This linguistic textbook by Peter Culicover provides a broad introductory overview of various topics in the study of syntax. Its major objectives are to show how natural language makes use of various syntactic and morphosyntactic devices, and to lay out the conceptual structures that correspond to particular aspects of linguistic form (xi). The title of the book is thought-provoking: rather than Introduction to syntax, Syntactic theory or the like, Culicover chose the title Natural language syntax, which can be interpreted twofold. First, it may be regarded as a direct response to recent research in syntax, especially the Minimalist program (Chomsky Reference Chomsky1995), emphasizing that linguistic theory needs to place greater emphasis on accounting for actual language data rather than indulge in purely formalistic investigation. Another possible reading of the title is that the syntactic theory proposed in this book is natural, intuitive and simple, and that the title alludes to Culicover's previous book Simpler syntax (Culicover & Jackendoff Reference Culicover and Jackendoff2005). In general, the discussions in this book are framed in theory-neutral terms, supplemented with comparisons between Culicover & Jackendoff's ‘simpler syntax’ (SS) approach and so-called mainstream generative grammar (MGG).

Chapter 1, ‘Overview’, defines syntax as ‘the system that governs the relationship between form and meaning in a language’ (1); and states that the goal of linguistic theory is ‘to understand what the properties of human languages are, and why they are that way’ (3). Culicover compares the SS and MGG approaches for the treatment of displacement. Instead of pursuing a derivational analysis, as in MGG, SS directly specifies ‘correspondence rules’ between syntactic positions and meanings.

Chapter 2, ‘Syntactic categories’, provides a detailed description of various syntactic and morphosyntactic categories, and introduces the notational device of an attribute value matrix (also employed in Lexical Functional Grammar and Head-driven Phrase Structure Grammar), which displays the features and values of lexical items. To Culicover, a lexical item expresses the correspondences between mophosyntactic categories, semantic values and phonological representation, as exemplified in (1).

1. (1)

   

SS and MGG also differ with respect to the interpretation of the claim that ‘there is a small, finite number of universal categories that all languages choose from’ (48). Instead of following Chomsky (Reference Chomsky, Jacobs and Rosenbaum1970) in decomposing lexical categories into binary features ([±N], [±V]), Culicover holds that universal categories can only be semantically defined (e.g. by means of basic semantic constructs such as objects, actions, space, and properties), and that syntactic categories do not necessarily match perfectly with semantic categorization. It follows that grammatical theory cannot be ‘syntactocentric’ (Jackendoff Reference Jackendoff1997, Reference Jackendoff2002), that is, it cannot be defined with recourse only to syntactic categories. Instead, it is a language-specific correspondence rule between syntactic and semantic/conceptual categories that determines syntactic structure.

Chapter 3, ‘Basic sentential structure’, introduces notions such as phrase structure and constituency tests, as commonly found in syntax textbooks. The chapter includes a section on grammatical functions (e.g. subject, object) and how they are marked morphologically, which presents data from a number of different languages. Culicover claims that syntactic structure is ‘flat’, unless there is strong empirical evidence to the contrary. For instance, double object constructions have a flat V(erb)P(hrase), in which the main verb selects the two object noun phrases simultaneously.

Chapter 4, ‘Phrasal categories’, discusses the syntactic structure of various phrasal categories. An obvious departure from most other syntax textbooks is that only the ‘weak version’ of X′-theory is adopted here, i.e. phrase structure is endocentric, recursive and ‘as flat as possible consistent with the evidence’ (106). This return to pre-Principles and Parameters phrase-structural notions is echoed in Culicover's revival of Chomsky's (Reference Chomsky1975) classic analysis of English verbal morphology. Culicover allows the possibility of adopting a ‘strong version’ of X′-theory that makes a structural distinction between specifiers, heads, complements, and adjuncts, noting that this will ‘result in greater complexity … [with respect to] … processing, language acquisition and language change’ (105–106). As Culicover shows, strong X′-theory can be evoked in the analysis of projections such as C(omplementizer)P, I(nflection)P and D(eterminer)P.

Chapter 5, ‘Conceptual structure and the lexicon’, constitutes the core of the book. The chapter introduces the notion of ‘Conceptual Structure’ (CS), which mediates between a linguistic expression and its reference in the actual world. CS ‘reflects a mentalistic approach to meaning’ (140): Culicover is interested in the question of how grammar is construed within the mind of speakers, and ‘how the syntactic structure of an expression determines its conceptual structure’ (141). Concepts, according to Culicover, are the basic primitives of grammar whose combination gives rise to complex concepts in a systematic way. A simple CS is linked to a lexical item's phonological and syntactic representation by a language-specific correspondence rule, as is shown by the representation of book and the idiom kick the bucket in (2) and (3).

