained, direct PF–LF communication must be allowed, or alternatively, the Inclusiveness Condition must be violated by the presence of a syntactic feature [+Focus] whose role is to encode the accent–focus correspondence. But this state of affairs does not reflect an ‘amended notion of syntactocentrism’ (134), as Trotzke puts it; rather, it is a mighty ‘throw of our gun into the grain’: the admission that as far as the prosody–focus correspondence is concerned, the Jackendovian architecture is empirically more adequate than the Chomskian syntactocentric view.
