Andrew Carnie's Constituent structure is a very timely addition to Oxford University Press's series of surveys, appearing, as it does, at a point when numerous themed international conferences demonstrate the heightened interest both in the formal properties of syntactic theories and in diverse approaches to capturing constituency.
The book is organised in three parts: part 1, ‘Preliminaries’ (chapters 1–4); part 2, ‘Phrase structure grammars and X-bar theory’ (chapters 5–7); and part 3, ‘Controversies’ (chapters 8–11). The brief first chapter, ‘Introduction’, demonstrates the hierarchical organisation of clauses and previews the book's content. Chapter 2, ‘Constituent structure’, reviews classic arguments against an account of constituency based on simple concatenation and shows that regular grammars (finite state automata) are similarly inadequate. The chapter then briefly discusses constituency tests and the role of compositionality. Chapter 3, ‘Basic properties of trees: Dominance and precedence’, introduces tree diagrams as representations of constituent structure, and the basic formal concepts used to describe them. The primitive relation of dominance is used to formulate different axioms that rule out structures which are usually assumed to have no linguistic relevance, such as, for example, multiply rooted trees. The relation of precedence is defined, in a top-to-bottom fashion, in terms of the primitive relation of sister precedence. Chapter 4, ‘Second order relations: C-command and government’, deals with the motivation for and the origins of the c-command relation and considers different attempts to derive its effects.
The first chapter of part 2, chapter 5, ‘Capturing constituent structure: Phrase structure grammars’, presents Chomsky's (1957) interpretation of the structuralist system of immediate constituent (IC) analysis in terms of his phrase structure (PS) grammar formalism. After taking the reader through a number of derivations to show, for example, that in Chomsky's earliest work a derivation is a set of strings generated through successive application of rewrite rules and that trees are constructed from such derivations, Carnie discusses an important conceptual shift regarding trees and PS rules. Initially, trees were a convenient expository device with limited theoretical status, and grammatical constraints and transformations were formulated directly over the strings that make up a derivation. In later literature, two alternative views emerged. One view interpreted PS rules as well-formedness conditions on tree structures, not as derivational instructions. The second view held them to be ‘projection rules’ (87), that is, bottom-to-top instructions on what category must dominate a given pair of sister categories. Following a summary of Chomsky's original arguments for transformations, chapter 6, ‘Extended phrase structure grammars’, outlines how complex categories can be used in unification-based frameworks, such as Generalised Phrase Structure Grammar (GPSG), to capture long-distance dependencies that had previously received transformational analyses. Carnie then shows how syntactic representations are linked with semantic structures in GPSG and Lexical-Functional Grammar (LFG). This chapter also contains a brief discussion of the decomposition of PS rules into separate precedence and dominance rules which was proposed in the early 1980s in both GPSG and LFG. Chapter 7, ‘X-bar theory’, develops an extended argument for a particular version of X-bar theory, using examples from English and drawing heavily on Carnie (Reference Carnie2007). The chapter concludes with an account of the development of X-bar theory up to the late 1980s.
Part 3 of the book begins with chapter 8, ‘Towards set-theoretic constituency representations’. First, Carnie discusses proposals that X-bar-theoretic projection levels be derived from the structural context of a node, considering the implications for adjunction. He then turns to Kayne's (Reference Kayne1994) antisymmetry hypothesis and Chomsky's (Reference Chomsky and Webelhuth1995) Bare Phrase Structure theory. After a brief summary of Phillips's (Reference Phillips2003) argument that certain constituency tests suggest that derivations should proceed from left-to-right, rather than the familiar bottom-to-top, Carnie argues that Irish data showing movement of complex phrasal constituents to a head position support the claim that bar-levels are not primitives. Chapter 9, ‘Dependency and constituency’, focuses on approaches that do not rely on purely structural primitives, namely LFG, Categorial Grammar, Tree Adjoining Grammar (which is introduced as a ‘variant on Categorial Grammar’, 183), Relational Grammar, Dependency Grammar, Head-driven Phrase Structure Grammar, Role-and-Reference Grammar and Cognitive Grammar/Construction Grammar, describing each of these approaches in between one and five pages. Chapter 10, ‘Multidominated, multidimensional, and multiplanar structures’, deals with proposals that do not assume that constituency is represented by a single, single-rooted, linearly ordered tree structure. Different accounts of free word order and movement phenomena are explored. The final section of this chapter considers models in which constituency is represented by several more or less independent levels of representation. Finally, chapter 11, ‘Phrasal categories and cartography’, discusses the substantive content of structures, which sets it apart from the other chapters. Carnie argues that ‘almost every major constituency-based theory – whether formalist or functionalist – seems to have converged on the idea that the “backbone” of clauses consist of at least three major parts’ (221), which correspond to the levels of the Verb Phrase (VP), Inflection Phrase (IP), and Complementiser Phrase (CP) in Principles-and-Parameters (P&P) syntax. After a thorough discussion of the evidence for and against the universality of a VP constituent, the chapter considers each of the three clausal domains, (mostly) from a P&P perspective.
It should be clear from this overview that this book covers a lot of ground. However, while the book addresses a wide range of approaches to constituency, it does not provide a balanced comparison of the different theoretical frameworks that are mentioned at various points (and in varying detail). One way of reading this book is as an interesting study of the development of X-bar theory, its implementation in P&P and the theoretical and technical issues and proposals that arose in this context. Chapter 9, the only chapter devoted (mostly) to approaches outside the Chomskyan PS grammar tradition, could be dropped from the book without greatly affecting its internal coherence – although this would of course have an impact on the book's intended function as a survey of different approaches to constituent structure.