1. (2)

   

1. (3)

   

CS also encodes the thematic structure of the verb. Assuming that CS, syntactic representation and grammatical functions (GFs) constitute three distinct representational levels, Culicover contends that the constructs of each of the three tiers are licensed by the default linking rules. For instance, the agent corresponds to the subject, and the patient to the object, as schematically represented in (4).

1. (4) Mary writes a book.

   

Languages differ in the way the linking rule between thematic roles and grammatical functions is stated, as well as in the correspondence rules that relate GF and syntactic representation. Whether a given expression is grammatical depends primarily on whether all parts of the tripartite structure are properly matched, that is, whether the correspondence is complete.

Chapter 6, ‘Argument correspondences’, applies Culicover's theory of linking CS to grammatical functions and syntactic structures to much-studied syntactic phenomena, such as passives, applicatives, dative shift, antipassives, dummy subjects, null pronouns, and clitics. In passives, the CS argument that corresponds to the logical subject is suppressed (if expressed, it becomes an oblique argument). Applicatives and dative shift involve the same mechanism, which renders the default correspondence between direct object/direct argument and indirect object/oblique argument inapplicable. For dummy subjects, while there exists a correspondence between the syntactic and grammatical subject, there is no corresponding argument within CS that relates to either of the other two tiers. Null pronouns and clitics usually involve a morphological correspondence between the verbal morphology and the semantic features of a CS argument, akin to the phenomenon in English, whereby plural verbal agreement is used with scissors even if scissors is not overtly expressed (e.g. if one picks up a pair of scissors, one can utter the sentence These are really sharp). Other formal features of nouns, such as person and gender, can also initiate verbal agreement. In this chapter, Culicover also provides a detailed discussion of the approaches of SS and MGG to passives.

Chapter 7, ‘Complex clauses: Raising and control’, illustrates how Culicover's correspondence theory accounts for control and raising constructions. For control structures, Culicover argues that the complement is an infinitival VP rather than a full clause and that ‘the argument of the complement that corresponds to Subject is not syntactically realized … [but] … coindexed with the CS argument that corresponds to the subject of the higher verb’ (241).

1. (5)

   

For raising constructions, Culicover argues that raising predicates such as appear or seem select a clausal complement. In his non-transformational analysis, the matrix subject does not correspond to a CS argument of the matrix verb; instead the CS argument of the complement that corresponds to the subject GF of the infinitival complement is linked also to the subject GF of the raising verb (244).

1. (6)

   

The chapter provides a detailed comparison of Culicover's approach with that of MGG. Demonstrating the pitfalls of analyzing control constructions in terms of structural configurations (i.e. by making reference to c-command, the minimal distance principle, etc.), Culicover concludes that the special properties of verbs with respect to control are straightforwardly accounted for ‘if we take a lexical approach to control’ (281).

Chapter 8, ‘Predication’, considers secondary predicates (e.g. Kim went to bed hungry), resultatives (e.g. Kim hammered the metal flat), and small clauses (e.g. John considered Kim innocent). Analogous to control predicates, Culicover analyzes secondary predication as involving the secondary argument having the same referential index as the matrix subject. The various secondary predication types are accounted for by postulating different CS representations. For instance, the CS of the resultative verb hammer is HAMMER (X, Y^α; RESULT: PROPERTY (α)) (304), while read in Sandy read the newspaper naked has the CS representation READ (X^α, Y; WHILE: PROPERTY (α)) (304), in which a particular property (here, ‘naked’) corresponds to the sentential subject during the whole reading process (i.e. WHILE).

Chapter 9, ‘A′ constructions’, considers A′-dependencies, as in, for example, wh-questions, topicalization and relative constructions, clefts and pseudo-clefts. In Culicover's theory, wh-questions involve a question operator at the level of CS that takes a predicate within its scope. No wh-movement is assumed: the clause-initial wh-word in the syntactic representation, which bears a referential index, is related to the operator at the level of CS, which then suffices to interpret the identity of the filler gap. The difference between wh-moving and wh-in-situ languages rests upon the language-specific correspondence rule. Further in this chapter, Culicover addresses the transformational treatment of A′ constructions (e.g. in terms of overt/covert movement and feature checking) and issues that arise in such analyses (e.g. constraints on movement, barriers, and subjacency).

Chapter 10, ‘Coreference and binding’, discusses coreferential relations, which, in Culicover's theory, arise from the fact that arguments bear identical referential indices at the CS level. For example, the sentences Mary swims and Mary runs and Mary swims and she runs have the CS representations given in (7).

1. (7)

   1. (a) SWIM (AGENT: MARY^α) & RUN(AGENT:MARY^α)

   2. (b) SWIM (AGENT: MARY^α) & RUN(AGENT:α )

Culicover introduces the set of binding principles employed in the Principles and Parameters framework, and goes on to argue that binding conditions cannot be adequately stated at the syntactic level, but must be formulated at the CS level and the GF level. While CS-binding holds between arguments at CS, GF-binding holds between arguments that correspond to GFs, see (8)–(10).

1. (8) α CS-binds β if and only if

   1. (a) α and β are coindexed;

   2. (b) α, β are contained within the scope of a CS relation R.

      (416, ex. (54))

1. (9) α locally CS-binds β if and only if

   1. (a) α CS-binds β;

   2. (b) α, β are arguments of a CS relation R.

      (416, ex. (55))

1. (10) α GF-binds β if and only if

   1. (a) the GFs corresponding to α and β are coindexed;

   2. (b) the GF of α is higher in the hierarchy than the GF of β;

   3. (c) the GFs correspond to the same clause.

      (416, ex. (56))

Moreover, Culicover postulates the two conditions in (11), which regulate the distribution of anaphors and pronouns.

1. (11)

   1. (a) An anaphor must be locally GF-bound.

      If it corresponds to a CS argument, that argument must be CS-bound.

      (417, ex. (62))

   2. (b) Pronouns cannot be locally CS-bound.

      (418, ex. (65))

The way in which these conditions can account for ungrammatical structures is exemplified in (12).

1. (12)

   1. (a) *Himself_i loves John_i. (anaphor is not GF-bound)

   2. (b) *Mary expects that John_i will excuse herself_i.

      (anaphor is not locally CS-bound)

   3. (c) *John_i told Mary about him_i. (pronoun is CS-bound)

Chapter 11, ‘Fragments’, presents an analysis of the syntactic properties of sentential fragments, focusing on bare argument ellipsis (BAE) and related constructions, such as sluicing, VP-ellipsis and gapping. Culicover reiterates the view espoused in Culicover & Jackendoff (Reference Culicover and Jackendoff2005: chapter 7) that BAE (e.g. A: John has been drinking something; B: Yeah, scotch) does not receive a satisfactory syntactic account under a deletion approach (i.e. Yeah, John has been drinking scotch). His argument is that BAE involves a pragmatic discourse relation to the antecedent rather than a copy of the antecedent. The discourse relation can be described at the level of CS via indirect licensing. That is, in Sandy can't speak German, but Robin can, indirect licensing is established between the target and the fragment, as illustrated in (13).

1. (13)

   

Given the above correspondence relation, the fragments are indirectly licensed as if they appeared in the larger syntactic structure. The arguments of the fragments then substitute their correspondents at the CS level, as is shown in (14).

1. (14)

   
   (418, ex. (48))

Culicover justifies his approach empirically by listing a number of examples that show that fragments can only be related to the target pragmatically and not syntactically.

This book covers almost all basic concepts of syntax that a syntax student needs to grasp. Culicover provides a commendable outline of the historical developments in syntactic theory, while at the same time introducing his own, non-mainstream generative theory, founded on Culicover & Jackendoff (Reference Culicover and Jackendoff2005) and related work on conceptual semantics (e.g. Fodor Reference Fodor1975, Jackendoff Reference Jackendoff1983), which he believes to be more empirically grounded. Each topic of discussion is fully illustrated with attested examples from a number of different languages, and each chapter ends with a list of further research questions. At first blush, it is hard to see whether Culicover succeeds in rendering syntax ‘simpler’ by shifting the explanatory burden to the level of conceptual structure. To settle this question, more empirical studies are needed, as well as a clearer understanding of what conceptual structure refers to (the definition of CS as ‘a central system of the mind’, found in Culicover & Jackendoff (Reference Culicover and Jackendoff2005: 20), is of only limited usefulness). To conclude, Culicover's innovative syntactic analyses in this book forcefully demonstrate that many years of research from the generative enterprise, led by Noam Chomsky, have provided the background to almost all syntactic frameworks on the market, and that ‘the old wisdom’ remains a valuable resource, which we cannot afford to ignore.