Yet, Carnie's X-bar-centred approach proves illuminating. The historical perspective adopted by the author draws attention to the remarkable similarity of the structural representations assumed at different stages of the theory's development. Representations from the earliest PS-rule-based phrase markers to the most recent Merge-derived structures share (most of) the formal properties enshrined in the standard axioms discussed in chapter 3 (see also Partee et al. Reference Partee, ter Meulen and Wall1990), most notably, the non-tangling condition, which rules out crossing lines, and the exclusivity condition, which excludes two nodes from being ordered through both dominance and precedence. (These assumptions are shared by many approaches outside the P&P tradition.) A PS grammar is a system designed to generate structures that have these properties, reflecting the structuralist IC analyses of the first half of the twentieth century, and subsequent developments within P&P theory inherited these properties (see Matthews Reference Matthews1993 on the continuity of IC-analysis and X-bar theory). Given the standard definition of constituent as ‘a set of [terminal] nodes exhaustively dominated by a single node’ (37), an important consequence of the inherited axiom that lines in a structure cannot cross is that there can be no discontinuous constituents. Chametzky (Reference Chametzky2000: 71f.) presents an argument that this particular prediction is both suspect on conceptual grounds and empirically questionable. He argues against precedence as a basic ordering relation alongside dominance, and Carnie himself discusses empirical arguments for discontinuous constituents in detail in chapter 9. Note that discontinuous constituents are in fact compatible with an earlier definition of constituent in the book, whereby constituents are understood to be ‘groups of words that function as units with respect to grammatical processes’ (18). The discussion concerning discontinuous constituents should remind us that the basic assumptions of a theory are ultimately empirical hypotheses and are therefore subject to review. A first step towards reviewing a theory's assumptions is knowing what they are, and Carnie's accessible presentation makes this considerably easier.
A clear understanding of a theory's formal properties may also shed light on its relation to other theories. Moving beyond Carnie's discussion, recall that PS grammars and their descendents were designed to generate structures that meet the exclusivity condition (EC). However, the EC (as well as the closely related assumption of a distinction between word-level and phrase-level categories) is also an empirical hypothesis, derived from IC-based systems of analysis. Thus – and this is particularly apparent for systems that do not use PS rules to generate structures – there is no a priori reason why representations should meet this condition. If the EC is dropped, representations can contain words that dominate other words, even if they are also linearly ordered relative to them. Structures of this kind are commonly used in Dependency Grammar approaches, which suggests that one of the major divisions between dependency-based and PS-based models of constituency is the EC. In fact, unrelated to the EC, Carnie's overview of Dependency Grammar briefly considers one system within the P&P tradition (Brody Reference Brody2000) whose representations violate the EC (178).
The previous paragraphs have illustrated two lines of investigation that familiarity with the content of this book and its sources can facilitate. In the remainder of this review, I will briefly comment on the material covered in the book and on editorial aspects. Obviously, space is limited in a book, but the central role of constituency tests in syntactic argumentation would make it appropriate to go beyond the discussion of standard textbook examples in chapter 2 and the arguments for intermediate projections in chapter 7. It would have been useful to see constituency tests applied to languages other than English as well as a more detailed discussion of the significance of situations in which tests either produce conflicting results or underdetermine the structure of a given string (see, for example, Miller Reference Miller1992 on tests for the constituent structure of noun phrases in English and French). On a related note, the book's ability to serve as a ‘bridge builder’ across different approaches would have been enhanced by a more systematic discussion of how different types of evidence for constituent structure are handled in the various non-X-bar-theoretic approaches, highlighting their respective strengths and weaknesses. Similarly, how do approaches that do not have a c-command relation deal with apparently c-command-based phenomena?
The book is written with a clarity and style that many readers may recognise from the author's successful textbook (Carnie 2007). There are just a few occasions, such as the discussion of basic X-bar theory in chapter 7, where it might have been appropriate to adjust the presentation to the more advanced target audience of the series in which this book appears. The book generally provides plenty of references to primary and secondary sources. Rare exceptions are the section on ‘Functional equations, f-structures, and metavariables in LFG’, which contains no references to the LFG literature whatsoever, and the lack of any references to work in Relational Grammar. The number of typos – over 150 – is surprisingly high for an Oxford University Press publication (and readers are recommended to check the author's website for a list of errata and information on a possible corrected printing). Often the typos are mainly cosmetic (missing italics on cited words, spelling errors, wrong numbering of examples, etc.), but there are also many more substantive errors (missing brackets in definitions and in Categorial Grammar derivations are particularly frequent). The book also includes some plain mistakes. For example, in a section explaining the relation of immediate precedence (41), R is said to immediately precede S in a tree corresponding to [M [N Q R] O [P S]]. Another mistake occurs in a discussion of how a particular definition of c-command (‘[n]ode A c-commands node B if neither A nor B dominates the other and the first branching node dominating A dominates B’; 53, ex. (20)) is affected by the properties of the dominance relation. Here the argument that, assuming reflexive dominance, a branching node would not c-command the nodes contained within its sister constituent is followed by the incorrect statement that the node ‘would only c-command its own daughters’ (53). Such errors are unfortunate if the book is to be used as a reference tool. It is easy to imagine students who are not yet used to finding mistakes in esteemed scholarly publications struggling to understand a definition against the accompanying prose.
Such editorial glitches notwithstanding, the book is clearly a great resource, with its main strength being the creation of a historical context for current P&P approaches to constituency, and libraries should have a copy. Most readers will learn something new from this book, and I think that overall Carnie succeeds in his aim to ‘give both the beginning syntactician and the experienced old hand pause to think about the nature of the relationship between words and the representation of these relationships in terms of constituent structure’ (7).
